Potent organo-osmium compound shifts metabolism in epithelial ovarian cancer cells.

Potent organo-osmium compound shifts metabolism in epithelial ovarian cancer cells JessicaM.Hearna,b,Isolda Romero-Canelónb,AlisonF.Munroc,YingFub,d,AnaM.Pizarrob,e,MathewJ.Garnettf, UltanMcDermottf,NeilO.Carragherc,andPeterJ.Sadlerb,1 aWarwickSystemsBiologyCentre,UniversityofWarwick,CoventryCV47AL,UnitedKingdom;bDepartmentofChemistry,UniversityofWarwick,Coventry CV47AL,UnitedKingdom;cEdinburghCancerResearchCentre,UniversityofEdinburgh,EdinburghEH42XR,UnitedKingdom;dGeorgetownUniversity MedicalCentre,Washington,DC20057;eInstitutoMadrileñodeEstudiosAvanzadosenNanociencia(IMDEANanociencia),CampusUniversitariode Cantoblanco,28049Madrid,Spain;andfWellcomeTrustSangerInstitute,CambridgeCB101SA,UnitedKingdom EditedbyVivianWing-WahYam,TheUniversityofHongKong,HongKong,China,andapprovedJune8,2015(receivedforreviewJanuary23,2015) The organometallic “half-sandwich” compound [Os(η6-p-cymene) PE01andPE04(SIAppendix,SIMaterialsandMethodsandFig. (4-(2-pyridylazo)-N,N-dimethylaniline)I]PF is 49× more potent S1)andinvivoinacolorectalxenograftmodel(15).Compound 6 than the clinical drug cisplatin in the 809 cancer cell lines that 1wasalsoscreenedin809cancercelllinesaspartoftheSanger we screened and is a candidate drug for cancer therapy. We Institute’sCancerGenomeProjectwithpromisingresults(mean investigate the mechanism of action of compound 1 in A2780 GI 0.75 μM, cf. 36.7 μM for cisplatin, SI Appendix, Fig. S2). 50 epithelial ovarian cancer cells. Whole-transcriptome sequencing Activity comparisons are made with cisplatin because it is the identifiedthreemissensemutationsinthemitochondrialgenome most widely used metal-based drug in the clinic. The MOA of of this cell line, coding for ND5, a subunit of complex I (NADH compound 1 is not associated with activation by hydrolysis and dehydrogenase) in the electron transport chain. ND5 is a proton DNA binding, but instead appears to involve redox mediation pump,helpingtomaintainthecouplinggradientinmitochondria. (11, 13). We use a systems biology approach to investigate the The identified mutations correspond to known protein variants MOA of compound 1, using pathway analysis to identify the (p.I257V,p.N447S,andp.L517P),notreportedpreviouslyinepithelial sourceofredoxmodulation.Wecomplementwhole-transcriptome ovarian cancer. Time-series RNA sequencing suggested that os- sequencing with temporal proteomic (reverse-phase protein mium-exposed A2780cells undergoa metabolicshuntfrom gly- microarray) and phenotypic (high-content imaging and flow colysistooxidativephosphorylation,wheredefectivemachinery, cytometry) profiling technologies to demonstrate how com- associated with mutations in complex I, could enhance activity. pound 1 may exploit mitochondrial deficiencies detected in Downstream events,measuredbytime-seriesreverse-phase pro- A2780 cells. tein microarrays, high-content imaging, and flow cytometry, showedadramaticincreaseinmitochondriallyproducedreactive Results oxygen species (ROS) and subsequent DNA damage with up- We used time-series RNA sequencing to determine differential regulationofATM,p53,andp21proteins.Incontrasttoplatinum gene expression in A2780 EOC cells in response to treatment drugs, exposure to this organo-osmium compound does not cause withalowdoseofcompound 1.Thiswascomplementedwith significant apoptosis within a 72-h period, highlighting a different reverse-phaseproteinmicroarrays(RPPA)tostudycellularlevels mechanismofaction.Superoxideproductioninovarian,lung,colon, ofkeyproteinsinvolvedinDNAdamagerepairandflowcytometry breast,andprostatecancercellsexposedtothreeotherstructurally relatedorgano-Os(II)compoundscorrelatedwiththeirantiprolifer- Significance ative activity. DNA damage caused indirectly, through selective ROSgeneration,mayprovideamoretargetedapproachtocancer therapyandaconceptfornext-generationmetal-basedanticancer Platinum-based metallodrugs are the most widely used anti- drugsthatcombatplatinumresistance. canceragents.Theirreducedeffectivenessafterrepeatdosing (resistance) constitutes a major clinical problem. We study a organometallicanticancercompound | cancermetabolism | mitochondria | potentorgano-osmiumcompoundwithimprovedactivityover transcriptomesequencing | proteinmicroarrays cisplatinandnocross-resistanceinplatinum-resistantcancers. This compounddisrupts metabolismin A2780human ovarian Platinum (Pt) compounds are the most widely used drugs for cancercells,generatingreactiveoxygenspeciesanddamaging DNA. We identified mutations in complex I of the electron cancer chemotherapy, including epithelial ovarian cancer transport chain in A2780 cells and suggest that the osmium (EOC). Although clinical response rates to Pt drugs can exceed compound may exploit these mutations to exert a potent 60% in late stage patients, relapse and the subsequent de- mechanismofaction.Suchactivityincreasesselectivitytoward velopment of resistance leads to treatment failure and low sur- cancer cells, given that normal-functioning cells can better vivalrates.Assuch,thereismuchinterestinthediscoveryofnew adapttodrug-inducedmetabolicperturbations.Therefore,this metal-based drugs with alternative target sites and mechanisms ofaction(MOA)towhichresistancecannotreadilydevelop(1–6). report highlights a promising strategy to drive thefuturede- velopmentoforganometallicanticancercompounds. Osmium,alsoathird-rowtransitionmetal,offerspromisefor distinctmetal-basedanticanceragents(7).Itexhibitsantiproliferative activityinavarietyofoxidationstates(8–10),particularlyOs(II) in Authorcontributions:J.M.H.,I.R.