Tracking of antibiotic resistance transfer and rapid ... · of antibiotic resistance plasmids within and across multiple species. Starting in 2009 we experienced an outbreak with

1

Tracking of antibiotic resistance transfer and rapid plasmid evolution in ahospitalsettingbyNanoporesequencing

Silke Peter1,2*, Mattia Bosio3†, Caspar Gross4†, Daniela Bezdan3, Javier Gutierrez5, PhilippOberhettinger1,2, Jan Liese1,2,WichardVogel6, DanielaDörfel6,7, Lennard Berger4,MatthiasMarschal1,2,MatthiasWillmann1,2, IvoGut5,8,MartaGut5,8, IngoAutenrieth1,2,andStephanOssowski3,4,8*

1) Institute of Medical Microbiology and Hygiene, University of Tübingen, Tübingen,Germany

2) GermanCenterforInfectionResearch(DZIF),partnersiteTübingen,Tübingen,Germany.3) Centre for Genomic Regulation (CRG), The Barcelona Institute of Science and

Technology,Barcelona,Spain4) InstituteofMedicalGeneticsandAppliedGenomics,UniversityofTübingen,Tübingen,

Germany5) CNAG-CRG, Centre for Genomic Regulation (CRG), Barcelona Institute of Science and

Technology(BIST),Barcelona,Spain6) Medical Center,DepartmentofHematology,Oncology, Immunology,Rheumatology&

Pulmonology,UniversityofTübingen,Tübingen,Germany.7) Medical Center,DepartmentofHematology,Oncology, Immunology,Rheumatology&

Pulmonology, University of Tübingen, Tübingen, Germany; Clinical Collaboration UnitTranslational Immunology, German Cancer Consortium (DKTK) and German CancerResearchCenter(DKFZ),partnersiteTübingen,Tübingen,Germany.

8) UniversitatPompeuFabra(UPF),Barcelona,Spain

*Corresponding authors: [email protected], [email protected]

†Theseauthorscontributedequally

Abstract

Background: Infectionofpatientswithmultidrug-resistant(MDR)bacteriaoften leaveverylimited or no treatment options. The transfer of antimicrobial resistance genes (ARG)carrying plasmids between bacterial species by horizontal gene transfer represents animportant mode of expansion of ARGs. Here, we evaluated the application of Nanoporesequencingtechnologyinahospitalsettingformonitoringthetransferandrapidevolutionofantibioticresistanceplasmidswithinandacrossmultiplespecies.

Results: In 2009 we experienced an outbreak with an extensively multidrug resistant P.aeruginosaharboringthecarbapenemaseenzymeblaIMP-8,andin2012thefirstCitrobacterfreundii and Citrobacter werkmanii harboring the same enzyme were detected. UsingNanopore and Illumina sequencing we conducted a comparative analysis of all blaIMP-8

.CC-BY-ND 4.0 International licensenot certified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which wasthis version posted May 17, 2019. . https://doi.org/10.1101/639609doi: bioRxiv preprint

https://doi.org/10.1101/639609

http://creativecommons.org/licenses/by-nd/4.0/

2

bacteria isolated in our hospital over a 6-year period (n = 54). We developed thecomputationalplatformspathoLogicandplasmIDentforNanopore-basedcharacterizationofclinical isolatesandmonitoringofARG transfer, comprisingde-novoassemblyofgenomesandplasmids,polishing,QC,plasmidcircularization,ARGannotation, comparativegenomeanalysisofmultipleisolatesandvisualizationofresults.UsingplasmIDentweidentifieda40kbplasmidcarryingblaIMP-8 inP.aeruginosaandC. freundii,verifyingthatplasmidtransferhad occurred. Within C. freundii the plasmid underwent further evolution and plasmidfusion, resulting in a 164 kb mega-plasmid, which was transferred to C. werkmanii.Moreover, multiple rearrangements of the multidrug resistance gene cassette weredetectedinP.aeruginosa,includingdeletionsandtranslocationsofcompleteARGs.

Conclusion:Plasmidtransfer,plasmidfusionandrearrangementofthemultidrugresistancegenecassettemediatedtherapidevolutionofopportunisticpathogensinourhospital.Wedemonstrated the feasibility of tracking plasmid evolution dynamics and ARG transfer inclinicalsettingsinatimelymanner.Theapproachwillallowforsuccessfulcountermeasurestocontainnotonlyclonal,butalsoplasmidmediatedoutbreaks.

Background

Theincreaseofmulti-drugresistant(MDR)bacteriahasledorganizationssuchastheWorldHealthorganization(WHO)andtheUSCentersforDiseaseControlandPrevention(CDC)tocategorize MDR bacteria a major public health problem [1]. Infection or colonization ofpatientswithMDRbacteriaoftenleaveonlyverylimitedorevennotreatmentoptions,thusposingapotentially life-threatening risk to the individualpatient inparticularon intensivecareunits[2,3].Inaddition,infectioncontrolmeasurestopreventspreadingarerequired,resulting in increased efforts in patient care and costs for healthcare providers and thepublic healthcare systems [4, 5]. Although action is needed on different national andinternational levels,understandingcolonization, infectionandtransmissionroutesoftheseMDR resistantbacteria in the localhospital setting representa crucial initial step towardsimplementationofharmonized, successful strategies to combat infections causedbyMDRbacteria[1,4].

Next-generationsequencing(NGS)hasbecomewidelyavailableandwasusedsuccessfullytoresolve outbreaks anddetermine transmission routes (e.g. reviewed in [6]).However, notonly the clonal transmissionofMDRbacteria, but also the spreadofmultidrug resistanceplasmids by horizontal gene transfer between different bacterial species represents animportant mode of expansion of antimicrobial resistance genes [7]. Although multidrugresistance plasmids and plasmid transfer have been studied in hospital settings, theirinterrogation is not part of routine infection control practice. Moreover, plasmidcharacterization and comparison based on short-read sequences is error prone andunreliable particularlywhen larger plasmids (>50 kb) are involved [8], while long-read denovo assembly based plasmid analysis is currently limited to large centerswith access toPacificBiosystems(PacBio)Sequelsequencers(e.g.[9,10]).Recently,theMinIONlong-readsequencer (Oxford Nanopore Technologies, ONT) has become more widely available,


https://doi.org/10.1101/639609


3

facilitating fast and inexpensive analysis ofmultidrug resistance plasmids, horizontal genetransfer and evolution of plasmid-born antimicrobial resistance (AMR) [11, 12]. Thus, thetechnology is potentially suitable for application within the hospital setting. In recentpublications, Dong et al examined themicroevolution of KPC carbapenemase plasmids inthreeclinical isolatesapplyingNanoporetechnology [13],whileLemonetaloptimizedtheNanoporesequencinglaboratoryworkflowandanalyzedplasmidsfromthreeclinicalisolates[11]. Long-read sequences substantially increase the contiguity of de novo assemblies byspanningrepeatregions,resultinginfinishedmicrobialgenomeandplasmidassemblies[14].However, due to thehigh error ratesofNanopore sequencing, hybrid assemblers such ashybridSPAdes[15]andUnicycler[16]combine longandshortreadstoachieveahighbaselevelaccuracyneededforthecorrectidentificationofAMRrelatedgenesandvariants.Inthepresentstudy,weaimedtoevaluatetheapplicationofNanoporesequencingtechnologyinahospitalsettingandtodemonstratethefeasibilityofmonitoringtransferandrapidevolutionofantibioticresistanceplasmidswithinandacrossmultiplespecies.

