2012_intra-tumour heterogeneity- a looking glass for cancer

7/27/2019 2012_Intra-Tumour Heterogeneity- A Looking Glass for Cancer

1/12

Whereas phenotypic heterogeneity of cells withintumours has been noted since the earliest days of can-cer biology, the experimental exploration of the subjecthad been limited by a lack of conceptual framework andstudy tools. The situation changed with the discoverythat the formation of tumours is dependent on the acqui-sition of oncogenic mutations, and with the realizationthat cancers arise through a Darwinian-like clonal evolu-tion. These findings have underlined the developmentof a gene-centric perspective, in which the existence ofheterogeneity in clinically important traits was attributedto genetic diversity resulting from clonal evolution1. The

validity of this perspective has been supported by recenthigh-resolution genome-wide studies that have revealedstartling genetic heterogeneity within individual cancers,

including population diversity in mutations involv-ing putative driver loci2,3. However, the dominance ofgene-centric views has been challenged with the rapiddevelopment of research within the cancer stem cell(CSC) hypothesis, which posits that differences betweenfunctionally important properties of individual tumourcells arise from differences in their differentiation sta-tus, thus bringing non-genetic sources of phenotypic

variability into focus.Thus, the concepts of clonal evolution and CSCs con-

stitute two major frameworks for interpreting the causesof phenotypic heterogeneity4,5. Clearly, these two con-cepts are not mutually exclusive, as evidenced by recent

reports on substantial intra-tumour genetic hetero-geneity in populations of putative CSCs3,6,7. Furthermore,non-genetic sources of phenotypic heterogeneity are notlimited to differences that can be mapped to differentia-tion hierarchies, and noise in gene expression leads tosubstantial cell-to-cell variability within seemingly pheno-typically homogenous subpopulations. In this Review,we discuss genetic and non-genetic causes of phenotypic

variability(BOX 1) and argue that contributions of geneticheterogeneity, as well as changes in non-genetic sourcesof heterogeneity resulting from oncogenic transforma-tion and abnormalities in the tumour microenviron-ment, make intra-tumour phenotypic heterogeneity adistinct phenomenon. We further discuss the impactthat phenotypic heterogeneity has on clinical diagnos-

tics and therapeutic resistance, suggesting that it needsto be incorporated into clinical practice to improve themanagement of cancer patients.

Genetic heterogeneity

The initiation and progression of cancers are dependenton the acquisition of multiple driver mutations that acti-

vate oncogenic pathways and that inactivate tumour sup-pressors8. However, unlike experimental tumours thatcan be cleanly induced by a few defined genetic events9,spontaneous tumours arise through a much more com-plex process of Darwinian-like somatic evolution, dur-ing which the acquisition of necessary driving mutations

1Department of Medical

Oncology, Dana-FarberCancer Institute, and

Department of Medicine,

Harvard Medical School,

Boston MA 02215, USA.2Department of Medical

Oncology, Hospital Clnic,

Institut dInvestigacions

Biomdiques August Pi i

Sunyer, Barcelona 08036,

Spain.

Correspondence to K.P.

e-mail: Kornelia_Polyak@

dfci.harvard.edu

doi:10.1038/nrc3261

Published online 19 April 2012

Intra-tumour heterogeneity:a looking glass for cancer?

Andriy Marusyk1, Vanessa Almendro1,2 and Kornelia Polyak1

Abstract | Populations of tumour cells display remarkable variability in almost every

discernable phenotypic trait, including clinically important phenotypes such as ability to seed

metastases and to survive therapy. This phenotypic diversity results from the integration of

both genetic and non-genetic influences. Recent technological advances have improved

the molecular understanding of cancers and the identification of targets for therapeuticinterventions. However, it has become exceedingly apparent that the utility of profiles based

on the analysis of tumours en masseis limited by intra-tumour genetic and epigeneticheterogeneity, as characteristics of the most abundant cell type might not necessarily predict

the properties of mixed populations. In this Review, we discuss both genetic and non-genetic

causes of phenotypic heterogeneity of tumour cells, with an emphasis on heritable

phenotypes that serve as a substrate for clonal selection. We discuss the implications of

intra-tumour heterogeneity in diagnostics and the development of therapeutic resistance.

REVIEWS

NATURE REVIEWS |CANCER VOLUME 12 | MAY 2012 |323

2012 Macmillan Publishers Limited. All rights reserved
mailto:[email protected]:[email protected]:[email protected]:[email protected]


2/12

Phenotypic plasticity

The ability of cells to change

phenotype stochastically or in

response to changes in the

environment.

is associated with the occurrence of many thousands ofsomatic genetic alterations that do not seem to conferselective fitness advantage (so-called passengers). Thelarge numbers of proliferation cycles that are requiredfor the formation of macroscopic tumours and theincreased mutation rates commonly seen in cancers(mostly in the form of genomic instability) allow for sub-stantial genetic diversification of populations of tumourcells. Furthermore, as discussed below, clonal evolutionthat underlies tumour progression probably proceeds ina branching rather than in a linear manner, which mightlead to substantial clonal diversity, further contributingto genetic heterogeneity within tumours.

Obviously, not every genetic difference between cellsin a tumour translates into differences in phenotype, andonly limited numbers of these phenotypic differences arerelevant to clinically important traits. However, given thereported intra-tumour heterogeneity in putative driver

mutations2,3 and in loci implicated in therapeutic resist-ance10, genotypic diversity is expected to translate intoheterogeneity in clinically relevant heritable phenotypes.In addition to straightforward phenotypic manifesta-tions, the accumulation of large numbers of passengermutations can affect cellular phenotypes in more subtleways. As elaborated on below, mutations that are nor-mally silent might potentially increase the phenotypicplasticity of cancer cells by deregulating gene expres-sion networks and by altering epigenetic landscapes.Furthermore, because tumour cells accumulate largenumbers of genetic mutations, closer attention shouldbe paid to possible genetic interactions among these

mutations11. It should also be noted that the classificationof mutations into passengers and drivers can be a mis-leading oversimplification, as the fitness effect of a givenphenotype is context-specific, and, as tumours changeover time, the selective value of a given mutation canchange substantially. Therefore, although genetic hetero-geneity is unlikely to be the major contributor to pheno-typic heterogeneity in general, it underlies heritabledifferences and thus provides an essential substrate forfuelling tumour evolution during tumour progressionand therapeutic resistance.

Non-genetic heterogeneity

Non-genetic heterogeneity in normal tissues. As all thecells in an individualshare the same genotype (withthe exception of immunoglobulin rearrangements inimmune cells and random somatic mutations), thephenotypic identities of normal cells are defined bynon-genetic mechanisms. Non-genetic heterogeneity ofphenotypes can be separated into two types: determin-istic and stochastic (BOX 1). Deterministic heterogene-

ity in normal tissues corresponds to the heterogeneityof cellular phenotype that can be mapped to a tissue-specific differentiation hierarchy(FIG. 1). However, owingto stochastic heterogeneity, cellular phenotypes withingiven deterministic states are not static, as examinationof individual cells within populations inevitably revealssubstantial cell-to-cell variability emanating from thestochastic character of biochemical processes within thecells12. The most well-studied aspect of these fluctuationsis transcriptional noise, but stochastic fluctuations arelikely to involve many other aspects of cell physiology,such as translation and the assembly of signalling com-plexes13. In fact, some researchers in this field equate thisnoise-driven variability with heterogeneity1416. Below,we provide an overview of how non-genetic mecha-nisms that contribute to phenotypes in normal tissuesare altered in cancers and how these changes might affectphenotypic diversity.

Differentiation hierarchies in cancers. The boom in CSCresearch promoted the CSC perspective as the dominantframework in which to consider non-genetic sourcesof phenotypic heterogeneity of cells within tumours5.According to this paradigm, phenotypic heterogeneityin cancers is a reflection of differentiation hierarchiesthat exist in normal tissues. Indeed, phenotypic hetero-geneity that can be mapped to distinct differentiation

statesseems to be dominant over the effects of oncogenictransformations, as gene expression profiles of more dif-ferentiated and stem cell-like subpopulations in breastcancers cluster more closely to their counterparts innormal tissues than they do to each other 7,17.