-C.,A.F.M.,A.M.P.,N.O.C.,andP.J.S.designedresearch; J.M.H.,I.R.-C.,A.F.M.,Y.F.,M.J.G.,U.M.,andN.O.C.performedresearch;J.M.H.,I.R.-C., organometallic half-sandwich arene compounds (11). There is A.F.M.,Y.F.,M.J.G.,U.M.,N.O.C.,andP.J.S.analyzeddata;andJ.M.H.,I.R.-C.,A.F.M.,Y.F., particularinterestinOsinvolvementincellularoxidativestress, A.M.P.,M.J.G.,U.M.,andP.J.S.wrotethepaper. potentiallyamultitargetedmechanismforwhichresistancecould Theauthorsdeclarenoconflictofinterest. behardertodevelop (12). ThisarticleisaPNASDirectSubmission. Compound1[Os(η6-p-cym)(NMe 2 -azpy)I]PF 6 (Fig.1A)exhibits Datadeposition:ThesequencesreportedinthispaperhavebeendepositedinArrayExpress higher in vitro antiproliferative activity in A2780 ovarian cancer (accessionE-MTAB-2758). cells, compared with cisplatin, and is equipotent in cisplatin- 1Towhomcorrespondenceshouldbeaddressed.Email:p.j.sadler@warwick.ac.uk. resistantEOCcells(13,14).Itisalsohighlyactive(EC :0.85–2.14 50 Thisarticlecontainssupportinginformationonlineatwww.pnas.org/lookup/suppl/doi:10. μM) in patient-derived high-grade serous ovarian cancer cells 1073/pnas.1500925112/-/DCSupplemental. www.pnas.org/cgi/doi/10.1073/pnas.1500925112 PNASEarlyEdition | 1of6 SULPSANP YRTSIMEHCOIB A 50 Compound Contro * 40 * 30 20 10 0 0 4 12 24 48 and high-content imaging to investigate activation of oxidative usedanextensiveliteraturedatabase(IngenuityKnowledgeBase) stress[productionofreactiveoxygenspecies(ROS)]andapoptosis. to integrate biological and chemical pathway perturbation in- formation (www.ingenuity.com/products/ipa). Of the 20,713 RNA Sequencing and Mitochondrial DNA Mutations. To study the DEGs, 14,384 were mapped, and ateach time point we exam- globaleffectofcompound1,A2780EOCcellswereexposedtoa inedthedownstreamcanonical pathways activated/inhibited. GI concentration(150nM),over48h.At0,4,12,24,and48h; 50 whole-cellRNAwassequencedforbothcontrol-andcompound Oxidative Stress. Some of the most prominently activated path- 1-exposedcellswith∼30million50-bppaired-endreadsgenerated wayswereinresponsetooxidativestress.Genesassociatedwith persample(SIAppendix,TablesS1andS2).Sequencereadswere a vitamin C antioxidant response (P value 1.58 × 10 −3 at 24 h, mapped to the hg19 (GRCh37) human genome (ArrayExpress hypergeometictest)andanNRF2(nuclearfactorerythroid-derived accessionE-MTAB-2758). 2-like 2) oxidative stress response (P value 3.74 × 10 −17 at 4 h) Mutations in mitochondrial DNA (mtDNA) are well docu- wereactivated. mented in cancer cells (16). Therefore, before exploring differ- TheNRF2transcriptionfactor(encodedbytheEFE2L2gene) entialgeneexpression,weinvestigatedmitochondrialchromosome isinvolved inanimportantantioxidant responsepathway,bind- (ChrM)sequencereadsinA2780cells.Thishighlightedmultiple ing to the hARE (human antioxidant response element) cis- mtDNA mutations in all samples, with ≥99% homology in elementtotransactivatedetoxifying/antioxidantgenes.TheAP-1 ∼1,200–2,000 counts. The MT-ND5 gene, coding for the ND5 complexes C-FOS/C-JUN and FRA-1/c-JUN, encoded by FOS, membrane-bound subunit of complex I (CI) in the electron FOSL1, and JUN genes, compete with NRF2 for binding to transport chain (ETC), had three missense mutations at posi- hAREtoactivateantioxidant genes(23). tions m.13106A > G, m.13677A > G, and m.13887T > C (SI Fig. 2A shows some of the components of this oxidative re- Appendix, Fig. S3). These mutations translate to amino acid sponse pathway with transactivation of eight response genes. substitutionsatp.I257V(isoleucine>valine),p.N447S(aspargine> UpondetectionofROS,NRF2dissociatesfromactinandKEAP1, serine), and p.L517P (leucine > proline). p.I257V (dbSNP variant translocates to the nucleus, and interacts with cofactor proteins rs2853501), p.N447S, and p.L517P are known variants of ND5; (e.g.,MAF)tobindtohARE.Fig.2BshowstheDEG(FDR≤ however,theyhavenotbeenreportedinovariancancer(17).The 0.10)ofcomponents inthispathway. ND5subunitisresponsibleforprotonpumpingfromthematrixto NRF2(NFE2L2)andoneofitscofactors,MAF(MAF),were theintermembranespaceofthemitochondria;therefore,mutations up-regulatedat4handremainedsothroughoutthetimeseries, inCImachinerymayimpactcellularrespiration.Pointmutationsin showingamaximumlevelofDEat48h(logFC1.57,FDR1.17× CI machinery have been found in breast, colon, prostate, thyroid, 10 −17 and LogFC 1.18, FDR 2.57 × 10 −19, respectively). Of the andEOC;however,mostreportedmutationsfocusontheD-loop twoAP-1complexes,onlyFRA-1isup-regulatedthroughoutthe ofChrM(18–22). timeseries,showingsignificantDEafter4h(LogFC1.51,FDR 1.54 × 10 −10). C-FOS (FOS), which is part of the second AP-1 Differential Sequence Mapping to the Mitochondrial Chromosome. complex,issignificantlydown-regulatedafter12h(LogFC−0.73, Weexplored the differential sequence mapping across the time FDR4.08×10 −3),suggestingthatthisAP-1complexdoesnotplay series. Fig. 1B shows a heat map of the mean number of reads aroleincompound1-inducedoxidativestressresponse(23). mappedtochromosomes1−22,X,andMforeachsampletype: Fig.2Bshowsthatonlyaselectionofantioxidantgenesdown- controlat0handcontrolandcompound1-exposedat4,12,24, stream of the transcriptional activators are up-regulated. De- and 48 h. The mean read counts for each sample type are rep- toxificationgenesUGT(UDP-glucuronosyltransferase)andGST resentedaslog FPKM(fragmentsperkilobasepermillionreads). (GST)andantioxidantgenesCAT(catalase)andEPHX(epoxide 10 The log FPKM values for mapping to ChrM were significantly hydrolase) were all down-regulated in response to compound 1. 