Starting in 2009 we experienced an outbreak with an extensively multidrug resistant P.aeruginosacloneinourhospital[17].Thestrainharboredacarbapenemaseenzyme(blaIMP-

8),whichrendersallbeta-lactamsresistant,includingcarbapenems,anantibioticclassoflastresort.Extensiveinfectiousdiseaseinterventionsandtheestablishmentofarectalscreeningprogramtoidentifycolonizedpatientsledtoareductionofcases.However,inMarch2012we detected the first Citrobacter freundii harboring the same enzyme blaIMP-8carbapenemase[18],approximately2.5yearsafterthefirstP.aeruginosablaIMP-8hadbeendetected. Shortly after, the carbapenemasewas detected inCitrobacter werkmanii. SincethisenzymeisrarelyencounteredinEurope,wehypothesized,thathorizontalgenetransferbetweenthedifferentGram-negativebacteriahadoccurredwithinourhospital.Therefore,we conducted a sequencing study including allmultidrug resistant bacteria harboring theenzyme blaIMP-8 isolated in our hospital over a 6-year period including patient andenvironmental isolates.Wedevelopedandestablishedabioinformaticspipeline inorder i)todeterminethesequenceoftheblaIMP-8harboringplasmidandcharacterizeallcontainedAMRgenesii)toidentifypotentialtransmissioneventsoftheplasmidbetweenspeciesandiii)tocharacterizetheevolutionarydynamicsoftheplasmid.

Results

Comprehensiveanalysisplatformforantibioticresistancegenecarryingplasmids

Wehave developed a comprehensive computational platform for the genomic analysis ofclinical isolates and the monitoring of antibiotic resistance gene transfer. PathoLogiccomprisesofahybridde-novoassemblypipelinegeneratingfinishedgenomesandplasmids,genome polishing, quality control, annotation, comparative genome analysis of multipleisolates, aswell as visualization of results (Fig 1). Furthermore,pathoLogic integrates theplasmIDentmethod,whichconfirmsthecircularityofputativeplasmidsbyring-closureusinglong reads, performs AMR gene annotation, calculates various characteristic plasmidproperties, and creates a circular visualization of the annotated plasmid. Finally, the


https://doi.org/10.1101/639609


4

sequencesofplasmidsfrommultipleisolatesofthesameordifferentspeciesarecomparedin order to identify horizontal gene transfers, structural variations (e.g. AMR genepresence/absence) and pointmutations,which can further be utilized for phylogenetic ortransmissionanalysis.PathoLogic,plasmIDentandagraphicaluserinterface(GUI)arefreelyavailableongithub(plasmIDentpipeline:https://github.com/imgag/plasmIDent,pathoLogicpipeline:https://github.com/imgag/pathoLogic).


https://doi.org/10.1101/639609


5

Figure 1. Schematic diagram of the data analysis workflow used in this study. The pathoLogicplatform was created using the Nextflow [19] environment to chain different tools and scripts,represented here as circular nodes. Connecting lines indicate data flow between the separateprocesses, dashed lines show tools that arenotdirectly included in thepipelineandneedmanualdatahandling. InpathoLogic, theassemblystep (*)canbeperformedbyUnicycler [16],Canu [20],miniasm[21],hybridSPAdes[15]orflye[22,23].

Characterizationofstudyisolates

Inourstudy,weincludedallblaIMP-8AMRgenepositivestrainsisolatedinourhospitalfrompatients or patient-related environmental water sources in the haemato-oncologydepartment over a 6 years’ period (n=54). This also comprised the previously reportedP.aeruginosa outbreak clones (n=34) [17] and oneC. freundiiblaIMP-8 isolate [18], forwhichIllumina short-read data is available (Tab. S1). In order to obtain finished genomes andcircularized plasmids long read Nanopore sequencing was conducted of all Citrobacterfreundii (n=8)andCitrobacterwerkmanii (n=1) isolatesandselectedP.aeruginosa isolates(n=5) representing different time points (Tab. S2). Applying the pathoLogic pipelinedescribedaboveenabledustogeneratehighqualitygenomesforallsamples.In5ofthe14sampleswewereabletogenerateasinglecircularchromosomealongwithseveralcircularplasmids(Tab.S3).Allotherassembliesalsohavefew,largecontigs,asindicatedbyahighNG75 values. Samples with a lower depth of coverage of Nanopore reads (e.g. isolate9_E_CF)alsoresultedinmorefragmentedassemblies.

Plasmidcontentandphylogenyofthestudyisolates

Within the first2.5yearsweonlyobservedP.aeruginosablaIMP-8 involvingcolonizationorinfectionof26patients,before the firstdetectionofC. freundiiandC.werkmaniicarryingblaIMP-8asshown inthetimeline(Fig.2A).TheplasmidswithrelevancetothedynamicsoftheblaIMP-8plasmidevolutionaredisplayedinfigure2B.ThecompleteplasmidcontentofallisolatesissummarizedintableS4.

InP.aeruginosaisolates,wedetecteda40kbplasmidcarryingtheblaIMP-8gene(plasmidA,blue). InC. freundii isolates thesameblaIMP-8plasmidAwas found inaddition toan88kbplasmid(plasmidB,green)withoutacarbapenemasegene.Surprisingly,intheC.werkmaniiisolate a large 164 kb plasmid harboring the blaIMP-8 gene was detected (plasmid C, red)withoutanyevidence for thepresenceof theplasmidsAorB.The structuresandcircularnature of the three plasmids were confirmed by remapping the long-read sequencesresultingincontinuousreadcoveragealongtheplasmidswithoutbreakpoints.

PhylogeneticanalysisshowedthatallP.aeruginosastrainsarecloselyrelatedandbelongtoa single cluster indicating clonal spread (Fig. S1). In contrast, the maximum likelihoodphylogenyof theCitrobacter isolates revealedaphylogeneticallymorediversepicture (Fig2C).TheC.freundiiisolatesformedtwoclustersCf1(n=5)andCf2(n=3),whichwereclearlydistinct (Fig. 2C).Both clusters contained theplasmidsAandB. Isolatesof the clusterCf2contained an additional plasmid G (Tab. S4) and a transposon region localized on the


https://doi.org/10.1101/639609


6

chromosome absent in cluster Cf1, which is further described below. The C. werkmaniiisolateclusteredwithtworeferencestrains forC.werkmaniiNBRC105721andDSM17579,asexpected.

