Chromothripsis and beyond: rapid genome evolution from complex chromosomal rearrangements.

2013·0.74 MB·English

by  Zhang, Cheng-Zhong

REVIEW
Chromothripsis and beyond: rapid genome
evolution from complex chromosomal
rearrangements
Cheng-Zhong Zhang,1,5 Mitchell L. Leibowitz,2,3,5 and David Pellman1,2,3,4,6
1BroadInstituteofHarvardandMassachusettsInstituteofTechnology,Cambridge,Massachusetts02142,USA;2Departmentof
Pediatric Oncology, Dana-Farber Cancer Institute, Boston, Massachusetts 02115, USA; 3Department of Cell Biology, Harvard
MedicalSchool,Boston,Massachusetts02115,USA;4HowardHughesMedicalInstitute,Boston,Massachusetts02115,USA
Recent genome sequencing studies have identified sev- Another potential advantage of all-at-once or ‘‘punctu-
eral classes of complex genomic rearrangements that ated equilibrium’’ (Gould and Eldredge 1977) mecha-
appear to be derived from a single catastrophic event. nisms is that intermediate steps in tumor development
These discoveries identify ways that genomes can be can actually be deleterious. Forexample, oncogene acti-
altered in single large jumps rather than by many in- vationcantriggerstressresponses,suchasp53activation,
cremental steps. Here we compare and contrast these that can inhibit proliferation or promote cellular senes-
phenomena and examine the evidence that they arise cence(Loweetal.1994;Fearnheadetal.1997;Kastanand
‘‘all at once.’’ We consider the impact of massive chro- Bartek2004;Halazonetisetal.2008;JacksonandBartek
mosomalchangeforthedevelopmentofdiseasessuchas 2009).Inprinciple,anall-at-oncemutationalmechanism
cancer and for evolution more generally. Finally, we could circumvent this problem by simultaneously gen-
summarize current models for underlying mechanisms erating an active oncogene together with changes that
and discussstrategies for testing thesemodels. eliminatethestressresponse.
Althoughall-at-oncemassivegenomicalterationsmight
acceleratetheacquisitionofgrowth-promotingmutations,
Theevolutionofanormalcelltoacancercellisgenerally they have an obvious downside: They could also poten-
thought of as a sequential accumulation of many in- tially generate massive deleterious effects that could
dependentlesionstothegenome(Nowell1976;Greaves overwhelm any growth-promoting mutations. One way
and Maley 2012; Yates and Campbell 2012). In addition to minimize this collateral damage would be to have
to somatic point mutations, these lesions include copy themassivegenomicchangeberestrictedtoaportionof
numberalterations,suchaschromosomaldeletionsordu- thegenome.Althoughitiscommontothinkofmutation
plications (Fig. 1A,B), and balanced structural rearrange- rates as numbers that apply evenly across the genome
ments, such as translocations and inversions (Mitelman (Durrett 2008; Bozic et al. 2010; Tomasetti et al. 2013),
etal.2007;Rabbitts2009;Stephensetal.2009;Stratton thereisagrowingappreciationoflocalizedvariabilityin
et al. 2009). By chance, these somatic alterations in the theserates(HodgkinsonandEyre-Walker2011;Hodgkinson
genomewillhitcancergenes—disruptingtumorsuppres- etal.2012;Robertsetal.2012;Schuster-Bo¨cklerandLehner
sors or activating proto-oncogenes—affecting cells and 2012). Because of transcription-coupled DNA repair,
causing their progeny to gradually accumulate the nec- transcriptionally silent portions of the genome have ap-
essary ‘‘hallmarks of cancer’’ (Hanahan and Weinberg proximately threefold higher mutation rates (Fousteri and
2011).Althoughthereisampleevidencethatcancercan Mullenders 2008; Lawrence et al. 2013). Likewise, late-
developthroughthistypeofgradualmutationalaccrual, replicating regions of the genome exhibit a similar fold
therearereasonstothinkthattherecouldbeadvantages increaseinmutationrates(Stamatoyannopoulosetal.2009;
iftheoncogenicchangescouldbeacquiredallatonce(at Chen et al. 2010; Koren et al. 2012; Donley and Thayer
least from the vantage point of the aspiring cancer cell). 2013). Localized hypermutation has also been associated
One obvious advantage is speed, with the possibility of withchromosomalrearrangements.Inbuddingyeast,DNA
rapidly accruing a large phenotypic effect from combi- breaksrepairedbymitoticgeneconversionareaccompanied
natorial acquisition of smaller-effect genetic changes. bysurprisinglyhighmutationrates(>1000-foldhigherthan
[Keywords:chromothripsis;chromoplexy;chromoanasynthesis;genome (cid:2)2013Zhangetal. ThisarticleisdistributedexclusivelybyColdSpring
evolution;cancer;chromosomaltranslocation;copynumberalteration] Harbor Laboratory Press for the first six months after the full-issue
5Theseauthorscontributedequallytothiswork. publicationdate(seehttp://genesdev.cshlp.org/site/misc/terms.xhtml).After
6Correspondingauthor sixmonths,itisavailableunderaCreativeCommonsLicense(Attribution-
[email protected] NonCommercial3.0Unported),asdescribedathttp://creativecommons.org/
Articleisonlineathttp://www.genesdev.org/cgi/doi/10.1101/gad.229559.113. licenses/by-nc/3.0/.
GENES & DEVELOPMENT 27:2513–2530 Published by Cold Spring Harbor Laboratory Press; ISSN 0890-9369/13; www.genesdev.org 2513
Zhangetal.
Figure1. Characteristicsofsimpleandcomplexchromosomalrearrangements.(A)Copynumberlossesorgainscanbedetectedby
array-basedorsequencing-basedtechnologiesfromabruptchangesinthereaddepthsignal(sequencing)ortheintensityoffluorescent
probes(array).Lossorgaingenerallyaffectsonlyonehomolog;therefore,acopynumberlossalsoresultsinLOH.(B)Copynumber
alterationsareoftengeneratedbysimplechromosomalrearrangements,suchasdeletionsortandemduplications.(C)Chromothripsis
is characterized by an alternating copy number profile with loss and retention of heterozygosity. Minus signs in C and D denote
chromosomalsegmentsthatareinverted.Fluorescentinsituhybridization(FISH)characterizationconfirmsthatonlyonechromatid
(allele B) isaffected. (D) Chromoanasynthesisis characterizedbyacopynumber profile thatalternates betweeneuploid and higher
ploidy.Incontrasttochromothripsis,chromoanasynthesisdoesnotnecessarilyexhibitLOHbutreflectsresynthesisofsegmentsfrom
onechromatid.Chromoanasynthesisalsocontainsfrequentinsertionsofshortsequencesbetweentherearrangementjunctions(red
arrows) that are copied from the rearranged segments (templated insertions). Both DNA resynthesis and templated insertions are
suggestiveofreplication-basedmechanisms.(E)Chromoplexyischaracterizedbyaclosedchainoftranslocations,withlittleornocopy
numberalteration.