Furthermore, phenotypic heterogeneity that is under-lined by differentiation states has been linked to cruciallyimportant clinical outcomes such as prognosis, resist-ance to therapy and ability to seed metastases18. In gen-eral, higher expression of stem cell markers, presumablyreflecting the higher frequency of these cells in tumours,is associated with worse clinical outcome, leading tothe commonly accepted interpretation that CSCs are

At a glance

Primaryhumantumoursconsistofcellsthatdifferinclinicallyimportantphenotypic

features.Thisphenotypicheterogeneityisaresultoftheinterplaybetweengenetic

andnon-geneticfactorsthatshapecellularphenotypes.

Genomicinstability,whichisfrequentlyobservedinhumancancers,incombination

withthelargenumbersofcelldivisionsrequiredfortheformationofmacroscopic

tumours,leadstoinevitablegeneticdiversityinpopulationsoftumour cells.

SomaticevolutionthatdrivestumourprogressionischaracterizedbycomplexdynamicsarisingfromtheDarwiniannatureoftheprocess.Asaresult,individual

tumourshaveauniqueclonalarchitecturethatisspatiallyandtemporally

heterogeneous.

Thecancerstemcellperspectivecanexplainonlysomeofthenon-geneticvariability

intumourcellphenotypes. Amorecomprehensiveexplanationofnon-genetic

sourcesofphenotypicheterogeneitynecessitatestheconsiderationofmechanisms

thatunderliecellularphenotypes.

Bothdeterministicandstochasticdeterminantsofcellularphenotypescanbe

substantiallyaffectedduringoncogenictransformationandtumourprogression,

contributingbothtoabnormalphenotypesandtoanincreaseddegreeofphenotypic

plasticity.

Phenotypicandgeneticheterogeneitywithintumoursimpedesclinicaldiagnostics:

owingtotopologicalheterogeneityinthedistributionofdiagnosticallyimportant

phenotypesevenmultiplesamplingmightnotprovideadequateinformation.Atthe

sametime,giventhelinkbetweenahighdegreeofgeneticheterogeneityandpoor

prognosis,ameasureofheterogeneitybyitselfmaybeusefulasaprognostic marker.

Phenotypicheterogeneityintumourcellpopulationsthatresultsfrombothgenetic

andnon-geneticdeterminantsconstitutesamajorsourceoftherapeuticresistance.

Initialphenotypicheterogeneityandchangesincellularphenotypesresultingfrom

adaptationtoresponseandselectionforresistantphenotypesneedtobeaccounted

forinordertoachievesubstantialimprovementsintherapeuticoutcomes.

R E V I E W S

324 | MAY 2012 | VOLUME 12 www.nature.com/reviews/cancer



3/12

Effective population sizes

The numbers of tumour cells

capable of passing their

genotypes to next generations.

more treatment-resistant and metastatic than non-CSCcancer cells.However, alternative explanations includelarger effective population sizes that increase the prob-ability of genetic and non-genetic variability regardlessof stemness.

Despite the obvious attractiveness and potential util-ity of the CSC perspective, unambiguous description ofcancer cell phenotypes simply based on their differentia-tion states might not be realistically achievable19 owingto the altered identity of differentiation states in cancer.The relationship between normal and tumour differen-tiation hierarchies can be aptly described using a similethat describes tumours as like caricatures of correspond-ing normal tissues. This blurring of identity should notbe surprising, as the massive epigenetic abnormalitiesacquired by cancer cells, as well as the multiple driverand passenger mutational events, can potentially resultin unique deterministic phenotypes that might notbe directly mapped to similar states in normal tissues.Thus, owing to the unique evolutionary trajectory ofevery tumour resulting in distinct patterns of mutational

abnormalities, every caricature could be expected to haveat least some unique features.

In addition to the altered identity of differentiationstates, the CSC perspective faces the additional caveatof phenotypic plasticity of CSCs. Whereas the abilityto dedifferentiate into a stem cell state is not limited tocancer cells, oncogenic transformation increases theprobability of such a transition20. Furthermore, stemnessof cancer cells can be dramatically affected by the par-ticularities of the functional assays that are used for their

definition. For example, the strain of immunodeficientmice used for xenograft transplantations of primaryhuman melanoma influenced the frequency of CSCs byseveral orders of magnitude21. Whereas phenotypic iden-tification of CSCs faces substantial challenges, the cleardefinition of non-stem cell states is even more problem-atic, and in most cases cancer cells that do not displaya CSC phenotype are placed in the non-CSC category,which can include phenotypes with biologically andclinically important distinctions (FIG. 1).

Epigenetic landscapes and expression noise. A morecomprehensive view of non-genetic sources of hetero-geneity on tumour cell phenotypes can be achieved byborrowing the systems biology concepts of epigeneticlandscapes and gene-regulatory networks (GRNs)(BOX 2). The epigenetic landscape allows visualization ofstable and semi-stable cellular phenotypes as valleys inthe landscape; these valleys correspond to GRN statesof higher probability. Whereas rigorous treatment of thesubject is outside the scope of this Review, the epigenetic

landscape concept can provide a useful framework forthe consideration of non-genetic changes that occurduring cancer development.

From this perspective, there are two ways to changecellular phenotypes: either by noise-driven transition-ing between distinct attractor states in the given land-scape or by changing the topology of the landscape.Fluctuations in gene expression and other cellular pro-cesses shift the position of cells on the epigenetic land-scape. If a fluctuation exceeds a certain threshold a cellcan transition to another attractor state, resulting in achange of phenotypic state (in the epigenetic landscapethis would amount to jumping to an adjacent valley)(FIG. 2). Unlike unicellular organisms, in which noise-based phenotypic switching can sometimes be a hedg-ing strategy for enhanced survival in harsh or changingenvironments22, in complex multicellular organisms thenoise-driven phenotypic transitions during homeostasisare held under tight control through multiple regulatorymechanisms23. However, many of these mechanisms arelikely to be affected by oncogenic transformation, geneticperturbations and abnormal microenvironments.

The landscape topology is sculpted both by cell-intrinsic factors (genetic makeup and pre-existing differ-entiation state) and by cell-extrinsic factors: the varioussignals that cells receive from their microenvironment.In normal tissues, the genetic makeup of cells is mostly

identical, and the environment is well structured. Bycontrast, both factors can be dramatically modified intumours (FIG. 2). Genotypic differences between tumourcells will translate to corresponding differences in thetopology of epigenetic landscapes even in a homo-geneous environment. Mutational inactivation of non-essential genes can subtly change the architecture ofGRNs24, and thusthe panoply of passenger mutationsthat accumulate during tumour progression mightalso contribute to both landscape topology and prob-abilities of noise-driven transitions. Moreover, as wediscuss below, microenvironments in tumours are bothabnormal and less structured, which can create distinct

Box 1 | Determinants of phenotypic heterogeneity

Genetic heterogeneityInnormaltissues,geneticheterogeneityisverylow,andalmostallphenotypic

diversityisattributabletonon-geneticsources.Incancers,branchingevolutionand

increasedgenomicinstabilityleadtoasubstantialdiversificationofgenotypes.

Whereasonlyasmallproportionofmutationswillhavedirectphenotypic

manifestations,thecoexistenceoflargenumbersofotherwiseneutralmutationsis

expectedtoleadtosomenovelphenotypesowingtogeneticinteractionsbetween

thesemutations.Exceedingbufferingcapacityoftheheatshockproteinresponsecan

furtherincreasethephenotypicdiversificationthatarisesfromgeneticdiversity.