10 higher than those of the other chromosomes, a result of copy- Thismay suggest that the C-FOS/C-JUN AP-1 complex controls numbervariationsandlengthbiasbecauseChrMismuchshorter activationofthesegenesandnotNRF2norFRA-1/C-JUN. thantheotherchromosomes(∼16,600bp). GSR (glutathione reductase), MRPs (multidrug resistance- However,Fig.1CandSIAppendix,Fig.S4,showasignificant associatedproteins),andNQO[NADP(H)quinoneoxidoreductase] differentialmappingtoChrMbetweencontrol-(blue)andcompound were up-regulated after 12 h, highlighting the delay between 1-exposed(green)samples.Thisisparticularlyevidentat12and24h, transcription/translation of NRF2 and AP-1. Although these wheretherearesignificantlymorereadsmappedincompound genes are activated after 12 h, they remain expressed at low 1-exposedversuscontrolsamples(P=0.030and0.016,respectively), levels, even after 48 h. suggestingthat compound1 inducesamitochondrial response. We confirmed production of ROS in A2780 cells exposed to compound1byflowcytometry(Fig.2CandSIAppendix,Table DifferentialGeneExpressionandPathwayAnalysis.Weinvestigated S4).Fig.2CshowsROSmeasurementsafterexposureto150nM thecell-wideeffectsofcompound1bystudyingdifferentialgene of compound 1 for 24 h. The FL1 channel detected total ROS, expression (DE) across a 48-h RNAseq time course. We found including peroxides, peroxynitrites, and hydroxyl radicals, and significantDEafterjust4hexposure,withthelargestnumberof the FL2 channel superoxide levels. SI Appendix, Table S4, lists differentially expressed genes (DEG) at 48 h (SIAppendix, Fig. thepercentagesofcellspopulatingeachquadrant.Fig.2Cshows S5). Pathway analysis with Ingenuity Pathway Analysis software thatcompound 1 inducesapproximately thesamelevel ofROS )301x( sdaer dezilamroN Control 50 XX MM 11 2 3 4 40 2222 5 21 6 30 20 Chr map 7 19 8 20 18 9 17 1 6 11 1 0 10 15 14 13 12 0 0 4 12 24 48 Duration (h) 4 3 2 1 0 B C Compound 1 Fig.1. Differentialmappingofsequencereadsto themitochondrialchromosome.(A)Chemicalstruc- tureofthecationincompound1(theanionPF 6 −is not shown). (B) Heat map showing chromosomal Chromosome mappingofsequencereads(log10FPKM).Frominside map out(arrow),valuesarerepresentedasthetriplicate meanforeachsampletype:0h,control-exposed;4, 12,24,and 48h,control- and compound-exposed. (C)Graphshowingtotalnumberofreadsmappedto 44 L loo gg 11 0 0 FF P P KK MM 00 t e h x e po C se h d rM sam (F p P le K s M a ) cr f o o ss r t c h o e n t t i r m o e l- se a r n ie d s. c G o r m ay po li u n n e d = - mean;*P<0.05(Welchtwo-samplettest). 2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1500925112 Hearnetal. Neggative control 1044 KKEeAapP11 ROS NRF2 1022 CytC osy ot lo sol Nucleus 1000 FL1 (Total ROS) NRF2 100 102 104 GGS G SR S R R PPoossiittiivvee ccoonnttrrooll Compound 1 M N M Q RM NR OQ PR O P P 104 HNOTQX-1NO UHGHMTO -O1 X1 UGT MMAaFf C-FOS FRA-1 C G UA SC TG TA TT 102 NRF2 CJ-JuUnN CJ-JuUnN EC G PG E HA SP STX HT T X hARE EPHX 100 100 102 104100` 102 104 as the positive control (pyocyanin), with 95% of the cell pop- foundontheinnermitochondrialmembraneandisresponsiblefor ulationintheFL1+/FL2+quadrantafterexposuretocompound detoxificationofmitochondrialsuperoxide(O ·− )producedinthe 2 1. This highlights the poor antioxidant response in A2780 cells matrix by increasing the influx of protons and reducing electron after24h,especiallygiventhatNFE2L2,FOSL1,andC-JUNwere leakage(28).Up-regulationofUCP2hasbeenshownduringdrug- allup-regulatedafter4handcontinuedtoincreaseupto48h. induced oxidative stress, and when functional, can drastically re- We hypothesized that the induction of superoxide might be ducethelevelsofmitochondrialO ·− .Thecontinualup-regulation 2 directly correlated to the antiproliferative activity of compound of thisgene (SI Appendix, Fig. S8) could suggest that this protein 1. To confirm this, we measured the production of ROS and cannotcontroltheexcesslevelsofO ·− .SIAppendix,Fig.S8,also 2 superoxide in A2780 cells exposed to three other Os(II) com- shows down-regulation of nuclear-encoded mitochondrial genes, poundsstructurallyrelatedtocompound1,butwithawiderange particularly OGDH (2-oxoglutarate dehydrogenase) and GPX8 ofantiproliferative activities: compound 2 ([Os(η6-p-cym)(azpy) (glutathione peroxidase 8). GPX8 is another antioxidant protein, Cl]PF ; IC >100 μM), compound 3 ([Os(η6-bip)(Cl-azpy)Cl]PF; which catalyzes the reaction of glutathione with HO as a de- 6 50 6 2 2 IC >100μM)andcompound4([Os(η6-p-cym)(OH-impy)I]PF ; toxificationstep.GPX8 wasdown-regulated after 12 h, suggesting 50 6 IC =30±2μM)(SIAppendix,Fig.S6andTableS5).Wealso thatH O detoxificationisdisrupted. 