2514 GENES&DEVELOPMENT
Genomeevolutionandcomplexrearrangements
spontaneousmutations)inthevicinityofthebreak(Hicks Vogelsteinetal.2013).Despiteacceleratedtumorevolu-
et al. 2010; Malkova and Haber 2012). Similarly, localized tion, individual mutation events are usually regarded as
regionsofhypermutation(‘‘kataegis’’)thatoftenoccurredat independent of each other, acquired randomly and grad-
thesiteofsomaticrearrangementswerediscoveredinbreast ually. In addition to site-specific mutation events (point
cancer genomes (Nik-Zainal et al. 2012; Alexandrov et al. mutations, chromosomal translocations, and copy num-
2013). Even more dramatic local genomic change can be ber variations), doubling of the genome content due to
generated from critical telomere shortening, resulting in cytokinesis failure is a one-off event that can generate
breakage–fusion–bridge(BFB)cycles(McClintock1941)that aneuploidy and structural alterations of chromosomes
occur over several cell division cycles (and hence are not (Fujiwaraetal.2005;HollandandCleveland2009;Carter
‘‘all-at-once’’ events). Analysis of genome copy number etal.2012;GanemandPellman2012;Zacketal.2013).
alterations in cancer also identified ‘‘firestorms’’ of local Instudyingthetumorgenomeofapatientwithchronic
amplificationsanddeletionswhose origin isuncertainbut lymphocytic leukemia (CLL), Stephens et al. (2011) ob-
mightoriginatefromBFBcycles(Hicksetal.2006). servedastrikingpatternofchromosomalrearrangements
Recently, whole-genome sequencing has led to the that was at odds with all known models of genomic
discovery of three new classes of complex catastrophic alteration. The tumor from this patient contained 45
chromosomal rearrangement: chromothripsis, chromo- tumor-specifictranslocations,42ofwhichwereconcen-
anasynthesis,andchromoplexy(Fig.1C–E;Liuetal.2011; trated on the long arm of chromosome 4. Furthermore,
Stephensetal.2011;Rauschetal.2012;Bacaetal.2013). therearrangementeventswereclusteredwithinmultiple
Althoughtheunderlyingmechanismsforthesephenom- focal regions in chromosome 4, raising the question of
ena are unknown and likely to be different, all of these how the damage could be so highly localized. Did this
patterns of rearrangements appear to have originated resultfromagenuinelynewmechanismforgenomedam-
from a single event, simultaneously generating tens or age,or was the genome center playing the conventional
evenhundredsoftranslocationsand/orDNAcopynum- gameofrandomchancebuthadbeendealtararehand?
ber changes (Meyerson and Pellman 2011; Tubio and Several linesofevidencesuggestedthis was,in fact,a
Estivill 2011; Forment et al. 2012; Jones and Jallepalli novel phenomenon. First, despite the large number of
2012;MaherandWilson2012).Toencompassthepoten- rearrangements,SNParraycopynumberanalysisshowed
tiallyall-at-onceoriginofthesemutationalmechanisms, that the altered region of chromosome 4 existed in only
a unifying term, chromoanagenesis, has been proposed twocopynumberstates,withmanytransitionsbetween
(Holland and Cleveland 2012). Two of these classes of thesetwostates.(AsinthestudybyStephensetal.(2011),
chromosomal rearrangement, chromothripsis and chro- weusethetermcopynumberstatesbecauseofthehighly
moanasynthesis, affect limited portions of the genome aneuploid nature of cancer genomes, where each of the
(oftenasinglechromosomeorchromosomearm),raising parental haplotypes can be present in multiple copies.)
themechanisticquestionofhowextensivechromosomal Thispatternofhavingonlytwocopynumberstatesisin
rearrangementscanbesohighlylocalized.Thefrequency sharp contrast to conventional clusters of complex rear-
of these types of events in cancer remains unclear, but rangements(Hicksetal.2006;Bignelletal.2007;Zhang
manyrecentstudiessuggestthattheymaybemorecom- et al. 2009a). Known causes of complex rearrangements
monthaninitiallysuspected(Malhotraetal.2013;Yang such as BFB cycles characteristically result in multiple
et al. 2013; Zack et al. 2013). These new classes of mu- amplifiedcopynumberstatesthatareassumedtoderive
tagenesis are not restricted to human cancer and have fromeventsoccurringover manyroundsofcelldivision
turnedupinsuchexoticplacesastheTasmaniandevil’s (McClintock1941;Starketal.1989;Gisselssonetal.2000;
transmissiblefacialtumor(Deakinetal.2012)andhuman Bignelletal.2007;Greenmanetal.2012b).
congenital diseases (Kloosterman et al. 2011a, 2012; Liu Thealternationbetween twocopy numberstateswas
et al. 2011; Chiang et al. 2012; Nazaryan et al. 2013). accompanied by loss and preservation of heterozygosity
Although these new classes of chromosome rearrange- (Fig. 1A,C). Whole-genome sequencing further revealed
menthavemostlybeenfoundinthecontextofdisease,it a very complex pattern of segmental rearrangements at
seems likely that they reflect very general mechanisms the boundaries of high copy number states, with events
forgenomealterationthatcouldbeimportantfororgan- thatwereheadtotail(deletiontype),headtohead/tailto
ismalevolution. tail(inversiontype),andtailtohead(tandemduplication
type) (Fig. 2A). However, despite this chaotic pattern of
rearrangement,locithatpreservedparentalcopynumber
also preserved heterozygosity of the parental genotypes
A newgame atthe genomecasino
(Fig.1C).Asdiscussedbelow,thismaintenanceofislands
Genome instability is not only a hallmark but also an ofheterozygosityamongmanylargeregionswithlossof
enabling characteristic of cancer (Negrini et al. 2010; heterozygosity (LOH) is difficult to explain by multiple
Hanahan and Weinberg 2011; Gordon et al. 2012). Dis- independentdeletions.
ruption of genome maintenance and elevated rates of Athirdfeatureoftheobservationwasthatthepattern
somaticmutations(e.g.,duetodisruptionofDNArepair of end-joining strongly suggested an origin from a DNA
machineries or increased damage) accelerate the trans- double-strandbreak(DSB).Ashighlightedinlaterstudies
formation from precancerous tissue into malignant tu- (Kloostermanetal.2011a,2012;Chiangetal.2012),the
mors (Gorgoulis et al. 2005; Halazonetis et al. 2008; preponderanceofrearrangementsinvolvedligationofboth
GENES&DEVELOPMENT 2515
Zhangetal.
