Mutational Heterogeneity in Human Cancers: Origin and ...depts.washington.edu/loeblabs/pdf/2009_Salk.pdf · University of Washington, Seattle, Washington 98195; email: [email protected]

AREV403-PM05-03 ARI 11 December 2009 9:31

Mutational Heterogeneityin Human Cancers:Origin and ConsequencesJesse J. Salk, Edward J. Fox, and Lawrence A. LoebJoseph Gottstein Memorial Cancer Research Laboratory, Department of Pathology,University of Washington, Seattle, Washington 98195; email: [email protected]

Annu. Rev. Pathol. Mech. Dis. 2010. 5:51–75

First published online as a Review in Advance onSeptember 10, 2009

The Annual Review of Pathology: Mechanisms ofDisease is online at pathmechdis.annualreviews.org

This article’s doi:10.1146/annurev-pathol-121808-102113

Copyright c© 2010 by Annual Reviews.All rights reserved

1553-4006/10/0228-0051$20.00

Key Words

evolution, cancer genome, DNA sequencing

AbstractCancer recapitulates Darwinian evolution. Mutations acquired duringlife that provide cells with a growth or survival advantage will prefer-entially multiply to form a tumor. As a result of The Cancer GenomeAtlas Project, we have gathered detailed information on the nucleotidesequence changes in a number of human cancers. The sources of mu-tations in cancer are diverse, and the complexity of those found to beclonally present in tumors has increasingly made it difficult to identifykey rate-limiting genes for tumor growth that could serve as potentialtargets for directed therapies. The impact of DNA sequencing on futurecancer research and personalized therapy is likely to be profound andmerits critical evaluation.

51

Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


INTRODUCTION

The early historical reliance on surgical exci-sion for treatment of cancers suggests that tu-morigenesis was believed to be a fundamen-tally irreversible process. The multiplicity ofchromosomal aberrations associated with ab-normal cellular morphology in many humantumors was noted by pathologists more than

a

b

-9-1-14

-3-21-19

-11-17-4

-1

-14

1

6

13

19 20 21 22 X Y

14 15 16 17 18

7 8 9 10 11

2 3 4

Figure 1Heterogeneity in cancer. (a) Chromosomal heterogeneity. Spectral karyotype from an acute myelogenousleukemia cell demonstrating aneuploidy and multiple chromosomal rearrangements. Reproduced courtesy ofDr. Karen Swisshelm, Department of Pathology, University of Colorado, Denver. (b) Morphologicheterogeneity. Hematoxylin and eosin section from a large-cell, undifferentiated lung cancer demonstratinga highly pleiomorphic cellular population. Reproduced courtesy of Dr. Ray Monnat, Department ofPathology, University of Washington, Seattle.

a century ago and has continued to serve asa means of identifying malignant cells andstratifying the aggressiveness of certain cancers(Figure 1). As early as 1902, Boveri (1) sug-gested that the intercellular cooperation re-quired between cell types during embryo-genesis was disrupted in tumors as a resultof chromosomal aberrations. Currently, weno longer consider chromosomes, or even

52 Salk · Fox · Loeb

Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


individual genes, as the irreducible genetic unitsof cancer. Instead, we can identify single-basechanges buried among billions of faithfullyreplicated nucleotides in the human genome.Modern biochemical approaches allow us notonly to identify mutated cancer genes, but alsoto infer how specific mutations affect gene func-tion. As a result, we can catalog unique nu-cleotide changes that contribute to malignantphenotypes or that increase the susceptibilityof individuals to develop specific tumors.

The primacy of DNA as the critical macro-molecule involved in the etiology of cancer isstrongly supported by the inherited human dis-eases that are associated with, and increase theincidence of, specific cancers (2). In addition,multiple genetic changes in cancer cells havefrequently been documented by microscopicexamination of chromosomes or by hybridiza-tion with specific probes. These changes in-clude deletions, insertions, amplifications, andtranslocations, and they frequently involve mil-lions of nucleotides. Chromosomal studies onadenocarcinoma of the breast (3) and ovary (4),as well as in leiomyosarcoma (5), have docu-mented tumors harboring more than 20 dif-ferent chromosomal alterations. Measurementsof loss of heterozygosity in tumors using poly-merase chain reaction (PCR)-amplified genefragments have revealed an even greater num-ber of changes (6). Klein et al. (7) used bothof these techniques and demonstrated the pres-ence of multiple alterations within single tumorcells (7). These chromosomal changes have of-ten been considered to result from chromoso-mal instability and in some tumors may occursequentially (8).

The technology for dissecting chromo-somes into their finest nucleotide elements hasexponentially improved in recent years, bothin terms of throughput and cost-effectiveness.It is now possible to analyze the processes thatgenerate mutations in normal and malignantcells and begin the ambitious task of catalogingcancer-associated nucleotide changes by DNAsequencing. For some, the surprise has beenthe unexpectedly large number and diversity ofmutations present in human tumors. In light of

this emerging mutational complexity, it seemstimely to review mechanisms that guaranteethe high fidelity of DNA replication in normalhuman cells, to consider how mechanismsfor preventing mutations may be altered intumors, and to interpret the recently reportedresults on mutations identified in differenthuman cancers.

THE ACCURACY OF DNAREPLICATION IN NORMALHUMAN CELLS

DNA is a living molecule; it continuallybreathes and is exposed to modifications. Yet,in normal cells it is faithfully copied duringeach division cycle. Each human cell containsmore than 6 billion nucleotides that are repli-cated with exceptionally high accuracy. Approx-imately one point mutation is introduced intoDNA during each division cycle. Most muta-tion rate measurements have been carried outat the hgprt locus because it is present as a singlecopy on the X chromosome. Spontaneous mu-tations in this gene render a cell dominantly re-sistant to the toxic effects of the nucleoside ana-log 6-thioguanine and form countable coloniesunder appropriate culture conditions. A tabula-tion of data derived from hgprt studies indicatesthat the overall mutation frequency in mam-malian cells varies from 10−5 to 10−7 mutationsper gene, or approximately 10−9 to 10−10 substi-tutions per DNA nucleotide (9–12). It should benoted that cancers likely arise in stem cells, anddetailed studies of mouse embryonic stem cellsindicate that mutation rates in these pluripotentprogenitors are as much as 100-fold lower thanthose observed in cultured fibroblasts derivedfrom adult tissues (13). Based on the conserva-tive assumption that the accuracy of DNA repli-cation in stem cells is similar to that in othercells, it can be estimated that each stem cellwould amass, on average, only one to two mu-tant genes during 100 cell divisions in a normalhuman life span (14).

This remarkably high accuracy resultsfrom sequential processes, each contributinga 100- to 1000-fold increase in the fidelity

www.annualreviews.org • Mutational Heterogeneity in Human Cancers 53

Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


of DNA replication. First, based on simplethermodynamics, the difference in free energyof hydrogen bonding between complementaryand noncomplementary base pairings duringDNA synthesis can provide an accuracy of baseselection down to approximately one error per102 nucleotides incorporated (15–16). Second,DNA polymerases are believed to undergo al-losteric transformations with each nucleotide-addition step that tightens the bonding ofcomplementary nucleotides at the active siteon the polymerase (17). Third, replicative DNApolymerases have an associated “proofreading”3′ → 5′ exonuclease activity that preferen-tially excises noncomplementary nucleotidesprior to incorporation of the next nucleotide(18). Fourth, remaining noncomplementarynucleotides are removed by mismatch repairafter the replication fork has passed (19).Together, these processes have the potential tosynthesize DNA in vitro with an accuracy thatapproximates the fidelity of DNA replicationobserved in vivo. Notably, however, experi-mental values are based on reactions carriedout under simplified, optimal conditions thatmay not exist in cells; other components of thereplicative machinery are likely to play a role inreplication fidelity. Also notable is that the er-ror rates of DNA polymerases are proportionalto the concentrations of noncomplementaryto complementary nucleotides in the reaction(20). This finding suggests that size of cellularnucleotide pools has a significant impact on theaccuracy of DNA replication. Alterations in theaccuracy of DNA polymerases by mutation,damage, or imbalances in nucleotide poolscould therefore have profound effects on theoverall fidelity of DNA replication (17).

