selecting short dna fragments in plasma improves detection ...plasma dna resulted in a median of 1.5...

Selecting short DNA fragments in plasma improvesdetectionofcirculatingtumourDNA

Florent Mouliere1,2,*, Anna M. Piskorz1,2,*, Dineika Chandrananda1,2,*, ElizabethMoore1,2,3,*, James Morris1,2, Christopher G. Smith1,2, Teodora Goranova1,2, KatrinHeider1,2,RichardMair1,2,AnnaSupernat1,2,4,IoannisGounaris1,2,3,SusanaRos1,2,JonathanC. M. Wan1,2, Mercedes Jimenez-Linan2,3, Davina Gale1,2, Kevin Brindle1,2,5, Charles E.Massie1,2,ChristineA.Parkinson1,2,3,6,7,JamesD.Brenton1,2,3,6,7,#,NitzanRosenfeld1,2,#

*co-firstauthors;

#correspondingauthors:[email protected];[email protected]

1. CancerResearchUKCambridgeInstitute,UniversityofCambridge,Cambridge,UK2. CancerResearchUKMajorCentre–Cambridge,CancerResearchUKCambridge

Institute,Cambridge,UK3. CambridgeUniversityHospitalsNHSFoundationTrust,Cambridge,UK4. MedicalUniversityofGdansk,Gdansk,Poland5. DepartmentofBiochemistry,UniversityofCambridge,Cambridge,UK6. DepartmentofOncology,Hutchison/MRCResearchCentre,UniversityofCambridge,

Cambridge,UK7. NIHRCambridgeBiomedicalResearchCentre,Cambridge,UK

Introductoryparagraph:

Non-invasive analysis of cancer genomes using cell-free circulating tumour DNA (ctDNA) is beingwidelyimplementedforclinicalindications.ThesensitivityfordetectingthepresenceofctDNAandgenomic changes in ctDNA is limited by its low concentration compared to cell-free DNA of non-tumour origin. We studied the feasibility for enrichment of ctDNA by size selection, in plasmasamples collected before and during chemotherapy treatment in 13 patientswith recurrent high-grade serous ovarian cancer. We evaluated the effects using targeted and whole genomesequencing. Selecting DNA fragments between 90-150 bp before analysis yielded enrichment ofmutated DNA fraction of up to 11-fold. This allowed identification of adverse copy numberalterations, includingMYCamplification,otherwisenotobserved.Sizeselectionallowsdetectionoftumour alterations masked by non-tumour DNA in plasma and could help overcome sensitivitylimitationsofliquidbiopsyforapplicationsinearlydiagnosis,detectionofminimalresidualdisease,andgenomicprofiling.

certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was notthis version posted May 5, 2017. ; https://doi.org/10.1101/134437doi: bioRxiv preprint

https://doi.org/10.1101/134437

Text:

AnalysisofcirculatingtumourDNA(ctDNA)bynon-invasivesamplingofcell-freeDNAfromplasmaisnowbecominganimportanttoolinoncologyformolecularstratification,monitoringtumourburdenandanalysisofgenomicevolutionduringtreatment1–3.AnalysisofctDNAistechnicallychallenging,asctDNAisoftenpresentinlowconcentrationsinplasmaandismixedwithcell-freeDNA(cfDNA)ofnon-cancerous origin, which is generally present at much higher concentrations. In patients withadvancedcancers, themedianconcentrationofctDNAcanreach10%ormoreof the totalcfDNA,but this fraction ismuch lower in earlier stage cancer, and ctDNAmay rapidly decrease followinginitiationofsystemictreatmentorsurgery1,2,4,5.Variousstrategieshavebeenproposedto improvethesensitivityofctDNAanalysis2.Suchmethodsgenerally focusona small subsetof thegenome,such as hot-spot PCR-based assays or ultra-deep sequencing across gene panels2,6–9. Recentobservations that ctDNA fragmentsmay be shorter than non-tumour cfDNA in plasma has led tosuggestions that these differences may be exploited to enrich for the tumour-specific signal inplasmaDNA10–14.In-silicoanalysisofctDNAsizedifferenceshasbeenusedtoenhancethesignalforchromosomal changes13. However, physical size selection to filter out non-tumour DNA prior tolarge-scalegenomicsequencinghasnotbeendemonstrated. Therefore,wetested thehypothesisthat selecting DNA fragments of specific sizes could improve the sensitivity of detecting genomicalterationsincfDNAfromplasmaofcancerpatients,enablingthedetectionofpointmutationsandcopynumberalterationsthatarepreviouslyundetectable11.