Figure 2. Signatures of chromosomal rearrange-
mentsfrompaired-endsequencing.(A)Fourmajor
types of chromosomal rearrangement signatures
deducedfrompaired-endsequencing:deletiontype,
tandemduplicationtype,inversiontype,andlong-
range type (including interchromosomal). Each
type of read pair shows aberrations from proper
pairs when aligned to the reference assembly,
whichimpliesanewjunctioninthetestchromo-
some. Adeletion (readpairs shownas red arrows)
or tandem duplication (shown as green arrows)
only introduce one fusion and hence can be un-
ambiguously inferred from a single type of signa-
ture (in the absence of other overlapping events).
Inversion events, however, introduce at least two
rearranged junctions (shown here is a simple in-
version);hence,therearrangedgenomecanonlybe
constructedbyconsidering allrelatedevents.The
sameistrueforlong-rangeeventswithpairmates
residingindifferentchromosomesorseparatedby
centromeres, assembly gaps, or other rearrange-
ments. In general, it is necessary to incorporate
copynumber information together with rearrange-
ment signatures to derive the structure of the
rearrangedgenome(foracomputationalframework,
seeGreenmanetal.2012a).(B)Examplesofcomplex
chromosomal rearrangements exhibiting multiple
overlapping rearrangement signatures that do not
reflectthesameeventsasthe‘‘typing.’’Thediscor-
dant read pair signatures (color arrows) are shown
bothintherearrangedgenomeandaccordingtotheiralignmenttothereferencegenome.(i)Acomplexrearrangementresultinginthe
deletionofsegmentAandswappingofsegmentsBandCgeneratestwo‘‘tandemduplication’’andone‘‘deletion’’readpairsignatures,
noneofwhichcorrespondstoasimpleevent.(ii)Twoinversionsalteradeletionsignatureintoatandemduplicationsignature(green
arrows).
ends of a single DNA break to other genomic segments almostcertainlyproduceafinaloutcomewithmorethan
(endsbeingoperationallydefinedassequenceswithin;1 twocopynumberstates.Finally,maintenanceofislands
kbofeachother;hereafterreferredtoasapairingofends). ofheterozygosityinterspersedbetweenregionsthathave
This contrasted with complex genomic rearrangements lostheterozygosityishardtoexplainbymultiplesequen-
that arise by replication-based mechanisms (Lee et al. tialevents.Thisisillustratedbythefactthatadeletion
2007; Carvalho et al. 2009, 2011) where the rearrange- early in the process will remove sequences that cannot
ment is often seen only on one side of the initial lesion be regained. Thus, if the observed pattern resulted from
(HaberandDebatisse2006;seebelowformoredetails). sequential mutation events, deletions would need to be
The final defining feature of this new mutagenesis restricted to a later time period, when they could not
eventwasthenonrandom,highlylocalizedclustering interrupt the acquisition of balanced rearrangements—a
of chromosomal breakpoints. These four hallmarks— complex ‘‘programmed’’ event for which it is hard to
alternatingcopynumberprofiles,lossandpreservation envisionamechanism(oneexampleisshowninFig.3).
of heterozygosity, pairing of the breakpoint ends, and Incontrasttothegradualmodel,Stephensetal.(2011)
clusteringofrearrangements—definetheclassofchro- proposedamuchmorecompellingcatastrophicmodelof
mosomal alteration termed chromothripsis (Greek for chromothripsisthatoffersaparsimoniousexplanationof
shatteringofchromosomes). theobservation:Onechromosome(orregionsfromoneor
morechromosomearms)isshatteredintosegmentsofdif-
ferentlengthsinasinglecatastrophicevent;thesegments
All atonce?
are then stitched together, most likely by nonhomolo-
The defining hallmarks of chromothripsis cannot be gous end-joining, resulting in a derivative chromosome
explained by a gradual accumulation of independent containingasubsetoftheshatteredsegmentsinrandom
events.First,trulyindependentchromosomalrearrange- orientations.Theretainedsegmentsmaintaintheparen-
mentsareexpectedtoberandomlydistributedacrossthe tal copy number and heterozygosity, yielding the high
genomeratherthanclusteredinfocalregions(Beroukhim copy number state. The segments that are not incorpo-
et al. 2007, 2010). Second, the gradual accumulation of rated into the derivative chromosome are either lost,
suchalargenumberofoverlappingrearrangements,both yieldingthelowcopynumberstate,orincorporatedinto
copynumber neutralandcopynumberchanging,would adouble-minutechromosome,asdiscussedbelow.
2516 GENES&DEVELOPMENT
Genomeevolutionandcomplexrearrangements
Figure3. Reconstructionofalocalcomplex
rearrangement and its mutational history
from read pair signatures and copy number
information.(A)Copynumberalterationscan
bedetectedbyarraymethodsorreaddepths
insequencingdata;rearrangementsignatures
canbedetectedfromdiscordantreadpairsin
sequencingdata.(B)Bycombiningcopynum-
ber information and rearrangement signa-
tures,onecanconstructtheconfigurationof
therearrangedsegmentsinthechromothrip-
ticchromosome.(C)Thehistoryofmutation
events that could have generated the out-
come inB is oftennot unique. However,all
solutionshavetobreakfourjunctionsinthe
reference assembly (LjA, AjB, BjC, and CjR)
and generate three new junctions (LjC, CjB,
and BjR). A catastrophe model can easily
solve this puzzle because all of the breaks
andtranslocationsoccurallatonce.Bycom-
parison,asolutionfromthegradualmodelof
simple events requires at least three simple
eventstogeneratethefinalconfiguration(one
foreachnew junction).Aseacheventcauses
twobreakpoints,threeeventsneedsixbreak-
points, which requires two additional breaks
inadditiontothefouroriginaljunctionstobe
broken. These two additional breaks arise
fromeitherwithinthelostAsegment(losthistory)orhittingthesamejunctionmultipletimes(recurrentbreaks).Shownhereisone
possiblesolutionthatinvolvesrecurrentbreaksatbothBjCandCjRjunctions.Bothlosthistoryandrecurrentbreaksarehighlyunlikely.