Figure2A.Timelineof isolationofblaIMP-8Gram-negatives inthehemato-oncologydepartmentover6years.“P”indicatesanisolatefromapatient.Forpatientisolatesthelengthofhospitalstayisdisplayed.Patient21wasseenonly in theoutpatientdepartment,markedwith“o”.Environment isolatesarenamedwithan“E”.Thedateofisolationisshownasan“x”.Theintroductionofarectalscreeningprogramismarkedwithablackarrow. PA:P. aeruginosa, CF:Citrobacter freundii, CW:Citrobacterwerkmanii.B. Overviewof plasmidswithrelevance to the evolution of the blaIMP-8 plasmid found in P. aeruginosa, C. freundii and C. werkmanii. C.Maximum likelihood phylogeny of Citrobacter species included in the study (n=9). The Citrobacter freundiistrainsformedtwoclustersCf1(n=5)andCf2(n=3).StrainsofclusterCf2harboredchromosomaltransposonregion (black triangle) homologous to the regions of plasmid C. C. werkmanii clusteredwith the respective


https://doi.org/10.1101/639609


7

reference strains NBRC105721 and DSM17579. The scale bar shows the expected number of nucleotidechangespersite.

Comparativegenomicanalysisandannotationofplasmids

Next,weperformedmultiplesequencealignmentofthegeneratedreferencesequencesofplasmidsA,BandC(Fig.3).Tobetterunderstandthechronologicalorderofthehorizontalgenetransfer(HGT)andfusioneventswefirstperformedanin-depthannotationofplasmidfeatures,includingantimicrobialresistancegenes,transposons,originofreplicationandGC-content(Fig.3).PlasmidA,whichcontainstheblaIMP-8gene,hadanaverageGCcontentof59%(Fig.3greeninnercircle).TheblaIMP-8waslocatedonaclass1integrontogetherwitheightadditionalantimicrobialresistancegenes(Fig.3bottom).ThisresistancegenecassettehadalowerGCcontentcomparedtotheotherpartsofplasmidA,indicatingthatthispartmighthave been acquired from a different organism. Plasmid B had a size of approx. 88 kb, asubstantiallylowerGCcontent(<50%)whencomparedtoplasmidA,andlackedtheblaIMP-8

integron.The largestplasmidCwithasizeof164kbcomprisedofplasmidA includingtheantimicrobial resistance gene cassette and plasmid B, as well as two large stretches oftransposaseelementsTn3andIS6.Therefore,plasmidCresultedmostlikelyfromafusionofplasmidsAandB.ThetwofusionregionsbetweenplasmidsAandBharboredaduplicatedregion(markedinfigure3withablackarrow)composedofonetransposonTn3,threeIS6elementsandseveralAMRgenes.TwoadditionalTn3copiesinterspersedbyadditionalAMRgenesextendedoneofthefusionregions.


https://doi.org/10.1101/639609


8

Figure 3. Detailed alignment of plasmid A, B and C. Plasmid A (blue, outer circle) harbored a multidrugresistance cassette including blaIMP-8 on a class 1 integron and shows high GC content (inner circle, green).PlasmidB(green,outercircle)harborednoblaIMP-8resistancegeneandshowslowerGCcontent(red).PlasmidC(red)wasthelargestandcomprisedplasmidA,plasmidB,aduplicatedtransposonregion(blackarrows)andaunique extension by two Tn3 elements in one fusion region. The class 1 integron consists of 9 AMR genesincludingaminoglycosides,beta-lactams,sulphonamidesandfluoroquinolones.AdditionalAMRandamercuryresistanceoperon[24]arefoundwithintheduplicatetransposonregion.

Plasmidcontentofisolatesandplasmidfusion

Inordertodeterminetheplasmidcontentofallstudied isolateswerealignedthe Illuminashort-read sequencesusing the assembledplasmidC as a reference,which comprises thesequencesofplasmidsAandBandtheduplicatetransposonregion(Fig.3).ThecoverageforeachstrainisdisplayedinFigure4.AllP.aeruginosa isolatescontainedonlyplasmidA,not


https://doi.org/10.1101/639609


9

plasmid B or C. Sequencing reads ofP. aeruginosa thatmapped to a small section of thetransposonregionmostlikelyoriginatefromthechromosome.Thepictureismorecomplexfor the Citrobacter species, which could be divided into three groups. The C. werkmanii(28_P_CW)straincontainedthecompleteplasmidC,whichwashomogenouslycovered.TheC. freundii isolates formed two groups, one group with Plasmid A and B and the secondgroupcontainingplasmidAandBaswellascoverageonthetransposonregions.Thetwogroupsare identicalwiththeclustersCf1andCf2distinguishedbyphylogeneticanalysisofthechromosomes.

We further investigated the read coverage distribution for the C. freundii isolates todetermineiftheyharboronlycopiesofplasmidsAandB,oracombinationofcopiesofA,BandC.NocontinuousshortorlongreadscouldbedetectedspanningthegapbetweentheplasmidAorBsequenceandthetransposonregionsineitherofthetwoC.freundiiclusters(Fig. 4, red lines indicating the breakpoints, Fig. S2 showing alignments to one examplebreakpoint), suggesting that the short reads mapping between A and B originate from achromosomal integration of the transposon elements. Annotation of the assembledchromosomesofC. freundii andC.werkmanii isolates confirmed that cluster Cf2 containsthe transposon sequence within the chromosomal scaffold, while cluster Cf1 and C.werkmanii do not (Fig. S3). We conclude that both Cf1 and Cf2 harbor only copies ofplasmidsAandB,butnotofplasmidC,andthatCf2harborsacopyofthetransposonregioninthechromosome.


https://doi.org/10.1101/639609


10

Figure4.A.Coverageplotbasedonshort-readIlluminadatamappedagainstthereferenceplasmidCfoundinCitrobacterwerkmanii.PlasmidC(redbar)comprisesofthesequencesofplasmidsA(bluebar)andB(greenbar) fused by two transposon-rich regions. Red lines indicate breakpoints, which are characterized by theabsenceofreadsspanningthebreakpoint.Whiteareasindicatetheabsenceofcoverageandhenceabsenceofsequenceinagivenisolate,whichinsomecasescouldbetheresultofadeletioneventorindicateboundariesbetween scaffolds. B.Comparisonof the resistancegenecassette configurationofRSC1andRSC2, showingputativeAMRgenetranslocationeventsandadeletionofqacH.

OnlyisolatesofclusterCf2showacomplete‘smear’ inthecoverageplotacrossthewholetransposon region (13_E_CF, 34_P_CF, 38_P_CF). In isolates of cluster Cf1, however, weobserveonlypartialcoverageofthetransposonregion(Cf1.1:3isolates,9_E_CF,29_P_CF,27_P_CF)oralmostnocoverage(Cf1.2:2isolates32_P_CFand30_P_CF).Interestingly,theregions distinguishing these two subclasses Cf1.1 and Cf1.2 harbor a mercury resistanceoperon [24] present inCf1.1but absent inCf1.2. Pairwise alignment to the full genomesusing thenucmer alignerconfirmed that thesegenesare locatedonplasmid J (Tab.S4) inisolate29_P_CFandonnon-circularcontigs intheothertwoisolates(9_E_CF,27_P_CF)ofgroupCf1.1(Fig.S3).


https://doi.org/10.1101/639609


11

In summary, our phylogenetic analysis aswell as the comprehensive plasmid annotationsindicated that theC. freundii isolates in the clusters Cf1 and Cf2 originate from differentclones and should be treated as separate entities in the identification of plasmid-bornhorizontalgenetransfers.