Previousreportsusingpaired-endsequencingreadsrevealedthatcell-freeDNAismostlydistributedaround amode at 167bp, a length that could correspond to the chromatosome (corehistones +linker)15,16.Thissizedistributionpatternischaracteristicofacaspase-dependantcleavage,thereforeit was hypothesized that apoptosis releases a large fraction of cfDNA into the bloodstream14,15,17.Previous studies in non-invasive prenatal testing have explored the potential of size selection forenrichingthefetalDNAfractioninmaternalplasmawithbothphysicaland in-silicomethods14,18–20.However, this analysis of the size distribution cannot be easily generalised to tumour-derivedfragments as the characterisation of ctDNA-specific patterns requires analysing fragment sizes ofDNA with tumour-derived alterations13. Plasma samples from xenografted animal models haveshownctDNAtobehighlyfragmentedbelow167bp10andthisdistributionwithamodeat145bphas been then confirmed with PCR, atomic force microscopy and recently with whole-genomesequencing12,21. If the distribution differs between tumour-derived and non-tumour derivedfragments, sieving fragmentsby their size could reduce the (oftenoverwhelming) fractionofnon-tumourcfDNAandimprovethesignaltonoiseratioindownstreamanalysis.

Wefirstanalysedpaired-endreadsfromshallowwholegenomesequencing(sWGS)ofplasmaDNAfrom animal models xenografted with a human ovarian cancer cells, and confirmed that thedistributionofctDNAdifferedinthismodelfromnon-tumourcfDNA(Fig.1a,b).ctDNAinthismodelwasenrichedinthesizerangebetween90and150bp,whilenon-tumourcfDNAwasdominantatsizesgreaterthan150bp,andpeakedat~166bp,similartopreviousobservationsinanimalmodelsandinhumansamples12,15,21,22.Basedontheseobservationsweusedanautomatedelectrophoresisagarose gel selectionmethod to isolate DNA fragments between 90 and 150 bp for downstreamanalysis.Wesequencedsize-selectedDNAusingbothsWGSandtagged-amplicondeepsequencing(TAm-Seq23), in 26 samples collected from 13 patients with high-grade serous ovarian cancer(HGSOC),andcomparedtheresults to thoseof thesamesampleswithoutsizeselection.Foreach


https://doi.org/10.1101/134437

patient, two plasma samples were collected: one pre-treatment, when levels of ctDNA weregenerallyhigh,andanotherseveralweeksafterthestartoftreatment,whenlevelsofctDNAwereoftenmuchlowerduetotreatment24.Analysisofthedistributionoffragmentlengthsafter in-vitrosizeselectionandsWGS indicatedthat96%ofresultingreadswere intheselectedrange(Fig.1c).Fortwoofthepatients,thedistributionoffragmentsizesobtainedbysWGSwithoutsize-selectionexhibitedadegradedpatternofcfDNA,withoutaprominentpeakat166bporthe10-bpincrementpeaks,whichhasbeenobservedpreviously12,13,15,and ispresent inall theothercases inthisstudy(Suppl.Fig.1).

Analysis of somatic copy number aberrations (SCNAs) was carried out with sWGS on all plasmasamplesbeforeandaftersizeselectionofDNAfragmentsbetween90and150bp(Fig.2a).OnecaseisexemplifiedinFig.2band2c(seefurtherdatainSupplementaryFig.2).Withoutsizeselection,asmallnumberofSCNAsweredetectedwithsWGS inDNA isolated fromplasmacollected3weeksafter initiationof treatment frompatientOV04-83 (Fig.2b). These included focal amplifications inchromosomes 8p, 14p, 17, and 19q that were observed in this sample at very low levels (Suppl.Table2).AnalysisofthesameDNAsamplefollowingsize-selectionforshortDNAfragmentsrevealedanincreaseinthelevelofthesedetectedSCNAs,inadditiontomultiplesotherSCNAsthatwerenotobservedwithout size selection (Fig. 2c). The samepatternof SCNAsand focal amplificationswasobserved inDNA fromplasmacollected from the samepatient before initiationof the treatment,whenthefractionoftumourDNAintheplasmawashigher(Suppl.Fig.2).