50 2 2 compared compound 1 with the less active compound 4 in five The mitochondrion is the major source of superoxide pro- different cell lines of ovarian, lung (A549), colon (HCT116), duction in the cell, particularly from CI. Superoxide dismutase breast(MCF7),andprostate(PC3)origin(SIAppendix,Fig.S7 (SOD) normally converts superoxide to water and hydrogen and Table S6). In all cases, the level of superoxide induced peroxide,whichisthenbrokendownbycatalase(CAT)towater correlated with the observed antiproliferative activity, further and oxygen. Our analysis showed that, even with detection of highlightingROSproductionasanimportantaspectoftheMOA ROS,neitherCATnorSODgeneswereup-regulatedafterexpo- in this series of Os(II) compounds. Importantly, the activity of suretocompound1(Fig.2B).Infact,CATwasexpressedathigher thesecompoundsislinkedtosuperoxideconcentrationandless levels in control samples, highlighting the oxidatively stressed na- totheother ROS.Thisfurther supportstheinvolvement ofthe ture of A2780 cells and that no response to H O was mobilized 2 2 mitochondria, which are responsible for the majority of super- afterexposuretocompound1.Generally,boththeantioxidantand oxideproduction insidecells. detoxifying systems work in synergy to remove damaging ROSin cells. The down-regulation of key response genes suggests that MitochondrialDysfunction.Wepreviouslyshowedthatcompound compound1bothinducesROSproductionandreducestheability 1perturbsthepolarizationofthemitochondrialmembraneafter ofA2780cellstorespondtoexcessiveROS. just4 hofexposure(24).Here, inaddition tothe regulation of antioxidant pathways, compound 1 also differentially regulated DNA Damage Response. We studied the downstream effects of genesrelatedtomitochondrialdysfunction(P=7.78×10 −15at oxidativestressandROSproduction,giventhatROScantarget 4 h). SI Appendix, Fig. S8, shows the DEG (FDR ≤ 0.10) of many biological components within the cell, including DNA. nuclear-encoded mitochondrial genes. There is an increase in Zeptosens RPPAs were used to study the cellular levels of key DE through the time series, with a large number of DEGs at proteins involved in DNA damage repair (DDR) in compound 48h,in contrast to the mitochondria-encoded genes with maxi- 1-exposed cells versus controls (29). RPPA measures the abun- mumDEat24h(Fig.1B). dance of total protein levels and phosphorylated proteins using Up-regulation of SNCA (α-synuclein), MAOA (monoamine epitope-specific fluorescently tagged antibodies (SI Appendix, oxidase), and UCP2 (mitochondrial uncoupling protein 2) sup- Table S3). Fluorescence measurements for each protein were portsmitochondria-specificROSproduction.SNCAisprimarily normalizedtoahouse-keepingproteinineachsample(prohibitin) located in the cytoplasm, but can localize in mitochondria and before the ratios of fluorescence between compound 1-exposed be involved in controlling mitochondrial morphology and cell and control samples were determined (4, 24, 48, and 72 h), to death (25). MAOA is found in the outer mitochondrial mem- quantify relative fluorescence intensity (RFI) values relative to brane and produces H O from deamination of dietary amines vehicle(DMSO)control. 2 2 (26). It is generally implicated in Parkinson’s disease; however, Fig.3AdescribesthelocationofRPPA-measuredproteinsin recentstudies showan overlapwith cancer(27).Significant up- DDRpathways,wherecomponentswithanasterisk(*)represent regulation of MAOA, particularly at 48 h, highlights a route to phosphorylated proteins (SI Appendix, Table S3) (30). Fig. 3B HO productionbythemitochondria,detectedinFig.2C.UCP2is shows RFI values for proteins in Fig. 3A; for RFI < 1, protein 2 2 K e a p 1 N R F 2 M a f 1 F R A - c-F o s Ju n G S R M R P N Q O H O -1 U G T C A T G S T E P H X )edixorepuS( 2LF A B C --11..55 LLooggFFCC 11..55 KKeEapA1P 1 NNRFFE22 L2 MMaAfF FFROAS-1L 1 cF-FOOSS JJuUn N GGSSRR MMRRPP NNQQOO 1 HHOM-1O X1 UUGGTT CCAATT GGSSTT EEPPHHXX 4 h 12 h 24 h 48 h Fig.2. NRF2-andAP-1–mediatedoxidativestressresponseinA2780cells.(A)DiagramshowingthedifferentiallyexpressedgenesateachtimepointintheNRF2 oxidativestressresponsepathwayfromthetranscriptomeanalysis.(B)TheDEGofeachbiologicalcomponent[asfoldchange(FC),FDR<0.10]issplitintoexpression valuesat4,12,24,and48h,fromlefttoright.(C)FlowcytometrymeasurementsofROSafterA2780cellswereexposedto150nMofcompound1for24h. Hearnetal. PNASEarlyEdition | 3of6 SULPSANP YRTSIMEHCOIB fortheDEGsrelatedtoapoptoticcelldeath.Caspasesassociated withapoptosisincludecaspase-3,-7,and-9asdownstreammarkers and caspase-8 as an upstream marker of extrinsic apoptosis (SI Appendix, Fig. S9). Fig. 4A shows significant down-regulation of CASP-9after4h,CASP-6,-2,and-9after12h,andCASP-3and-7 after48h.Genesencodinginhibitorsofapoptosis(IAP)proteins, which inhibit the activation of caspase proteins, like BIRC2 and BIRC3,wereup-regulatedafter4h(LogFC2.06,FDR1.53×10 −12). BIRC5, which encodes survivin, another IAP, was down-regulated until 24 h (LogFC −0.43, FDR 1.03 × 10 −5), suggesting specific modulationofapoptoticsignalsthroughcaspaseinhibition. However,Fig.4Ashowsup-regulationofBID[LogFC0.90(48h), FDR1.77×10 −14]andBAX[LogFC0.50(24h),FDR3.46×10 −6], which both code for apoptotic initiators, BID from extrinsic ac- tivationand BAXfromintrinsicactivation.Thissuggeststhat apo- ptotic signals are activated in response to compound 1, but that thesesignalsdonottranslatethroughtoexpressionofeffector caspases. The only caspase