An Accessible and Unique Insight into Metastasis Mutational Content Through Whole-exome Sequencing of Circulating Tumor Cells in Metastatic Prostate Cancer

An

Accessible

and

Unique

Insight

into

Metastasis

Mutational

Content

Through

Whole-exome

Sequencing

of

Circulating

Tumor

Cells

in

Metastatic

Prostate

Cancer

Vincent

Faugeroux

a,b

,

Ce´line

Lefebvre

a

,

Emma

Pailler

a,b

,

Vale´rie

Pierron

c

,

Charles

Marcaillou

d

,

Se´bastien

Tourlet

c

,

Fanny

Billiot

b

,

Semih

Dogan

a

,

Marianne

Oulhen

b

,

Philippe

Vielh

b

,

Philippe

Rameau

b

,

Maud

NgoCamus

e

,

Christophe

Massard

e

,

Corinne

Laplace-Builhe´

b

,

Arian

Tibbe

f

_,

_Me´lissa

_Taylor

b

_,

_Jean-Charles

_Soria

a,e

_,

_Karim

_Fizazi

a,e

_,

_Yohann

_Loriot

a,e,1

_,

Sylvia

Julien

c,1

,

Franc¸

oise

Farace

a,b,1,

*

a_{INSERMU981,InstitutGustaveRoussy,UniversitéParis-Saclay,Villejuif,France;}b_{INSERMUS23AMMICA,InstitutGustaveRoussy,UniversitéParis-Saclay,} Villejuif,France;c_{IPSEN-Innovation,LesUlis,France;}d_{IntegraGenSA,Evry,France;}e_{DepartmentofCancerMedicine,InstitutGustaveRoussy,Université} Paris-Saclay,Villejuif,France;f_{VyCAPBV,Deventer,TheNetherlands}

a v a i l ab l e a t w w w . s c i e n c e d i r e c t . c o m j o u r n a l h o m e p a g e : e u o n c o l o g y . e u r o p e a n u r o l o g y . c o m Articleinfo Articlehistory: AcceptedDecember7,2018 AssociateEditor: PaulNguyen Keywords:

Circulatingtumorcells Liquidbiopsy

Prostatecancer

Whole-exomesequencing

Abstract

Background: Genomicanalysisofcirculatingtumorcells(CTCs)couldprovidea uniqueandaccessiblerepresentationoftumordiversitybutremainshinderedby technicalchallengesassociatedwithCTCrarityandheterogeneity.

Objective: ToevaluateCTCsassurrogatesamplesforgenomicanalysesin meta-staticcastration-resistantprostatecancer(mCRPC).

Design,setting,andparticipants: Threeisolationstrategies(filterlaser-capture mi-crodissection,self-seedingmicrowellchips,andfluorescence-activatedcellsorting) weredevelopedtocaptureCTCswithvariousepithelialandmesenchymalphenotypes andisolatethematthesingle-celllevel.Whole-genomeamplification(WGA)andWGA qualitycontrolwereperformedon179CTCsamples,matchedmetastasisbiopsies,and negativecontrolsfrom11patients.Allpatientsbutonewerepretreatedwith enzalu-tamideorabiraterone.Whole-exomesequencing(WES)of34CTCsamples,metastasis biopsies,andnegativecontrolswereperformedforsevenpatients.

Outcome measurements and statistical analysis: WES of CTCs was rigorously qualified in terms of percentage coverage at 10 depth, allelic dropout, and uncoveredregions.SharedsomaticmutationsbetweenCTCsandmatched metas-tasisbiopsieswereidentified.Acustomizedapproachbasedondeterminationof mutationratesforCTCsampleswasdevelopedforidentificationofCTC-exclusive mutations.

Resultsandlimitations: SharedmutationsweremostlydetectedinepithelialCTCs andwererecurrent.Fortwopatientsforwhomadeeperanalysiswasperformed,a fewCTCswere sufficient torepresent halfto one-thirdofthe mutationsinthe

1 _{Theseauthorscontributedequallytothiswork.}

*Correspondingauthor.InstitutGustaveRoussy,Universite´ Paris-Saclay,114rueE´douardVaillant, F-94805Villejuif,France.Tel.:+33142115198.

E-mailaddress:francoise.farace@gustaveroussy.fr(F.Farace).

https://doi.org/10.1016/j.euo.2018.12.005

2588-9311/©2020TheAuthors.PublishedbyElsevierB.V.onbehalfofEuropeanAssociationofUrology.Thisisanopenaccessarticle undertheCCBY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction

Advances in the development of precision medicine for cancer patients rely on accurate identification of the genomicfeaturesunderlyingapatient’stumor[1].Asthe numberofgenomictargetswithmatchedtherapiesgrows, thecurrenthurdleremainstheavailabilityoftumortissue thatcanreflectaconstantlyevolvingdisease[2–4]. Recog-nition thatmetastatic lesions haveadiscordantgenomic fingerprint compared to primary tumor has led to recommendations for invasive biopsies. However, tumor heterogeneity poses the predicament that single-site biopsiesmaynotcapturethegenomicalterationsrelevant totargetedtherapy[5].Thisisespeciallytrueinmetastatic prostatecancer(mPC),forwhichbonemetastasesarising severalyearsafterdiagnosisandtherapeuticinterventions are either inaccessible or yield insufficient material for genomicprofilingofthedisease[6,7].Inaddition,biopsies of bone metastases are painful and cannot be repeated multipletimes during thedisease course, calling for less invasivemethodsformolecularcharacterizationofmPCand formonitoringdiseaseprogressionduringtherapy.