The relative copy number signals in plasmaDNA, across a list of 29 genes frequentlymutated inHGSOC, were compared with and without size selection of DNA from plasma samples collectedduringtreatmentacrossthecohortof13patients.Thisshowedthata largenumberofSCNAsthatwere not observedwithout size selection, could be detected after size selection for shorter DNAfragments,notablyasamplifications inkeygenessuchasNF1,PARP2andMYC (Fig.2dandSuppl.Fig.3).MoreSCNAscouldbedetectedaftersizeselectionin11/13patients,andtheabsolutelevelofthelog2ratiowassignificantlyincreasedaftersizeselection(t-test,p=7.72.10-9).The2patientsforwhomtheSCNAsignaldidnotincreaseexhibitedadegradedpatternofcfDNA,whichcouldexplainwhythesizeselectionhadnotenrichedforctDNA(Fig.2dandSuppl.Fig.2).

WenextassessedthedetectionofSCNAsandpointmutations,inplasmasamplesofthe13patientscollected at baseline (before treatment initiation) and 3 weeks after initiation of chemotherapytreatments,withandwithoutsizeselectionoftheplasmaDNA(Fig.3aandSuppl.Table1).Thedatafrom the baseline samples, where ctDNA levels are generally higher24, was used to identify andconfirmgenomicchanges,whichwerethenstudiedinthesamplesafterinitiationoftreatment,withgenerally, lower levels of ctDNA. The amplitudes of the absolute log2ratio for the SCNAs werehigher,andtheconcordanceofthealterationsdetectedbetweenthebaselineandpost-treatmentsampleswereimproved,withsizeselectionoftheplasmaDNA(Fig.3bandSuppl.Fig.3).

UsingsWGSdata,weconvertedtheamplitudeoftheCNAsintoaquantitativemetriccalledt-MAD(trimmedMean Absolute Deviation from copy-number neutrality, seeMethods). Size selection ofplasmaDNAresultedinamedianof1.5fold(n=26)increaseinthet-MADscore(figure3cand3d).However, in thesamplescollectedafter the initiationof treatment,whenctDNAcontentwas low,thegenomewideenrichmentwashigher(Fig.3d),withamedianincreaseofthet-MADscoreof2.9fold(range:0.6-4.5fold),exceptfortwosamples.Forthosetwosamplesweobservedadecreasein


https://doi.org/10.1101/134437

thet-MADscore,andananalysisofthesizeprofilebeforeselectionforthesesamplesrevealedthattheDNAhadbeenheavilyfragmented,whichcouldexplainwhythesizeselectiondidnotresult inenrichmentforthesecases(Suppl.Fig.1).Inthisdataset,wedidnotidentifyadifferentialeffectofsize selection on the recovery of specific alterations, suggesting that there is a global genomeenrichmentpostsizeselection.Additionally,sizeselectionnotably ledtoan increaseddetectionofdeletions(Suppl.Fig.2).Analysiswithgreatersequencingdepth,integrationofsamplesfromothercancertypesanddifferentstagesofthediseasewouldhelptoextendourobservationsandexpandfurtherourunderstandingofctDNAbiologyandfragmentation.

In order to confirm that enrichment for tumour DNA could be observed irrespective of thesequencingapproach,wefurtheranalysedthemutantallelefractionsinthesamplesusingTagged-Amplicon Sequencing (TAm-Seq). We detected a relative enrichment in the ctDNA fraction in allsamplesexhibitingatypicalpatternofcfDNAfragmentation,withamodeoffragmentdistributionat166bp, before size selection (Suppl. Fig. 2). The enrichment effect was below 2-fold in samplescollectedpre-treatment,whenthectDNAfractionswerehighinplasma(20%-50%allelefractionsforTP53mutationsasdetectedbyTAm-Seq)(Fig.3eand3f).Enrichmentofthetumourfractionbysizeselection was much greater, between 5-fold and 11-fold for most samples, in samples collectedapproximately3weeksafter initiationof treatment,when levelsof ctDNA (without size selection)werelow(rangingfrom

librarypreparationforsequencingmayenablemoreeffectiverecoveryofshortDNAfragmentsfromplasma samples, which have led to new observations on cfDNA size distributions28,30,32. SuchmethodsshouldbefurtherinvestigatedtodetermineifthesecouldhelprecovermorectDNA.