Asthenumberofrearrangedfragmentsincreases,theneedtoincludethesetypesofunlikelyeventsfurtherincreases,stronglyarguing
againstthegradualmodel.
Historiography ofthe cancer genome et al. 2007; Campbell et al. 2008; Medvedev et al. 2009)
has a completely different interpretation under a rear-
Thesegeneralargumentsmakeintuitivesensebutideally ranged genome when additional rearrangement events
wouldbebackedupbyadetailedmolecularassemblyof have occurred between the pair mates (Fig. 2B). This
theevolutionarypathtakentogeneratethechromothrip- technical issue is less problematic for simple genomes
ticchromosome.Stephensetal.(2011)facedtheobvious where rearrangements seldom overlap and can be de-
challengeofinferringpriorhistoryfromaterminalsnap- coupledandinterpretedunambiguously(Greenmanetal.
shot of the present, which is already difficult for cancer 2012a,b; Quinlan and Hall 2012). However, it becomes
genomeswithsimplerearrangementpatterns(Greenman prohibitively complex in chromothripsis, where tens or
etal.2012a)andquicklybecomesimpracticalforacom- hundredsofrearrangementsoverlap,resultinginanenor-
plex case such as chromothripsis (Yates and Campbell mousnumberofpossiblepermutations.Thisprovidesan
2012). importantcautionarynoteforaspiringgenomeanalysts:
The first complication for inferring an evolutionary Anaccurateinterpretationofahighlyrearrangedcancer
trajectory is purifying selection: The mature tumor is genome from paired-end sequencing data requires an
biasedtopreservegenomicchangesthatarebeneficialto integrative analysis of rearrangement signatures (such
tumordevelopmentandtohavelostdeleteriouschanges asdeletiontypeortandemduplicationtype)(Fig.2A)and
(Beroukhim et al. 2007; Greenman et al. 2007; Bignell copynumberdata.Inotherwords,contextmatters.
etal.2010).Byviewingonlytheendpointofselection,we The following back-of-the-envelope calculations il-
couldmissintermediatestepsinthedevelopmentofthe lustrate the magnitude of the problem. The number of
rearrangedchromosomewhosesignatureswerelostduring differentoutcomescreatedbyrandompermutationsof10
tumor evolution. An additional technical complication overlapping rearrangements is 10! = 3,628,800 and the
forsolvingthegenomicRubik’scubeforthechromothrip- number grows to ;1064 for 50 rearrangements. Typical
tic chromosome is that one cannot interpret paired-end chromothripsis harbors tens or even hundreds of over-
sequencesinasimplestraightforwardwaywhentheyoc- lapping rearrangements, and it is clearly impossible to
cur in the context of other overlapping rearrangements. enumerate all possible outcomes. Additional complica-
Forexample,adiscordantpairofsequencingreadsthatis tions are caused by intermediate events being obscured
inferredtobe a ‘‘tandem duplication type’’ event when by later events; e.g., a later deletion event that removes
directly mapped onto the reference genome sequence the signature of an earlier translocation. This estimate
(Fig.2A;Raphaeletal.2003;Tuzunetal.2005;Korbel also does not account for rearrangements that are not
GENES&DEVELOPMENT 2517
Zhangetal.
detectableduetotechnicalissuessuchasuncertaintiesin approach,discussedfurtherbelow,potentiallyprovidesa
thereferencegenomeassemblyorinaccuraciesinshort- complementary method to validate the idea that chro-
readalignment. mothripticlesionsoccurredatthesametime.
The most difficult challenge stemming from the vast Insummary,chromothripsisisbestexplainedbyacat-
number of outcomes due to permutation complexity, astrophic all-at-once model and cannot be reconciled
however, is finding not just one solution, but the most with the gradual model. The catastrophic model yields
likely solution basedon a certain backgroundstatistical asimpleexplanationofthelownumberofobservedcopy
modeloftumorevolution.Ultimately,wewanttoknow number states. It also provides an explanation for the
thelikelihoodofthemultieventsolutiontochromothrip- manyoverlappingtandemduplicationtypeeventsthatdo
sisbasedonthebackgroundmodeloftumorevolutionto notincreasecopynumber(seeFig.2B,i,foranexample):
dismissthegradualmodelinfavorofasinglecatastrophe. These duplication type events are instead explained by
(Fig. 3 illustrates this problem by a simple example.) To therandomorientationsatwhichthesegmentsarestitched
answerthislastquestionwithoutenumeratingallpossi- together.Similarexplanations can apply todeletion and
ble outcomes of a gradual model, Stephens et al. (2011) other‘‘type’’events(Fig.2B).
focused on the number of copy number states during
genomeevolution,essentiallyreducingthemultidimen-
Whenis it chromothripsis?
sional space of chromosome configurations to a single
dimensionof‘‘copynumber.’’Theysetupsimulationsin The discovery of chromothripsis in cancer immediately
which individual rearrangement events were applied to raisesthequestionofhowfrequentlyitoccurs.Afterthe
the genome sequentially, starting from the reference se- initial discovery, there have been a plethora of studies
quence.Insteadofbruteforceenumerationofallgenomic confirming that chromothripsis is indeed a widespread
configurations, the eventual number of copy number and previously unappreciated phenomenon in both pri-
states was scored, making a testable prediction about marytumors(Kloostermanetal.2011b;Magrangeasetal.
thestructureofthechromothripticchromosomewithout 2011; Berger et al. 2012; Rausch et al. 2012) and cancer
theneedforacompletemolecular reconstructionofthe cell lines such as HeLa (Adey et al. 2013; Landry et al.
rearrangedchromosome.Thegradualmodelrapidlygen- 2013). Many cancer genomes reported in earlier studies
eratedmultiplehighcopynumberstatesbecausegradual also exhibited features that resemble chromothripsis
accumulationallowsforoverlappingtandemduplications (Stephens et al. 2009; Campbell et al. 2010). So how are
toaffectthesameregionmultipletimes.Thediscrepancy we to recognize chromothripsis? How many rearrange-
betweenthelargecopynumbercomplexityimpliedbythe ments ‘‘make’’ the definition, and how focal is focal?
gradualmodelandthehallmarktwocopynumberstates Becausecopynumberdataarethemostavailable,acom-
observed in chromothripsis provided a strong statistical monsenseapproachwastoadoptoperationaldefinitions
argumentagainstagradualmodelforchromothripsis.Itis that would require, say, 10, 20, or 50 oscillating copy
worth noting that chromothriptic chromosomes can numberchanges(Magrangeasetal.2011).Insomecases,
sometimesexhibitmultiplecopynumberstates(Stephens correctionswereappliedtonormalizethenumberoffocal
et al. 2011; Rausch et al. 2012). One explanation is the events to the general frequency of events on that chro-
overlapbetweenchromothripsisandotherindependently mosome(Kimetal.2013).