Whereasgeneticheterogeneityisnotlikelytobetheprimarycontributorto

intra-tumourphenotypicheterogeneity,ingeneral,geneticchangesareheritable

and,therefore,essentialforsomaticevolutionduringtumourprogressionandthe

developmentoftherapyresistance.

Deterministic heterogeneity

Deterministicheterogeneitydescribestheexistenceofmultiplefairlystable

phenotypicstates.Thesestatescanbevisualizedasvalleysonepigeneticlandscapes.

Innormaltissues,allofdeterministicheterogeneitycanbemappedtodistinctstagesin

tissue-specificdifferentiationhierarchies.Incancers,substantialgeneticchangesand

epigeneticaberrations,aswellasabnormalandheterogeneousmicroenvironments,

mightleadtoincreasesindeterministicheterogeneity,includingtheappearanceof

phenotypicstatesthathavenodirectcorrespondencetonormaltissues.

Stochastic heterogeneity

Stochasticheterogeneityreferstothetransientdifferencesinphenotypesbetween

isogeniccellsthatsharethesamedeterministicphenotypicstate.Thesedifferences

resultfromthestochasticnatureofbiochemicalprocesseswithincells,andfrom

burst-likepatternsofgeneexpressionrelatedtochromatinorganization.Stochastic

heterogeneitycanmediateshort-termtherapeuticresistance.Furthermore,stochastic

processescanmediatetransitionsbetweendistinctdeterministicphenotypicstates.

R E V I E W S




4/12

Tissue stem cell

Committedprogenitors

Differentiated

cells

Normal

Stem cells

Non-stemcells

Cancera b

Allelic imbalances

Changes in the copy number

of alleles as a result of

chromosomal amplification

or deletions.

Heat shock protein

response

A response to cellular stress

involving the activation of

molecular chaperones called

heat shock proteins (HSPs).

HSPs help to maintain cellular

homeostasis and to promote

survival in the face of stress.

landscape topologies that are not accessible for cells innormal tissues. Changes in GRNs can both increase theprobability that a given f luctuation in gene expressiontransfers the cell into a different existing stable state (forexample, by changing the steepness of the landscape),and create new stable phenotypic states that weretransient in the normal tissue.

Abnormalities in gene expression that are seen incancers are not limited to transcriptional changes.Oncogenic transformation and abnormal micro-environments induce stress responses in cancer cells; infact, cellular stress has been proposed to be one of thehallmarks of cancer25. Cellular stresses, allelic imbalancesand the expression of certain oncogenes can activate theheat shock protein response, the saturation of which canelevate molecular noise at the protein level and increasethe probability of phenotypic manifestations of silentmutations26. We are not aware of any studies that directlycompare noise-driven transitions between normal andcancer cells; however, the demonstration that HRAS-driven transformation increases the occurrence of spon-taneous dedifferentiation in human mammary epithelialcells20 suggests that noise-driven phenotypic transitionsmight be enhanced following oncogenic transformation.

Effects of heterogeneous tumour environments. Tumourprogression involves dramatic changes in tissue micro-environments, including changes in the components27and properties of stromal cells28,29, as well as remodellingof the extracellular matrix (ECM)30. The recent surge ofinterest in the tumour microenvironment is associ-ated with the recognition that microenvironmentalalterations are not just passive consequences of geneticevolution occurring in cancer cells, but that they areactive participants in tumorigenesis. As many excellentreviews summarize progress in this area27,29, we focuson the effects of microenvironmental alterations on thephenotypic heterogeneity of malignant cells. Tumour

microenvironments can shape tumour cell phenotypesboth by providing abnormal contexts that augmentinternal variability (by inducing stress responses andgenomic instability) and through the higher variabilityof microenvironmental contexts. Furthermore, math-ematical modelling has suggested that heterogeneity intumour microenvironments could lead to the selectionof more invasive phenotypes31,32.

Tumour cells are continuously exposed to abnormalmicroenvironments, such as direct contact with fibro-blasts, inflammatory cells, hypoxia, acidity and remod-elled ECM. These abnormal environments can havepronounced influences on the epigenetic landscapes ofcells, translating into abnormal phenotypes, includingepithelial-to-mesenchymal transitions33 and vascularmimicry34. Normalization of the tumour microenviron-ment has been linked to the normalization of cellularphenotypes, and destabilization of normal tissue organiza-tion can translate into an increased risk of tumorigenesis30.Furthermore, abnormal tumour microenvironmentscan result in increased genomic instability35, further

diversifying tumour cell populations.Another important aspect of the tumour microenvi-

ronment is the loss of normal anatomical organization.Normal tissues and organs are highly organized, consist-ing of well-defined structural units (FIG. 2). Therefore, nor-mal cells experience limited numbers of environmentalhabitats (niches). This orderly organization is lost in can-cers, especially in later stages of progression. For example,neovascularization, which is required for macroscopictumour growth, results in highly abnormal vasculature,which consists of spatially irregular vascular architecturethat is composed of leaky and inefficient blood vessels. Asa result, even abundantly vascularized tumours often havehypoxic regions, and, in contrast to cells in normal tis-sues, tumour cells experience a range of conditions fromnormoxic to hypoxic, leading to variations in the supplyof oxygen and nutrients to the tumour. Moreover, owingto inconsistencies of blood flow through the tumour

vasculature, some tumour cells might experience fluc-tuations of oxygen supply over time36. This higher spatialand temporal variability of tumour microenvironmentsmight translate into a higher phenotypic heterogeneity oftumour cells. Heterogeneity in the tumour microenviron-ment is not limited to that of the vasculature but can beextended to most of its elements. Inter-tumour differencesin stroma have been recognized to have a major effect ontumour biology29 and clinical outcome37,38. However, dif-

ferences in microenvironments are frequently observednot only between tumours or between primary and meta-static lesions, but also within the same tumour. This intra-tumour heterogeneity of the microenvironment affectsthe phenotypic heterogeneity of tumour cells, poses aformidable barrier to cancer therapy and influences theevolutionary trajectory of cancers by providing differentselective pressures within the same tumour.

Heterogeneity in the clonal evolution of tumours

Since the original publication by Peter Nowell39, theidea that tumour progression is a genetic evolutionthat follows Darwinian logic (the diversification of

Figure 1 | Differentiation hierarchies in normal tissues and cancers. a |Thetypical

differentiation hierarchy in normal tissues is shown: stem cells self-renew and give rise

to committed, lineage-restricted progenitors that are capable of transient amplification

but that eventually differentiate. Under appropriate circumstances, committed

progenitors can dedifferentiate into stem cells (dashed arrow), but the probability of

this event ocurring is low. b | Differentiation hierarchies in cancers are shown. Typically,

tumour cells are separated into two categories: cancer stem cells and non-cancer

stem cells. Both categories can include phenotypically distinct subpopulations with

substantial biological and clinical differences. Furthermore, the probability of trans-

differentiation from the non-stem cell into the stem cell category is enhanced compared

with normal tissues.

R E V I E W S




5/12

c

a b

Probability

Parameter X

Attractor A

Attractor B

Low

Low

High

High

Probability

Parameter A

Parameter B

Attractor BAttractor A

Attractor C

Clone

A group of cells that share a

common ancestor and that are

genetically identical. Whereas

all of the cells in the body

originate from a single

ancestor, every new mutation

creates a new clone (a

subclone).

heritable types tested by natural selection) becamewidely accepted in cancer biology40. Genetic diversityis essential for evolution both in nature and in tumours.However, as selection is oblivious to the molecularmechanisms underlying new phenotypes, heritable non-genetic changes can also serve as a substrate for somaticevolution. Although epigenetic changes are not strictlyirreversible, the selective effect of some epigeneticevents, such as the inactivation of tumour suppressorgenes by promoter methylation, is frequently observedin cancer41, and it is indistinguishable from genetic inac-tivation. Furthermore, even epigenetic marks that canbe erased over multiple rounds of cell division can stilllubricate the machinery of natural selection (REF. 42).