DeletionandtranspositionofAMRgenesinP.aeruginosa

WhileP.aeruginosaisolateshomogeneouslyonlycontainedplasmidA,weobservedthatforsomeisolatestheresistancegenecassettewasnotcontinuouslycoveredwithshortreadsinthe referencealignment shown in figure4and figureS4.Usingshortand long readbasedstructural variant detection methods we identified two types of re-arrangement events.First,wefoundvariousdeletionsofresistancegeneswithintheresistancegenecassettein12strains,indicatedbyzerocoverage(Fig.S4,whiteareasflankedbyredbrackets).AnalysisoftheresistancegenesannotatedbyResFinderorCARDontherespectiveplasmidscaffoldsconfirmed that thesedeletionscorrespond tomissingAMRgenes in the respective strains(Tab.S5).Furthermore,alldeletionsspanexactlyfrom5’tothe3’endoneAMRgeneplustheflankingISsequence.

Moreover,when comparing the class 1 integron regionofP. aeruginosa isolates 37_P_PAand39_P_PAwedetectedbreakpointswithoutacorrespondingdropofcoverageinflankinggenes, indicating translocation events. We therefore performed a multiple-sequencealignment of the resistance gene cassettes of the 5 P. aeruginosa isolates for whichNanopore sequenceswere generated, as the long-read data facilitates highest confidenceassemblies. Indeed,we identifiedtwostructurallydifferentversionsoftheresistancegenecassette, termedRSC1andRSC2,the latter likelybeingtheresultofmultipletranspositionanddeletionevents(Fig.4B).Fourisolatesharboredthewild-typecassetteRSC1,whileoneisolate harbored RSC2. Finally,we aligned the short reads of all 49P. aeruginosa isolatesagainstthebreakpointsdistinguishingRSC1andRSC2.Weidentified21isolatesmostsimilartoRSC1and9isolatesmostsimilartoRSC2,while10isolatescouldnotbeuniquelyassignedtooneortheother,pointingtoathirdcassetteconfiguration(FigS4).Ourresults indicatethat AMR genes on plasmids are subject to strong selective pressure and are frequentlyremoved,likelyduetothehighcostoftranscribingmultipleresistancegenes.

Rapidplasmid-mediatedadaptation:acquisitionandlossofAMRgenesbyhorizontalgenetransferandstructuralrearrangementevents

Our findings generated multiple evidences that the rapid evolution of opportunisticpathogens inourhospitalwasmediatedbyplasmid transfer,mergingand rearrangement,andevolvedovermultipledistinguishablestages(Fig.5)followingthemostlikelysequenceofevents:

i) Plasmid A (40 kb) harboring blaIMP-8 and multiple other AMR genes was transferredbetween P. aeruginosa and C. freundii. Although the direction of transfer cannot bedeterminedwith certainty, the fact thatP. aeruginosa blaIMP-8was isolated approximately2.5yearsbeforethefirstCitrobacterblaIMP-8strainwasdetected,suggestsatransferfromP.


https://doi.org/10.1101/639609


12

aeruginosa to Citrobacter species. Moreover, the higher GC content of plasmid A pointstowards an origin of the plasmid from a background with a high GC content such as P.aeruginosa(averageGCcontentof66%).However,anunknownintermediatehostservingasareservoirforplasmidAcannotbecompletelyruledout.

ii) In C. freundii, the plasmid underwent further evolution resulting in the fusion of theacquiredplasmidAandaresidentplasmidBtothemega-plasmidCultimatelyrecoveredinC. werkmanii. We hypothesized that this happened by plasmid fusion, since plasmid Ccontainsregionswithgenetichomologyofcloseto100%acrossthefulllengthofplasmidAandplasmidB.Inaddition,plasmidCcontainedtransposonregions,whichwerealsopresentinthechromosomeofC.freundiiclusterCf2strains,indicatingthatthisorganismwasmostlikelythehostoftheplasmidfusion.However,aplasmidfusion inC.werkmaniicannotberuledout(Fig.5greyarea).

iii)We speculate thatC. freundii Cf2 strains ‘distributed’ plasmid A toC. freundii Cf1 andplasmid C to C. werkmanii. However, it is also possible that Cf1 and Cf2 independentlyacquiredplasmidA fromP.aeruginosaor thatCf2acquiredplasmidA fromCf1.Althoughless likely, the plasmid fusion resulting in plasmid Cmight have occurred inC.werkmaniiafter independent transfer of plasmids A and B from any of the other three bacteria.However, C. werkmanii is also lacking a copy of the transposable element region in itschromosome,whichispresentinclusterCf2,makingafusioninC.werkmaniihighlyunlikely(Fig.S3).Figure5depictsallpossibletrajectoriesof theadaptationprocessesmediatedbyplasmid HGT leading to three bacterial species and four clones with multi-antibioticresistancesinasinglehospitalwithinafewyears.

iv) In parallel, inP. aeruginosa the class 1 integron harboring the antimicrobial resistancegenes including blaIMP-8 underwent various re-arrangements such as deletions andtranspositionsofAMRgenes. In 12of theP. aeruginosa isolatesoneormoreAMRgeneswerelost(Tab.S5)andatleast9strainsshowatranspositionoftwoAMRgeneswithintheintegron(Fig.S4).


https://doi.org/10.1101/639609


13

Figure5.Conceptofplasmidevolutionandtransmissionacrossthreebacterialspecies.P.aeruginosablaIMP-8

wasisolatedapprox.2.5yearspriortothefirstisolationofCitrobacterspeciesharboringblaIMP-8,leadingtothehypothesis of a transfer of plasmid A from P. aeruginosa to Citrobacter species. The C. freundii and C.werkmaniiblaIMP-8startedtooccuratthesametime,thusthetimelinedoesnotsuggestaspecificdirectionofthe transfer. However, the existence of the transposable element region in the chromosome of C. freundiiclusterCf2(markedwithablacktriangle)makesitthemostlikelyhostofthemergeofplasmidAandB,whicharelinkedbytwocopiesofthetransposableelementregion.Solidarrowsrepresentthetransmissionsequenceresultingfromthishypothesis.

Inconclusion,wedemonstratedthesuccessfulapplicationofNanoporesequencingtotrackthetransmissionandrapidevolutionofanantibioticresistanceplasmidswithinandbetweenmultiple bacterial species in a comprehensive and systematic collection of multi-drugresistant Gram-negative bacteria obtained from a large cohort of high-risk patients andcorrespondingenvironmentsamples.

Discussion

Understanding the evolution and spread of multidrug resistant organisms has become amajor challenge in the medical field, necessitating the development of novel diagnosticmethods in order to effectively combat increasing numbers of infections with theseorganisms. The clinical importanceof an antimicrobial resistance gene is determinedby i)theclassofantibioticsthatisrenderedresistant,ii)thepathogenicityofthebacteriumandiii) thegenetic locationoftheAMRgene.The localizationofoneormoreAMRgenesona


https://doi.org/10.1101/639609


14

mobile genetic element, e.g. a plasmid, strongly increases the riskof resistance spreadingbetweendifferentbacterialgenera,includingwelladaptedandsuccessfulhumanpathogensestablishedinthehospitalenvironment.