generatedcopynumbergains;anotherpossibilityisthat Although an operational criterion is reasonable for a
chromothripsisthatoccurredduringorafterDNArepli- first-pass screening and many instances with multiple
cation can result in a shattering and assembly of both oscillations between two copy number states likely are
sisterchromatidstogeneratecopynumbergains.Ineither bonafidechromothripsis,thefieldismovingawayfrom
scenario, however, the low copy number states (three or operational definitions and adopting criteria that incor-
four)arestillinstarkcontrasttotherapidaccumulationof porateinformationfrompaired-endsequencing.Through
multiplecopynumberstates,asthegradualmodelsuggests. more precise analysis of the structure of the derivative
The focal pattern of breakpoints in chromothripsis chromosome, one can better distinguish chromothripsis
providesyetanotherargumentagainstthegradualmodel. fromalargenumberofprogressivelyacquiredrearrange-
Independentrearrangementsaregenerallyassumedtobe ments in tumors with high levels of genetic instability.
randomlydistributedacrossthegenome.Thelargesizeof Because an excellent review by Korbel and Campbell
thehumangenomeimpliesthatitisextremelyunlikely (2013) has recently covered this topic in depth, we only
toseetwoeventsthatareveryclosetoeachother(;1kb brieflyreviewthemainelementsofthesecriteriahere.
apart).Thelikelihoodofseeingtensorhundredsofbreak- Asdiscussedabove,onehallmarkofchromothripsisis
points within one chromosome (or chromosome arm) theclusteringofrearrangements.Thedegreeofclustering
withfeweventselsewhereinthegenomeisvanishingly canbeassessedbysimplestatisticalmethodsbecausethe
small if these events were independent and not due to null hypothesis of random breakpoints predicts that the
specific targeted mechanisms. This statistical argument distance between breakpoints should be distributed ex-
suggeststhatclusteredrearrangementsinchromothripsis ponentially(Beroukhimetal.2010;Malhotraetal.2013).
verylikelyoccurredsimultaneously.Ageneralizationof However,itisimportanttonotethatclusteringofbreak-
this statistical argument into a systematic way of in- pointscanalsobegeneratedbyotherprocessesinvolving
ferring co-occurring translocations (‘‘chained’’transloca- multiplestepsoverseveralcelldivisioncycles.Forexam-
tions)wasrecentlyproposed(Bacaetal.2013).Thisnew ple,clusteringcanresultfromBFBcycles(Gisselssonetal.
2518 GENES&DEVELOPMENT
Genomeevolutionandcomplexrearrangements
2000),chromosomefragilesites(Debatisseetal.2012),and structural rearrangements (e.g., simple deletions/ampli-
strongpositiveselectionforlocaloncogeneamplifications fications/translocations) layered on top of chromothrip-
(Bignelletal.2007).Thus,clusteringbyitselfisanecessary sis, tumor heterogeneity, and aneuploidy. All of these
but not sufficient criterion for chromothripsis. An addi- technical factors may lead to an underestimate of the
tionalcriterionisprovidedbythefactthatchromothripsis frequencyofchromothripsis.
ischaracterizedbyregionswithinterspersedretentionand
LOH, which can be verified from genotype information
Howfrequently does chromothripsis happen?
fromeitherSNParrayorsequencing.Finally,ifrestricted
toasinglechromosome,theregionsofrearrangementand So what, then, is our best guess for the frequency of
segment loss should be confined to a single parental chromothripsis?Unfortunately,thelargestvolumeofcan-
haplotype because the shattering and stitching back cer genomic data (such as in The Cancer Genome Atlas
togetheroccurs(mostly)onasinglechromatid.Thiscan [TCGA]project)isarray-basedanddoesnothavethepower
be directly verified by cytogenetic methods, such as toprovideanaccurateestimateoftheoverallfrequencyof
fluorescenceinsituhybridization(FISH),orinferredfrom chromothripsis.However,itisbecomingevidentthatthe
phased haplotypes in trio family studies (Browning and frequency of chromothripsis tends to vary between dif-
Browning2011). ferent tumor tissue types. In a survey of whole-genome
Additional information from sequencing can evaluate sequencingdatafor64samplesofseventumortypesfrom
the model that a chromosome has been shattered and TCGA, Malhotra et al. (2013) found no chromothriptic
randomly stitched together in more detail. First, the rearrangementsinbasal-likebreastcancers,colonadeno-
model predicts that the order of the fragments on the carcinomas,ovariancancers,orrenaladenomasbutfound
derivativechromosomewouldberandomornearrandom rearrangementsconsistentwithchromothripsisintwoof
withrespecttotheirpositionsonthereferencegenome. six (33%) lung adenocarcinomas, two of 13 (15%) lung
Second, if reassembly of the derivative chromosome is squamouscellcarcinomas,and39%ofglioblastomas(n=
random, the orientations of fragments on the derivative 18).Inarecentsurveyof4934cancersfromTCGA,Zack
chromosome should also be random. This means that etal.(2013)employedmultiplecriteriatomodelchromo-
thereshouldbenobiasforfragmentstobejoinedheadto thripsis by SNP array copy number alone and suggested
tail (deletion type), tail to head (tandem duplication thatchromothripsisoccurredin5%ofallsamples,with
type), head to head, or tail to tail (inversion type). Such frequenciesrangingfrom0%inheadandnecksquamous
randomness in the orientation of fragments contrasts carcinomatoamaximumof16%inglioblastoma.Inad-
with, for example, BFB cycles, which are characterized dition, chromothripsis has been observed in colorectal
byinvertedsegmentsbiasedtobeheadtoheadortailto cancer (Kloosterman et al. 2011b), melanoma (Berger
tail (Bignell et al. 2007). Finally, if the shattering and et al. 2012; Hammond et al. 2013; Hirsch et al. 2013),
restitching indeed happen all at once, then each rear- small-cell lung cancer (Iwakawa et al. 2013), prostate
rangedfragmentwillhaveauniqueheadandtailsequence cancer (Wu et al. 2012), and a variety of brain tumors
(with respect to the reference sequence). Thus, when (Molenaar et al. 2012; Northcott et al. 2012; Brastianos
‘‘walking’’ down a chromothriptic chromosome, there et al. 2013) and blood malignancies (MacKinnon and
will be an invariant alternation between head and tail Campbell 2013; Morin et al. 2013; Nagel et al. 2013). It
(ortailandhead)sequences,reflectingalinearconnection isinterestingtopointoutthattherateofchromothripsis
(Korbel and Campbell 2013). This contrasts with the seemstobeunrelatedtotheoverallrateofsomaticcopy
prediction for a mechanism involving sequential acqui- numberalterations(SCNAs)(Zacketal.2013).