A predominant, textbook model of somatic tumourevolution is that of clonal succession: tumour evolutionis thought to occur as a series of clonal expansions, eachof which is triggered by the acquisition of a driver muta-tion that confers large fitness gain, allowing the mutantclone to outcompete and outgrow the cells that lackthis mutation43. The implication of this model is that

tumours are homogeneous for mutations that function-ally matter (with the exception of a less advanced paren-tal clone that is in the process of being outcompeted).However, this view is in marked contrast to the currentunderstanding of evolutionary processes in asexualcell populations44,45 the closest natural analogues

of the evolutionary processes that occur in cancer.Furthermore, an increasing number of studies havedescribed the coexistence of genetically distinct clonalsubpopulations in primary cancers (reviewed in REF. 46).These clonally distinct subpopulations sometimes dis-play substantial genetic divergence47,48, which indicatescoexistence over long periods of time. Importantly, thedifferences between these clonal populations are notlimited to neutral loci, but they involve driver lesionsthat are expected to provide fitness gains3,49,50. Multiplemechanisms are probably involved in limiting the clonalhomogenization of tumours (BOX 3). Thus, evolutionarydynamics of somatic evolution in populations of tumourcells is likely to result in the coexistence of clonal vari-ants with genetic or epigenetic differences that can havesubstantial biological and clinical effects.

Heterogeneity in metastases

Cancer is a systemic disease, as disseminated tumourcells and numerous micrometastatic and macrometa-static lesions can be found throughout the body, thus

considerations of intra-tumour phenotypic hetero-geneity should not be limited to primary tumours. Thisis especially important given that distant metastases areresponsible for the majority of cancer-related mortal-ity, and that the therapeutic decisions in most cases arebased on the analysis of primary tumours. Therefore,

Box 2 | Epigenetic landscape

Theconceptoftheepigeneticlandscape,introduced

byWaddington102(seethefigure;parta),wasoriginally

developedasavisualrepresentationofcell

differentiation.Inthisrepresentation,acellisdepicted

asaballrollingonaruggedsurface.Thecoordinatesonthissurfacerepresentcellularstates;forexample,the

concentrationsofdifferentexpressedgenesthat

collectivelyleadtoadefinedphenotype.Theheightof

thesurfacerepresentsapotentialthatindicatesthe

relativestabilityofdifferentstates.Valleyscorrespond

tostablestates(alsotermedattractors)andridges

providecanalizationintodistinctstates.Althoughthis

landscapeishighlymulti-dimensional,itcanbe

schematicallyvisualizedinatwo-dimensionalorathree-dimensional

format(seethefigure;partsb and c).Theepigeneticlandscape,which

dictatesthestabilityofphenotypicstates,isacompoundresultofthe

interactionsandfeedbacksbetweendifferentexpressedgenesinwhat

areoftencalledgeneregulatorynetworks(GRNs)103.Thetopologyand

dynamicbehaviourofthesenetworksdependonbothcell-intrinsicand

cell-extrinsicfactors.IntrinsicfactorsincludeDNAmutations,aswellasepigeneticmodificationsofDNA(methylation)andhistones(primarily

acetylationandmethylation)104.Suchchanges,specificallyinpromoter

regions,couldleadtoarewiringofGRNs,directlyaffectingthe

landscape.Extracellularfactorsincludeautocrine,paracrineand

endocrinesignalling,andarethushighlydependentonthecell

microenvironment.Althoughtheoriginalmotivationoftheconceptof

theepigeneticlandscapewastoexplainhowidenticalgenotypescan

leadtodifferentdifferentiationhierarchiesofcells,ithasalsobecomea

usefultooltoconceptualizetheeffectsofcellularnoiseingene

expression.Astochasticchangeintheconcentrationofacertaingene

cantransferacelltoadifferentstablephenotypedependingonthe

steepnessoftheepigeneticlandscape.Partaisreproduced,with

permission,fromREF. 102(1957)Taylor&Francis.

R E V I E W S




6/12

Noise: high Microenvironment: disorganizedGenotypes: heterogeneous Network architecture: noisy

Attractor A

Noise: low Microenvironment: structuredGenotypes: homogeneous Network architecture: robust

Normal tissue: low phenotypic heterogeneity

Tumour: high phenotypic heterogeneity

Attractor B

Attractor A Attractor B

differences between primary tumours and metastaticlesions can lead to inaccurate diagnosis and treatment.Indeed, discordance in diagnostic markers between pri-mary and metastatic tumours has been reported in manystudies51. Given the clinical consequences of discord-ance, understanding the extent of differences betweenprimary and metastatic tumours is crucial for the propermanagement of cancer patients. However, experimentalevidence on this subject is still fragmentary.

Historically, views on relationships between primaryand metastatic lesions have been dominated by theclonal succession model of tumour evolution, in whichthe acquisition of metastatic ability is the final stage(FIG. 3). From this perspective, metastases are seeded bythe most advanced and aggressive clone that should alsodominate the primary tumour52. Furthermore, compari-sons of expression profiles between primary and meta-

static lesions have demonstrated a close similarity53,54,suggesting that primary tumours can indeed serve asadequate diagnostic proxies. Conversely, the analysisof disseminated tumour cells in cancer patients and inanimal models, and mathematical modelling based onthe timing of metastatic outgrowth, led to a competingidea of parallel progression of primary and metastatictumours associated with early spread of metastatic dis-ease and substantial genetic divergence both betweenprimary and metastatic lesions and among metastaticlesions55. Ultimately, the answers must come from theanalysis of clinical specimens. Multiple studies based onunbiased, but limited resolution, approaches, such as

comparative genomic hybridization (CGH) and singlenucleotide polymorphism (SNP) arrays, have revealeda mixed picture: although the majority of cases demon-strate a very close clonal relationship between primaryand metastatic tumours, some cases display markeddifferences (reviewed in REFS 46,55).

Recent advances in DNA-sequencing technolo-gies56 have allowed higher resolution queries of theclonal relationships between primary and metastatictumours. Analysis of metastatic pancreatic cancersrevealed that, although metastatic lesions were closelyrelated to primary tumours, they also contained addi-tional mutations57. Interestingly, clones that gave rise tometastases were also present in the primary tumoursas small subpopulations. The comparison of non-synonymous coding mutations between primary lobularbreast cancer and a metastatic lesion that developed

9 years after initial diagnosis revealed that mostof the mutations in the metastasis were not present inthe primary tumour or that they had a substantiallylower frequency, indicating considerable geneticdivergence despite common ancestry58. By contrast,the comparative analysis of a primary basal-like breastcancer and metastatic lesions that developed 8 monthsafter diagnosis demonstrated a much more similargenetic composition59. However, the metastatic lesionsdisplayed several de novo mutations, as well as alteredfrequency of some mutations present in the primarytumours. Interestingly, a xenograft derived fromthe primary lesion displayed changes in mutational

Figure 2 | Factors that shape cellular phenotypes: normal tissues versus tumours. Cellular phenotype represents

the integration of several inputs. Indirect evidence suggests that oncogenic transformation can lead to increased

stochastic fluctuations (noise) of cellular processes, increasing the probability of traversing a ridge in an epigenetic

landscape, leading to a phenotypic switch. Genotypes are homogeneous in normal tissues but heterogeneous in cancers

(owing to on-going genomic instability and clonal diversity). Microenvironments are highly structured in normal tissues,

resulting in limited numbers of distinct niches; but they are disorganized in tumours, leading to abnormalities and larger

numbers of less well-defined niches. Stable gene regulatory network (GRN) states in normal tissues are limited by intact

controls over gene expression, but multiple alterations of gene expression in cancers can lead to a loss of robustness.

The integration of these influences might lead to a larger variety of less stable cellular phenotypes in cancers compared

with normal tissues.