DNA DAMAGE BY ENDOGENOUSAND ENVIRONMENTAL AGENTS

DNA is subject to attack by both endoge-nous and exogenous reactive molecules. A ma-jor source of DNA modification in human cellsis reactive oxygen species (ROS) arising as a by-product of energy metabolism in mitochondria.(22). On the basis of measurements of 8-oxo-dG

and other oxidative modifications, it has beenestimated that more than 10,000 nucleotideresidues in DNA are altered by ROS per cell perday (23). Other modifications include methyla-tion, alkylation, inter-/intrastrand cross-links,and apurinic sites; there may be as many as50,000 alterations per cell per day that resultsolely from normal cellular metabolism (24).Many human cancers arise in the setting ofchronic inflammation (25–28), where extracel-lularly derived ROS are likely to contribute tothe burden of DNA damage in affected tissues(29). Mutagens are ubiquitous in our environ-ment, and it is important to recognize theircontribution to spontaneous human cancer. Ex-posure to high concentrations of mutagens hasfrequently been associated with an increased in-cidence of cancer (30–32), and the recognitionof tobacco smoke as a human carcinogen (33–34) may have led to the most significant andsuccessful effort at reducing cancer incidencein human history (35).

THE REPAIR OF DNA DAMAGEIN HUMAN CELLS

Working against this onslaught of DNA dam-age is an armamentarium of repair systems withoverlapping specificities (Figure 2). These sys-tems continuously monitor the genome andcorrect many forms of DNA damage. So far,more than 100 repair genes have been iden-tified (36). Pathways of DNA repair includebase-excision repair, nucleotide-excision repair,transcription-coupled repair, mismatch repair,double-strand-break repair, and even direct re-versal of adduct-mediated lesions (37). DNAdamage by environmental agents is predom-inantly a stochastic process. Recognition ofdamage is generally dependent on the natureof the lesion and is less a function of sequencecontext. Small adducts on bases are excised byboth short-patch and long-patch pathways, andresynthesis of the excised segment is carried outby DNA polymerase β and presumably DNApolymerases δ and ε, respectively (17). Bulkyadducts such as thymine dimers, resulting fromultraviolet irradiation, or benzo[a]pyrene, re-sulting from tobacco products, are subject to


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


nucleotide-excision repair. In the presence ofDNA damage, many sensing systems signal tocell-cycle control genes, such as TP53, to arrestthe cycle and allow additional time for DNArepair (38). The high efficiency of these mech-anisms for DNA repair guarantees that only afew of the tens of thousands of DNA lesionsgenerated persist prior to DNA replication andhave the potential to cause mutations.

INHERITED CANCERSFREQUENTLY INVOLVEMUTATIONS INDNA-REPAIR GENES

A number of rare inherited diseases caused bygermline mutations in genes involved in DNArepair are associated with elevated risks ofspecific cancers. Investigation of many of thesediseases has been instrumental in decipheringthe different cellular mechanisms for DNArepair. The seminal findings on the defectsof ultraviolet-induced DNA-damage repairin patients with xeroderma pigmentosumhighlighted the association of DNA repair withsuppression of carcinogenesis and provided apowerful tool for identifying genes involved innucleotide-excision repair (2). Subsequently,inherited defects in members of several otherDNA-repair pathways have been shown tounderlie a variety of cancer syndromes, includ-ing mismatch repair [hereditary nonpolyposiscolorectal cancer (39)], base-excision repair[MYH-associated polyposis (40)], homologousrecombination [early-onset breast cancer (41)],nonhomologous end-joining [LIG4 syndrome(42)] and translesion synthesis [xerodermapigmentosum variant (43)].

Hereditary mutations in other DNA main-tenance enzymes are also associated with can-cer. Defects in genes encoding members of theRecQ helicases are found in Bloom and Wernersyndromes—rare inherited diseases character-ized by developmental abnormalities and ahigh incidence of specific cancers (44–45).Mutations in TP53 are found in Li–Fraumenisyndrome (46), a highly cancer-prone condi-tion most frequently associated with sarcomas

Ma

n-m

ad

e

Ind

ust

ria

l

Na

tura

l

Fo

od

De

pu

rin

ati

on

Fre

e r

ad

ica

ls

DN

A p

oly

me

rase

err

ors

Na

tura

l

BER

NER

TCR

MMR

DSBR

Mu

ltip

le r

epai

r p

roce

sses

MisincorporatednucleotidesUnrepaired

DNA damageDNA breaks

Endogenoussources

CarcinogensRadiation

Mutations

Figure 2Mutational homeostasis. In each human cell, DNA is damaged thousands oftimes per day by both exogenous and endogenous sources. Most alterations arecorrected by cellular mechanisms, including base-excision repair (BER),nucleotide-excision repair (NER), transcription-coupled repair (TCR),mismatch repair (MMR), and double-strand-break repair (DSBR). Lesions thatescape repair have the potential to cause mutations during DNA replication.

and breast adenocarcinomas. Additionally,polymorphisms in a wide variety of DNA-repairgenes, including OGG1 and XRCC1, are in-creasingly being considered as risk factors forcancer (47). Why defects in particular DNA-repair genes result in specific types of cancerremains largely unresolved.

PREVENTION OF MUTATIONSBY GROWTH LIMITATION,CLONAL HIERARCHY, ANDPROGRAMMED CELL DEATH

Despite these extensive DNA-repair mech-anisms, every time a cell divides there re-mains an opportunity for fixation of new mu-tations through miscopying across unrepaireddamage, missegregation of replicated chromo-somes, and/or failure to recognize improperly


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


repaired sequences. Thus, a key mechanism forpreventing the accumulation of DNA muta-tions is limiting the number of cell divisionsthat occur. The importance of proliferation inoncogenesis has been demonstrated by exper-iments showing that liver regeneration is as-sociated with an increased incidence of cancer(48). The initiation of skin cancer in mice bymutagens was markedly accelerated by the sub-sequent application of phorbol esters that pro-mote cell proliferation (49). Many of the molec-ular mechanisms that control cellular growthcontrol were first identified in the context oftheir disruption in cancer. The identificationof replication-enhancing avian viral oncogenes(50), and the discovery that these represent con-stitutively active versions of endogenous genesirreversibly activated through mutation in somecancers (51), led to the characterization of oneof the earliest known growth-control pathwaysin human cells. Extensive work during the pastthree decades has revealed the cellular networkof defenses preventing unregulated prolifera-tion to be staggeringly complex, with many re-dundant protections (52).

The necessity of replacing worn out or dam-aged tissues must be carefully balanced againstthe risk of proliferation-induced mutations. Toallow cellular repopulation with minimal riskof mutation, tissues in the body are frequentlyorganized hierarchically, whereby the ability tocontinuously proliferate is relegated to a spe-cialized subset of cells (53, 53a). Stem cells arebelieved to have an inherently lower rate ofmutation than the majority of their daughtercells, which have only limited replication po-tential (13). Among the most studied examplesof this hierarchical organization is colonic ep-ithelium. Here, a small number of long-livedstem cells reside at the base of each crypt andproduce progeny that migrate luminally to pop-ulate the upper levels of the crypt—first as tran-siently amplifying cells, then as terminally dif-ferentiated colonocytes destined to slough offafter several days (54). Because the majority ofmutations that arise during division occur inshort-lived daughter cells, most mutant cells arerapidly purged from the population. It has been

hypothesized that the same so-called immortalstrand of DNA is maintained in the parentalstem cell and that the (potentially imperfect)newly replicated strand is always transferredto the daughter (55), although at least one re-cent study (21) suggests that this is not thecase. Whether tumors derive from abnormallyreplicating stem cells, or from dedifferentiatedprogeny, remains an open question (56).