Inseveralstudies,thespreadofcarbapenemaseharboringplasmidshasbeendemonstratedto happen in the hospital environment (e.g. [10]). For example, Conlan et al providedevidenceforhorizontalgenetransferofcarbapenemaseblaKPCharboringplasmidsbetweenKlebsiella pneumoniae, Enterobacter and Citrobacter [10]. Interestingly, the Citrobacterstrain from their study (CFNIH1), which was isolated from the hospital environment,contained a 272 kb plasmid harboring the carbapenemase gene KPC and clustered verycloselywithourstudyC.werkmaniiisolateP28(Figure2C),whichharboredthelargeplasmidC.ThismightindicatethatthegeneticbackgroundofthisCitrobacterstrainsallowsforlargeplasmiduptakeor formationofmegaplasmids in this species.The formationofmulti-drugresistance megaplasmids has also been noted in other Enterobacteriaceae. For example,Desmet et al. analysed two clinical isolates harboring blaOXA-427 carbapenemase andidentifieda321kbmegaplasmidwhichresultedfromacointegrationoftheMDRplasmidinanotherplasmidbackground[25].Afurtherstudydemonstrated,thatafusionplasmidhadoccurred as a result of recombination in a clinical Escherichia coli isolate containing thecarbapenemase blaNDM-5 [26]. Of note, this blaNDM-5 megaplasmid harbored duplicatetransposon regions in the plasmid,which resulted from the plasmid fusion. On the otherhand,thefusionplasmidwasnotstable,whentransferredtoanE.colirecipientstrain[26].This is in line with our observation. Fusion plasmid C contained a duplicate transposonregion,mostlikelyasaresultofrecombination.WhileplasmidCwasstablewithinourstudyisolates, itwas never detected afterwards, suggesting that the plasmidwas not positivelyselectedinthehospitalenvironment.However,furtherstudiesareneededtoelucidatethefactorsinvolvedinmegaplasmidevolutiondynamics.

AlthoughtheimportanceofplasmidevolutionandhorizontalgenetransferforthespreadofMDR bacteria has clearly been documented, the epidemiological surveillance withinhospitals isnot commonlyperformedand remains limited tovery fewcenters. Short-readsequencing technology, which is available in many hospitals, cannot reliably distinguishbetweenplasmids and chromosomesandoften leads to fragmentedgenomeandplasmidassemblies. Long-read sequencing technologies on the other hand allow for high quality,finished assemblies of plasmids.With the emergence ofNanopore sequencing, a fast andinexpensive alternative technology for de novo assembly of multidrug resistant bacteriaisolates became available [11, 13, 27]. Here, we demonstrated that the application ofNanopore sequencing in combination with Illumina short reads and epidemiological dataallowed for detailed tracking of plasmid evolution on a comprehensive consecutivecollection ofblaIMP-8 harboringmulti-drug resistant Gram-negative bacteria. In addition tomultipleplasmid-basedhorizontal gene transfers,wewere able todetect rearrangementswithinthemultidrugresistancegenecassette,aswellasfusionoftwoplasmidstoamega-plasmid. While the presence and absence of antimicrobial resistance genes can bepostulatedbasedonIlluminashortreadassemblies,their locationonmobileelementsand


https://doi.org/10.1101/639609


15

the determination of the structure of multidrug resistance gene cassettes remainschallenging due to difficulties with assembling repetitive regions. In the P. aeruginosagenomes assembled using Nanopore data, we could readily detect continuous readsconfirmingthecircularityoftheplasmidandtheexactorderoftheresistancegenecassette,andcoulddistinguishbetweenbacteriaharboringthemega-plasmidorthetwoindependentplasmids, further emphasizing the power of long reads for determination of structures ofmobilegeneticelements.

In the context of this study,wedeveloped aNanopore-based analysis pipeline for clinicalisolates(termedpathoLogic)includingtheplasmidanalysismethodplasmIDent, inordertoobtain richly annotated circularized plasmids and possible transmission, fusion orrearrangementevents.Importantly,thepipelineisfullyautomatedandrequirednomanualcuration of plasmid sequences. It supports pure Nanopore and Nanopore-Illumina hybridassemblyapproaches,evaluatesthecircularityofplasmids,annotatesresistancegenesandperforms a comparative genome and plasmid analysis between any number of hospitalisolates suspected to harbor a specific resistance gene or resistance gene cassette. Thechosenmethod for tracking ofMDR resistance plasmids and their evolutionary dynamicsrepresents a powerful approach, which could be applied for real-time infection controlsurveillance, in small scale using Nanopore MinION or in large-scale using NanoporePromethIONmachines.Therapiddetectionofresistancespreadviaplasmidswillallowforsuccessfulcountermeasuresandefficientcontainmentofhospitaloutbreaks.

Conclusion

TheapplicationofNanoporesequencingandtheestablishmentofacomputationalpipelinefor genome and plasmid assembly, annotation and comparative analysis enabled us toinvestigateplasmid-drivenadaptationandemergenceofmultidrugresistantbacteriausingacomprehensive strain collection including patient and environment isolates. UsingNanopore-baseddenovoassemblies,wedemonstratedthathorizontalgenetransferviaamultidrugresistanceplasmidbetweenP.aeruginosa,C.freundiiandC.werkmanii,plasmidfusion resulting inamegaplasmidandevolutionof themultidrug resistancegenecassettehad occurred within the short period of 3 years within our hospital. In summary, wedeveloped and showcased a novel pipeline forde novo bacterial genome assembly, AMRgeneandplasmidcharacterisationandcomparativeanalysisacross species,enabling rapidtrackingofAMRtransmissionviaplasmidsinhospitalsettings.


https://doi.org/10.1101/639609


16

Methods

Studyisolates

Intotal,54hospitalstrainswereincludedinthestudy,comprisingP.aeruginosa (n=45),C.freundii (n=8) and C. werkmanii (n=1). Strains were obtained from patient specimensincludingrectalscreeningculture(n=40)andwater-relatedenvironmentsources(toilet,sink,n=14).Allisolateswereculturedandidentifiedfollowingstandardmicrobiologyprotocolsasdescribedbefore[28]andwerepositivefortheblaIMP-8geneasdeterminedbyPCR[29].Allisolateswere recovered fromsamplesof thehaemato-oncologydepartmentbetween July2009 and July 2015. During this time the sampling strategy for screening cultures andenvironmental surveillance was adjusted as a consequence of the P. aeruginosa blaIMP-8

outbreak. Between July 2009 and October 2010 only clinical specimens were obtained.Weekly rectal screening programs of all haemato-oncology patients and environmentscreeningoftoilets,sinksandshowersina14-dayturnwereintroducedinOctober2010.

NanoporeandIlluminasequencing

NanoporesequencingwasperformedonOxfordNanoporeTechnologiesMinIONdevicewiththreedifferentchemistries(versions6,7and8)andflowcellversions(FLO-MAP103versionPk.1,FLO-MIN105versionR9andFLO-MIN106versionR9.4).Anoverviewofchemistryandflowcellversionsusedforeachsampleisshowninsupplementarytable2.