The development of “liquid biopsies” presents new opportunitiesfornoninvasivemonitoringofclonal heteroge-neity. Circulating tumor cells (CTCs) captured as a “liquid biopsy”arecurrentlyregardedasanoninvasiveand repeat-able source of tumor material that could overcome the samplingchallengesformetastaticdisease[8–11].CTCsare likelytoarisefromdistinctmetastaticsites,andmaybetter representtumorheterogeneityinbothspaceandtime.Liquid biopsiesmight allowgenomiccharacterizationofmPCand routinemonitoringofmetastaticspread,drugresistance,and diseaserelapseduringthecourseoftreatment[9–11]. How-ever,itisstill unknownwhether using liquidbiopsies can provideamorecompleteprofileofmPCclonaldiversity.

CTCs often exhibit combinations of epithelial and mesenchymal traits, highlighting the role of epithelial-mesenchymaltransition(EMT)intheprocessof intravasa-tion and cancer cell dissemination [12]. Therefore, non-epithelial-basedCTCenrichmentmethodsareessentialto capturethephenotypicheterogeneityofCTCsinterms of EMTmarkerexpression[13–15].Reportsofsingle-cell high-dimensional analyses of CTCs are scarce and limited to

EpCAM-positive CTCs [16–18]. The rarity and biological heterogeneityofCTCshaveimposedtechnicalchallengesto their isolation and analyses at single-cell level, and impacted the success of robust processing of complex andcostlydownstreammethodologies.Onceisolatedatthe single-celllevel,CTCsmustundergowhole-genome ampli-fication (WGA), a mandatory process to identify CTC somaticvariantsbutpronetoamplificationbias, polymer-aseerrors,andallelicdropout(ADO)[19].

Ourcentralhypothesiswasthatthegenetic heterogene-ityofCTCsassessedatthesingle-celllevelviawhole-exome sequencing(WES)reflectstheirphenotypicheterogeneity in mPC and may help to resolve the clonal relatedness betweenCTCsandmetastasisinmPC.Here,wereportfor thefirsttimeisolationatthesingle-celllevel,qualification, andWESofCTCs harboringvariousepithelial, mesenchy-mal,and/ormorphologicalcharacteristicsfrom11patients withmetastaticcastration-resistantPC(mCRPC).AllmCRPC patients were included in the MOSCATO01 clinical trial evaluating the potential clinical benefit of screening metastasisbiopsiesviahigh-throughputgenomicanalyses toidentifyactionablealterationsinpatientswithadvanced cancer [1]. Somaticmutations sharedbetween CTCs and matchedmetastaticbiopsies,andCTC-exclusivemutations (exclusivelyidentifiedinCTCsandnotinmatchedbiopsies) were comprehensively explored. Our results reveal that CTCs can be used as surrogate samples for mutational analyses of mCRPC and provide a genomic picture of disseminatingclonesinvolvedinmetastasis.

2. Patientsandmethods

2.1. Patients

PatientswithmCRPCwereenrolledintheMOSCATO01prospectivetrial (IDRCB2010-A00841-40;NCT02613962)[1].Thestudywasauthorized bytheFrenchnationalregulationagencyANSMandapprovedbythe ethicscommittee.BloodwascollectedintoEDTAandCellSavetubes.

2.2. CTCenrichment,detection,andisolation

Isolationbysizeofepithelialtumorcells(ISET)filtration,immuno fluo-rescencestaining,andscanningof_filterswereperformedaspreviously

matchedmetastasisbiopsy.CTC-exclusivemutationswereidentifiedinboth epi-thelialandnonepithelialCTCsandaffectedcytoskeleton,invasion,DNArepair,and cancer-drivergenes.Some41%ofCTC-exclusivemutationshadapredicted delete-rious impact on protein function. Phylogenic relationships between CTCs with distinctphenotypeswereevidenced.

Conclusions: CTCscanprovideuniqueinsightintometastasismutationaldiversity andrevealundiagnosedgenomicaberrationsinmatchedmetastasisbiopsies.

Patient summary: Our results demonstrate the clinical potential of circulating tumorcellstoprovideinsightintometastaticeventsthatcouldbecriticaltotarget usingprecisionmedicine.

©2020TheAuthors.PublishedbyElsevierB.V.onbehalfofEuropeanAssociationof Urology.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense( http://creati-vecommons.org/licenses/by-nc-nd/4.0/).

reported [20].CTCs and CD45-positivecells were isolated via laser microdissection. CTCs were detected using CellSearch as previously described [10,20] and isolated using self-seeding microwell chips [21]. Enrichment via RosetteSep was performed according to the manufacturer_’s protocol(StemCell Technologies,Vancouver,Canada). RosetteSep-enrichedCTCswerepermeabilized,stainedviaimmuno flu-orescence,andisolatedusingfluorescence-activatedcellsorting(FACS). ThemethodsaredescribedintheSupplementarymaterial.

2.3. WGA,qualitycontrols,andpurification

WGAwasperformed usingan Ampli1WGAkitversion1 (Menarini SiliconBiosystems,SanDiego,CA,USA)andevaluatedviaaqualitative polymerasechainreaction(qPCR)assaytodetermineaqualityscore(QS; Supplementarymaterial).

2.4. IsolationofgenomicDNAfrombloodandtumorbiopsies DNAfromformalin-_fixed,paraf_{fin-embedded}tumorbiopsyspecimens andconstitutionalDNAwerepurifiedasdescribedintheSupplementary material.

2.5. WESandbioinformaticsanalysis

WES and bioinformatics analysis are described in detail in the Supplementarymaterial.RawdataareavailableonEGAataccession numberega-box-1082.