R E V I E W S




7/12


8/12

Dominant clone

Late dissemination

Minor cloneEarlydissemination

Primarytumour

Metastasis

Metastasis

Metastasis

Stepwise progression

Parallel progression

Bloodvessel

cure. Unfortunately, the heterogeneity of tumour cellsprevents this from happening in most cases.

For any therapy to work, it first needs to reachtumour cells at an effective dose. However, the hetero-geneity of the tumour vasculature, increased hydro-static pressure and other microenvironmental changesmake drug delivery inefficient, and in some regions ofthe tumour a therapeutic dose may not be achieved.This inequality of therapy gives tumour cells time todevelop resistant phenotypes. In addition to provid-ing sheltering from drugs, components of the micro-environment can actively protect tumour cells fromtreatment through secreted factors and cell contact-mediated pro-survival stimuli68. Furthermore, hetero-geneity in the tumour microenvironment translatesinto heterogeneity of tumour cell phenotypes, and sosome tumour cells might be intrinsically less sensitiveto the therapy. For example, hypoxic conditions have

been linked to more aggressive tumour cell features69,70;therefore, tumour cells in hypoxic regions might react

very differently to treatment compared with those innormoxic regions.

The most well-understood and documented modeof therapeutic resistance involves genetic alterationsthat can lead to the initial lack of response and relapsein multiple ways. Mutations can prevent the binding ofa drug to its target. For example, mutant forms of theBCRABL fusion protein71 and KIT72 have been impli-cated in the relapse of chronic myelogenousleukaemias(CMLs) and gastrointestinal stromal tumours (GISTs),respectively, that are treated with imatinib mesylate.

Mutations can also provide alternative means of acti-vating signalling pathways that are targeted by ther-apy, as exemplified by the resistance of BRAF-mutantmelanomas to the BRAF inhibitor PLX4032 owing tomutant KRAS73, or can activate alternative pro-survivalsignalling pathways, as exemplified by the amplifica-tion ofMETin lung cancers treated with epidermalgrowth factor receptor (EGFR) inhibitors74. An inter-esting mechanism of mutational resistance that restoreswild-type protein function has been implicated in therelapse of BRCA-mutant tumours that are treated withpoly(ADP-ribose) polymerase (PARP) inhibitors. Theimpairment of homologous recombination-mediatedDNA repair that results from the inactivation of BRCA1and BRCA2 sensitizes mutant cells to PARP inhibitortreatment. Resistant tumours display in-frame deletionsin BRCA that partially restore its DNA repair func-tion, allowing the outgrowth of these cells even in the

presence of PARP inhibitors75,76.Selective pressures that are exerted by cytotoxic

therapy can lead to the expansion of resistant clonesthat either existed before the onset of treatment or thatformed as a result of new mutations that were gainedduring the treatment. Whereas sampling and detectionsensitivity issues often limit the ability to distinguishbetween these two possibilities, multiple reports havedemonstrated that relapsed clones could be traced to

variants present as minor clones before the start of ther-apy7779. Therefore, the degree of genetic heterogeneityof a tumour is likely to be an important determinant oftherapeutic outcome.

Figure 3 | Tumour heterogeneity in metastatic spread. Two hypothetical models of tumour progression during

metastases are stepwise progression, in which metastases are seeded at the latest, clinically diagnosable stage of tumour

progression, and parallel progression, in which tumour dissemination occurs at earlier stages of tumour progression,

allowing for substantial divergences between metastatic lesions. In the case of parallel progression, the genetic

divergence between primary and metastatic tumours makes the primary tumour unsuitable for diagnostic purposes and

the design of effective therapies. However, even in the stepwise progression model, intra-tumour clonal diversity can

lead to a situation in which metastases are shed by a minor subclone, leading to inaccuracy of diagnostic and treatment

decisions that are based on the analysis of the primary tumour that do not consider intra-tumour heterogeneity.

R E V I E W S




9/12

Inter-tumourheterogeneity

Intra-tumourheterogeneity

Mixed dominanceDominance of clone 2Dominance of clone 1

Resistant phenotypes are not necessarily associatedwith genetic changes. In many cases, therapeutic resist-ance can be linked to altered gene expression patternswithout associated changes in DNA sequence80,81. Arecent study demonstrated that drug resistance can bea consequence of a stochastic phenotype switch that canpersist for multiple generations82. Whereas epigeneti-cally mediated drug-resistant states are not strictly per-manent, the stochastic nature of the transition and itsheritability within therapeutically relevant time framesallows for the emergence of resistant clones owing toselection. However, phenotypes that are associatedwith non-genetic resistance can also be a consequenceof stochastic heterogeneity. Populations of geneticallyidentical cells in a homogeneous environment showsubstantial cell-to-cell variability in response to cyto-toxic83 or ligand-based84 apoptotic stimuli. This vari-ability has been linked to noise-driven variability in theexpression levels of proteins that are involved in apop-tosis. Interestingly, even this resistance mechanism canbe heritable for a few population doublings, thereby

providing a pool of resistant cells that can potentiallyacquire resistance via more stable genetic or epigeneticmechanisms.

In summary, both genetic and non-genetic sources ofheterogeneity limit the ability of therapies to kill tumourcells, whereas clonal diversity feeds therapeutic relapseby the outgrowth of heritable genetic or epigenetic

variants that are resistant to therapy.

Conclusions and future directions

In this Review, we have considered distinct sources ofphenotypic heterogeneity in tumour cell populations.Tumour cell phenotypes are the result of the integra-tion of inputs from genotype, environmental stimuliand stochastic processes that occur within cells (FIG. 2).Genetic and epigenetic changes that arise during onco-genic transformation and tumour progression alter anddiversify cellular phenotypes, posing a major obstacleto the understanding and clinical management of can-cers. We suggest that the phenomenon of intra-tumourphenotypic heterogeneity, especially aspects that arerelated to clonal diversity, deserves to be recognized andaccounted for during the analysis of primary tumours,building of experimental models and design of thera-peutic approaches. Furthermore, because tumourscontain phenotypically distinct populations of bothtumour and stromal cells that interact in a dynamic andreciprocal manner, these interactions are likely to resultin the emergence of networks of interactions the prop-erties of which can be understood from an ecological

perspective40,46,85,86.How can intra-tumour heterogeneity be accounted

for in our quest to understand and treat cancers?We see several major possibilities. First, it is worthinterrogating whether intra-tumour phenotypicheterogeneity is linked to clinically importantaspects of primary human cancers, such as subtype,prognosis, risk of metastases and therapeutic resist-ance. Indeed, the degree of intra-tumour geneticheterogeneity has been associated with poor prog-nosis in oesophageal cancer 65 and breast cancer 87.However, the subject remains mostly unexplored.Furthermore, given the major contribution of non-genetic sources to phenotypic heterogeneity, it maybe worth exploring the link between non-geneticphenotypic diversity and clinical outcomes.

Whereas the quantitative measures of clonaldiversity can be adopted from other fields88, adequatemethods allowing unbiased and cost-effective interro-gation of tumours remain to be developed. Recentdevelopments in sequencing technologies hold prom-ise in this regard89. In addition, advances in the stud-ies of clonal diversity will require the developmentand refinement of sampling techniques46. Althoughbiologically and clinically relevant insights can begained from the analysis of multiple spatially distinctregions of the same tumour without prior knowledge

of clonal composition2, the application of methodsthat allow the separation of clonal populations beforethe analyses, such as fluorescence-activated cellsorting based on ploidy status10,90, can provide fur-ther improvements in resolution. Multi-parameterhigh-throughput analysis at the single-cell level isthe most desirable approach; however, despite suc-cessful application of single cell-based analysis inseveral recent studies47,91,92, the widespread applicabil-ity of this approach will require additional technicalimprovements.