sition of rearrangements, particularly those involving Thehighfrequencyofchromothripsisincertaintumor
amplifications. In this latter case, rearranged segments types suggests that chromothripsis depends on the ge-
could be ‘‘reused’’ in ways that would generate heads neticandenvironmentalbackgroundofcancers.Among
followedbyheadsandtailsfollowedbytailswhenampli- the most frequently mutated genes in human cancer is
ficationsarenested. TP53 (Olivier et al. 2010), which plays central roles in
Clearly, within a short period of time, the field has maintaining genome stability (Zhu et al. 2002; Meek
advanced far beyond ‘‘know it when you see it’’ opera- 2004; Thompson and Compton 2010). The correlation
tional definitions for chromothripsis. The above-delin- between TP53 mutation and chromothripsis was first
eatedcriteriaeachaddressadifferenthallmarkfeatureof demonstratedinSonicHedgehog(SHH)medulloblastoma.
chromothripsisandcanbesubjectedtorigorousstatisti- Rauschetal.(2012)appliedhigh-throughputwhole-genome
cal evaluation. Nevertheless, challenges remain. How sequencing and long-fragment mate-pair sequencing to
exactly to combine multiple criteria and weigh them characterize genomic rearrangements in four patients of
properly to make the most probable case for or against SHHmedulloblastomawhoalsohadgermline-inactivat-
chromothripsisneedsfurtherdevelopment.Therearealso ingTP53mutations(Li-Fraumenisyndrome).Sequencing
numeroustechnicalchallenges.Shortsequencereadscan ofeachofthesetumorsnotonlyconfirmedclusteringof
onlybealignedtothemappableregionsofthegenome,and chromosomal breaks indicative of chromothripsis but
the interesting question of what happens within highly alsorevealedhigh-levelamplificationsduetodouble-minute
repetitive sequences remains to be resolved. Complica- chromosomes that originated from the shattered frag-
tions alsocomefromother factors, suchasinaccuracies ments. Thus, TP53 mutations are permissive or condu-
in short-read alignments, the obscuring effect of other cive for either the initiating chromothriptic event, the
GENES&DEVELOPMENT 2519
Zhangetal.
selective advantage of cells bearing the rearranged chro- ing mechanism in the formation of high-level focal
mosome, or both. Rausch et al. (2012) further identified amplifications that incorporate multiple disconnected
strongcorrelationsbetweenTP53mutationandchromo- segments(Iwakawaetal.2013;Sanbornetal.2013).The
thripsis in acute myeloid leukemia, demonstrating that multiple focal amplificationscanbeeither extrachromo-
the effect of TP53 mutation can manifest in multiple somal (double minutes) or intrachromosomal (homoge-
tissuetypes. neously staining regions [HSRs]) due to chromosomal
How accurate are the estimates of the frequency of integration of double minutes (Storlazzi et al. 2010;
chromothripsis?Althoughtherecouldbesomeoveresti- Gibaud et al. 2013). Indeed, data linking the formation
mationduetoambiguitiesintheinterpretationofgeno- of double minutes to the formation of derivative chro-
micdata,wemustalsoconsiderunderestimationdueto mosomes containing complex rearrangements were
thefactthattheoverwhelmingmajorityofchromothrip- known prior to the first description of chromothripsis.
ticeventsarealmostcertainlylostbecauseofdeleterious Gibaud et al. (2010) characterized five different extra-
fitness effects. Negatively selected events in normal or chromosomal amplifications of double-minute chromo-
premalignant tissues could be detected only at a single- somes in a glioma sample: One double-minute chromo-
cell level immediately after the chromothriptic rear- someconsistedofmultiplesegmentsfromchromosome5
rangement occurs,whichiscurrentlynotfeasibleonthe and chromosome 9, which, in retrospect, was likely to
appropriate scale. Ultimately, understanding the mecha- haveresultedfromchromothripsis.FISHanalysisverified
nisms that generate chromothripsis will be integral to thattheextrachromosomalampliconsequenceswerenot
estimatingitsfrequency.Iftheunderlyingcellulardefects present in the derivative chromosomes derived from
could lead to, but not cause, chromothripsis every time chromosome5andchromosome9.Wefurthernotethat
they occur, then the observed cases of chromothripsis copy number profiles of both chromosomes showed the
onlyrepresentthetipoftheicebergforlessdramaticbut hallmark copy number patterns suggestive of chromo-
possiblymorecommonmutationalevents. thripsis. In addition, the rearrangements also created
oncogenic lesions. The double minute contained the
TERT (telomere reverse transcriptase) gene from a frag-
Doeschromothripsis cause cancer?
ment in chromosome 5. TERT is amplified or overex-
Ingeneral,weexpectthatthedisruptionofchromosomal pressedinmanycancertypes(Beroukhimetal.2010).The
integrityinacatastrophesuchaschromothripsiswillbe functional significance of TERT overexpression is re-
negatively selected or even trigger apoptosis. However, markably underscored by the recent discovery that
this does not exclude the possibility that rare catas- ;80% of glioblastomas harbored mutations in the 59
trophes could promote cellular transformation, which flanking region of TERT that activate its expression
couldhappeninatleasttwowaysthroughcopynumber (Horn et al. 2013; Huang et al. 2013; Killela et al. 2013).
alterations.First,themassiveDNAlosscanincludeone Interestingly,thegliomasamplestudiedbyGibaudet al.
or more tumor suppressor genes (Stephens et al. 2011; (2010) also had deletions encompassing the CDKN2A
Yangetal.2013).IntheirsurveyoftheTCGASNParray locus on the chromosome 9 derivatives. Thus, one
data,Zacketal.(2013)foundthat72%ofchromothripsis episode of chromothripsis appears to have generated
events resulted in SCNAs in regions that are frequently two critical oncogenic lesions (TERT amplification and
disrupted in cancer, with the most dramatic case in CDKN2Adeletion).