3. Results

Eleven patients were included in the current analysis (Table 1).At the time of CTC collection, allpatients had metastasesfromCRPCandallbutonehadbeenpreviously treated with either abiraterone or enzalutamide, while threehadbeentreatedwithpriordocetaxel.

WeinitiallydevelopedaCTCenrichmentstrategybased onfiltrationbyISETbeforeamultistepprocessthatincludes immunofluorescentstainingwithepithelialand mesenchy-malmarkersforCTCidentificationandlaser microdissec-tion of individualCTCs (Fig.1A). CTCs fromfive patients (P1–P5)were isolatedinthis way.We detected veryfew

epithelialCTCs,especiallyintwoofthepatients(P2andP3), compared tohigher CellSearch counts,asconfirmedbya recentstudy [13] (Supplementary Table 1). In agreement withpublishedresults[15,22],ISETrevealedhybridcells,as wellaspopulationsoflarge(nuclei 16

m

m)and mesen-chymal candidate CTCs, of possible interest for further analysis by WES (Fig. 1B). We developed two other strategiesto captureepithelial andothercandidateCTCs. In the second strategy, CTCs were enumerated via CellSearchandindividuallyisolatedusingtheself-seeding microwellchipsplatform[21](Fig.1A).Ourthirdapproach involvedhematopoieticblood-celldepletionandindividual CTC isolation via FACS (Fig. 1A). These three strategies allowed identification of five phenotypic subtypes of candidate CTCs on the basis of both epithelial and mesenchymal marker expression and/or morphological characteristics (Fig. 1B). These initial results led us to explore and isolate CTCs in six more mCRPC patients (Table1andSupplementaryTable1).Aglobalworkflowwas establishedbasedonthesestrategies,WGAand qualifica-tion, and WES of CTCs and matched negative controls (CD45+ cells; Fig. 1C). We hypothesized that pools of a limitednumberofindividualCTCscouldresultin higher-qualityWGAandcouldyieldhighersensitivityforvariant detectionthansingleCTCs.Weisolated162poolsoftwoto ten CTCs (each pool containing CTCs of the same phenotype),17singleCTCs,andmatchedCD45+_cells.The

WGAqualitywascontrolledusingaqPCRassay,andaQS3 was set as the criterion for WES eligibility, which was reached by 87/179 WGA samples (49%; Supplementary Table2,SupplementaryFig.1AandB).Ashypothesized,the probabilityofreachingQS3washigherinpoolsthanin singleCTCs(Fig.1D).Amongthe87WGACTCsampleswith QS3,31fromsevenpatientswereselectedforWES.Two supplementary epithelial CTC pools (P7-E-7 and P7-E-8) from10CTCsandasingleCTCbearingahybridphenotype, allwithQS<3,wereincludedforpatientP7toaddressCTC heterogeneity(SupplementaryTable3).WESforthe34CTC sampleswassubjectedtorigorousqualitycontrol(Fig.1E,

SupplementaryFig.1).Ofthe34WESsamples,28(82%)had

Table1_–Clinicalcharacteristicsofpatientswithmetastaticcastration-resistantprostatecancer

Case Age(yr) TNMa _{Metastaticsites PSA}b_{(ng/ml) Gleasonscore TTB(mo) Biopsysite Previoustreatment}

CTx CTN ENZ ABI

P1 77 TxNxM1 BonesandLNs 333 7(4+3) 46 LN No Yes No Yes

P2 77 T1N0M0 BonesandLNs 101 7(3+4) 32 LN No Yes No Yes

P3 73 T3N0M0 LNs 249 7(4+3) 22 LN No Yes No Yes

P4 46 TxN0M1 Bonesandliver 85 8 14 Liver No Yes No Yes

P5 68 T3N0M0 LNs 14 9 24 LN Yes Yes No Yes

P6 65 T3N0M0 LNs 283 8(4+4) 13 LN No Yes No Yes

P7 60 T3N0M0 BonesandLNs 1715 7(4+3) 26 LN Yes Yes Yes Yes

P8 66 T3N0M0 BonesandLNs 51 9 2 LN No Yes No Yes

P9 67 TxN1M1 LNsandbones 27 8(4+4) 12 LN Yes Yes No No

P10 71 TxNxM1 Bones,liver,andLNs 10 NA 72 Liver No Yes Yes No

P11 76 T3N0M0 Bones 163 NA 24 Bone No Yes No Yes

LN=lymphnode;PSA=prostate-specificantigen;NA=notavailable;TTB=timefrommetastaticprostatecancerdiagnosistobiopsy;CTx=chemotherapy; CTN=castration;ENZ=enzalutamide;ABI=abiraterone.

a

Atdiagnosis.

b

Fig.1–ExperimentalprocessforWESofCTCsatthesingle-celllevel.(A)Schematicofthethreeexperimentalstrategiesusedforenrichment, detection,andisolationofCTCsatthesingle-celllevel.(B)NomenclatureforCTCcandidatesaccordingtoepithelialandmesenchymalmarker expressionandtheexperimentalstrategyusedforisolation.EpithelialCTCs(E)wereCD45-negativeandpositiveforEpCAMandpan-cytokeratin(green staining)andHoechst33342(bluestaining).HybridCTCs(H)wereCD45-negativeandpositiveforEpCAMandpan-cytokeratin,vimentin(redstaining), andHoechst33342.LargeCD45-negative,Hoechst33342-positivecells(L)withanucleus16mmandepithelialandmesenchymalexpressionbelow thethresholdoftheassaywereselectedforWESonthebasisofcytomorphologicalfeaturesreportedinpreviousstudies[20].Largemesenchymal CD45-negative,Hoechst33342-positivecells(LM)withanucleus16mmandvimentinexpressionwereselectedforWESonthebasisof