genes up-regulated were CASP-1 and CASP-12, inflammatory caspases, that were activated by IL-1β. ThisislikelycontrolledbythehighlevelsofIAPproteinsat24and 48hthatdisruptthecaspasecascadeandinhibitapoptosis. We measured the levels of 18 apoptotic proteins by RPPA levels were lower in compound 1-exposed cells versus the con- over 72 h (Fig. 4B) at concentrations of 150 and 450 nM of trol;forRFI=1,levelsdidnotchange;andforRFI>1,protein compound 1. Fig. 4B shows the RFI values for 13 proapoptotic levels increased. The RFI values for p21* and ATM* were not markers and 5 prosurvival markers (SI Appendix, Table S3). At measuredat72h. both concentrations, nearly all of the apoptotic markers were DuringcellularresponsetoDNAdamage,ATM,amongother down-regulated,bothpro-andanti-apoptotic.However,caspase-3 proteins,ismobilizedandauto-phosphorylatedtoATM*,which (CASP3) and cleaved caspase-3 (CASP3*) levels increased be- inturnphosphorylatesCHK2toCHK2*.Atthispoint,theresponse tween 24 and 48 h before dropping again at 72 h. PARP is pathway branches into a rapid response via CDC25A and a cleaved by caspase-3* (PARP*) and showed similar levels to delayedresponseviap53andp21.Fig.3Bshowsthatcompound1 thoseofCASP3*—lowat4and72handhigherat24and48h. causesasignificantup-regulationofATM*,p53*,andp21*between Theonlysignificantlyup-regulatedproteinsthroughoutthetime 24and48h,highlightingthedetectionofDNAdamage. series were XIAP and survivin, both inhibitors of caspase-3, -7, and-9activation (32). CellDeathMechanisms.Activation of p53 is commonly linked to Interestingly, the genes for BID and BAX were up-regulated inductionofapoptosis.ROSandthep21proteinhavealsobeen after4h(Fig.4A),yetFig.4Bshowsminimaldetectionofboth implicated in activating apoptosis, both of which increased in proteins by RPPA. The levels of BID protein fluctuate and the responsetocompound 1(Figs.2Cand3B)(31). levels of BAX slowly increase, both at a maximum after 72 h. We investigated the activation of apoptosis by compound 1, This again suggests that molecular mechanisms associated with using DEGs, RPPA, and flow cytometry. Fig. 4A is a heat map apoptosiswereinitiated.However,thesignificantlyhighlevelsof -1.5 LogFC 1.5 CASPC4A SP BCL2BLC1L12 CASPC8A SP PTENP TEN BIRCB5I RC5 CASPC6A SP CASPC2A SP CASPC9A SP CASPC7A SP CASPC3A SP PARPP1A RP BBC3B BC3 PARPP9A RP BIRCB2I RC2 HSPBH1S PB BID BID BAX BAX CASPC1A SP BIRCB3I RC3 CASPC1A2S P 4 12 24 48 (h) )edidoi muidiporP( 2LF RFI 104 104 0.3 1 1.7 BIM PARP BCL-X 102 102 Survivin XIAP D D BAD* BAD(*) BAK A A BAX 100 100 100 102 104 100 102 104 BCL-2 BCL-2* FL1 (Annexin V) BID CASP3 CASP3* PARP* PTEN PTEN* PUMA Survival Apoptosis A B C 104 104 102 102 D D A A 100 100 100 102 104 100 102 104 D Control Compound 1 4 24 48 72 4 24 48 72 (h) 0.15 0.45 Compound 1 (μM) )edidoi muidiporP( 2LF A B RFI DDNNAA ddaammaageg e 0.3 1.3 p21 AATTMM* * p ATM A CDK1-Cyc pp53 CCHHKK22** CCDC25C* pp5533** CCDDCC2255AA* CCDC25A pp53* pp21* pp2211 AATM* 4 24 48 72 CCDDKK22 ++ 44/6/6 Exposure (h) Fig.3. Induction ofDNAdamageinA2780cells.(A)Diagramofprotein interactions controlling DNA damage and cell cycle progression. (B) Heat mapshowingtheproteinsasrelativefluorescenceintensity(RFI)detectedby RPPA,normalizedtothehouse-keepingproteinprohibitinandtothecontrol. C FL2 (Annexin V) Negative control Compound 1 Fig.4. Nonapoptoticresponsetocompound1inA2780cells.(A)Heatmapshowingthedifferentialgeneexpression(asfoldchange,logFC)ofapoptotic markers(FDR<0.10)atfourtimepoints.(B)Heatmapshowingthedifferentialproteinlevelsofapoptoticmarkersinresponsetocompound1attwodosesat fourtimepoints.(C)Flowcytometrymeasurementsfordetectingapoptoticcelldeath:negativecontrol(Left)andafterexposureto150nMofcompound1 for24h(Right).(D)High-contentmicroscopyimagesofcellsexposedto2.5μMofcompound1for48husingtheNucViewsubstrate.Greenfluorescence identifiescaspase3activationofNucViewprobe,withblueDAPIdyehighlightingcellnuclei. 4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1500925112 Hearnetal. caspase inhibitors ensure that, within a 72-h time frame, apo- treatedcellsrespondedwithashuntofmetabolismintoOXPHOS, ptosis is unlikely to occur. We measured the population of ap- increasingsensitivitytochemotherapeuticagents(38). optotic A2780 cells after a 24-h exposure to 150 nM of Point mutations in mtDNA can affect the function of compound1(Fig.4CandSIAppendix).Duringearlyapoptosis, OXPHOS machinery; therefore, OXPHOS is not an optimal membrane-bound phosphatidylserine proteins, which ordinarily process in cancer cell division. Such mutations are common in faceintothecytoplasm,fliptofaceoutsideandallowannexinV cancercellsandarethoughttobepartlyresponsibleforfavoring binding. Once the cell has lost viability, its membrane becomes glycolysis(36).ThethreemutationsincomplexIoftheETCin permeable to propidium iodide (PI), which fluoresces upon A2780 cells might translate to impairment of the ability of mi- bindingtoDNA.InFig.4C,theFL1channelmeasuredannexin tochondriatomaintainaprotongradientduringOXPHOS(37). V fluorescence and the FL2 channel PI fluorescence. Early ap- The efficiency of maintaining this coupling gradient is directly optotic cells should exhibit high annexin binding and low PI proportional to the efficiency of ATP production by the last binding. Late apoptosis should give high annexin