glioblastoma showing that 20 out of 22 samples with Besides copy number alterations, translocations in
chromothripsis had one event disrupting the CDKN2A chromothripsiscanalsoleadtogenefusionsordysregu-
locus. Second, and perhaps most importantly, shattered latedgeneexpression.Mehineetal.(2013)suggestedthat
chromosomalsegmentsnotincorporatedintothederiva- chromothripsis is a frequent driver event in uterine
tivechromosomecanbelinkedtoformadouble-minute leiomyomas,resultinginincreasedexpressionofthetrans-
(circular)chromosome.Thesedoubleminutesarereadily located HMGA1 and HMGA2 genes. However, the proba-
amplified, which commonly occurs if they harbor an bilityofgeneratingin-framegenefusions—themajorclassof
oncogene.Anillustrationofthiscomesfromtheoriginal oncogenic translocations—through random chromosomal
reports on chromothripsis (Stephens et al. 2011; Rausch translocationsisgenerallysmall(Stephensetal.2009),and
et al. 2012). Stephens et al. (2011) described a small-cell theobservedfrequencyofin-framefusionsduetochromo-
lung cancer cell line where a chromothriptic event gen- thripsis is comparable with those resulting from simple
erated a derivative chromosome 8 and a double minute rearrangementevents(Stephensetal.2011).Nonetheless,
containing MYC and other segments of chromosome 8. although each rearranged segment in chromothripsis has
Most notably, the sequences on the derivative chromo- asmallprobabilityofgeneratingafunctionallyimportant
some and the double minute were exclusive, providing fusion gene (Teles Alves et al. 2013), it is reasonable to
compellingevidencethatearlyinthedevelopmentofthis speculatethatchromothripsis,becauseitgeneratesmany
tumor, chromosome 8 was shattered, with some bits rearrangements,willincreasethefrequencyofsuchfusions.
being contemporaneously incorporated into the deriva- Thisisnotbecausechromothripsishasanyuniqueability
tiveandothersgoingintothedoubleminute. togenerategenefusionsbutsimplyreflectsthefactthat
Having recognized the association between chromo- inchromothripsis,moreshotsaretakenongoal.
thripsis, double minutes, and cancer-causing mutations, Although there is good reason to think that chromo-
onecaninferchromothripsisasthemostlikelyunderly- thripsiscangenerateoncogenicalterationsandpotentially
2520 GENES&DEVELOPMENT
Genomeevolutionandcomplexrearrangements
multiplemutationsatonefellswoop,wearejuststartingto Chromoanasynthesis
understandtheimpactofchromothripsisoncancerpatho-
genesis or prognosis. Recent studies have shown that A different pattern of constitutional complex genome
chromothripsis can independently arise in either the pri- rearrangementswasreportedbyLiuetal.(2011)intheir
marytumororitsmetastasis(Kloostermanetal.2011b)and investigation of patients with congenital developmental
does not necessarily predict resistance to drug treatment defects. The investigators combined aCGH and FISH to
(Bassaganyas et al. 2013). Even though chromothripsis characterize copy number alterations and chromosomal
hasbeensuggestedasapoorprognosticfactor(Molenaar rearrangementsinthesepatientsandsequencedthePCR
et al. 2012; Hirsch et al. 2013), direct experimental products at rearrangement junctions to resolve break-
characterization of chromothripsis-derived mutations pointsequences.Incontrasttochromothripsis,Liuetal.
islacking. (2011) observed frequent copy number gains with inter-
Oneimportantquestioniswhenchromothripsisoccurs spersed regions containing duplication or triplication of
during tumor development: Can it be an early driving oneparentalallelebutnoLOH(Fig.1D).Breakpointjunc-
event,orisitjustanotherconsequenceofgenomicchaos tion sequences exhibited microhomology but also in-
in late stage cancers? The latter possibility is intuitive, cluded templated sequence insertions (100–1000 base
butthereisalsointerestingevidenceforchromothripsis pairs [bp]) on the rearranged chromosome. The short
occurring in the absence of massive genetic instability. template insertions suggested a replication-based mech-
First, allele frequencies of chromothriptic rearrange- anismforincorporatingthesesegmentsduringtherepair
ments suggest that chromothripsis generally occurs ear- ofDNADSBs(Leeetal.2007;Hastingsetal.2009).The
lier than simple rearrangements (Malhotra et al. 2013). observationofmicrohomologyfurtherledLiuetal.(2011)
Moreover, the high frequency of chromothripsis identi- to propose microhomology-mediated breakage-induced
fied in tumors with low to moderate mutation rates repair (MMBIR) as the major mechanism for generating
(neuroblastoma, medulloblastoma, glioblastoma, and theobservedtranslocations(Carvalhoetal.2009;Zhang
bloodandbonemarrowmalignancies)suggeststhatchro- etal.2009b;Liuetal.2012).Inthismodel,theextraallele
mothripsisdoesnot,inanyobligatoryway,requireaback- copiesareincorporatedduetotemplate-switchingevents
groundofelevatedgenomicinstability.Ontheotherhand, in DNA replication instead of ligation (via end-joining)
tumortypeswithfrequentcopynumberalterations,such of shattered fragments. To emphasize the difference in
as ovarian or lung cancers, exhibit a larger number of mechanisms, Liu et al. (2011) termed this phenomenon
chromosomal alterations, but they are often indepen- chromoanasynthesis (synthesis of new chromosomes).
dentlyacquiredratherthancorrelatedasinchromothrip- Theyalsosuggestedthatthesamereplicativemechanism
sis(Malhotraetal.2013;Zacketal.2013).Additionally, mayplayaroleinchromothripsisduetosimilarfeatures
as described in more detail below, de novo complex in both phenomena, such as localized clusters of rear-
rearrangements that resemble chromothripsis have also rangementsandcopynumberalternations.
been observed in patients with congenital disorders. So One of the hallmark features of chromothripsis is
far, there has been no evidence that these patients have interspersedlossandpreservationofheterozygosity,which
a cancerpredisposition ordefectsinthemaintenance of suggestsa‘‘shatter-and-stitch’’processofonechromatid.