cytomorphologicalfeatures.CD45-negative,Hoechst33342-positivesmallcellcandidates(S)withcytokeratinexpressionbelowthethresholdofthe assayandcontaininganucleusmorphologicallydistinctfromsurroundingCD45-positivecellsweretestedviaWES.(C)Globalworkflowfor sequencingofmetastasisbiopsies,individualCTCs,andCD45+

controlcells.Somaticmutationswereidentifiedaccordingtofiltersdescribedinthe

Supplementarymaterial.Commonsingle-nucleotidepolymorphisms(SNPs),patientSNPs,andvariantscommonbetweenCTCsandpairedCD45+_cell

sampleswereremoved.(D)PercentageofWGAproductswithQS3accordingtothenumberofCTCspersample.CTCsampleswereclassified accordingtothenumberofCTCspersample.ThepercentageWGAwithQS3wasdependentonthenumberofCTCspersample.(E)Coverageat10T sequencingdepthofthe34selectedCTCsamplesfromtheWESexperimentalset.ThenomenclatureforCTCsisdetailedinSupplementaryTable4.Of the34CTCsamples,27(79%)hadmorethan50%coverageat10T.Themeanpercentagecoverageat10Tis54.3%(range12–71%).Thelower percentagecoverageat10TwasbecauseofWGAwithQS<3.WES=whole-exomesequencing;CTCs=circulatingtumorcells;WGA=whole-genome amplification;QS=qualityscore;ISET=isolationbysizeofepithelialtumorcells.

<30%ADOanduncoveredregions,and27(79%)had>50% coverage at 10 depth (10 was the minimum level of coverageconsidered,butcoveragewasinpracticetypically inexcessof10).

We analyzed mutations shared between CTCs and matched metastasis biopsy samples. Among the 34 CTC samplessequenced,15fromfourpatients(P1,P7,P10,P11) sharedmutationswiththecorrespondingmatchedbiopsy (Fig. 2A). A total of136 sharedmutations for 65 unique mutationsweredetected(SupplementaryFigs.2and3A).Of these65,61(93.9%)weredetectedexclusivelyinepithelial CTCs, while four (6.1%) were detected in CTCs bearing variousphenotypicalcharacteristics(Fig.2B).Forpatients P7andP11,forwhomalargernumberofCTCsampleswith various characteristics were sequenced, 59% and 33%, respectively, of mutations detected in biopsies were identifiedinCTCs(Fig.2C).ForpatientP7,sharedmutations were mostly identified in two CTC pools with high QS (Fig.2D).ForpatientP11,31%ofmutationssharedwiththe matchedbiopsywerepresentinonepooloffive andone singleepithelial CTC with QS 5 (SupplementaryFig. 4). Mostofthesharedmutations,includingthoseinGRM8and TP53, were identified in several epithelial CTC samples (Table2,Fig.2B,SupplementaryTable4).Overall,24/53of shareduniquemutations(45%)hadapredicteddeleterious impacton proteinfunction. Ourresultsdemonstrate that WESofasmallnumberofpoolsorsingleepithelialCTCs, selectedusingarigorousqualificationprocess,can reveal significantmutationalsimilaritiestothemetastasisbiopsy. Shared mutations were recurrent in CTCs, suggesting a possibleimportantroleinmetastaticspread.

ToreliablyidentifysomaticCTC-exclusivemutations,we excluded variants shared between CTCs and matched biopsies and established a customized approach that involves variant classification according to recurrence in CTC samples (Supplementary Fig. 3B). We determined a mutation rate per Mb per CTC sample for criteria of observation in one,two, or three samples, as shownfor patientP11inFig.3A.MutationsidentifiedinCTCsamples withamutationrateperMboflessthanthatformetastasis biopsies were considered as CTC-exclusive mutations. Among the 240 variants identified in two CTC samples, 31 were known mutations with a predicted deleterious impact on protein function, including variants in cancer genessuchasHSP90AB1andKDM5B(SupplementaryFig.2, SupplementaryTable5).Whenvariantswererequiredtobe observedinthreeCTCsamples,themutationrateperMb per samplewaslessthan values for matchedmetastases (Fig.3A).Nineteensomaticvariantswerepresentinatleast threeCTCsamples,ofwhichtenvariantswerereportedin databases(Table2).Sevenofthe19(37%)wereidentifiedin genesinvolvedincytoskeleton(eg,MACF1)orinvasion(eg, NEDD9)andimpactproteinfunction.ForpatientP11,20%of CTC-exclusivevariantswereidentifiedinatleastthreeCTC samples(Fig.3B). Whenweexaminedthedistribution of these19exclusivevariantsaccordingtoCTCphenotype,11 (58%)weredetectedexclusivelyinepithelialCTCs,andeight (42%) were presentin CTCs harboring various character-istics (Fig. 3C). These data demonstrate the existence of

exclusive mutations in CTCs that were not detected in matchedbiopsies.

ToevaluatetherelationshipbetweenCTCswithdistinct characteristics in patients P7 and P11, we performed hierarchicalclusteringforsharedandCTC-exclusive muta-tionsobservedinatleastthreeCTCsamples(Fig.4AandB). The existenceof common mutations between CTCs with distinct characteristics supports their phylogenetic rela-tionshipbutdivergentevolution.