CANCER AS A SOMATICEVOLUTIONARY PROCESS

November, 2009 marks the 150th anniversaryof the publication of Darwin’s seminal work, Onthe Origin of Species. Therein, Darwin postulatedthat heritable phenotypic variation underliesnatural selection and is responsible for adapta-tion as well as for the emergence of new species.Although it was initially used to describe howorganisms evolve across generational time,the idea that evolutionary forces also driveintraorganismal neoplastic development hasfrequently been noted (57–59). In this model,individual cancer cells, rather than completemetazoans, are considered the reproductiveunits within a population (Figure 3). New mu-tations are acquired somatically, and genetic al-terations that bestow a growth advantage uponthe cancer cell enable them to clonally expand.Additional mutations that arise in the expand-ing population generate further selectable phe-notypes, such as the ability to invade adjacenttissues, recruit a blood and lymphatic supply,overcome nutritional deficiencies, and resistimmune attack. After bypassing all antineo-plastic defense mechanisms, tumor growth maycontinue indefinitely until the death of the host.

THE NUMBER OF MUTATIONSTO CANCER

Given that carcinogenesis may be viewedas an evolutionary process that sequentiallyincreases neoplastic cell fitness through a seriesof (epi)genome-modifying events, an impor-tant question arises: How many mutations areneeded to produce a tumor? The increased


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


incidence of most human cancers as the fifth orsixth power of age (60) has been taken to indi-cate that there are five or six events (presumablymutations) that drive the carcinogenic process,each event increasing the probability of thenext. Exceptions include (a) certain pediatrictumors, such as retinoblastoma, in which signif-icantly fewer mutations appear to be necessary(61), and (b) some late-onset adult tumors,such as those of the prostate, that may requireas many as 10 to 12 events (62). Weinberg andcolleagues (63) have demonstrated that at leastthree or four altered genes are required forthe expression of the malignant phenotype incultured cells. Passage in tissue culture or asimplanted xenografts may, in itself, select foradditional mutations as highlighted by the re-cent work of Mahale et al. (64). Thus, if cancerrequires as many as 12 different rate-limitingmutations to arise, if the normal per-divisionmutation rate of human stem cells is as low ascalculated, and if the number of long-lived stemcell divisions is limited, how can a cancer pos-sibly occur within the human lifetime? It is hy-pothesized that early in the neoplastic process atleast one, and likely several, of the mechanismsfor preventing mutations must be reduced.

ANTIMUTATIONAL DEFENSES:PRIMARY VERSUS SECONDARYMECHANISMS

From a simplified perspective, one may divideantimutational processes that suppress theemergence of new genetic diversity into twoclasses. The first class consists of primarymechanisms, which act at the level of DNAto prevent genetic and epigenetic mutationsfrom occurring or persisting until cell division,such as proofreading by DNA polymerases,DNA-repair processes, ability to quench ROS,and other means of limiting the per–cell divi-sion mutation rate. The second class includessecondary mechanisms, which in contrastprevent the accumulation of mutations in thepopulation at large by limiting the total numberof divisions in long-lived cells. These mecha-nisms include means of controlling cell growth,

DNAdamage

Tumormass Metastasis

Years0 20

Self-sufficientgrowth

InvasionAngiogenesis

Figure 3Cancer recapitulates evolution. Within a developing tumor, mutationsaccumulate over time as a result of unrepaired DNA damage. Most of thesemutations are either neutral or detrimental; only a small number bestowgrowth and survival benefits upon a cell. Cells with these beneficial variantspreferentially multiply and additional mutations occur that may undergofurther selection and expansion. Advantageous phenotypes for tumor growthinclude, among others, the ability to divide independently of extracellularsignals ( yellow), the ability to recruit a blood supply ( green), and the ability toinvade adjacent and distant tissues (blue).

confinement of reproductive abilities to stemcells in a well-protected niche, and culling of ir-reversibly damaged cells through programmedcell death. Secondary mechanisms do not affectthe per-division rate at which individual cellsaccumulate mutations; instead, they limit theoverall number of different mutations in thepopulation as a whole.

For evolution to occur in a somatic setting,(a) there must be heritable genetic or epige-netic variation within a population of cells, and(b) these variants must be able to undergo selec-tion (i.e., through more efficient division andsurvival) in response to advantageous environ-ments. If either of these features is rate-limitingto the process, and if the factor responsible forthe limitation is genetically encoded within theevolving cells, disruption of the responsiblegenes or regulatory sites through mutationaccelerates the adaptive process. The questionof which parameter is most limiting in differentstages of tumorigenesis is complex and, rightly,has been the subject of extensive debate (65–66). Given that both therapeutic interventionsand preventative measures might be better


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


Mutator phenotype:an increased (per–celldivision) mutation rateresulting fromheritable cellulardefects, generally inDNA maintenancemachinery

Clonal expansion:multiplication of asingle cell to produce apopulation ofgenetically relatedprogeny

directed if this was known, the question is notmerely of academic interest.

Within the view that somatic evolutiondrives carcinogenesis, at least two schools ofthought have emerged regarding the relativeimportance of overcoming primary and sec-ondary antimutational defenses. One hypoth-esis asserts that the mutation frequency exist-ing within nonmalignant cells provides suffi-cient genetic diversity to fuel tumorigenesis (65,67–68). Some of these variants in a normal pop-ulation can confer a phenotype of increasedproliferative abilities that enables subsequentclonal outgrowth and consequent generation ofmore mutations, upon which further selectionmay act. Not only does escape from growthlimitations facilitate a cell acting upon pre-existing genetic diversity through selection, italso enhances the production of additional vari-ants by increasing the number of fallible DNA-replication cycles. This school of thought con-ceptually favors the importance of defects insecondary mechanisms but also relies upon thefact that primary mechanisms are imperfect innonneoplastic cells. Eventually, after enoughsuccessive rounds of mutation, selection, andexpansion, the threshold number of events re-quired to drive carcinogenesis is reached.

A second school of thought emphasizesthe importance of defects in the primary classof antimutational mechanisms for fuelingneoplastic evolution (69–70). Deficits in thesecellular components render DNA replicationmore error prone and increase the numberof potentially advantageous variants producedper generation. It is argued that, even withadditional mutations generated from increasedcell division, the number of variants in thepopulation will still be a limiting parameterfor evolutionary adaptation. Proponents ofthis view suggest that mutation-prone variantswould overcome this limitation and thereforebe likely to emerge during clonal selection. Agenetically encoded mutator phenotype neednot, in itself, be a driver that directly increasesreproductive fitness. Instead, by virtue ofincreasing the probability of advantageous newmutations, this phenotype would be passively

carried along on the resulting clonal expansionas a passenger. Because of this unique position,generation of a mutator phenotype is unlikelyto be the very first “hit” on the path to cancer.As with inherited deficiencies of DNA repair,it is only when the phenotype is expressed ina plurality of cells that it can meaningfullyincrease the total number of genetic variantsin the population. Thus, defects in secondaryantimutational defenses leading to increasedproliferation remain a critical component ofthis model of tumor development.