genome stability like patients with germline mutations This is in sharp contrast to chromoanasynthesis, which
intheTP53orBRCAgenes.Together,thesefindingsimply reflectsextrasynthesisofDNAsegments.Althoughextra
thatchromothripsiscanoriginateincellswithnormalor synthesiscanleadtoalternatingDNAcopynumbers,it
near-normalgeneticstability. doesnotnecessarilyresultinLOH:LOHoccursbyDNA
deletionorgeneconversion.Anotherdifferencebetween
chromoanasynthesisandchromothripsisisthatadjacent
Chromothripsis incongenitaldisorders
chromosomal ends in chromothripsis (especially in the
Strong selection during embryonic development en- germline)oftenresultfromoneDNADSB,whereasseg-
suresthatonlymildgenomicalterationscanbetolerated mental breaks in chromoanasynthesis are separated by
through development. It was therefore very surprising longer distances (0.01;1 Mb) and appear to result from
when recent studies reported massive constitutional template-switching events. Although replication-based
chromosomal rearrangements similar to those of chro- DNA repair can contribute to chromothripsis, as evi-
mothripsis in cancer in some patients with congenital denced from examplesof templated insertion sequences
developmentaldisorders(Kloostermanetal.2011a,2012; (Chiangetal.2012),itisdifficulttoexplaintheproximity
Chiangetal.2012;Nazaryanetal.2013).Incomparison ofadjacentchromosomalbreaksbasedpurelyontemplate
withchromothripsis incancer,theseinstancesofappar- switching without DNA DSBs. Finally, chromoanasyn-
entgermlinechromothripsisshowedminimalDNAloss, thesisresultsinoneaberrantchromosomewithduplicated
presumably reflecting the selection against haploinsuffi- segments due to extra DNA synthesis; in contrast, chro-
ciency or complete loss of gene function due to mono- mothripsisgenerally givesrisetoone or multiple deriva-
allelicgeneexpression.Array-basedcopynumberanalysis, tive chromosomes with either massive DNA loss (in
suchasarraycomparativegenomehybridization(aCGH) cancer)or noapparentDNAcopy number alterations(in
or SNP arrays, cannot, as expected, detect these copy- the germline). Based on these differential characteristics,
neutral chromosomal alterations, necessitating whole- weagreewiththeconclusionthatchromoanasynthesisis
genomesequencing. adistinctphenomenonfromchromothripsis(Hollandand
GENES&DEVELOPMENT 2521
Zhangetal.
Cleveland 2012; Liu et al. 2012; Kloosterman and signaturesofchromoplexysuggestthatallofthepartner
Cuppen2013). segments involved in the translocations originate from
The distinction between chromothripsis and chromo- DNADSBs,withbothendsofeachbreakcontributingto
anasynthesis is also evident in a more recent report of the daisy chain cycle of translocations (Fig. 4A). This
germline genomic rearrangements. Kloosterman et al. organizationofbreakpointendsbearssomesimilarityto
(2012) identified eight cases that resemble chromothrip- thepairingoftheendsdescribedaboveinchromothripsis.
sis,exhibitingrandomstitchingofsegmentsfrommulti- As with chromothripsis, the pairing pattern strongly
ple chromosomes with little DNA loss, and two cases implies that chromoplexy is another example of an all-
withapparentgainordeletionofDNAsegmentsthatare at-once mechanism for chromosomal rearrangement: If
suggestiveofreplicationerrors.Oneinterestingobserva- the breakpoint ends originate from DSBs and are joined
tionisthattherearrangedallelesinbothchromothripsis intoanintegratedpatternoftranslocations,thenthebreaks
and chromoanasynthesis are preferentially of paternal musthaveoccurredsimultaneously.
origin; this was attributed to errors that result from the To put the intuitive argument into a more rigorous
higher number of mitotic divisions in gametogenesis in framework, Baca et al. (2013) developed an algorithm
malesrelativetofemales(KloostermanandCuppen2013). calledChainFindertoidentifythechainedtranslocations
In contrast, most human germline aneuploidies are ma- of chromoplexy. The assumption is that the probability
ternally derived (Handyside et al. 2012; Nagaoka et al. that the observed translocations of two adjacent break-
2012), and de novo chromosomal instability in early em- point ends occurred simultaneously is related to the
bryonic development does not show a preferred origin probability that two breakpoint ends originate from a
(Vannesteetal.2009;Voetetal.2011). singleDSB.TheinferenceofaDSBisnotstraightforward
because the breakpoint ends originating from a single
DSB due to resection and other processing are not pre-
Chromoplexy
ciselyconnectedinthereferencesequence(seeFig.4Bfor
Chromothripsis and chromoanasynthesis are two exam- anillustration);anullmodelforindependentlygenerated
ples of complex genomic rearrangement triggered or breaksisthereforenecessaryforastatisticalinference.To
generated by a single event. A single catastrophic event create the null model of independently generated chro-
wasalsoinvokedtoexplainarecentlydescribedpattern mosomalbreaks,ChainFinderfirstestimatesthedensity
of rearrangement found in prostate cancers (Fig. 1E; of chromosomal breaks at any locus by averaging over
Bergeretal.2011;Bacaetal.2013).Termedchromoplexy multipletumorsamplesandthenmultiplyingthisbythe
(Greek for braid of chromosomes), this event is charac- total number of chromosomal breaks detected in a par-
terized by ‘‘chains’’ of translocations involving multiple ticular sample. This generates a sample-specific, locus-
chromosomes—uptoeightsofardescribed.Thesequence dependentfrequency ofbreaks. Basedonthenull model
Figure4. ChromoplexyandChainFinder.(A)Chromo-
plexy is an extension of balanced rearrangements to
multiplepartners.Thedifferentcolorsoflinkeddiscor-
dantpairmatesreflectdifferentchromosomesorlocito
whicheachmateisaligned.DNAresectionduringthe
repairofDSBsoftenresultsinlossofasmallsegment
between the two ends. (B) Adjacent translocations of
two breakpoint ends can arise in two different ways.
(Left) In a single-event model, a single DNA DSB
generates two broken ends with a small deletion due
toDNAresection;joiningofthetwobrokenendswith
ends from other DNA breaks produces the observed
translocations. (Right) In a multiple-event model, two
independent balanced translocations occur indepen-
dentlyatdifferenttimes;theinnerDNAsegmentcould
havebeenlostinasubsequentdeletionevent(shownin
thedashedbox)orduetoDNAresection.Thedistance
between the two adjacent translocation ends, D, pro-
vides a statistical measure of the likelihood of either
scenario.Thetwo-eventmodelimpliesthattwoindepen-
dentbreaksmusthaveoccurredwithindistanceD.This
likelihood is approximately given by 1 (cid:2) exp((cid:2)mD),
wheremistheaveragedensityofchromosomalbreaks
that can be estimated for each sample at any given
locus (&10(cid:2)6 per base pair). When the ends are very
close (100–1000 bp), this probability is vanishingly
small, favoring the single-event model and rejecting the multievent model. The central idea of ChainFinder is to find ‘‘chained’’
translocationsthatformaclosedcycleandwherethemultiple-eventmodelcanberejectedforanytwopairsofadjacenttranslocations
(i.e.,noreciprocaltranslocationcanbepostulatedasanintermediatestate,connectingtwosmallerchains).
2522 GENES&DEVELOPMENT