The size selectionprocesswedemonstratedhere isbasedon inherent characteristicsof ctDNA incomparison to cfDNA and does not require alteration of these fragments. The enrichment weobserved is therefore compatible with any downstream genomic analysis, from locus-specific towider genomic sequencing. This work shows that sWGS (and by extension, whole exomesequencing)canbeperformedonplasmaDNAsampleswith lowctDNAcontent,andthat thiscanfacilitatethecharacterisationofmutationspresentinplasmaatlowerallelefractionsandwithlowersequencing depth. The compatibility of the cfDNA fragment size selection with wide-scale andsensitivegenomicanalysiscouldunlockthepotentialofliquidbiopsiesforthediagnosisofcanceratanearlierstage,andforthedetectionofminimalresidualdisease.

Methods:

Patientsandsamplepreparation.13patientswererecruitedaspartofprospectiveclinicalstudiesatAddenbrooke’sHospital,Cambridge,UK, approvedby the local researchethics committee (RECreferencenumbers07/Q0106/63,08/H0306/61and07/Q0106/63).Written informedconsentwasobtainedfromallpatientsandbloodsampleswerecollectedbeforeandafterinitiationoftreatmentwith chemotherapeutic agents. DNA was extracted from 2 mL of plasma using the QIAampcirculatingnucleicacidkit(Qiagen)accordingtothemanufacturer’sinstructions.

Size selection. Between 8-10 ng of DNAwere loaded into a 3% agarose cassette (HTC3010, SageBioscience) and size selection was performed on a PippinHT (Sage Bioscience) according to themanufacturer’sprotocol.

TAm-Seq. Tagged-Amplicon Sequencing libraries were prepared as previously described23, usingprimersdesignedtoassesssinglenucleotidevariants(SNV)andsmallindelsacrossselectedhotspotsandtheentirecodingregionsofTP53.LibrariesweresequencedusinganHiSeq4000(Illumina).

sWGS. Indexed sequencing libraries were prepared using a commercially available kit (ThruPLEX-PlasmaSeq,RubiconGenomics).Librarieswerepooled inequimolaramountsandsequencedonaHiSeq4000(Illumina)generating150-bppaired-endreads.Sequencedatawasanalysedusinganin-housepipelinethatconsistsofthefollowing;Pairedendsequencereadswerealignedtothehumanreference genome (GRCh37) using BWA-mem following the removal of contaminating adaptersequences33. PCR andoptical duplicatesweremarkedusingMarkDuplicates (Picard Tools) featureand thesewereexcluded fromdownstreamanalysis alongwith readsof lowmappingquality andsupplementaryalignments.

SCNA analysis: Copy number analysis was performed in R using a modification of the QDNAseqpipeline34,asfollow:sequencereadswereallocatedintoequallysized(50kbp)non-overlappingbinsthroughout the length of the genome. Read counts in each bin were corrected to account forsequenceGCcontentandmappability,andbinsoverlapping ‘blacklisted’regions (ENCODEproject)were excluded from downstream analysis. After median normalisation of the counts, bins were


https://doi.org/10.1101/134437

segmentedusingbothCircularBinarySegmentationandLocus-awareCircularBinarySegmentationalgorithms,andanaveragedlog2Rvalueperbinwascalculated.Thet-MADscoreiscalculatedastheaveragedabsolutedeviationfromlog2R=0afterfirsttrimmingbincountsgreaterthan4standarddeviationsfromthemeancountacrossallgenomicregions.