and high PI complexintheETC,ATPasecomplexV(CV),whereCIaccounts binding. The percentages of cells found in each quadrant are for40%oftheprotongradientgeneratedinNADHoxidation(39). listedinSIAppendix,TableS7.Fig.4Cshowsthatthemajorityof Wetherefore proposethatthe inductionofROSinresponse cells were still viable and that a small percentage of nonviable to compound 1 may result from a shift in metabolism, forcing cellshaddiedthroughnonapoptoticmechanisms.Increasingthe A2780 cells to use dysfunctional metabolic machinery. Some concentration 3× made little difference to the RPPA data for metal-based drugs are known to disrupt the redox balance in apoptotic proteins, except that BCL-2 levels were higher, sug- cancercells,oftenthroughtargetingROSscavengers(40).Metal gestingastrongerinitiation ofapoptosissignaling. compounds themselves can catalyze ROS production when ac- Previously,wehaveshownbyflowcytometrythatcompound1 tivatedinthereducingenvironmentofcancercells(41,42).The does not induce apoptosis in A549 lung cancer cells at 500 nM production of ROS by the mitochondria has a cyclic effect on (3× concentration used here) after 24 h (24). We therefore in- theirfunction,potentiatingtheproductionofhigherandhigher creasedtheconcentrationofcompound1further,to2.5μM,and levels of ROS (43). Mitochondrially targeted anticancer drugs measured the levels of caspase-3/7 after 48 h at the single-cell that target the bioenergetics of cancer cells and increase ROS level by high-content microscopy. Activated caspase-3/7 cleaves are becoming a focus for anticancer drug development (44). NucView,asubstratethatintercalateswithDNAandfluoresces ROS induced by compound 1 activate oxidative stress response uponcleavage.TheimagesinFig.4Dshowthat,athighercon- pathways, which we propose are responsible for DNA damage. centrations, the proportion of cells with detectable caspase ac- DNAisawell-knowntargetforROS,particularlyguaninebases tivity (green) increased and the number of cell nuclei (blue) (45).Compound1isequipotentinPt-resistantcancercelllines, decreased.Thus,weobserveacomplexdose-dependentpathway with ROS-centered DNA damage not repairable by enhance- response from both pro- and anti-apoptotic proteins following mentofPt-induced DDR(13). exposuretocompound1.Asaresult,weseeatemporalwaveof The cytotoxic effects of cisplatin are often not selective for apoptosisdetectedattheproteinlevelbyRPPAbetween24and cancer cells and can be just as damaging to normal cells (46). 48handbymicroscopy at48h. DNA damage caused indirectly by compound 1, through selec- tiveROSgeneration,providesanelementoftargetingofcancer Discussion cells through exploiting deficient mitochondria. Indeed, the Thediscoveryanddevelopmentofchemotherapeuticagentswith antiproliferative selectivity of compound 1 for A2780 ovarian newMOAsareimportantgoalsincancerresearch.Wearefocusing cancer cells versus normal fibroblast cells is ∼3× greater than on the development of organometallic compounds that may be that of cisplatin(SIAppendix, TableS8). Compound1 therefore activetowardcisplatin-resistantcancersandhavefewersideeffects appears to exploit the metabolic deficiencies in cancer cells, compared with the platinum drugs. We discovered through probably potentiated by the effects of mutations in the compo- screeningthatcompound1haspotentantiproliferativeactivityina nentproteinsoftheETCoftheEOCcelllineusedinthiswork. widerangeofcancercelllines(11)(SIAppendix,Fig.S2)andisalso Our findings should stimulate future work to investigate the activeinvivo(12).Withthemovetowardpersonalizedmedicine,it functional implications ofthesemutations. becomesvitaltoelucidatetheMOAfornewmetal-baseddrugsso Hanif and coworkers reviewed the promising activity of thattreatmentscanbematchedtothepredictedresponseofpar- osmium-based anticancer compounds, highlighting their poten- ticularcancercells.Hence,weadoptedasystemsbiologyapproach tialasROSinducers,DNAbinders,andproteinkinaseinhibitors togaininsight intotheactivity ofcompound1 (33). (7). Our work has revealed the rapid timescale of ROS pro- Ourdatasuggestadualmechanismofactioninvolvingmodulation ductioninovariancancercellsandtheconsequentswitch-onof of metabolicprocessesandanincreasein thelevel of ROS and/or antioxidantresponsepathways,togetherwithdownstreamDNA lackofcontroloftheirdestruction.Inaddition,wehaveidentified damageandinitiation ofapoptosis (24and48h). threemutationsinCIoftherespiratorychainoftheA2780EOC Ouruseoftranscriptomics,proteomics,andimagingprovides cells,whichmaycontributetothepotencyofcompound1. a promising approach for obtaining insights into the cellular Mitochondrialgenesplayanimportantroleindrugresponse, MOA of metal-based antiproliferative compounds. RNA se- and the mitochondrion is fast becoming an important target in quencing and RPPA have rarely been used in such studies, cancer research (34). Targeting this organelle directly or in- possibly because of the complexity of data analysis and inter- directly provides an element of selectivity for therapeutic anti- pretation. Mitochondrial mutations such as those that we have cancer agents. The metabolic profile of cancer cells is often discoveredinthisA2780cancercelllinearewelldocumentedin primarily glycolytic, compared with normal cells that produce variouscancertypes,but,asfarasweareaware,havenotbeen