Second, the intra-tumour heterogeneity of tumourcells,stromal cells and non-cellular components of

Figure 4 | Tumour heterogeneity in diagnostics. Similar to inter-tumour

heterogeneity, intra-tumour heterogeneity of cellular phenotypes that result from

genetic and non-genetic influences can complicate definitive diagnostics and can

obstruct therapeutic decision-making. First, spatial phenotypic heterogeneity

can lead to a situation in which a biopsy does not provide an adequate reflection of

the phenotypic composition of the whole tumour. Second, decisions made based on

scoring the dominant phenotype in a given sample might be misleading if they do not

account for minor subpopulations with clinically and biologically important distinct

features.

R E V I E W S




10/12

microenvironments probably results in the develop-ment of a network of interactions that cannot be deci-phered from the analysis of each of these individualcomponents alone. Although the importance of inter-actions between distinct subpopulations of tumourcells has been implicated in therapeutic resistance93 andmetastases94, this area remains unclear.

Third, as heterogeneity is a major obstacle to ther-apeutic success and as, in many cases, mechanismsunderlying drug resistance seem to be non-genetic,there is a potential for reducing the phenotypicheterogeneity of tumour cells towards limited num-bers of drug-sensitive states. The burgeoning of epige-netics research has generated an ever-increasing levelof understanding of the regulation of cellular pheno-types through the modification of DNA and histonesand has provided therapeutic tools for its modifi-cation via modulating the activity of enzymes thatare involved in these processes. Histone deacetylaseinhibitors and other compounds that affect epigeneticstates are already being investigated as anticancer

agents95, and the number of such drugs will probablyincrease dramatically in light of the recent discoveryof recurrent somatic mutations in epigenetic modu-lators in many human cancer types96. In addition, areduction of phenotypic heterogeneity can poten-tially be achieved through modulating the tumourmicroenvironment. An obvious example for this isanti-angiogenic therapy: although it is inefficientand potentially counterproductive as a monotherapy,it enhances the ef fect of conventional chemotherapyby normalizing the tumour vasculature, thus reduc-ing spatial and temporal heterogeneity in the bloodsupply and providing drugs with better access to can-cer cells36. Whereas considerations of tumour hetero-geneity were not behind the original rationale for the

development of therapies that target the epigeneticmodifications of cancer cells and the tumour micro-environment, accounting for heterogeneity can fur-ther inform the design of therapeutics. Importantly,given that inf luences of non-genetic modulators ofphenotypes can be dominant even over genetic influ-ences associated with oncogenesis7,17,97, it is likely thatthe reduction of non-geneticphenotypic heterogene-ity can be achieved despite a certain degree of geneticheterogeneity.

Finally, an adequate understanding of drivingforces in tumour initiation and progression neces-sitates evolutionary considerations40,98100. Heritablephenotypic heterogeneity that is associated withgenetic and non-genetic causes is a key parameterthat mediates evolutionary dynamics in cancers.Understanding evolutionary dynamics could inturn lead to the development of novel therapeuticapproaches. For example, given that in many casestherapy-resistant phenotypes are associated withreduced fitness under normal conditions, it was pro-

posed that the development of therapeutic resistancecould be avoided if a therapy is administered in away that prevents the resistant subpopulation fromcompletely outcompeting drug-sensitive cells101 .Unfortunately, our current understanding is mostlybased on mathematical modelling, whereas estimatesof variability from primary tumours are in short sup-ply, and animal models to interrogate evolutionarydynamics in tumour progression and therapy aremostly lacking. We believe that the recognition ofthe importance of intra-tumour heterogeneity com-bined with the rapid development of technologiesallowing its interrogation will improve our under-standing of tumours and the design of therapeuticinterventions.

1. Fidler, I. J. Tumor heterogeneity and the biology of

cancer invasion and metastasis. Cancer Res.38,

26512660 (1978).

2. Gerlinger, M. et al. Intratumor heterogeneity and

branched evolution revealed by multiregion

sequencing. N. Engl. J. Med.366, 883892

(2012).

This paper demonstrated substantial spatial

genetic heterogeneity within tumours, including

synonymous driving mutations independently

arising in distinct clones. Furthermore, it reported

distinct diagnostic signatures from different

biopsies of the same tumour.

3. Anderson, K.et al. Genetic variegation of clonal

architecture and propagating cells in leukaemia.Nature 469, 356361 (2011).

This paper demonstrated substantial genetic

variability in putative driver mutations between

leukaemia stem cells, suggesting complex patterns

of clonal evolution.

4. Campbell, L. L. & Polyak, K. Breast tumor

heterogeneity: cancer stem cells or clonal evolution?

Cell Cycle6, 23322338 (2007).

5. Shackleton, M., Quintana, E., Fearon, E. R. &

Morrison, S. J. Heterogeneity in cancer: cancer stem

cells versus clonal evolution. Cell138, 822829

(2009).

6. Notta, F. et al. Evolution of human BCR-ABL1

lymphoblastic leukaemia-initiating cells. Nature 469,

362367 (2011).

This paper demonstrated substantial genetic

diversity between leukaemia-initiating cells,

suggesting complex clonal architecture.

7. Shipitsin, M. et al. Molecular definition of breast

tumor heterogeneity. Cancer Cell11, 259273

(2007).

This paper demonstrated that presumed stem

cell and non-stem cell subpopulations in breast

tumours can have different patterns of

mutations.

8. Vogelstein, B. & Kinzler, K. W. Cancer genes and the

pathways they control. Nature Med.10, 789799

(2004).

9. Rangarajan, A., Hong, S. J., Gifford, A. & Weinberg,

R. A. Species- and cell type-specific requirements for

cellular transformation. Cancer Cell6, 171183

(2004).

10. Ruiz, C. et al. Advancing a clinically relevantperspective of the clonal nature of cancer. Proc. Natl

Acad. Sci. USA108, 1205412059 (2011).

11. Ashworth, A., Lord, C. J. & Reis-Filho, J. S. Genetic

interactions in cancer progression and treatment. Cell

145, 3038 (2011).

12. Raj, A. & van Oudenaarden, A. Nature, nurture, or

chance: stochastic gene expression and its

consequences. Cell135, 216226 (2008).

13. Niepel, M., Spencer, S. L. & Sorger, P. K. Non-genetic

cell-to-cell variability and the consequences for

pharmacology. Curr. Opin. Chem. Biol. 13, 556561

(2009).

14. Altschuler, S. J. & Wu, L. F. Cellular heterogeneity: do

differences make a difference? Cell141, 559563

(2010).

15. Huang, S. Non-genetic heterogeneity of cells in

development: more than just noise. Development136,

38533862 (2009).

16. Eldar, A. & Elowitz, M. B. Functional roles for noise in

genetic circuits. Nature467, 167173 (2010).

17. Dalerba, P. et al. Single-cell dissection of

transcriptional heterogeneity in human colon tumors.

Nature Biotechnol.29, 11201127 (2011).

18. Clevers, H. The cancer stem cell: premises,

promises and challenges. Nature Med.17,

313319 (2011).

19. Rosen, J. M. & Jordan, C. T. The increasing complexity

of the cancer stem cell paradigm. Science324,

16701673 (2009).

20. Chaffer, C. L. et al. Normal and neoplastic nonstem

cells can spontaneously convert to a stem-like state.

Proc. Natl Acad. Sci. USA. 108, 79507955

(2011).21. Quintana, E. et al. Efficient tumour formation by single

human melanoma cells. Nature456, 593598

(2008).

22. Balazsi, G., van Oudenaarden, A. & Collins, J. J.

Cellular decision making and biological noise: from

microbes to mammals. Cell144, 910925 (2011).

23. Arias, A. M. & Hayward, P. Filtering transcriptional

noise during development: concepts and mechanisms.

Nature Rev. Genet.7, 3444 (2006).

24. Hughes, T. R. et al. Functional discovery via a

compendium of expression profiles. Cell102,

109126 (2000).

25. Luo, J., Solimini, N. L. & Elledge, S. J. Principles of

cancer therapy: oncogene and non-oncogene

addiction. Cell136, 823837 (2009).