ONT chemistry version 6: Sequencing libraries were prepared with the Genomic DNASequencing SQK–MAP006 kit using 1.5 μg of gDNA as startingmaterial. Briefly, the nick-repaired DNA (NEBNext FFPE DNA Repair Mix, NEB) was sheared in a Covaris g-TUBE(Covaris,Inc.),followedbyend-repairanddA-tailing(NEBNextUltraIIEndRepair/dA-TailingModule,NEB).The leaderandhairpinsequencingadapters (ONT)were ligatedusingBluntTALigase(NEB).Aftertetheraddition,thefinallibrarywaspurifiedwithMyOneStreptavidinC1Beads(ThermoFisher).TheMinIONflowcell(FLO-MAP103,ONT)wasprimedandloadedwith the library for 48h runwith 24h interval of adding newpre-sequencingmix, runningbufferandFuelMix(ONT).

ONTchemistryversion7and8:LibrarieswerepreparedwiththeGenomicDNASequencingKitSQK_NSK007andSQK-LSK108startingwith1.5μgofgDNAshearedinaCovarisg-TUBE(Covaris, Inc.) and nick-repairedwithNEBNext FFPEDNARepairMix (NEB). Subsequently,DNAwasend-repairedandadenylated(NEBNextUltraIIEnd-Repair/dA-tailingModule,NEB)followedbyligationofadaptor(ONT)usingNEBBlunt/TAMasterMix(NEB).Afterflowcellspriming,FLO-MIN105forkitSQK_NSK007andFLO-MIN106forkitSQK-LSK108librarieswereloadedandrunfor48hfollowingthemanufacturerprotocols(ONT).

Illuminasequencing:Duetotheadvancesinsequencingtechnologyavailableoverthestudyperiod,differentprotocolswereused toobtainshort-readsequences,asdescribedbefore[17,28,30].Inbrief,earlyisolatesweresequencedwith2x50bpsonaIlluminaHiSeq2000


https://doi.org/10.1101/639609


17

[17],orby2x300bpsonaIlluminaMiseq[30]orusing2x250bpsonaIlluminaMiseq[28].Supplementarytable2providesadetailedoverviewonthesequencingprotocolsapplied.

Hybriddenovoassemblypipelineusinglongandshortreads

Toachievecompletedenovogenomeassemblieswedevelopedacustompipeline(termedpathoLogic, see figure 1) consisting of individual steps for read pre-processing, hybrid denovoassembly, quality control and generationof assembly statistics. First, longNanoporereads are subjected to adapter trimming with Porechop(https://github.com/rrwick/Porechop), quality filtering with Filtlong(https://github.com/rrwick/Filtlong) and quality control (QC) using Nanoplot [31].Adapter trimming and QC for short reads is performed using SeqPurge [32]. Webenchmarkedmultipleassemblyapproachesimplementedinpathologic.Unicycler,ahybridassemblerusingshortandlongreadreads[16],producedthelongestcontigsathighandlowreadcoverageandwasthereforeusedinthisstudy.Finally,assemblystatisticsarecalculatedand contigs shorter than 2000 bp are removed. Application specific parameters aredocumentedinthepublishedsourcecodeandconfigurationfile.Alltoolsareincludedintheprovided Docker image (release v1.0) available on github (plasmIDent:https://github.com/imgag/plasmIDent,pathoLogic:https://github.com/imgag/pathoLogic).

Phylogeneticanalysis

AssemblyoftheshortreadIlluminadataforallstudiedisolateswasperformedusingSpadesversion3.7.0[33],followedbyalignmentusingProgressiveMauve(version2.3.1)[34]withalocally collinear block size of 1000 bp. Phage content was removed using Phast [35]. Forphylogenycalculation IQ treeversion1.6.3wasappliedusing theUFbootmodewith1000bootstraps and the modelFinder [36-38]. For visualisation Figtree version v1.4.2(http://tree.bio.ed.ac.uk/software/figtree/) was applied. For calculation of the Citrobactermaximum likelihood phylogeny, the 9 study isolates and 14 reference genomes wereincluded (Citrobacter amalonaticusY19 (CP011132_1), Citrobacter braakiiGTA-CB01(JRHK01000001.1),C. braakiiGTA-CB04 (JRHL01000001.1),Citrobacter farmeriGTC 1319(NZ_BBMX01000031.1),C. freundiiCFNIH1(NZ_CP007557_1),CitrobacterrodentiumICC168(NC_013716_1),Citrobacter sedlakiiNBRC 105722 (NZ_BBNB01000030.1), C.werkmaniiNBRC 105721 (NZ_BBMW01000009.1),Citrobacter youngaeATCC 29220(NZ_GG730308.1),Citrobacter koseriATCC BAA-895 (CP000822.1),C. freundiiATCC 8090(JMTA01000001.1),C. rodentiumATCC 51459 (JXUN01000001.1), C. amalonaticusL8A(JMQQ01000001.1) andC. werkmanii DSMZ17579 [30]. For P. aeruginosa, the 45 studyisolatesandoneP.aeruginosablaVIMoutgroupstrainwereincludedintheanalysis.

Plasmiddetectionandannotation

Formostisolatestheassemblyproducedoneorfewlargechromosomalscaffoldsalongwithseveral shorter contigs between 10 kb and 200 kb in length. The latter could either stemfrom complete circular plasmids or from fragments of the chromosome or plasmids.Wetherefore developed the plasmIDent tool, which uses long reads to ascertain whether a


https://doi.org/10.1101/639609


18

scaffold is circular, identifies all antibiotic resistance genes and calculates characteristicmetrices such asGC content and read coverage.PlasmIDent takes assembled genomes infasta format and nanopore reads in fastq format as input. First, contig ends are fused inorder tomimic a circular layout. Next,minimap2 is used to alignNanopore reads to theputativeplasmidandtheend-endfusionsite. If longreadscontinuouslycoverthescaffoldand the artificially closed gap, we assume that the sequence originates from a circularplasmid.Furthermore,suddenchangesofmedianGCcontentwithintheplasmidareusedtopredictancestralfusionsoftwoormultipleplasmids.Finally,plasmIDentsupportsdiscoveryofresistancegenesusingtheCARDdatabase.

Genomeannotations

Assembled FASTA files were uploaded to the ResFinder tool(https://cge.cbs.dtu.dk//services/ResFinder/) applying a 98 % identity threshold and aminimum overlapping length of 60%. Additionally, CARD-based annotations automaticallygeneratedbyplasmIDentweremergedwiththeResFinderresults.Finally,weusedtheRASTWebservice to get complete genome and plasmid annotations for all isolates and theISFinderWebservicetospecificallyidentifytransposableelements.

Comparativegenomeandplasmidanalysisacrossspecies

Whole genome alignment (WGA): Multiple whole genome alignments of all assembledplasmids were generated with progressiveMauve in order to find highly similar regions.Plasmidswithhighlyhomologousregionswereadditionallycomparedbypairwisesequencealignment using nucmer (e. g. Fig S3), resulting in a pairwise identity score and theannotation of homologous regions. We used dotplots (pathoLogic utility scripts) of thepairwise alignments to visually identify rearrangements in plasmids. Homologous regionsbetween plasmids and chromosomal scaffolds were identified using pairwise alignment(nucmer)betweenaplasmidofinterestagainsttheconcatenatedsequenceofallscaffoldsinanisolate’sgenomeassembly.Morespecifically,weidentifiedhomologoussequencesofthetransposableelementregionfoundinplasmidC,butnotAandBinordertoascertain, ifaCitrobacterisolatecontainsonlyplasmidsAandBandthetransposonregioninsertedinthechromosome,orifitcontainsplasmidCwiththetransposonregionintheplasmid.