Investigators on both sides of the debatehave made arguments for (69) and against(71–72) the necessity of a primary mutatorphenotype in carcinogenesis on the basis of cal-culations that assume the number of mutationsneeded for cancer, the mutation rate in normalcells, and the estimated number of divisionsthat occur between conception and a late-stagetumor. Given that these values are, themselves,not easily quantified, the disparate results arenot surprising. An alternate approach taken byour group is to consider, instead, the relativeefficiency of mutator and nonmutator pathwaysto cancer (73). Modeling of fitness landscapessuggests that in spite of the cost of an extrastep to produce the phenotype, a primarymutator pathway is generally a more efficienttrajectory to cancer, as long as the total numberof mutations required exceeds three to five(73a). Moreover, although mutator lineagesare more likely to suffer deleterious mutationsthat terminate their lineage through negativeclonal selection (74), this negative effect doesnot predominate until mutation rates becomevery high (73a). An additional considerationthat has been frequently overlooked is thatnewly arising mutations, including those witha fitness advantage, have a high probabilityof becoming extinct from random drift (58).Depending on the population size and precisefitness advantage, a given mutation may have toarise on multiple independent occasions beforeit can expand to a clinically meaningful size.Hence, calculating the mutation rate requiredfor a defined number of genetic events to occuronce per tumor leads to underestimation of the


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


rate required for each mutation to arise andexpand.

Making a sharp distinction between primaryand secondary mechanisms of mutation sup-pression is conceptually interesting but ulti-mately artificial, given the mechanisms’ inti-mate link within the cell. It is likely that bothmechanisms are operative; the relative con-tribution of each may depend on the tumor.Cell-cycle progression, DNA repair, and pro-grammed cell death are coordinately regulated.TP53, for example, the most commonly mu-tated gene identified to date in human cancer,encodes a multifunctional protein that acts as anetwork hub to integrate information about thegenome state from more than a dozen sources(75). Many types of DNA damage can trig-ger activation of p53, which may lead to cell-cycle arrest, upregulation of DNA-repair pro-cesses or activation of programmed cell deathif damage is severe. TP53 therefore exempli-fies a gene that is directly involved in both pri-mary and secondary antimutational pathways.It has been argued that for such dual-functiontumor-suppressor genes the pressure for loss inneoplastic cells stems from the immediate pro-liferative advantage gained, rather than from anincreased mutation frequency (65). Althoughthis is likely to be the case given that a primarymutator phenotype is not itself selectably ad-vantageous, disruption of such genes neverthe-less increases the genetic diversity of a develop-ing cancer through both primary and secondarymutational pathways and facilitates continuedevolution.

DETERMINISTIC VERSUSPLASTIC TUMORIGENESIS

DNA damage by chemical agents and physicalprocesses is predominantly stochastic. For themost part, damage occurs randomly through-out the genome, and mechanisms governing thecorrection of affected nucleotides are specifiedprimarily by the nature of the chemical alter-ations and not by the surrounding nucleotidesequence. Exceptions to this random hitmodel are nucleotide sequences that can form

alternative DNA structures (76) that are resis-tant to DNA repair and highly repeated se-quences of the genome (77)—both of whichconstitute mutational hot spots. The specificrole of histones and other DNA-associated pro-teins in preventing DNA damage has not beenfully delineated. Similarly, transcriptional status(78) as well as local replication timing (79) havegeneral, yet incompletely predictive, influenceson local mutation rate. Only a small fraction ofall mutations that occur confer a selectable fit-ness advantage with the ability to initiate or pro-mote tumor growth. Given that different indi-viduals acquire a unique, yet random, selectionof mutations prior to and during tumorigenesis,the order and specific nature of advantageousvariants are determined stochastically. Just howsimilar are two tumors with respect to specificgenes mutated and the order in which the se-lected mutations arise? Do tumors evolve alonga pathway characterized by sequential rounds ofmutation and selection of a defined set of tar-gets, or is the process more variable, selectinginstead for phenotypes that may be encoded bymany different loci (Figure 4)?

There is considerable evidence for se-quential mutations during tumor progression.In melanomas (80), colon cancer (8), andesophageal (25) cancer, a series of genetic alter-ations frequently characterize different phasesof neoplastic progression. These changeshave been most extensively documentedin adenocarcinomas of the large intestine,where the order of DNA alterations has beencorrelated with tumor grade and stage (8).From a clinical perspective, a chronologicallyordered series of mutations driving malignancyis particularly attractive, as it implies thatthe evolutionary process must bottleneckthrough a defined set of genes that couldbe therapeutically targeted. Unfortunately,even in the most-studied model, early investiga-tions indicated that fewer than 10% of advancedcolon cancers simultaneously bear mutations inthe three most frequently mutated genes (81).

It has long been clear from traditional ge-netic and molecular methods that despite somecommonalities, the profile of clonal somatic


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


Genes

Tumor 1Tumor 2Tumor 3

Tumor 1

Tumor 2

Tumor 3

No

rma

l

Ca

nce

r

No

rma

l

Ca

nce

r

Phenotypes

a

b

A B C D E F

Figure 4Pathways to cancer. (a) The deterministic pathway. In this model, different tumors of the same cancer typeoccur reproducibly through sequential mutation of each gene within a defined series. Although mutationoccurs randomly, the order of selection is fixed. (b) The plastic pathway. In this model, different tumorsevolve along highly variable pathways, selecting for specific cancer phenotypes that may be achieved throughmutation or epimutation of many possible sites in the genome. Although some mutated loci may be sharedby different tumors, most are not, and the order of selection is predominantly stochastic.

mutations occurring in different tumors ishighly variable overall. The resolution of theseearly findings was inherently limited to a mod-erate number of genes by the biochemical toolsavailable. The complete human genome, how-ever, comprises more than 6 Gb of sequenceinformation and, until very recently, methodsfor high-throughput analysis were inadequatefor the task of whole-genome exploration.Many long-standing questions remain to beanswered: How frequently do different tumorsovercome hardwired barriers to neoplasia inthe same way, in terms of altering both specificgenes and specific pathways? Does this process

vary from one cancer type to another? What,if any, mutational changes differ betweenmetastatic lesions and their primary tumor?What is the relative importance of differ-ent types of mutations, such as single-basechanges versus deletions, rearrangements, andepigenetic phenomena? Can the mutationalspectrum of a tumor inform us of likely environ-mental contributors or specific dysfunctionalantimutational pathways? The emergence ofhigh-throughput capillary sequencing roboticsand more recent next-generation sequencingmethods provide an exciting opportunity todelve more deeply into these questions.


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


THE HUMAN CANCERGENOME ATLAS

Within the past two years, numerous DNA se-quences from human cancers have been pub-lished as part of The Cancer Genome Atlas.Included among these is the first completegenome of a human cancer and its paired nor-mal (82). With the passing of this milestone,it is important to consider the likely implica-tions of this data and how they might frameboth basic and clinical research in the near fu-ture. Prevailing models of tumorigenesis stressthat tumor progression is the result of sequen-tial mutations in a few key cancer genes, eachmutation driving a new round of clonal pro-liferation. The effort to systematically tabulatemutations found in different human cancers en-compasses the expectation that a cancer’s mostsignificant mutated genes will be potent targetsfor chemotherapy. This supposition has beenreinforced by the success of targeted treatmentsin some hematological cancers (83) and by thehope that identification of analogous key muta-tional events in solid tumors might allow spe-cific therapeutic targets to be similarly iden-tified and exploited. However, an increasinglycomplex picture has emerged from nearly 20studies detailing the genome of many solid tu-mors: The findings suggest that the extent ofprevalent, new targets may be far more limitedthan anticipated.