26. Whitesell, L. & Lindquist, S. L. HSP90 and the

chaperoning of cancer. Nature Rev. Cancer5,

761772 (2005).

R E V I E W S




11/12

27. Egeblad, M., Nakasone, E. S. & Werb, Z. Tumors as

organs: complex tissues that interface with the entire

organism. Dev. Cell18, 884901 (2010).

28. Allinen, M. et al. Molecular characterization of the

tumor microenvironment in breast cancer. Cancer Cell

6, 1732 (2004).

29. Tlsty, T. D. & Coussens, L. M. Tumor stroma and

regulation of cancer development.Annu. Rev. Pathol.

1, 119150 (2006).

30. Bissell, M. J., Kenny, P. A. & Radisky, D. C.

Microenvironmental regulators of tissue structure and

function also regulate tumor induction andprogression: the role of extracellular matrix and its

degrading enzymes. Cold Spring Harb. Symp. Quant.

Biol.70, 343356 (2005).

31. Anderson, A. R., Weaver, A. M., Cummings, P. T. &

Quaranta, V. Tumor morphology and phenotypic

evolution driven by selective pressure from the

microenvironment. Cell127, 905915 (2006).

32. Chen, J., Sprouffske, K., Huang, Q. & Maley, C. C.

Solving the puzzle of metastasis: the evolution of cell

migration in neoplasms. PloS ONE6, e17933

(2011).

33. Polyak, K. & Weinberg, R. A. Transitions between

epithelial and mesenchymal states: acquisition of

malignant and stem cell traits. Nature Rev. Cancer9,

265273 (2009).

34. Hendrix, M. J., Seftor, E. A., Hess, A. R. & Seftor, R. E.

Vasculogenic mimicry and tumour-cell plasticity:

lessons from melanoma. Nature Rev. Cancer3,

411421 (2003).

35. Radisky, D. C. & Bissell, M. J. Matrix

metalloproteinase-induced genomic instability. Curr.

Opin. Genet. Dev.16, 4550 (2006).

36. Goel, S. et al. Normalization of the vasculature for

treatment of cancer and other diseases. Physiol. Rev.

91, 10711121 (2011).

37. Farmer, P. et al. A stroma-related gene signature

predicts resistance to neoadjuvant chemotherapy in

breast cancer. Nature Med.15, 6874 (2009).

38. Finak, G. et al. Stromal gene expression predicts

clinical outcome in breast cancer. Nature Med.14,

518527 (2008).

39. Nowell, P. C. The clonal evolution of tumor cell

populations. Science194, 2328 (1976).

40. Merlo, L. M., Pepper, J. W., Reid, B. J. & Maley, C. C.

Cancer as an evolutionary and ecological process.

Nature Rev. Cancer6, 924935 (2006).

41. Jones, P. A. & Baylin, S. B. The epigenomics of cancer.

Cell128, 683692 (2007).

42. Brock, A., Chang, H. & Huang, S. Non-genetic

heterogeneity--a mutation-independent driving force

for the somatic evolution of tumours. Nature Rev.Genetics10, 336342 (2009).

43. Weinberg, R. A. The Biology of Cancer(Garland

Science, 2007).

44. Barrick, J. E. & Lenski, R. E. Genome-wide mutational

diversity in an evolving population of Escherichia coli.

Cold Spring Harb. Symp. Quant. Biol.74, 119129

(2009).

45. Desai, M. M., Fisher, D. S. & Murray, A. W. The

speed of evolution and maintenance of variation in

asexual populations. Curr. Biol.17, 385394

(2007).

46. Marusyk, A. & Polyak, K. Tumor heterogeneity: causes

and consequences. Biochim. Biophys. Acta1805,

105117 (2010).

47. Navin, N. et al. Tumour evolution inferred by single-

cell sequencing. Nature 472, 9094 (2011).

The first report that interrogated genomic-wide

diversity within tumours on a single-cell level.

48. Tsao, J. L. et al. Colorectal adenoma and cancer

divergence. Evidence of multilineage progression.Am.J. Pathol.154, 18151824 (1999).

49. Gonzalez-Garcia, I., Sole, R. V. & Costa, J.

Metapopulation dynamics and spatial heterogeneity in

cancer. Proc. Natl Acad. Sci. USA99, 1308513089

(2002).

50. Snuderl, M. et al. Mosaic amplification of multiple

receptor tyrosine kinase genes in glioblastoma. Cancer

Cell20, 810817 (2011).

51. Stoecklein, N. H. & Klein, C. A. Genetic disparity

between primary tumours, disseminated tumour cells,

and manifest metastasis. Int. J. Cancer. 126,

589598 (2010).

52. Fearon, E. R. & Vogelstein, B. A genetic model for

colorectal tumorigenesis. Cell61, 759767

(1990).

53. Ramaswamy, S., Ross, K. N., Lander, E. S. &

Golub, T. R. A molecular signature of metastasis in

primary solid tumors.Nature Genet.33, 4954 (2003).

54. Weigelt, B. et al. Molecular portraits and 70-gene

prognosis signature are preserved throughout the

metastatic process of breast cancer. Cancer Res.65,

91559158 (2005).

55. Klein, C. A. Parallel progression of primary tumours and

metastases.Nature Rev. Cancer9, 302312 (2009).

56. Mardis, E. R. Cancer genomics identifies

determinants of tumor biology. Genome Biol.11, 211

(2010).

57. Yachida, S. et al. Distant metastasis occurs late during

the genetic evolution of pancreatic cancer. Nature

467, 11141117 (2010).58. Shah, S. P. et al. Mutational evolution in a lobular

breast tumour profiled at single nucleotide resolution.

Nature 461, 809813 (2009).

The first report that used next-generation

sequencing to interrogate clonal diversity and

evolution during breast cancer progression.

59. Ding, L. et al. Genome remodelling in a basal-like

breast cancer metastasis and xenograft. Nature464,

9991005 (2010).

60. Perou, C. M., Parker, J. S., Prat, A., Ellis, M. J. &

Bernard, P. S. Clinical implementation of the intrinsic

subtypes of breast cancer. Lancet Oncol.11,

718719 (2010).

61. Sorlie, T. et al. Gene expression patterns of breast

carcinomas distinguish tumor subclasses with clinical

implications. Proc. Natl Acad. Sci. USA98,

1086910874 (2001).

62. Komaki, K., Sano, N. & Tangoku, A. Problems in

histological grading of malignancy and its clinical

significance in patients with operable breast cancer.

Breast cancer13, 249253 (2006).

63. Bertos, N. R. & Park, M. Breast cancer - one term,

many entities?J. Clin. Invest.121, 37893796

(2011).

64. Allred, D. C. et al. Ductal carcinoma in situ and the

emergence of diversity during breast cancer evolution.

Clin. Cancer Res.14, 370378 (2008).

65. Maley, C. C. et al. Genetic clonal diversity predicts

progression to esophageal adenocarcinoma. Nature

Genet. 38, 468473 (2006).

This paper linked higher extent of genetic clonal

diversity with poor clinical outcome.

66. Merlo, L. M. & Maley, C. C. The role of genetic

diversity in cancer.J. Clin. Invest.120, 401403

(2010).

67. Park, S. Y. et al. Heterogeneity for stem cell-related

markers according to tumor subtype and histologic

stage in breast cancer. Clin. Cancer Res.16, 876887

(2010).

68. Meads, M. B., Gatenby, R. A. & Dalton, W. S.

Environment-mediated drug resistance: a majorcontributor to minimal residual disease. Nature Rev.

Cancer9, 665674 (2009).

69. Ebos, J. M. et al. Accelerated metastasis after

short-term treatment with a potent inhibitor of tumor

angiogenesis.Cancer Cell15, 232239 (2009).

70. Paez-Ribes, M. et al. Antiangiogenic therapy elicits

malignant progression of tumors to increased local

invasion and distant metastasis. Cancer Cell15,

220231 (2009).