ATP primarily by oxidative phosphorylation (OXPHOS), the exploredinovariancancer.Ourworkthereforenotonlyprovides Warburgeffect(35).GlycolysisisgenerallylessefficientforATP unique insights into both the potency and selectivity of com- production; however, it is faster than OXPHOS, permits mac- pound1,butalso opensupnewpathwaysforfutureresearch. romolecule biosynthesis, andcan function underhypoxic condi- tions(36).ThisswitchfromOXPHOStoglycolysisisalsothought Conclusions to reduce the production of ROS, often by-products of the Theorganometallichalf-sandwichOs(II)arenecompound1isa OXPHOS pathway andpotentially deadly to cellular function. promising candidate for cancer treatment, being highly active The ChrM codes for important polypeptides in components of toward a range of cancer cells in vitro and showing promising theETCthatcontroltheprocessofOXPHOS(37).Theobserved activity in vivo (11, 15). Our combined transcriptomic and pro- up-regulation of these genes in response to compound 1 suggests teomic studies show that it hasadifferent mechanism of action that cells switch on, or enhance, the use of OXPHOS. A similar fromcisplatin.Ouranalysissuggestsanattackonglycolysisthat effectwasreportedwithglycolysisinhibitor2-deoxyglucose,where switches energy production toward OXPHOS in A2780 EOC Hearnetal. PNASEarlyEdition | 5of6 SULPSANP YRTSIMEHCOIB cells. This pathway may already be stressed by the three muta- Reverse-Phase Protein Arrays. Full details are in SI Appendix (47). Briefly, tions that we detected in the CI of the ETC. Not only does A2780cellswereseededat4×105andincubatedindrug-freemediumfor compound 1 induce ROS production, but it also reduces the 48 h. Compound 1 (150 and 450 nM) and vehicle control solutions were abilityofcellstorespondtoexcessiveROS,subsequentlyleading added,andafter4,24,48,and72hcellswerecollectedandproteinextracts todownstreamDNAdamage. were obtained. Normalized cell lysates were deposited onto microarray The chemical behavior of compound 1 differs significantly chips (ZeptoChip, Zeptosens-Bayer), and chips were blocked and washed. from cisplatin in that it does not readily undergo hydrolysis or Chipswereincubatedfor24hwithprimaryantibodies(SIAppendix,Table bindtoglutathioneorguanine(DNA,RNA).Thepossibilitythat S3),followedby2.5hincubationwithsecondaryAlexa-Fluor647-conjugated thereareunusualmechanismsforitsactivation insidecellsthat antibodydetectionreagent.ArrayswereimagedusingZeptoREADERand would allow it to attack components of the glycolytic pathway analyzedusingZeptoView3.1software. requires further investigation. Such activity by a metallodrug, involving mitochondria that are partly dysfunctional in cancer High-ContentImaging.A2780cellswereseededina96-wellplateat5,000 cells, may not only combat cancer cell drug resistance, but also cells/wellandincubatedfor48hbeforetreatmentwithcompound1or provideselectivityforcancercellsversusnormalcells,properties vehicle control for another 48 h. Before image acquisition, cells were thatarelikelytobevaluableinnext-generationanticancerdrugs. incubatedwith4μg/mLDAPI(SigmaD8417)and1mMNucView(Biotium) reagentfor30min.MicroscopicimagesofDAPI-andNucView-stainedcells MaterialsandMethods wereacquiredwitha10×objectiveandappropriateopticalfiltersusing anOlympusScanRhigh-contentimagingmicroscope.Mergedimagesof RNASequencing.FulldetailsareinSIAppendix.Briefly,A2780cellswereseeded DAPIandNucViewcellswerecreatedusingtheOlympusScanRimaging at3×106andincubatedindrug-freemediumfor24h.Compound1(150nM) softwareCell-IR. andvehiclecontrolsolutionswereaddedandafter4,12,24,and48h,cellswere collected and whole-cell RNA was extracted (RNeasy plus mini kit, Qiagen). Sampleswerequality-control–checkedusingaNanoDrop1000spectrophotom- ACKNOWLEDGMENTS. We thank Professor David Wild and Dr. Jonathan Mooreforhelpfuldiscussions.WethanktheBiotechnologyandBiological eterandanAgilent2100Bioanalyzer.Truseqlibrarieswerepreparedandsam- SciencesResearchCouncil(Grant324594,SystemsBiologystudentshipfor plesweresequencedonanIlluminaHiSeq2000with50-bppaired-endreadsand J.M.H.),theEuropeanResearchCouncil(Grants247450and324594),andthe ∼30millionpaired-endreadspersample.Readsweremappedtohg19using WellcomeTrust(Grants086357and102696)forsupport,aswellastheEuro- Tophat2,andHTSeqandedgeRwereusedfordifferentialexpressionanalysis. peanUnionCOSTActionCM1105. 1. BarryNPE,SadlerPJ(2013)Explorationofthemedicalperiodictable:Towardsnew 25.ZhuM,LiW,LuC(2012)Roleofalpha-synucleinproteinlevelsinmitochondrial targets.ChemCommun(Camb)49(45):5106–5131. morphologyandcellsurvivalincelllines.PLoSONE7(4):e36377. 2. FrezzaM,etal.(2010)Novelmetalsandmetalcomplexesasplatformsforcancer 26.MenazzaS,etal.(2010)Oxidativestressbymonoamineoxidasesiscausallyinvolved therapy.CurrPharmDes16(16):1813–1825. inmyofiberdamageinmusculardystrophy.HumMolGenet19(21):4207–4215. 3. KomedaS,CasiniA(2012)Next-generationanticancermetallodrugs.CurrTopMed 27.DevineMJ,Plun-FavreauH,WoodNW(2011)Parkinson’sdiseaseandcancer:Two Chem12(3):219–235. wars,onefront.NatRevCancer11(11):812–823. 4. MuhammadN,GuoZ(2014)Metal-basedanticancerchemotherapeuticagents.Curr 28.DallaPozzaE,etal.(2012)Roleofmitochondrialuncouplingprotein2incancercell OpinChemBiol19:144–153. resistancetogemcitabine.BiochimBiophysActa1823(10):1856–1863. 