INITIAL STUDIES ONNUCLEOTIDE VARIATIONWITHIN HUMAN CANCERS

The first large-scale efforts to systematicallyscreen individual tumors for somatic mutationsidentified remarkably few previously unknowngenes that were mutated in a significant propor-tion of specific cancers (84–86). The relativelylimited sequence coverage of these initial stud-ies prompted more comprehensive screens oflarger portions of the genome. The first com-plete sequencing of all predicted coding ex-ons, conducted in breast and colon cancer, con-cluded that these cancers respectively contain

The Cancer GenomeAtlas: the completecatalog of genetic andepigenetic alterationsfound in cancers of alltypes

Clonal mutation:mutation present inthe majority of cells ina tumor; detectable byconventionalsequencing

a median of 84 and 76 clonal mutations thatare likely to alter protein function (87–88). Al-though nearly one-tenth of the 18,197 genes se-quenced were detectably mutated in at least onespecimen, each tumor displayed a unique anddiverse profile of mutated genes. Other thanthose previously known, such as TP53, APC,and KRAS, no new prevalently mutated geneswere identified. The authors of these studiesproposed that the cancer genome may be con-sidered as a landscape composed of a handfulof commonly mutated gene “mountains” and alarger number of less frequently mutated gene“hills”—a view consistent with multiple path-ways to cancer. The authors additionally af-firmed that tumor-to-tumor heterogeneity ofclonally present mutations may underlie thewide variation in tumor behavior and respon-siveness to therapy.

The initial studies of Sjoblom et al. (87) andWood et al. (88) served to highlight severalimportant technical challenges faced by TheCancer Genome Atlas Project. First, with largeamounts of tumor sequence data comes signif-icant experimental noise that complicates de-tection of true clonal mutations. Such noisederives from PCR-introduced mutations, au-tomated base-calling errors, mutations arisingin the germline rather than somatically, andpreviously unknown single nucleotide poly-morphisms (SNPs). The most rigorous cancer-genotyping approach would entail sequencingof a matched nonneoplastic sample for everytumor to rule out germline variation as well asautomatic resequencing of every tumor-specificmutation identified. Unfortunately, even thehighest-throughput capillary sequencing sys-tems are cost- and time limited when used onthe scale of multiple human exomes. Thus,in these initial studies, as well as in severalthat followed, compromises to preferred pro-tocols have been made; for example, noncod-ing changes and known SNPs were eliminatedprior to confirming the small fraction of muta-tions remaining against a tumor’s correspond-ing control sample.

A second impediment to high-throughputcapillary sequencing is the substantial amount


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


Driver mutation:mutation that providesa selectable fitnessadvantage to the celland facilitates itsclonal expansion in thepopulation

Passenger mutation:mutation that has noeffect on a cell’s fitnessand that clonallyexpands in thepopulation as a resultof a different drivermutation

Exome: encompassesall coding sequenceswithin the genome

of DNA required for the hundreds of thousandsof PCRs. One way to overcome this limitationhas been to expand tumor cells in culture oras mouse xenografts. Such ex vivo passagingposes the possibility of artifactually introducingnew mutations as a result of artificial growthconditions (64). Direct biochemical methodsof whole-genome amplification have also beenused to extend DNA samples. It has been sug-gested that the differing mutational spectrumsreported within the discovery and validationphases of experiments of Sjoblom et al. (87) arelikely reflective of such differences in ex vivotreatments of samples (A.F. Rubin & P. Green,manuscript submitted).

A third, and arguably most significant,challenge to cancer genome sequencing liesin the complex problem of determining whichmutations in a tumor are causative and whichare merely present by chance. Mutagenesis islargely a random process; only a small subset ofmutations confer a proliferative advantage totheir host cell. Differentiating so-called drivermutations from neutral passenger mutations(or hitchhiker mutations) that happen toco-occur and be swept along with the sameexpanding genome poses a formidable chal-lenge to deciphering megabases of sequencingdata. The practical approach taken by theseand later studies has been to sequence multipletumors of the same variety and look for genesthat are commonly mutated. The appropriatestatistical methods to be used for determiningwhich genes are found clonally mutatedmore frequently than would be expected bychance alone has been heavily debated (90–93).More importantly, implicit in this approach isthe de facto assumption that a limited subsetof genes frequently drive tumorigenesis. Thealternate hypothesis—that a large number ofloci may combinatorially serve as weak driversand that any one may arise only infrequently—cannot easily be addressed by such methods,given that rare drivers are filtered out asprobable passengers. The number of tumorsneeding to be sequenced to resolve minoritydrivers increases substantially as the prevalenceof the driver declines. Researchers have also

attempted to identify likely driver mutationsby bioinformatically predicting the probablefunctional impact of specific mutations (98). Al-though such approaches are useful in a limitednumber of instances, the current technology foraccurately modeling the resulting changes toprotein activity remains limited. Despite thesecomplexities, the demonstrated ability of exomesequencing to reidentify the majority of previ-ously characterized genes known to play a rolein colorectal cancer was a noteworthy technicalachievement that set the stage for the cancergenome–sequencing studies that followed.

MULTIPLE MUTATIONS INDIVERSE TUMORS

Whereas the initial studies of Sjoblom et al.(87) and Woods et al. (88) focused on theexhaustive comparison of mutations in twotumor types, a second milestone projectanalyzed an extensive gene family in a widevariety of different cancers. Greenman et al.(94) sequenced 518 protein kinase genes in210 tumors of diverse origin, including breast,colorectal, lung, brain, and blood tumors.They observed 1007 likely driver mutations, ofwhich 921 were single-base substitutions. As inprevious studies (95–97), there was substantialvariation in the number of genes mutated pertumor regardless of type; again, few commonlymutated genes were found in any of thecancer types examined. Although these studiesgenerated an extensive catalog of somatic pointmutations, only a small number of prevalentlymutated genes were identified. The datareinforced the notion that mutational patternsof solid tumors evolve stochastically and arehighly diverse, in contrast to the relativelypredictable stepwise patterns of cytogeneticabnormalities in some hematological cancers.

Subsequent studies (Table 1) have increas-ingly relied on associating sequence data withother complementary genomic information.This trend has been paralleled by a shift to amore integrated interpretation of the signifi-cance of individual mutations: from one of spe-cific genes into one of pathways and processes.


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


Table 1 Cancer genome–sequencing studies

Study type

Number ofgenes

screened

Totalnumber ofmutations

Number ofgenes

mutated

Averagenumber ofmutationsper tumor

Estimatednumber of

drivermutations Reference(s)

ExomicBreast (n = 11) 18,191 1243 1137 84 140 87, 88Colorectal (n = 11) 18,191 942 848 76 140 87, 88Diverse (n = 210) 518 798 581 – 119 94Pancreatic (n = 24) 20,661 1163 1007 48 160 98Glioblastoma (n = 21) 20,661 748 685 47 155 102Glioblastoma (n = 91) 601 453 223 – 8 103Lung (n = 188) 623 1013 348 – 26 108GenomicAcute myeloid leukemia (n = 1) – 500–1000 10 Not applicable 10 82

One follow-up study focused on mutationsarising during the progression of adenoma tocarcinoma to metastasis (97a). No metastasis-specific mutations were detected in the vastmajority of specimens, and as expected, thenumber of mutations was markedly increasedin the carcinoma compared with its matchedprecursor adenoma. Building on the observa-tion that individual tumors express unique im-mune profiles (99), Segal et al. (100) demon-strated that the diverse mutational pattern ofbreast and colorectal tumors likely underliestheir immunological heterogeneity. Leary et al.(101) examined homozygous deletions and fo-cal amplifications in the breast and colorectalcancer genomes. Each of these studies furtherconfirmed the heterogeneity and intertumordiversity among breast and colorectal cancergenomes.