4. Discussion

We present the first genomic analysis atthe single-cell level of CTCs with various phenotypical characteristics. Our primary aim was to determine the feasibility of noninvasivemutationalcharacterizationofmPCpatients and todefine whether the mutation landscape of CTCs reflected thatofmatchedmetastases.Wedemonstrated thatWESofCTCsisfeasiblebytestingdifferentstrategies that enabled us to detect CTCs with multiple cellular phenotypes.Ourfirst CTCisolation strategyfellshortof identifying a fraction of epithelial CTCs, as recently confirmed[14].Thisledustodeveloptwosupplementary enrichment and isolation strategies that were used in conjunction with filtration and laser-capture microdis-sectiontoaddressthechallengeofCTCheterogeneity.Our results show the superiority of strategies 2 and 3 for futureclinicalapplicationsbutdonotallowustoidentify whichis the best. We determinedaglobal workflow to generatehigh-qualitylibrariesforWESinwhichrigorous qualification of WGA uniformity could reliably predict which CTC pool or single CTC was most likely toyield high-quality WES data. Our results differ from those reported by Lohretal. [16], whofoundthat ADO anda lackofsystematiccoverageinWGAproductsfromsingle EpCAM-positiveCTCsnecessitatedWESofmultiplesingle CTCs(n=19)forvariantdetermination.Althoughperhaps less pertinent for exploring tumor heterogeneity, the approach presentedhere,whichlimitedWEStoasmall numberofhigh-qualityCTCsamplesandintegratedCTC pools,istechnicallymorepracticalandcost-effectivefora clinicalsetting.

WeshowedthatWESforaverysmallnumberofpoolsor singleepithelialCTCsyieldedsignificantmutational similari-tiesfortheexomeofmatchedmetastasisbiopsies,including mutations inPCgenessuchasGRM8andTP53.Ourresults indicatethatCTCsequencing canprovideareasonableand feasibleproxyformetastasissamplingindisseminatedcancer. Interestingly,recurrentmutationsweresystematically iden-tified in both CTCs and matched metastasis biopsies, and sequencing more CTCs did not significantly extend the number of mutations identified. However, a fraction of mutations identifiedwithinthematchedmetastasisbiopsy werenotdetectedinCTCs,suggestingthatthesemutations maynothavetheabilitytodisseminateortransitinblood.By analogy, it could be hypothesized that these recurrent mutationsfoundinCTCsandsharedbymatchedmetastasis biopsies could play an important role in metastatic

progression.Alternatively,wecannotexcludethepossibility thatmutationsidentifiedwithinamatchedmetastasiscould bemissedinCTCsfordifferenttechnicalreasonsrelatedtothe sensitivityofourapproach.

Weestablishedastrictpipelineforvariantcallingandfor greaterreliabilityweonlyconsideredCTC-exclusive muta-tionsobservedinatleastthreeCTCsamples.CTC-exclusive mutationsincludedbothunknownmutationsthatwarrant

Fig.2–CharacterizationofvariantssharedbymetastasisbiopsiesandCTCs.(A)NumberofvariantssharedbymetastasisbiopsiesandCTCs.Fifteen CTCsamplesfromfourofthesevenpatientssharedatotalof136variantswithmatchedmetastasisbiopsysamples,ofwhich65wereunique.Foreach patientthesumofvariantssharedbytheirmetastasisbiopsyandCTCsisrepresentedbyhatchedbars.(B)Distributionofshareduniquevariants accordingtoCTCphenotype.Sixty-onevariants(93.9%)weredetectedinepithelialCTCs(E),twovariants(3.1%)wereonlydetectedinlargeCTCs(L), onevariant(1.5%)wasdetectedinbothinepithelialandlargeCTCs(E+L),andonevariant(1.5%)wasdetectedinepithelial,large,andhybridCTCs(E +L+H).(C)PercentageofvariantssharedbyCTCsandmetastasisbiopsies,oronlypresentinmetastasisbiopsies,forpatientsP7andP11.(D)Heatmap ofvariantsidentifiedinthemetastasisbiopsyandsharedbyCTCsinpatientP7.Some41%ofvariantswereexclusivelyidentifiedinthemetastasis biopsyand59%weresharedwithCTCs.TheP7-E-10pool(QS=7)exhibited51%sharedmutationswiththematchedbiopsy.ThepoolsP7-E-10and P7-E-9(QS=6)hadalmostthesamemutationsandshared54%ofmutationswiththematchedbiopsy.Thepositionsofvariantspresentinthemetastasis biopsyandnotcoveredinCTCsareshown.CTCs=circulatingtumorcells;QS=qualityscore.

furtherinvestigationandknownmutationswithapredicted deleterious impact on protein function in cancer-driver genessuchasHSP90AB1andKDM5B,aswellasingenesthat might have acrucial rolein thebiological specificity (eg, cytoskeletalremodeling)orsurvival(eg,DNArepair)ofCTCs. These CTC-exclusive mutations were mostly present in epithelialCTCsorinbothepithelialCTCsandCTCswithother morphologicalcharacteristics,aresultthatindirectly vali-datesthetruetumororiginofthesecells.Ourdatasuggest thatCTCsharboringtheseexclusivemutationsderivedfrom distinctmetastaticsites representedminorsubclonesthat wereundetectableinmetastaticbiopsies.

Whiletumortissuebiopsiesremainthegoldstandardfor validating or confirming genetic abnormalities found in CTCsorcirculatingtumorDNA(ctDNA),itisimportantto question the representability of a needle or single-site biopsy specimen. As illustrated here via analysis of CTC-exclusive variants,CTCs canbederived from lesionsthat were not biopsied and may contain a different genetic compositionthanthetumormaterialusedasareference.