References:

1. Siravegna,G.,Marsoni,S.,Siena,S.&Bardelli,A.Integratingliquidbiopsiesintothemanagementofcancer.Nat.Rev.Clin.Oncol.(2017).doi:10.1038/nrclinonc.2017.14

2. Wan,J.C.M.etal.Liquidbiopsiescomeofage:towardsimplementationofcirculatingtumourDNA.Nat.Rev.Cancer(2017).doi:10.1038/nrc.2017.7

3. Murtaza,M.etal.Non-invasiveanalysisofacquiredresistancetocancertherapybysequencingofplasmaDNA.Nature497,108–112(2013).

4. Bettegowda,C.etal.DetectionofCirculatingTumorDNAinEarly-andLate-StageHumanMalignancies.Sci.Transl.Med.6,224ra24-224ra24(2014).

5. Dawson,S.-J.etal.AnalysisofCirculatingTumorDNAtoMonitorMetastaticBreastCancer.N.Engl.J.Med.368,1199–1209(2013).

6. Diehl,F.etal.Detectionandquantificationofmutationsintheplasmaofpatientswithcolorectaltumors.Proc.Natl.Acad.Sci.102,16368–16373(2005).

7. Taly,V.etal.MultiplexPicodropletDigitalPCRtoDetectKRASMutationsinCirculatingDNAfromthePlasmaofColorectalCancerPatients.Clin.Chem.59,1722–1731(2013).

8. Newman,A.M.etal.IntegrateddigitalerrorsuppressionforimproveddetectionofcirculatingtumorDNA.Nat.Biotechnol.34,547–555(2016).

9. Khodakov,D.,Wang,C.&Zhang,D.Y.Diagnosticsbasedonnucleicacidsequencevariantprofiling:PCR,hybridization,andNGSapproaches.Adv.DrugDeliv.Rev.105,3–19(2016).

10. Mouliere,F.etal.HighFragmentationCharacterizesTumour-DerivedCirculatingDNA.PLoSOne6,e23418(2011).

11. Mouliere,F.&Rosenfeld,N.Circulatingtumor-derivedDNAisshorterthansomaticDNAinplasma.Proc.Natl.Acad.Sci.112,3178–3179(2015).

12. Underhill,H.R.etal.FragmentLengthofCirculatingTumorDNA.PLOSGenet.12,e1006162(2016).

13. Jiang,P.etal.LengtheningandshorteningofplasmaDNAinhepatocellularcarcinomapatients.Proc.Natl.Acad.Sci.112,E1317–E1325(2015).

14. Jiang,P.&Lo,Y.M.D.TheLongandShortofCirculatingCell-FreeDNAandtheInsandOutsofMolecularDiagnostics.TrendsGenet.32,360–371(2016).

15. Lo,Y.M.D.etal.MaternalPlasmaDNASequencingRevealstheGenome-WideGeneticandMutationalProfileoftheFetus.Sci.Transl.Med.2,61ra91-61ra91(2010).

16. Chandrananda,D.etal.High-resolutioncharacterizationofsequencesignaturesduetonon-randomcleavageofcell-freeDNA.BMCMed.Genomics8,29(2015).

17. Jahr,S.etal.DNAfragmentsinthebloodplasmaofcancerpatients:quantitationsandevidencefortheiroriginfromapoptoticandnecroticcells.CancerRes.61,1659–65(2001).


https://doi.org/10.1101/134437

18. Yu,S.C.Y.etal.Size-basedmoleculardiagnosticsusingplasmaDNAfornoninvasiveprenataltesting.Proc.Natl.Acad.Sci.U.S.A.111,8583–8(2014).

19. Lun,F.M.F.etal.NoninvasiveprenataldiagnosisofmonogenicdiseasesbydigitalsizeselectionandrelativemutationdosageonDNAinmaternalplasma.Proc.Natl.Acad.Sci.U.S.A.105,19920–5(2008).

20. Minarik,G.etal.UtilizationofBenchtopNextGenerationSequencingPlatformsIonTorrentPGMandMiSeqinNoninvasivePrenatalTestingforChromosome21TrisomyandTestingofImpactofInSilicoandPhysicalSizeSelectiononItsAnalyticalPerformance.PLoSOne10,e0144811(2015).

21. Mouliere,F.,ElMessaoudi,S.,Pang,D.,Dritschilo,A.&Thierry,A.R.Multi-markeranalysisofcirculatingcell-freeDNAtowardpersonalizedmedicineforcolorectalcancer.Mol.Oncol.8,927–941(2014).