PARADIGM SHIFT

This change of focus, away from the search forkey, sequentially mutated genes that governcancer progression toward a more systems-oriented description, is evident in recent studieson pancreatic cancer. Using a two-part dis-covery and prevalence-determination strategy,along with copy-number and transcriptomicanalyses, Jones et al. (98) concluded that pan-creatic cancers contain an average of 63 clonal

genetic alterations, of which the majority arepoint mutations. As with breast and colorectalcancer, there was considerable variation bothin the number of mutations and in the spe-cific genes mutated among different cancerspecimens; again, no new prevalently mutatedgenes were identified. Because nearly all of thepredicted protein-coding genes in the humangenome were evaluated, these data providedthe opportunity to investigate groups of genesoperating through specific signaling pathwaysand processes in a relatively unbiased manner.The authors concluded that pancreatic cancerresults from genetic dysregulation of 12 corepathways and processes, including apoptosis,DNA-damage control, and regulation ofinvasion. Although these 12 processes aregenetically altered in the majority of pancreaticcancers, the specific components mutated inindividual tumors were largely different. It wasproposed that agents be designed to target thephysiological effects of the altered pathwaysand processes rather than individual genes.Although this therapeutic logic is reasonable,and the analyses do demonstrate enrichmentfor specific cellular processes, the granularityof the results does not extend much beyond re-inforcing the general hallmarks of cancer (52).

Glioblastoma multiforme (GBM) was thefirst cancer type to be screened systematicallyby two independent groups (102–103). Both


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


Indel: a mutationentailing the insertionor deletion of one ormore bases

studies integrated gene sequencing, identifi-cation of focal amplifications and deletionsthrough comparative hybridization arrays, andexpression analysis to comprehensively inter-rogate the GBM genome. One study, focus-ing on a group of 601 selected genes in 92predominantly primary GBMs, found no novelcommonly mutated genes among different tu-mors (103). Interestingly, the number of genealterations in GBMs was smaller than thatpreviously reported for colorectal and breastcancers. The second study focused on exhaus-tively sequencing all likely coding exons in adiscovery screen, then determined the preva-lence of any identified variants in a secondaryscreen (102). The discovery of one novel re-current mutation [isocitrate dehydrogenase 1(IDH1)], mutated in 12% of all GBMs andstrongly associated with secondary GBMs inparticular, was cited as a validation of the util-ity of genome-wide genetic analysis of tumors.Indeed, as of this writing, this finding is amongthe most significant to be unearthed by can-cer genome sequencing. Two follow-up reportshave indicated that active site mutations inIDH1, and occasionally its homolog IDH2, arefound in more than 70% of certain central ner-vous system (CNS) tumors including grade IIand III astrocytomas and oligodendrogliomasand secondary glioblastomas, although they arerarely found in primary glioblastomas and havenot been found in any of the tested tumorsfrom outside the CNS (104–105). Reinforcedby functional studies of these mutations in cul-tured cells indicating lowered enzymatic ac-tivity (105–106), this work has unequivocallyidentified an important new pathway for a spe-cific subset of CNS tumors. Nevertheless, bothinitial sequencing studies independently con-cluded that for the bulk of GBMs dysregulationof three core pathways, based around the al-ready well-studied genes TP53, CDKN2A, andEGFR, is central to tumor progression.

Analysis of the lung cancer genome identi-fied 1013 mutations in 188 cases of lung cancer(107–108). Twenty-five cases harbored no mu-tations in the 623 genes analyzed, and only fourgenes had point mutations in more than 10% of

tumors. By examining the distribution of genesacross cellular pathways, the authors identi-fied five key pathways in which componentswere frequently mutated. By far the most com-monly affected of these was the MAPK path-way, in which 70% of tumors sequenced had atleast one mutation altering known MAPK pro-teins. This pathway, however, encompasses 56genes, most of which, individually, are mutatedin fewer than 1% of lung cancers. Again, themost significant findings of this study are themutational heterogeneity among tumors andthe absence of prevalently mutated genes.

THE FIRST COMPLETECANCER GENOME

The characterization of the first hematopoi-etic cancer genome represented an im-portant methodological milestone in cancergenomics—truly whole-genome sequencing ofa tumor specimen (82). Prior efforts at rese-quencing tyrosine kinase genes in acute myeloidleukemia (AML) had yielded few mutations(109–110). By exhaustively screening the en-tire genome of a paired set of cancer andnormal samples from a single AML patientthrough use of massively parallel sequencing,the authors identified 500–1000 nonsynony-mous somatic changes uniquely present in thecancer. Of these, only ten mutations occurredin protein-coding genes, including two previ-ously described indels (within FLT3 and NPM1)known to occur at high frequency in AML. Im-portantly, none of the eight newly identifiedgenes was found to be mutated in 187 additionalcases of AML.

This study marked a number of significantadvances in large-scale cancer genome sequenc-ing. First, unbiased whole-genome sequencingis inherently a more complete means of cat-aloging all the clonal alterations in a cancergenome. With the increasing recognition thatso-called intergenic “junk” DNA and intronicsequences contain functional elements such asregulatory regions and noncoding RNAs, exon-centered genotyping may be missing impor-tant drivers occurring in regions of the genome


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


not previously explored. Even if most of thesenoncoding mutations turn out to be non-causative, the spectrum and total number ofclonal mutations borne by a tumor may provideimportant information about sources of muta-tions as well as about a tumor’s life history.

Second, unlike previous studies in whichmatched normal DNA was only sequenced tospecifically validate candidate mutations, thisstudy was the first to simultaneously apply anunbiased analysis to paired normal DNA, al-beit with less sequence depth. One criticism ofthe earlier strategies was the exclusion of anysequence variant coincident with a previouslydescribed SNP, without considering it a pos-sible de novo event. New mutations in knownSNP sites, in fact, may represent some of themost likely selectable drivers given the strongfamilial component to many cancers for whichno specific genes have yet been implicated.Only by sequencing of paired normal DNAcan clonal mutations at polymorphic loci bescored.

Third, next-generation sequencing plat-forms have several advantages that will likelymake them the preferred technologies for fu-ture studies. The most obvious is the signifi-cantly lower cost per base pair sequenced. Inaddition, the minimal-input DNA requirementobviates the need for expanding tumor cellsin culture or as xenografts. Three such de-vices are presently in commercial production,and many more are on the horizon. Geno-typing is accomplished by randomly shearingthe genome into many pieces, clonally amplify-ing individual fragments on a solid matrix, andthen sequencing these immobilized clones us-ing one of a variety of chemistries. Althoughenriching for certain portions of the genomeis possible, typically the fragments to be se-quenced are generated randomly rather thanby user-selected sequencing primers. An im-portant benefit to these random fragments isthe greater ease with which breakpoints result-ing from large indels or other rearrangementscan be identified. Conventional targeted rese-quencing by standard capillary methods aloneis likely to miss many such events because PCR

Next-generationsequencing: refers tonew methods ofhigh-throughput DNAsequencing carried outon amplified singleDNA moleculesaffixed to a solid matrix

amplification of a given region cannot occurif one or more primer sites is lost or distantlyrelocated.

Fourth, and finally, the markedly lower costof next-generation methods also means that au-tomatic confirmation of mutant bases can bereasonably built into the sequencing protocol.With such platforms, this process is simplya matter of sequencing enough random frag-ments to have a high probability of genotypingevery base several times. The lower through-put of capillary methods has often necessitatedthe initial filtering of mutant calls such that onlythose deemed to have a high likelihood of beingdrivers are retested. For example, the first ma-jor study by Sjoblom et al. (87) triaged 260,000noncoding changes without further confirma-tion. Although this may have been unavoid-able from the logistical standpoint of tradi-tional sequencing, synonymous changes withintumors may be of importance, given that theycan influence both transcription and translation(89).