Theseresultsdemonstrateforthefirsttimethat sequenc-ing of CTCs at the single-cell level can reveal metastasis mutationalcontentanddiversitythatareotherwise inacces-sible. Exome sequencing of cell-free tumor DNA has also

Table2–ExamplesofvariantssharedbetweenCTCsandmatchedmetastaticbiopsies,andCTC-exclusivevariants

Gene Transcript Database Somaticsingle-nucleotide

variants Prediction (Polyphen2Sift) Numberof CTCsamples PID

dbSNP COSMIC Codonchange AAchange

VariantssharedbyCTCsandmatchedmetastaticbiopsies

GRM8 NM_000845.2 - - c.2207C>T p.Ala736Val D,T 4(P11-E-4c ;P11-E-5b ; P11-E-7b ;P11-E-8b ) P11

TP53 NM_000546.5 - COSM10656 c.742C>T p.Arg248Trp D,D 4(P11-E-4c

;P11-E-5b

; P11-E-7b_;P11-E-8b₎

P11

PTEN NM_000314.4 - - c.236C>G p.Ala79Gly B,NA 1(P7-E-10c

) P7

CTC-exclusivevariantsobservedinatleast3CTCsamples

PPAPDC1A NM_001030059.1 c.301_305delATTAA p.Ile101fs NA,NA 3(P7-L-2a

;P7-LM-3a ; P7-LM-5a₎ P7 INO80 NM_017553.2 c.2867C>T p.Pro956Leu B,T 3(P7-L-1a ;P7-L-2a ; P7-LM-3a ) P7 C3orf30 NM_152539.2 rs199919487 c.1369G>A p.Glu457Lys D,D 4(P11-L-2a ;P11-E-4c ; P11-E-5b_;P11-E-6b₎ P11 ANKRD32 NM_032290.3 c.2992G>A p.Glu998Lys B,T 4(P11-L-2a ;P11-E-3c ; P11-E-5b ;P11-E-7b ) P11 EIF4G3 NM_001198801.1 c.2317C>A p.Gln773Lys B,T 3(P11-L-2a ;P11-E5b ; P11-E-8b ) P11

OTUD4 NM_001102653.1 c.1575A>C p.Leu525Phe P,T 3(P11-L-2a

;P11-E-4c

; P11-E-8b

)

P11

ZFYVE1 NM_021260.3 COSM470205 c.488C>A p.Thr163Lys B,T 3(P11-L-2a_;P11-E-3c_;

P11-E-6b₎

P11

CELA3B NM_007352.2 rs202129706 c.803C>A p.Ala268Glu B,T 5(P11-E-3c

;P11-E-4c ; P11-E-5b ;P11-E-6b ;P11-E-8b ) P11

PGBD2 NM_170725.2 c.441_442insT p.Phe149fs NA,NA 4(P11-E-4c

;P11-E-5b

; P11-E-7b_;P11-E-8b₎

P11

MACF1 NM_012090.5 rs202091916 COSM128604 c.15886A>C p.Thr5296Pro D,NA 4(P11-E-5b

;P11-E-6b ; P11-E-7b ;P11-E-8b ) P11

ABCB11 NM_003742.2 rs200687717 c.1628A>C p.Asp543Ala B,T 3(P11-E-4c

;P11-E-5b ; P11-E-7b ) P11 QPCT NM_012413.3 c.152C>T p.Ser51Leu B,D 3(P11-E-5b ;P11-E-7b ; P11-E-8b ) P11

FAM161A NM_001201543.1 rs200331923 c.1573C>A p.Gln525Lys B,T 3(P11-E-5b_;P11-E-7b_;

P11-E-8b

)

P11

NEDD9 NM_001142393.1 c.1504A>C p.Thr502Pro D,T 3(P11-E-5b

;P11-E-6b

; P11-E-8b

)

P11

AKNA NM_030767.4 rs200970909 c.3287A>C p.His1096Pro P,D 3(P11-E-5b_;P11-E-6b_;

P11-E-8b₎ P11 C2CD3 NM_001286577.1 rs200719890 c.5233T>G p.Tyr1745Asp D,D 3(P11-E-5b ;P11-E-7b ; P11-E-8b ) P11

IFNGR2 NM_005534.3 c.100C>G p.Gln34Glu B,T 3(P11-E-5b_;P11-E-7b_;

P11-E-8b

)

P11

LANCL2 NM_018697.3 rs201180232 COSM453253 c.953T>G p.Val318Gly D,D 3(P11-E-6b

;P11-E-7b

; P11-E-8b

)

P11

NDUFV2 NM_021074.4 rs72935225 c.401T>C p.Val134Ala B,T 3(P7-E-9c_;P7-E-10c_;

P7-H-11c

)

P7

CTC=circulatingtumorcells;AA=aminoacid;PID=patientidentity.

a _{Isolatedusingstrategy1.} b

Isolatedusingstrategy2.

c

Fig.3–CharacterizationofCTC-exclusivevariants.(A)MutationrateperMbinCTCsfrompatientP11forcriteriaofvariantsobservedinone,two,or threeCTCsamples.ForthecriterionofvariantsobservedintwoCTCsamples,themutationratesperMbpersamplewereclosetothevaluesforthe correspondingmetastasis,althoughforsomeCTCsamples,thisvaluewasgreater,indicatingthatWGAerrorscouldbeconserved.Forthecriterionof variantsobservedinthreeCTCsamples,themutationratesperMbpersamplewerelessthanorequaltothevaluesformatchedmetastases.(B) HeatmapofCTC-exclusivevariantsinpatientP11.Of101CTCvariants,81(80.2%)weredetectedintwoCTCsamples,15(14.8%)inthreeCTCsamples, four(4%)infourCTCsamples,andone(1%)infiveCTCsamples.ThepositionsofvariantsnotcoveredinCTCsamplesareshown.(C)Distributionof uniqueCTC-exclusivevariantspresentinatleastthreeCTCsamplesaccordingtotheCTCsamplephenotype.Twenty-twouniqueCTC-exclusivevariants weredetectedinatleastthreeCTCsamples;55%hadanexclusivelyepithelialphenotype,while45%haddifferentphenotypes.CTCs=circulatingtumor cells;WGA=whole-genomeamplification.