22. Thierry,A.R.etal.OriginandquantificationofcirculatingDNAinmicewithhumancolorectalcancerxenografts.NucleicAcidsRes.38,6159–6175(2010).

23. Forshew,T.etal.NoninvasiveIdentificationandMonitoringofCancerMutationsbyTargetedDeepSequencingofPlasmaDNA.Sci.Transl.Med.4,136ra68-136ra68(2012).

24. Parkinson,C.A.etal.ExploratoryAnalysisofTP53MutationsinCirculatingTumourDNAasBiomarkersofTreatmentResponseforPatientswithRelapsedHigh-GradeSerousOvarianCarcinoma:ARetrospectiveStudy.PLOSMed.13,e1002198(2016).

25. Murtaza,M.etal.MultifocalclonalevolutioncharacterizedusingcirculatingtumourDNAinacaseofmetastaticbreastcancer.Nat.Commun.6,8760(2015).

26. Heitzer,E.etal.Tumor-associatedcopynumberchangesinthecirculationofpatientswithprostatecanceridentifiedthroughwhole-genomesequencing.GenomeMed.5,30(2013).

27. Belic,J.etal.RapidIdentificationofPlasmaDNASampleswithIncreasedctDNALevelsbyaModifiedFAST-SeqSApproach.Clin.Chem.61,838–849(2015).

28. Snyder,M.W.etal.Cell-freeDNAComprisesanInVivoNucleosomeFootprintthatInformsItsTissues-Of-Origin.Cell164,57–68(2016).

29. Ulz,P.etal.Inferringexpressedgenesbywhole-genomesequencingofplasmaDNA.Nat.Genet.48,1273–1278(2016).

30. Burnham,P.etal.Single-strandedDNAlibrarypreparationuncoverstheoriginanddiversityofultrashortcell-freeDNAinplasma.Sci.Rep.6,27859(2016).

31. Shema,E.etal.Single-moleculedecodingofcombinatoriallymodifiednucleosomes.Science(80-.).352,(2016).

32. Vong,J.S.L.etal.Single-StrandedDNALibraryPreparationPreferentiallyEnrichesShortMaternalDNAinMaternalPlasma.Clin.Chem.(2017).doi:10.1373/clinchem.2016.268656

33. Li,H.&Durbin,R.FastandaccurateshortreadalignmentwithBurrows-Wheelertransform.Bioinformatics25,1754–1760(2009).

34. Scheinin,I.etal.DNAcopynumberanalysisoffreshandformalin-fixedspecimensbyshallowwhole-genomesequencingwithidentificationandexclusionofproblematicregionsinthegenomeassembly.GenomeRes.24,2022–2032(2014).


https://doi.org/10.1101/134437

Notes:Acknowledgements:TheauthorswouldliketothankallmembersoftheRosenfeldLabandBrentonLab for help and constructive discussion, in particular FrancescoMarass,WendyN. Cooper, KevalPatel, JennyP.Y.Chan,MareikeThompson,LiseBarleboAhlbornand IrenaHudecovà.TheauthorswouldliketoalsothanktheCancerResearchUKCambridgeInstitutecorefacilitiesfortheirsupport,inparticular thegenomicsandbiorepository facilities.Wewould like toacknowledgeourpatientsand our caregivers, and the help and support of the research nurses, trial staff and the staff atAddenbrooke’s Hospital. In particular, we would like to acknowledge Charlotte Hodgkin, HeatherBiggs and Karen Hosking. We would like also to acknowledge the support of The University ofCambridge,CancerResearchUK(grantnumberA11906,A20240)andHutchisonWhampoaLimited.The research leading to these results has received funding from the European Research Councilunder the European Union's Seventh Framework Programme (FP/2007-2013) / ERC GrantAgreement n. 337905. This research is also supported by TargetOvarian Cancer and theMedicalResearchCouncilthroughtheirJointClinicalResearchTrainingFellowshipforDrMoore.Thefundershadnorole instudydesign,datacollectionandanalysis,decisiontopublish,orpreparationofthemanuscript.