IMPLICATIONS OF THE CANCERGENOME ATLAS: INTERTUMORHETEROGENEITY

The primary goal of cancer genome–sequencing studies has been the identificationof the genes and pathways that play a causalrole in the neoplastic process (98). It was theexpectation that sufficiently detailed geneticanalysis would lead to the identification of asmall set of commonly mutated genes thatdrive tumor progression and thus present newtherapeutic targets. Collectively, the initialstudies described above constitute the mostsystematic characterization of a disease genomeever undertaken and demonstrate the feasibilityof producing a compendium of clonally alteredsomatic sequences. Additional analyses that usecomplementary approaches, including thoseassessing rearrangements (111), deletions(112), and epimutations (113), as well as theimpact of mutations in noncoding sequences,are within the scope of our current technologyand will soon provide an even more complete


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


Subclonal mutation:mutation present in asingle cell or aminority of cells in atumor; not detectibleby conventionalsequencing

description of changes within the cancergenome.

The overarching conclusion to be drawnfrom completed cancer genome–sequencingstudies is that most cancer types display sub-stantial intertumor mutational heterogeneity.Individual solid organ tumors harbor, on aver-age, more than 50 nonsilent clonal mutationsin the coding regions of different genes, yetonly a small fraction of these genes are mutatedin a high proportion of tumors. Although cer-tain clonally disrupted genes are more preva-lently represented within specific types of can-cer, there remains a great deal of overlap. Thelarge number and breadth of diversity in genesmutated among individual tumor specimensemphasize the fundamentally stochastic natureof cancer evolution.

The therapeutic implications of these find-ings are considerable. Preliminary studies fo-cusing on kinases, the most druggable portionof the genome, explored the possibility of iden-tifying commonly mutated genes that might beexploited with targeted approaches (86). How-ever, after the number of samples profiled in-creased, it has become clear that any one ofthese genes is only mutated in a small fractionof tumors. To synthesize and test enough small-molecule inhibitors to combat even half of thekinase class of suspected tumor drivers wouldbe a daunting undertaking on a scale that isarguably beyond our current drug-developingand -regulatory capacities. The alternative op-tion of targeting general pathways rather thanspecific mutant proteins may be more feasible.

NONCLONAL MUTATIONSAND INTRATUMORHETEROGENEITY

A significant limitation of these studies lies notin the complexity of the clonal mutations theyare attempting to annotate, but in the fact that,by design, they are unable to address deeperheterogeneity within individual tumors. Mostinvestigations to date have been concerned onlywith identifying mutations in the dominantclone. A tumor is itself genomically heteroge-

neous, as each cell has a different mutationalsignature reflective of its unique lineage historywithin the evolving neoplasm. Analysis of dis-seminated single cells in minimal residual dis-ease has demonstrated a high level of genomicheterogeneity within individual lesions as wellas between primary tumors and metastatic cells(114). Irrespective of the predominant forcesgenerating mutations during tumorigenesis, itis of critical importance to recognize that muta-tions occur randomly and that only a tiny subsetof these are likely to be selectably advantageousat a given stage of development. This leaves amuch larger number of unexpanded mutationsto act as a dormant repository of genetic di-versity. A tumor is a dynamic entity that neverceases to evolve. The specific fitness of a cancercell depends on the context of its tumor en-vironment. As the environment changes overtime, new cellular stresses such as hypoxia, nu-trient depletion, and immune recognition arise,and cancer cells with the requisite phenotypesare selected. Thus, subclonal heterogeneity is ofparamount importance to tumor progression.

The clinical importance of subclonal muta-tions arises from the fact that genetic variantsencoding resistance to all single-target drugsare likely to preexist in a tumor cell population(70). Imatinib, the prototypical targeted ther-apy for chronic myelogenous leukemia (CML)bearing activating mutations in the ABL gene,frequently loses clinical efficacy due to theemergence of resistant clones (83). The basisfor this resistance is frequently attributable tonew point mutations in the ABL kinase domainthat decrease its affinity for the drug (115–116).It has been specifically demonstrated that resis-tance mutations can be found in CML prior tothe initiation of therapy (117). A method forovercoming the problem of resistance, whichhas long been standard in the clinic with ex-isting chemotherapy regimens, is to use mul-tidrug cocktails. A group of agents directed atmultiple, unrelated tumor-relevant pathways ismore likely to prevent or delay resistance, giventhat the combinatorial probability of a singlecell simultaneously bearing resistance muta-tions to each pathway is very low. Although this


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


concept is not new, the degree of intertumorheterogeneity revealed by cancer genome stud-ies hints that an even larger cocktail of drugsthan is presently used may be advantageous tocombat resistance arising from intratumor ge-netic variants (118).

The ability to quantify subclonal genetic di-versity may provide important clinical informa-tion about the likelihood of a tumor becom-ing resistant to specific therapies. Exactly howmany unexpanded random mutations are therein a tumor? This has historically been a diffi-cult question to answer because of the technicalchallenges facing low-frequency measurements(Figure 5). Standard capillary sequencing tech-nology measures average population genotypeand only detects minority clones down toapproximately 25% with routine, automateduse. Next-generation sequencing methods aremuch more sensitive, given that they genotypethe amplified product of individual molecules.Sequencing of many fragments from a given re-gion (i.e., deep sequencing) produces a digitalhistogram representing the frequency of dif-ferent genotypes in a population of molecules.However, because of imperfections in detectionhardware and chemistry, as well as the need foramplification steps by fallible polymerases, sen-sitivity is currently limited to approximately 1 in5000 under the most ideal circumstances (119)and probably nearer to 1 in 100 with routineuse. Although there do exist exquisitely sensi-tive methods of mutation detection, includingcell culture–based fluctuation assays (10) andsystems involving transgenic animals bearingreporter genes, neither method is amenable tothe direct examination of human tumors.

Our group has recently developed a methodfor the detection of random mutations that of-fers unprecedented sensitivity; one mutationcan be identified among 108 wild-type nu-cleotides in nuclear DNA (120). The system isbased on the concept that spontaneous muta-tions occurring in a noncoding, TaqI restric-tion endonuclease recognition site render itnoncleavable by this enzyme. After multiplerounds of enzymatic digestion, only the mutantsequences from a larger population remain

Clo

nal

ly d

eriv

ed tu

mo

r

Sin

gle

fou

nd

er c

ell

Capillary sequencing

Next-generation sequencing

Future sequencing technologies

Generation

Population size

0 1 2 3 12 20 30

0 2 4 8 4 x 103 106 109

Figure 5Limit of subclonal detection. Depicted is the clonal expansion of a single cellinto a population of 1 billion cells. In a hypothetical scenario in which no celldeath occurs, this expansion requires approximately 30 generations of division.Current capillary methods of DNA sequencing only detect mutations that haveclonally expanded to represent 25% or more of a tumor. Only mutations thatare present in the founding cell or that arise within the first two generationsof division can be identified. Deep sequencing on current next-generationsequencing platforms has been reported to detect subclonal mutations down toa frequency of 1 in 5000 (119) (i.e., those mutations that arise within the first 12generations after founding). Sensitivity is limited by the error rate of polymerasechain reaction amplification steps and that of the sequencing chemistry itself.Future technologies may eventually enable ultra-accurate, high-throughputdetection of mutations that arise during any stage of clonal expansion.

Deep sequencing:sequencing of manyindividual DNAfragments from anidentical portion of thegenome to identifysubclonal mutations

intact and are amplifiable by PCR primersflanking the restriction site. We (121) and oth-ers (122) have used this approach to demon-strate a markedly elevated frequency of ran-dom, unexpanded mutations in several types ofcancer. Although highly sensitive, this approachcan interrogate only four bases out of the entire


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


Percentageof tumors

Percentageof cells

b Landscape within individual tumors

a Landscape across multiple tumors

Chromosome X

Chromosome 1

q

p

Figure 6The mutational landscape of the cancer genome. (a) The cancer genome landscape proposed by Wood et al.(88) graphically represents the mutational heterogeneity among different tumors of a single cancer type. Theheight of each brown peak indicates the percentage of tumors found to carry a clonal mutation in a particulargene. The landscape comprises a small number of “mountains”—genes that are clonally mutated in a largefraction of individual cancers—and a significantly greater number of “hills”—genes clonally mutated in onlyone or a few tumors. Although 50 or more genes may be clonally mutated within the genome of anindividual tumor, most genes are rarely mutated in more than a few tumors. (b) An additional level of themutational landscape exists within individual tumors due to differences among the genomes of single tumorcells. Although a small number of mutations—“trees”—are clonally present in the majority of cells in anindividual tumor, an exponentially larger number—“seedlings”—exist subclonally in only one or a few cells.Among this vast reservoir of nonclonal mutations are many therapy-resistant variants.

genome at once. Hopefully the future will bringeven more sophisticated methods that combinethe throughput of present next-generation se-quencing platforms with the ultrahigh sensitiv-ity needed to accurately identify single mutants.