demonstrated concordance of mutations with metastasis biopsies [23]. Although ctDNA analyses may indeed offer advantages interms ofsimplicity, sequencing of CTCs can provideunique additional and complementary information such as EMT status and cancer stem-cell phenotype

[24,25]. The presence of different and multiple genetic

alterations canbe identified withinthe same CTC,offering thepossibilitytoexploreinter-andintratumorheterogeneity andevolution.ItisthereforeanticipatedthatCTCsandctDNA arecomplementaryintheirclinicalutility.Whiletumortissue biopsies remain the gold standard for confirming genetic abnormalitiesfoundinCTCs,itisimportanttoquestionthe representabilityofaneedleorsingle-sitebiopsy.Ourfindings emphasizethepotentialandclinicalutilityofdetectingsuch mutations in CTCs via minimally invasive blood draws for

optimaltherapyselection,precisionmedicine,andtreatment resistance options. A study quantifying digital pathology features ofCTCs recently showedthatlow CTCphenotypic heterogeneitywasassociatedwithbetteroverallsurvival(OS) among patients treated with androgen receptor signaling inhibitors, whereas highheterogeneity wasassociated with betterOSamongpatientstreatedwithtaxanechemotherapy

[26].Beyondidentifyingpotentialtargets,WESatthe single-celllevelcouldquantifyheterogeneityandhelpin decision-makingformCRPCpatients.

5. Conclusions

ThispioneeringstudyemphasizesthepotentialofCTCsto represent metastasis mutational content and tumor

Fig.4–HierarchicalclusteringofCTCsandmatched-metastasisbiopsies.(A)HierarchicalclusteringofCTCandmatchedmetastasisbiopsysamplesin patientP7accordingtomutationalsimilaritiesbetweenCTCsandthematchedmetastasisbiopsy,andbetweendifferentCTCsamplesforvariants observedinatleastthreesamples.Thenumberofsharedvariantsisrepresentedbytheredcolorgradientfromdarkred(highnumberofvariants)to white(nosharedvariant).(B)HierarchicalclusteringofCTCsandmatchedmetastasisbiopsyinpatientP11accordingtothesamerepresentation. CTCs=circulatingtumorcells.

diversitythatareotherwiseinaccessible.Byoffering real-time monitoring of a constantly evolving disease and detecting potentially critical mutations via minimally invasiveblooddraws,CTCsequencingcanserveanunmet needforoptimaltherapyselectionandprecisionmedicine.

Authorcontributions:FrançoiseFaracehadfullaccesstoallthedatain thestudyandtakesresponsibilityfortheintegrityofthedataandthe accuracyofthedataanalysis.

Studyconceptanddesign:Faugeroux,Massard,Fizazi,Loriot,Julien,Soria, Farace.

Acquisition of data: Faugeroux, Pailler, Marcaillou, Billiot, Oulhen, Rameau,Tibbe.

Analysisandinterpretationofdata:Faugeroux,Pailler,Pierron,Tourlet, Vielh,Loriot,Julien,Farace.

Draftingofthemanuscript:Faugeroux,Lefebvre,Taylor,Farace. Criticalrevisionofthemanuscriptforimportantintellectualcontent:Pailler, Pierron,Tourlet,Taylor,Soria,Fizazi,Loriot,Julien.

Statisticalanalysis:Lefebvre,Dogan.

Obtainingfunding:Pierron,Fizazi,Julien,Farace.

Administrative,technical,ormaterialsupport:Marcaillou,Billiot,Oulhen, NgoCamus,Laplace-Builhé.

Supervision:Farace,Loriot,Julien. Other:None.

Financial disclosures: Françoise Farace certifies that all conflicts of interest, including specific financial interests and relationships and affiliationsrelevanttothesubjectmatterormaterialsdiscussedinthe manuscript(eg,employment/affiliation,grantsorfunding, consultan-cies,honoraria,stockownershiporoptions,experttestimony,royalties, orpatents_filed,received,orpending),arethefollowing:None. Funding/Supportandroleofthesponsor:Thisstudywassupportedby IPSEN-Innovation,VincentFaugerouxissupportedbyFondationpourla Recherche Médicale (no. FDT20160435543). Céline Lefebvre was supported by SIRIC Socrate. Emma Pailler was supported by LabEx LERMIT (no. ANR-10-LABX-0033-LERMIT) and Fondation pour la RechercheMédicale(no.FDT20150532072).Thesponsorsplayedarole inthedesignandconductofthestudy;dataanalysisandinterpretation; andpreparation,review,andapprovalofthemanuscript.

Conflictofinterestdisclosurestatement:Authorsdisclosenopotential conflictsofinterest.

Acknowledgments:Wethankthepatientsandtheirfamilies,aswellasE. MartinforfruitfuldiscussionandM.Letexierforhelpinbioinformatics analysis.

AppendixA. Supplementarydata

Supplementarydataassociatedwiththisarticlecanbe found,intheonlineversion,atdoi:10.1016/j.eururo.2018. 11.052.