Readcoverage(density)analysis:WechosemegaplasmidCofisolate28_P_CWasreferenceplasmid, as it integrates both plasmids A and B involved in the studied horizontal genetransferofAMRgenes.Weusedbwa-memtorealignIlluminashortreadsofeachisolatetothereferenceplasmid,therebyidentifyingthepresenceorabsenceofspecificregionsbasedon read density (i.e. regions without read coverage are absent in a studied isolate, seefigures 4 and S4). We identified breakpoints, indicating structural variants or the end ofplasmids,basedonclip-orsplit-reads(seefigureS2).Wedefineddeletionsasregionswithvery lowdensity read coverage,with split- or paired-reads spanning the twobreakpoints.(Plasmid ends are identified by circularization as described before.) Furthermore, we


https://doi.org/10.1101/639609


19

evaluatedthatputativelydeletedresistancegeneswerealsoabsentfromtheplasmidAMRgeneannotationsbyResFinderandCARD.

AMRgene rearrangements:WGAof the resistancegenecassetteofall isolatesassembledwithNanoporereadsidentifiedtwohaplotypestermedRSC1andRSC2,distinguishedbytwotranslocationsofAMRgenes. Inorder toassignall sequenced isolates tooneor theothercassette configurationwealigned Illumina short reads to the4breakpointsperhaplotype(two breakpoints for each translocation event per cassette configuration). Then, wecomparedthenumberofalignedreadsspanningthefourbreakpointsinRSC1vs.RSC2andcomputedthelogtransformedfractionofbreaksinRSC1andRSC2,eachnormalizedbytherespectiveamountoftotalreads.Isolatesshowinglog-values>1wereassignedtoRSC1andlog-values<-1toRSC2,whileotherisolatesremainedunassigned.

DeclarationsEthicsapprovalandconsenttoparticipateThestudywasconductedinaccordancewiththelocalethiccommittee(EthiccommitteeMedicalfaculty,UniversityofTübingen,No.741/2016BO2and407/2013R).ConsentforpublicationNotapplicable.AvailabilityofdataandmaterialAllsequencedataisdepositedattheEuropeanNucleotideArchive(accessionnumberPRJEB31907,https://www.ebi.ac.uk/ena/data/view/PRJEB31907).PathoLogic(https://github.com/imgag/pathoLogic)andplasmIDent(https://github.com/imgag/plasmIDent)developedinthisstudyarefreelyavailableonGithub. CompetinginterestsTheauthorsdeclarethattheyhavenocompetinginterest.FundingThestudywasfundedbytheFacultyofMedicineoftheUniversityofTübingen,theSpanishMinistry of Economy and Competitiveness, the Centro de Excelencia Severo Ochoa, theCERCAProgramme /GeneralitatdeCatalunyaand the “laCaixa” Foundation. The fundershadnoroleinthedesign,analysis,interpretationandwritingthemanuscript.Authors'contributionsSP,DB,PO,JG,MM,MW,IG,MGgeneratedthelaboratoryandsequencingdata.JL,WVandDDgatheredepidemiologicaldata.SP,MB,CGandSOperformedthedataanalysis.CG,MBandSOdevelopedthebioinformaticsmethodsandpipelines,LBdevelopedtheGUI.SP, IAandSOdesignedthestudy.SP,CGandSOwrotethemanuscript.AcknowledgementsWethankNadineHoffmannandBarisBaderforexperttechnicalassistance.References


https://doi.org/10.1101/639609


20

1. RocaI,AkovaM,BaqueroF,CarletJ,CavaleriM,CoenenS,CohenJ,FindlayD,GyssensI,HeuerOE,etal:Theglobalthreatofantimicrobialresistance:scienceforintervention.NewMicrobesNewInfect2015,6:22-29.

2. MaraoloAE,CascellaM,CorcioneS,CuomoA,NappaS,BorgiaG,DeRosaFG,GentileI:Managementofmultidrug-resistantPseudomonasaeruginosaintheintensivecareunit:stateoftheart.ExpertRevAntiInfectTher2017,15:861-871.

3. KengklaK,KongpakwattanaK,SaokaewS,ApisarnthanarakA,ChaiyakunaprukN:ComparativeefficacyandsafetyoftreatmentoptionsforMDRandXDRAcinetobacterbaumanniiinfections:asystematicreviewandnetworkmeta-analysis.JAntimicrobChemother2018,73:22-32.

4. TacconelliE,CataldoMA,DancerSJ,DeAngelisG,FalconeM,FrankU,KahlmeterG,PanA,PetrosilloN,Rodriguez-BanoJ,etal:ESCMIDguidelinesforthemanagementoftheinfectioncontrolmeasurestoreducetransmissionofmultidrug-resistantGram-negativebacteriainhospitalizedpatients.ClinMicrobiolInfect2014,20Suppl1:1-55.

5. NaylorNR,AtunR,ZhuN,KulasabanathanK,SilvaS,ChatterjeeA,KnightGM,RobothamJV:Estimatingtheburdenofantimicrobialresistance:asystematicliteraturereview.AntimicrobResistInfectControl2018,7:58.

6. GilchristCA,CottonJA,BurkeyC,ArjuT,GilmartinA,LinY,AhmedE,SteinerK,AlamM,AhmedS,etal:GeneticDiversityofCryptosporidiumhominisinaBangladeshiCommunityasRevealedbyWhole-GenomeSequencing.JInfectDis2018,218:259-264.

7. PartridgeSR,KwongSM,FirthN,JensenSO:MobileGeneticElementsAssociatedwithAntimicrobialResistance.ClinMicrobiolRev2018,31.

8. Arredondo-AlonsoS,WillemsRJ,vanSchaikW,SchurchAC:Onthe(im)possibilityofreconstructingplasmidsfromwhole-genomeshort-readsequencingdata.MicrobGenom2017,3:e000128.

9. ConlanS,ParkM,DemingC,ThomasPJ,YoungAC,ColemanH,SisonC,ProgramNCS,WeingartenRA,LauAF,etal:PlasmidDynamicsinKPC-PositiveKlebsiellapneumoniaeduringLong-TermPatientColonization.MBio2016,7.

10. ConlanS,ThomasPJ,DemingC,ParkM,LauAF,DekkerJP,SnitkinES,ClarkTA,LuongK,SongY,etal:Single-moleculesequencingtotrackplasmiddiversityofhospital-associatedcarbapenemase-producingEnterobacteriaceae.SciTranslMed2014,6:254ra126.

11. LemonJK,KhilPP,FrankKM,DekkerJP:RapidNanoporeSequencingofPlasmidsandResistanceGeneDetectioninClinicalIsolates.JClinMicrobiol2017,55:3530-3543.