Authorcontributions:FM,AMP,DC,EM, JDBandNRconceptualisedanddesigned thestudy;FM,AMP,EM,KH,CGS,JCMW,DG,RM,TG,AS,IG,CAPhaveperformedexperimentsandcollecteddata;DChasconceptualisedanddesignedthet-MADindexandperformedsWGSbioinformaticsanalysis;JMperformedTAm-Seqbioinformatics analysis;RMandSRhavedesigned theanimalmodel;MJLperformedhistopathologyrevision;FM,AMP,DC,EM,JDBandNRhavewrittenthemanuscript;allco-authorshavecriticallyreviewedthemanuscript;FM,AMP,DC,JDBandNRsupervisedthestudy.

Author information: NR, JDB and DG are cofounders, shareholders and officers/consultants ofInivataLtd,acancergenomicscompanythatcommercialisesctDNAanalysis.InivataLtdhadnorolein the conceptualisation, study design, data collection and analysis, decision to publish orpreparation of the manuscript. NR and FM are co-inventors of patent applications that describemethods for theanalysisofDNA fragmentsandapplicationsof circulating tumourDNA.Otherco-authorshavedeclarednoconflictofinterests.


https://doi.org/10.1101/134437

Figure1:

Figure1:PlasmaDNAoriginatingfromtumourandnon-tumourcellshavedifferentsizes,enablingspecific enrichment for ctDNA. a. Using an animal model with xenografted cells enabled thediscrimination of DNA fragments released by the cancer cells (corresponding to the humanDNA)from the DNA fragments released by the wild-type cells (corresponding to the rat DNA). b. Sizedistribution,assessedbysWGS,ofDNAfragmentsfromaplasmasampleofaratxenograftedwithahumanglioblastomatumour.c.SizedistributionofDNAfragmentsfrom26plasmasamplesincludedin this study, assessed by sWGS. In green are the DNA fragments of the samples without size-selection,andinorangeaftersize-selection.Thetwodottedverticallinesindicatethesizeselectionrangebetween90bpand150bp.

sequencing andsequencing andalignmentalignment

human genome = tumour DNAhuman genome = tumour DNA

rat genome = wild-type DNArat genome = wild-type DNA

a

xenograftxenograft

selected fragment lengthsselected fragment lengthssize range enriched size range enriched for tumour DNAfor tumour DNA

b c

0.00

0.01

0.02

0.03

0.04

0.05

100 200 300Fragment size (bp)

Prop

ortio

n of

read

s

genome

human

mouse

0.00

0.01

0.02

0.03

0.04

0.05

100 200 300Fragment size (bp)

Prop

ortio

n of

read

s

statusno size selection

with size selection


https://doi.org/10.1101/134437

Figure2:

Figure 2: Recovery of short cfDNA fragments enriches for the representation of the cancergenome.a.Samplescollected from13patientswithHGSOCwereanalysedeitherwithorwithoutfilteringbysizeselection.b.SCNAanalysisbasedonthelog2ratiosofregionsalongthegenomeofDNAextractedfromaplasmasamplecollectedduringtreatmentforpatientOV04-83.c.Thesameanalysisof the samesamplewith size selectionof fragmentsbetween90bpand150bp. Inferredamplificationsareshowninblueanddeletionsinorange.d.SCNAanalysisofthesegmentallog2ratioacross a list of 29 genes frequently mutated in recurrent ovarian cancer, measured in plasmasamples collected during treatment for all 13 patients, without size selection (left) and with sizeselection(right).ThetwosampleswhichexhibitedadegradedpatternofcfDNAfragmentationwereOV04-292andOV04-300 (both labelledby“#”).e.Acomparisonof theabsolute levelof log2ratioacrossthe29genesofinterestindicatedasignificantdifferencebetweenthesamesampleswithoutand with size selection (p = 7.72.10-9). The 2 samples with the degraded pattern of cfDNAfragmentationhavebeenexcludedfromthisanalysis.