References

[1] MassardC,MichielsS,FerteC,etal.High-throughputgenomicsand clinicaloutcomeinhard-to-treatadvancedcancers:resultsofthe MOSCATO01trial.CancerDiscov2017;7:586–95.

[2] CancerGenomeAtlasResearchNetwork.Themoleculartaxonomy ofprimaryprostatecancer.Cell2015;163:1011–25.

[3]RobinsonD,VanAllenEM,WuYM,etal.Integrativeclinical geno-micsofadvancedprostatecancer.Cell2015;161:1215–28.

[4]WedgeDC,GundemG,MitchellT,etal. Sequencingofprostate cancersidentifiesnewcancergenes,routesofprogressionanddrug targets.NatGenet2018;50:682_–92.

[5]BoutrosPC,FraserM,HardingNJ,etal.Spatialgenomic heteroge-neity within localized, multifocal prostate cancer. Nat Genet 2015;47:736–45.

[6]GundemG,VanLooP,KremeyerB,etal.Theevolutionaryhistoryof lethalmetastaticprostatecancer.Nature2015;520:353–7.

[7]HongMK,MacintyreG,WedgeDC,etal.Trackingtheoriginsand driversofsubclonalmetastaticexpansioninprostatecancer.Nat Commun2015;6:6605.

[8]Alix-PanabieresC,PantelK.Challengesincirculatingtumourcell research.NatRevCancer2014;14:623–31.

[9]MiyamotoDT,SequistLV,LeeRJ.Circulatingtumour cells-monitor-ing treatment response inprostate cancer. NatRev Clin Oncol 2014;11:401_–12.

[10] AttardG,SwennenhuisJF,OlmosD,etal.CharacterizationofERGAR andPTENgenestatusincirculatingtumorcellsfrompatientswith castration-resistantprostatecancer.CancerRes2009;69:2912–8.

[11] deBonoJS,ScherHI,MontgomeryRB,etal.Circulatingtumorcells predictsurvivalbene_fitfromtreatmentinmetastatic castration-resistantprostatecancer.ClinCancerRes2008;14:6302_–9.

[12] Nieto MA, Huang RY, Jackson RA, Thiery JP. EMT 2016. Cell 2016;166:21–45.

[13] LecharpentierA,VielhP,Perez-MorenoP,etal.Detectionof circu-latingtumourcellswithahybrid(epithelial/mesenchymal) phe-notypeinpatientswithmetastaticnon-smallcelllungcancer.BrJ Cancer2011;105:1338_–41.

[14] MassardC,OulhenM,LeMoulecS,etal.Phenotypicandgenetic heterogeneity of tumor tissue and circulating tumor cells in patientswith metastatic castration-resistant prostate cancer:a report from the PETRUS prospective study. Oncotarget 2016;7:55069–82.

[15] KrebsMG,HouJM,SloaneR,etal.Analysisofcirculatingtumorcells inpatientswithnon-smallcelllungcancerusingepithelial marker-dependent and -independent approaches. J Thorac Oncol 2012;7:306–15.

[16] LohrJG,AdalsteinssonVA,CibulskisK,etal.Whole-exome sequenc-ingofcirculatingtumorcellsprovidesawindowintometastatic prostatecancer.NatBiotechnol2014;32:479–84.

[17] HeitzerE,AuerM,GaschC,etal.Complextumorgenomesinferred fromsinglecirculatingtumorcellsbyarray-CGHand next-genera-tionsequencing.CancerRes2013;73:2965–75.

[18]Polzer B,MedoroG, PaschS,etal.Molecularprofilingofsingle circulatingtumorcellswithdiagnosticintention.EMBOMolMed 2014;6:1371–86.

[19] AdalsteinssonVA,LoveJC.Towardsengineeredprocessesfor se-quencing-basedanalysisofsinglecirculatingtumorcells.CurrOpin ChemEng2014;4:97–104.

[20] Farace F, MassardC, VimondN, et al. A direct comparison of CellSearchandISETforcirculatingtumour-celldetectioninpatients withmetastaticcarcinomas.BrJCancer2011;105:847–53.

[21] SwennenhuisJF,TibbeAG,StevensM,etal.Self-seedingmicrowell chipfortheisolationandcharacterizationofsinglecells.LabChip 2015;15:3039–46.

[22] HofmanV,IlieMI,LongE,etal.Detectionofcirculatingtumorcells asaprognosticfactorinpatientsundergoingradicalsurgeryfor non-small-celllungcarcinoma:comparisonoftheefficacyofthe CellSearchassayandtheisolationbysizeofepithelialtumorcell method.IntJCancer2011;129:1651_–60.

[23] MurtazaM,DawsonSJ,TsuiDW,etal. Non-invasiveanalysisof acquiredresistancetocancertherapybysequencingofplasmaDNA. Nature2013;497:108–12.

[24] BaccelliI, Schneeweiss A,Riethdorf S, et al. Identificationof a populationofblood circulating tumor cells from breast cancer patientsthatinitiatesmetastasisinaxenograftassay.Nat Biotech-nol2013;31:539–44.

[25] PaillerE,AdamJ,BarthelemyA,etal.Detectionofcirculatingtumor cellsharboringauniqueALKrearrangementinALK-positive non-small-celllungcancer.JClinOncol2013;31:2273–81.

[26] ScherHI,GrafRP,SchreiberN,etal.Phenotypicheterogeneityof circulatingtumorcellsinformsclinicaldecisionsbetweenAR sig-nalinginhibitorsandtaxanesinmetastaticprostatecancer.Cancer Res2017;77:5687–98.