12. GeorgeS,PankhurstL,HubbardA,VotintsevaA,StoesserN,SheppardAE,MathersA,NorrisR,NavickaiteI,EatonC,etal:ResolvingplasmidstructuresinEnterobacteriaceaeusingtheMinIONnanoporesequencer:assessmentofMinIONandMinION/Illuminahybriddataassemblyapproaches.MicrobGenom2017,3:e000118.

13. DongN,YangX,ZhangR,ChanEW,ChenS:TrackingmicroevolutioneventsamongST11carbapenemase-producinghypervirulentKlebsiellapneumoniaeoutbreakstrains.EmergMicrobesInfect2018,7:146.

14. ChinCS,AlexanderDH,MarksP,KlammerAA,DrakeJ,HeinerC,ClumA,CopelandA,HuddlestonJ,EichlerEE,etal:Nonhybrid,finishedmicrobialgenomeassembliesfromlong-readSMRTsequencingdata.NatMethods2013,10:563-569.

15. AntipovD,KorobeynikovA,McLeanJS,PevznerPA:hybridSPAdes:analgorithmforhybridassemblyofshortandlongreads.Bioinformatics2016,32:1009-1015.


https://doi.org/10.1101/639609


21

16. WickRR,JuddLM,GorrieCL,HoltKE:Unicycler:Resolvingbacterialgenomeassembliesfromshortandlongsequencingreads.PLoSComputBiol2017,13:e1005595.

17. WillmannM,BezdanD,ZapataL,SusakH,VogelW,SchroppelK,LieseJ,WeidenmaierC,AutenriethIB,OssowskiS,PeterS:Analysisofalong-termoutbreakofXDRPseudomonasaeruginosa:amolecularepidemiologicalstudy.JAntimicrobChemother2015,70:1322-1330.

18. PeterS,WolzC,KaaseM,MarschalM,SchulteB,VogelW,AutenriethI,WillmannM:EmergenceofCitrobacterfreundiicarryingIMP-8metallo-beta-lactamaseinGermany.NewMicrobesNewInfect2014,2:42-45.

19. DiTommasoP,ChatzouM,FlodenEW,BarjaPP,PalumboE,NotredameC:Nextflowenablesreproduciblecomputationalworkflows.NatBiotechnol2017,35:316-319.

20. KorenS,WalenzBP,BerlinK,MillerJR,BergmanNH,PhillippyAM:Canu:scalableandaccuratelong-readassemblyviaadaptivek-merweightingandrepeatseparation.GenomeRes2017,27:722-736.

21. LiH:Minimapandminiasm:fastmappinganddenovoassemblyfornoisylongsequences.Bioinformatics2016,32:2103-2110.

22. KolmogorovM,YuanJ,LinY,PevznerPA:AssemblyofLongError-ProneReadsUsingRepeatGraphs.bioRxiv2018.

23. LinY,YuanJ,KolmogorovM,ShenMW,ChaissonM,PevznerPA:Assemblyoflongerror-pronereadsusingdeBruijngraphs.ProcNatlAcadSciUSA2016,113:E8396-E8405.

24. BoydES,BarkayT:Themercuryresistanceoperon:fromanorigininageothermalenvironmenttoanefficientdetoxificationmachine.FrontMicrobiol2012,3:349.

25. DesmetS,NepalS,vanDijlJM,VanRanstM,ChlebowiczMA,RossenJW,VanHoudtJKJ,MaesP,LagrouK,BathoornE:AntibioticResistancePlasmidsCointegratedintoaMegaplasmidHarboringtheblaOXA-427CarbapenemaseGene.AntimicrobAgentsChemother2018,62.

26. XieM,LiR,LiuZ,ChanEWC,ChenS:Recombinationofplasmidsinacarbapenem-resistantNDM-5-producingclinicalEscherichiacoliisolate.JAntimicrobChemother2018,73:1230-1234.

27. LiR,XieM,DongN,LinD,YangX,WongMHY,ChanEW,ChenS:EfficientgenerationofcompletesequencesofMDR-encodingplasmidsbyrapidassemblyofMinIONbarcodingsequencingdata.Gigascience2018,7:1-9.

28. PeterS,OberhettingerP,SchueleL,DinkelackerA,VogelW,DorfelD,BezdanD,OssowskiS,MarschalM,LieseJ,WillmannM:GenomiccharacterisationofclinicalandenvironmentalPseudomonasputidagroupstrainsanddeterminationoftheirroleinthetransferofantimicrobialresistancegenestoPseudomonasaeruginosa.BMCGenomics2017,18:859.

29. PitoutJD,GregsonDB,PoirelL,McClureJA,LeP,ChurchDL:DetectionofPseudomonasaeruginosaproducingmetallo-beta-lactamasesinalargecentralizedlaboratory.JClinMicrobiol2005,43:3129-3135.

30. PeterS,BezdanD,OberhettingerP,VogelW,DorfelD,DickJ,MarschalM,LieseJ,WeidenmaierC,AutenriethI,etal:Whole-genomesequencingenablingthedetectionofacolistin-resistanthypermutatingCitrobacterwerkmaniistrainharbouringanovelmetallo-beta-lactamaseVIM-48.IntJAntimicrobAgents2018,51:867-874.

31. DeCosterW,D'HertS,SchultzDT,CrutsM,VanBroeckhovenC:NanoPack:visualizingandprocessinglong-readsequencingdata.Bioinformatics2018,34:2666-2669.


https://doi.org/10.1101/639609


22

32. SturmM,SchroederC,BauerP:SeqPurge:highly-sensitiveadaptertrimmingforpaired-endNGSdata.BMCBioinformatics2016,17:208.

33. BankevichA,NurkS,AntipovD,GurevichAA,DvorkinM,KulikovAS,LesinVM,NikolenkoSI,PhamS,PrjibelskiAD,etal:SPAdes:anewgenomeassemblyalgorithmanditsapplicationstosingle-cellsequencing.JComputBiol2012,19:455-477.

34. DarlingAE,MauB,PernaNT:progressiveMauve:multiplegenomealignmentwithgenegain,lossandrearrangement.PLoSOne2010,5:e11147.

35. ZhouY,LiangY,LynchKH,DennisJJ,WishartDS:PHAST:afastphagesearchtool.NucleicAcidsRes2011,39:W347-352.

36. NguyenLT,SchmidtHA,vonHaeselerA,MinhBQ:IQ-TREE:afastandeffectivestochasticalgorithmforestimatingmaximum-likelihoodphylogenies.MolBiolEvol2015,32:268-274.

37. HoangDT,ChernomorO,vonHaeselerA,MinhBQ,VinhLS:UFBoot2:ImprovingtheUltrafastBootstrapApproximation.MolBiolEvol2018,35:518-522.

38. KalyaanamoorthyS,MinhBQ,WongTKF,vonHaeselerA,JermiinLS:ModelFinder:fastmodelselectionforaccuratephylogeneticestimates.NatMethods2017,14:587-589.


https://doi.org/10.1101/639609


Documents

Tracking of antibiotic resistance transfer and rapid ... · of antibiotic resistance plasmids within and across multiple species. Starting in 2009 we experienced an outbreak with