Sequencing, alignment and genomic analysis

no size selection

with size selection

0.01

1.00

no size selection with size selection

abso

lute

log2

ratio

no size selection

MSH

2M

SH6

APLF

PAX8

BARD

1FA

NCD2

MLH

1M

ECO

MTE

RT ID4

MYC

APTX

PTEN

CHEK

1BR

CA2

RB1

PARP

2FA

NCM

RAD5

1BPA

LB2

TP53

NF1

RAD5

1DCD

K12

BRCA

1RA

D51C

CCNE

1ZM

YND8

CHEK

2

with size selectionOV04−77OV04−83OV04−122OV04−141OV04−143OV04−180OV04−211OV04−226OV04−264OV04−292OV04−295OV04−297OV04−300

MSH

2M

SH6

APLF

PAX8

BARD

1FA

NCD2

MLH

1M

ECO

MTE

RT ID4

MYC

APTX

PTEN

CHEK

1BR

CA2

RB1

PARP

2FA

NCM

RAD5

1BPA

LB2

TP53

NF1

RAD5

1DCD

K12

BRCA

1RA

D51C

CCNE

1ZM

YND8

CHEK

2

log 2 ratio10.50−0.5

log 2 ratio10.50−0.5

a b

d e

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 202122

Chromosomes

0

-1

-2

1

2

3

log2

Rat

io

OV04-83 no size selection

c

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 202122

Chromosomes

0

-1

-2

1

2

3

log2

Rat

io

OV04-83 with size selection

***

#

#


https://doi.org/10.1101/134437

Figure3:

Figure 3:Analysis of the enrichment after size selection in 26 samples sequencedby sWGS andTaggedAmplicon Sequencing (TAm-Seq) revealed relative enrichment in tumour content. a. Foreach of 13 patients, we compared cfDNA from plasma samples collected before initiation ofchemotherapyand3weeksormoreafterinitiationofchemotherapy.Eachofthe26plasmasampleswasanalysedwithandwithoutsizeselection.b.Comparisonoftheabsolutevalueofthesegmentedlog2ratiooftheSCNAscalledfortheplasmasamplesofpatientOV04-83collectedbeforeandafterinitiationofthetreatment.Datafromthesampleswithoutsize-selectionisshowningreen,andwithsizeselectioninorange.c.Thet-MADscoredeterminedfromthesWGSwithsizeselection(vertical)washigherthanwithoutsizeselection(horizontal)formostsamples,includingthesamplescollectedat baseline (red circles) and after initiation of treatment (blue triangles). The 2 samples with noobserved enrichment are OV04-292 and OV04-300. d. The enrichment factor with size selection,

0.01

0.10

0.01 0.10t−MAD no size selection

t−M

AD w

ith s

ize

sele

ctio

n

statusbaselinepost−treatment

c

1

2

3

4

0.0 0.1 0.2 0.3t−MAD no size selection

t−M

AD fo

ld e

nric

hmen

t with

siz

e se

lect

ion


d

0.01

1.00

0.01 1.00MAF no size selection

MAF

with

siz

e se

lect

ion


e

0.01

1.00

baseline post−treatmentTimepoint

MAF

selectionno size selectionwith size selection

f

treatment

aa

0.0

0.5

1.0

1.5

2.0

2.5

0.0 0.5 1.0 1.5 2.0 2.5absolute log2ratio at baseline

abso

lute l

og2r

atio p

ost tr

eatm

ent

selectionno size selectionwith size selection

b


https://doi.org/10.1101/134437

determinedbyt-MAD,variedpersamplebutwas lowerforwassamplescollectedatbaseline(redcircles),which had high initial t-MAD score, compared to samples collected after treatment (bluetriangle).e.Themutantallelefraction(MAF)determinedbytargetedsequencingwithsizeselection(vertical) was higher thanwithout size selection (horizontal) formost samples, including samplescollectedatbaseline(redcircles)andafter initiationoftreatment(bluetriangles).Thedottedareahighlights samples with low MAF (5%andthereforeaccessibleforwide-scaleanalysis.f.ComparisonoftheMAFdetectedbyTAm-Seqbeforetreatmentandafterinitiationoftreatment,asassessedbytargetedsequencing,withsizeselection(yellowtriangles)andwithoutsizeselection(greencircles).

https://doi.org/10.1101/134437

selecting short dna fragments in plasma improves detection ...plasma dna resulted in a median of 1.5...

Documents