71. Deininger, M. Resistance to imatinib: mechanisms and

management.J. Natl Compr. Canc. Netw.3,

757768 (2005).

72. Corless, C. L. & Heinrich, M. C. Molecular

pathobiology of gastrointestinal stromal sarcomas.

Annu. Rev. Pathol.3, 557586 (2008).

73. Nazarian, R. et al. Melanomas acquire resistance to

B-RAF(V600E) inhibition by RTK or N-RAS

upregulation. Nature468, 973977 (2010).

74. Engelman, J. A. et al. MET amplification leads to

gefitinib resistance in lung cancer by activatingERBB3 signaling. Science316, 10391043 (2007).

75. Edwards, S. L. et al. Resistance to therapy caused by

intragenic deletion in BRCA2. Nature451,

11111115 (2008).

76. Sakai, W. et al. Secondary mutations as a mechanism

of cisplatin resistance in BRCA2-mutated cancers.

Nature451, 11161120 (2008).

77. Roche-Lestienne, C., Lai, J. L., Darre, S., Facon, T. &

Preudhomme, C. A mutation conferring resistance to

imatinib at the time of diagnosis of chronic

myelogenous leukemia. N. Engl. J. Med.348,

22652266 (2003).

78. Ding, L. et al. Clonal evolution in relapsed acute

myeloid leukaemia revealed by whole-genome

sequencing. Nature481, 506510 (2012).

79. Mullighan, C. G. et al. Genomic analysis of the clonal

origins of relapsed acute lymphoblastic leukemia.

Science322, 13771380 (2008).

80. Glasspool, R. M., Teodoridis, J. M. & Brown, R.

Epigenetics as a mechanism driving polygenic clinical

drug resistance. Br. J. Cancer94, 10871092

(2006).

81. Janne, P. A., Gray, N. & Settleman, J. Factors

underlying sensitivity of cancers to small-molecule

kinase inhibitors. Nature Rev. Drug Discov.8,

709723 (2009).

82. Sharma, S. V. et al. A chromatin-mediated reversible

drug-tolerant state in cancer cell subpopulations. Cell

141, 6980 (2010).

This study linked therapy resistance with a distinctepigenetic switch that arises through a

spontaneous phenotypic switch. Although this

phenotypic state is reversible, it can persist for

multiple cell generations and thus serve as a

substrate for Darwinian selection.

83. Cohen, A. A. et al. Dynamic proteomics of individual

cancer cells in response to a drug. Science322,

15111516 (2008).

84. Spencer, S. L., Gaudet, S., Albeck, J. G., Burke, J. M. &

Sorger, P. K. Non-genetic origins of cell-to-cell

variability in TRAIL-induced apoptosis. Nature459,

428432 (2009).

This study implicated noise-driven stochastic

non-genetic variability in short-term resistance to

therapy.

85. Heppner, G. H. Cancer cell societies and tumor

progression. Stem Cells11, 199203 (1993).

86. Bonavia, R., Inda, M. M., Cavenee, W. K. &

Furnari, F. B. Heterogeneity maintenance in

glioblastoma: a social network. Cancer Res.71,

40554060 (2011).

87. Park, S. Y., Gnen, M., Kim, H. J., Michor, F. &

Polyak, K. Cellular and genetic diversity in the

progression of in situ human breast carcinomas to an

invasive phenotype.J. Clin. Invest.120, 636644

(2010).

88. Merlo, L. M. et al. A comprehensive survey of clonal

diversity measures in Barretts esophagus as

biomarkers of progression to esophageal

adenocarcinoma. Cancer Prev. Res. (Phila)3,

13881397 (2010).

89. Russnes, H. G., Navin, N., Hicks, J. & Borresen-Dale,

A. L. Insight into the heterogeneity of breast cancer

through next-generation sequencing.J. Clin. Invest.

121, 38103818 (2011).

90. Navin, N. et al. Inferring tumor progression from

genomic heterogeneity. Genome Res.20, 6880

(2010).

91. Xu, X. et al. Single-cell exome sequencing reveals

single-nucleotide mutation characteristics of a kidney

tumor. Cell148, 886895 (2012).92. Hou, Y. et al. Single-cell exome sequencing and

monoclonal evolution of a JAK2-negative

myeloproliferative neoplasm. Cell148, 873885

(2012).

93. Miller, B. E., Machemer, T., Lehotan, M. &

Heppner, G. H. Tumor subpopulation interactions

affecting melphalan sensitivity in palpable mouse

mammary tumors. Cancer Res.51, 43784387

(1991).

94. Calbo, J. et al. A functional role for tumor cell

heterogeneity in a mouse model of small cell lung

cancer. Cancer Cell19, 244256 (2011).

95. Marks, P. A. & Dokmanovic, M. Histone deacetylase

inhibitors: discovery and development as anticancer

agents. Expert Opin. Investig. Drugs 14, 14971511

(2005).

96. Gregory, R. I. & Shiekhattar, R. Chromatin modifiers

and carcinogenesis. Trends Cell. Biol.14, 695702

(2004).

97. Bissell, M. J. & Hines, W. C. Why dont we get morecancer? A proposed role of the microenvironment in

restraining cancer progression. Nature Med.17,

320329 (2011).

98. Attolini, C. S. & Michor, F. Evolutionary theory

of cancer.Ann. NY Acad. Sci.1168, 2351

(2009).

99. DeGregori, J. Evolved tumor suppression: why are we

so good at not getting cancer? Cancer Res.71,

37393744 (2011).

100. Greaves, M. & Maley, C. C. Clonal evolution in cancer.

Nature481, 306313 (2012).

101. Gatenby, R. A., Silva, A. S., Gillies, R. J. & Frieden,

B. R. Adaptive therapy. Cancer Res. 69, 48944903

(2009).

This paper proposes adaptive therapy, which

takes into account clonal competition in order to

prevent therapy-resistant cells from taking over the

population.

R E V I E W S




12/12

102. Waddington, C. H. The Strategy of the Genes: a

Discussion of Some Aspects of Theoretical Biology

(Taylor & Francis, 1957).

103. Huang, S. & Kauffman, S. A. in Encyclopedia of

Complexity and SystemsScience (ed. Meyers, R. A.)

11801213 (Spr inger, 2009).

104. Fisher, A. G. Cellular identity and lineage choice.

Nature Rev. Immunol.2, 977982 (2002).

105. Bozic, I. et al. Accumulation of driver and

passenger mutations during tumor progression.

Proc. Natl Acad. Sci. USA107, 1854518550

(2010).106. Leedham, S. J. et al. Individual crypt genetic

heterogeneity and the origin of metaplastic glandular

epithelium in human Barretts oesophagus. Gut57,

10411948 (2008).

107. Kim, M. Y. et al. Tumor self-seeding by

circulating cancer cells. Cell139, 13151326

(2009).

AcknowledgementsWe thank members of our laboratory and F. Michor and S.

Itzkovitz for their critical reading of our manuscript and for

stimulating discussions. Tumour diversity research in the

authors laboratory is supported by US Army Congressionally

Directed Research W81XWH-07-1-0294 (K.P.) and BC087579

(A.M.), US National Cancer Inst itute PO1 CA80111 (K.P.),

Susan G. Komen Foundation (K.P.), Breast Cancer Research

Foundation (K.P.) and the Cellex Foundation (V.A.).

Competing interests statement

The authors declare no competing financial interests.

FURTHER INFORMATIONKornelia Polyaks homepage:

http://research4.dfci.harvard.edu/polyaklab/

ALL LINKS ARE ACTIVE IN THE ONLINE PDF

R E V I E W S

334 | MAY 2012 | VOLUME 12 / i /
http://research4.dfci.harvard.edu/polyaklab/http://research4.dfci.harvard.edu/polyaklab/http://research4.dfci.harvard.edu/polyaklab/http://research4.dfci.harvard.edu/polyaklab/

2012_intra-tumour heterogeneity- a looking glass for cancer

Documents