5. PatraM,GasserG(2012)Organometalliccompounds:Anopportunityforchemical 29.PawlakM,etal. (2002) Zeptosens’protein microarrays: Anovel high perfor- biology?ChemBioChem13(9):1232–1252. mancemicroarrayplatformforlowabundanceproteinanalysis.Proteomics2(4): 6. GasserG,OttI,Metzler-NolteN(2011)Organometallicanticancercompounds.JMed 383–393. Chem54(1):3–25. 30.AshwellS,ZabludoffS(2008)DNAdamagedetectionandrepairpathways:Recent 7. HanifM,BabakMV,HartingerCG(2014)Developmentofanticanceragents:Wizardry advanceswithinhibitorsofcheckpointkinasesincancertherapy.ClinCancerRes withosmium.DrugDiscovToday19(10):1640–1648. 14(13):4032–4037. 8. SuntharalingamK,etal.(2013)Bidentateligandsonosmium(VI)nitridocomplexescontrol 31.MasgrasI,etal.(2012)Reactiveoxygenspeciesandmitochondrialsensitivitytooxi- intracellulartargetingandcelldeathpathways.JAmChemSoc135(38):14060–14063. dativestressdetermineinductionofcancercelldeathbyp21.JBiolChem287(13): 9. MailletA,YadavS,LooYL,SachaphibulkijK,PervaizS(2013)AnovelOsmium-based 9845–9854. compoundtargetsthemitochondriaandtriggersROS-dependentapoptosisincolon 32.FlanaganL,etal.(2010)XIAPimpairsSmacreleasefromthemitochondriaduring carcinoma.CellDeathDis4(6):e653. apoptosis.CellDeathDis1:e49. 10.NiW-X,etal.(2011)Osmium(VI)complexesasanewclassofpotentialanti-cancer 33.SpreckelmeyerS,OrvigC,CasiniA(2014)Cellulartransportmechanismsofcytotoxic agents.ChemCommun(Camb)47(7):2140–2142. metallodrugs:Anoverviewbeyondcisplatin.Molecules19(10):15584–15610. 11.FuY,etal.(2011)Structure-activityrelationshipsfororganometallicosmiumarene 34.FuldaS,KroemerG(2011)Mitochondriaastherapeutictargetsforthetreatmentof phenylazopyridinecomplexeswithpotentanticanceractivity.DaltonTrans40(40): malignantdisease.AntioxidRedoxSignal15(12):2937–2949. 10553–10562. 35.WardPS,ThompsonCB(2012)Metabolicreprogramming:Acancerhallmarkeven 12.Romero-Canelón I, Sadler PJ (2013) Next-generation metal anticancer complexes: warburgdidnotanticipate.CancerCell21(3):297–308. Multitargetingviaredoxmodulation.InorgChem52(21):12276–12291. 36.ZhengJ(2012)Energymetabolismofcancer:Glycolysisversusoxidativephosphory- 13.FuY,etal.(2010)Organometallicosmiumarenecomplexeswithpotentcancercell lation(Review).OncolLett4(6):1151–1157. cytotoxicity.JMedChem53(22):8192–8196. 37.RichPR,MaréchalA(2010)Themitochondrialrespiratorychain.EssaysBiochem47: 14.FuY,etal.(2012)Thecontrastingchemicalreactivityofpotentisoelectroniciminopyr- 1–23. idineandazopyridineosmium(II)areneanticancercomplexes.ChemSci3(8):2485–2494. 38.PaloriniR,SimonettoT,CirulliC,ChiaradonnaF(2013)MitochondrialcomplexIin- 15.ShnyderSD,etal.(2011)Anti-colorectalcanceractivityofanorganometallicosmium hibitorsandforcedoxidativephosphorylationsynergizeininducingcancercelldeath. areneazopyridinecomplex.MedChemComm2(7):666–668. IntJCellBiol2013:243876. 16.ChatterjeeA,MamboE,SidranskyD(2006)MitochondrialDNAmutationsinhuman 39.HinchliffeP,SazanovLA(2005)Organizationofiron-sulfurclustersinrespiratory cancer.Oncogene25(34):4663–4674. complexI.Science309(5735):771–774. 17.CalvoSE,etal.(2010)High-throughput,pooledsequencingidentifiesmutationsin 40.JungwirthU,etal.(2011)Anticanceractivityofmetalcomplexes:Involvementof NUBPLandFOXRED1inhumancomplexIdeficiency.NatGenet42(10):851–858. redoxprocesses.AntioxidRedoxSignal15(4):1085–1127. 18.ParrellaP,etal.(2001)DetectionofmitochondrialDNAmutationsinprimarybreast 41.GrafN,LippardSJ(2012)Redoxactivationofmetal-basedprodrugsasastrategyfor cancerandfine-needleaspirates.CancerRes61(20):7623–7626. drugdelivery.AdvDrugDelivRev64(11):993–1004. 19.LarmanTC,etal.;CancerGenomeAtlasResearchNetwork(2012)Spectrumofsomatic 42.KovacicP,SomanathanR(2011)Recentdevelopmentsinthemechanismofanticancer mitochondrialmutationsinfivecancers.ProcNatlAcadSciUSA109(35):14087–14091. agentsbasedonelectrontransfer,reactiveoxygenspeciesandoxidativestress.An- 20.GuerraF,etal.(2012)MitochondrialDNAmutationinserousovariancancer:Implica- ticancerAgentsMedChem11(7):658–668. tionsformitochondria-codedgenesinchemoresistance.JClinOncol30(36):e373–e378. 43.HenchcliffeC,BealMF(2008)MitochondrialbiologyandoxidativestressinParkinson 21.AikhionbareFO,etal.(2008)mtDNAsequencevariantsinsubtypesofepithelial diseasepathogenesis.NatClinPractNeurol4(11):600–609. ovariancancerstagesinrelationtoethnicandagedifference.DiagnPathol3:32–43. 44.NeuzilJ,DongLF,RohlenaJ,TruksaJ,RalphSJ(2013)Classificationofmitocans,anti- 22.VanTrappenPO,etal.(2007)SomaticmitochondrialDNAmutationsinprimaryand cancerdrugsactingonmitochondria.Mitochondrion13(3):199–208. metastaticovariancancer.GynecolOncol104(1):129–133. 45.JenaNR(2012)DNAdamagebyreactivespecies:Mechanisms,mutationandrepair. 23.KatohY,etal.(2001)TwodomainsofNrf2cooperativelybindCBP,aCREBbinding JBiosci37(3):503–517. protein,andsynergisticallyactivatetranscription.GenesCells6(10):857–868. 46.PablaN,DongZ(2012)Curtailingsideeffectsinchemotherapy:AtaleofPKCδin 24.vanRijtSH,Romero-CanelónI,FuY,ShnyderSD,SadlerPJ(2014)Potentorgano- cisplatintreatment.Oncotarget3(1):107–111. metallicosmiumcompoundsinducemitochondria-mediatedapoptosisandS-phase 47.vanOostrumJ,etal.(2009)Tracingpathwayactivitieswithkinaseinhibitorsand cellcyclearrestinA549non-smallcelllungcancercells.Metallomics6(5):1014–1022. reversephaseproteinarrays.ProteomicsClinAppl3(4):412–422. 6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1500925112 Hearnetal.