The cancer genomic landscape has beenpreviously described from a multiple-tumorperspective, in which the abovementioned“mountains” and “hills” represent, respec-tively, frequently and infrequently mutatedgenes (Figure 6a) (87). We posit that a com-plete description of the cancer genome mustnecessarily include a provision for the intratu-mor heterogeneity of individual neoplasms. To

this landscape we add a small number of “trees”to represent clonally mutated genes present ina large number of cells and, surrounding these,a much larger number of “seedlings” repre-senting mutations present in only one or a fewcells (Figure 6b). We argue that it is this forest,undetected by the cancer genome–sequencingstudies described above, that provides muchof the basis for the wide variations in tumorbehavior and responsiveness to therapy andthat represents one of the most clinicallyimportant features of the cancer genome:When an old tree falls or is logged, manyseedlings are poised to grow and take its place.


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


FINAL THOUGHTSAND FUTURE DIRECTIONS

Currently we recognize the unidirectional fateof neoplastically transformed cells to be theresult of unrepaired mutational events thatbecome permanently fixed in the genomeand epigenome of subsequent generations ofprogeny. The path that has brought us to ourcurrent understanding of the genetic basis forcancer has been long and remarkable, punctu-ated by a breadth of discoveries. The ability ofmost cells in the human body to prevent thebuild-up of cancer-causing mutations is an im-pressive tribute to the billions of years of evo-lution leading up to the emergence of multi-cellular organisms. Mechanisms for mutationprevention and suppression are, nevertheless,imperfect; progressive accumulation of new ge-netic and epigenetic variants provides the fuelfor evolution on a cellular level and forms thebasis of tumorigenesis.

New technology has allowed us to beginto tabulate the mutations of cancer. First-generation cancer genome–sequencing studieswere driven by the expectation that clonal mu-tations in a limited set of key genes wouldbe commonly found in different tumors andmight provide new drug-responsive targets.However, the results so far indicate a morecomplex picture than initially hoped. Very fewgenes that have not been previously identifiedby other means are prevalently mutated in spe-cific cancers. Many genes likely to be involvedin driving tumorigenesis are altered in only asmall fraction of tumors. The presence of manythousands of clonally expanded passengers,

although playing no causal role in the can-cer, serve as a reminder of the invisible legionsformed by the exponentially larger number ofunexpanded variants, many of which are drugresistant and are awaiting the opportunity toselectively proliferate upon induction of newtreatments. In light of this emerging complex-ity, it is becoming increasingly difficult to en-vision how it will be possible to develop a real-istic number of targeted chemotherapies to bedirected against a discrete panel of commonlymutated cancer genes. The findings describedherein substantiate the concept that simulta-neous use of multiple agents against differentgeneral pathways may be the most efficaciousapproach.

Although sobering, the cancer genome stud-ies performed so far have established an impor-tant baseline of information from which to ad-vance. As technology improves and comes downin cost, large-scale genome-analysis meth-ods will become tractable to smaller researchgroups, who will be able to explore innova-tive and higher-risk approaches. As we moveforward in these endeavors, we must not losesight of the need to confirm the functional sta-tus of mutations identified. Although genomesequencing remains a powerful tool, it cannotaddress all the questions of cancer research,so we cannot neglect to spread our resourcesamong many complementary means of identi-fying novel features of the disease and ways toprevent and target them. Most importantly, weneed to recognize the many levels of hetero-geneity inherent to cancer and ensure that thisreality be integrated into future studies.

SUMMARY POINTS

1. Cancer is a disease of somatic evolution that occurs on a cellular level. Random mutationsoccur throughout an organism’s life, and only a small subset of mutations that bestowgrowth and survival benefits upon a cell clonally expand to form a tumor.

2. Mutations result from errors in DNA synthesis and failure of DNA repair. Mutationsaccumulate more rapidly during tumorigenesis from increased cell proliferation as wellas from defects in DNA maintenance pathways. The relative importance of these twomechanisms is likely to vary from one tumor to another and remains the subject of debate.


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


3. Identifying causative mutations within a tumor’s genome by DNA sequencing is a com-plex problem from both a technical and an analytical standpoint. Many mutations areclonally present by chance alone, and differentiating these neutral passengers from causaldrivers presents a significant challenge.

4. Recent large-scale cancer genome–sequencing studies have indicated a great deal ofvariation in the clonal mutations found among different tumors. Very few genes arecommonly mutated in any type of cancer, and this finding suggests that it will be difficultto design a limited number of widely usable targeted therapies that focus on specificgenes.

5. Current methods of DNA sequencing cannot accurately portray the many mutations ina developing cancer that are present in only a minority of tumor cells. These subclonalmutations constitute a tumor’s potential for overcoming therapeutic interventions, andtheir presence argues for multiple and simultaneous chemotherapeutic approaches fortumor ablation. Characterization of this intratumor heterogeneity will be of clinicalimportance.

DISCLOSURE STATEMENT

L.A.L. is a member of the advisory board for Stratos, Inc. The other authors are not aware of anyaffiliations, memberships, funding, or financial holdings that might be perceived as affecting theobjectivity of this review.

ACKNOWLEDGMENTS

Grant support: National Cancer Institute, CA102029 and CA115802 (L.A.L.); U.S. Department ofDefense, W81XWH-07-0029 (L.A.L.); National Institute of Aging, AG033485 ( J.J.S.); NationalCancer Institute/Health Research Board of Ireland, Joint Research Fellowship (E.J.F.).

LITERATURE CITED

1. Classic paper; first tospeculate about canceras a chromosomaldisease.

1. Boveri T. 1902. Uber mehrpolige Mitosen als Mittel zur Analyze des Zellkerns. Verh. Dtsch. Zool.

Ges. Wurzburg 35:67–90

2. First paper toestablish the associationbetween defects inDNA repair and humancancer.

2. Cleaver JE. 1968. Defective repair replication of DNA in xeroderma pigementosum. Nature

218:652–563. Kallioniemi A, Kallioniemi O-P, Piper J, Tanner M, Stokke T, et al. 1994. Detection and mapping of

amplified DNA sequences in breast cancer by comparative genomic hybridization. Proc. Natl. Acad. Sci.USA 91:2156–60

4. Riopel MA, Spellerberg A, Griffin CA, Perlman EJ. 1998. Genetic analysis of ovarian germ cell tumorsby comparative genomic hybridization. Cancer Res. 58:3105–10

5. El-Rifai W, Sarlomo-Rikala M, Knuutila S, Miettinen M. 1998. DNA copy number changes in devel-opment and progression in leiomyosarcomas of soft tissues. Am. J. Pathol. 153:985–90

6. Kerangueven F, Noguchi T, Coulier F, Allione F, Wargniez V, et al. 1997. Genome-wide search for lossof heterozygosity shows extensive genetic diversity of human breast carnomas. Cancer Res. 57:5469–74

7. Klein CA, Schmidt-Kittler O, Schardt JA, Pantel K, Speicher MR, Riethmuller G. 1999. Comparativegenomic hybridization, loss of heterozygosity, and DNA sequence analysis of single cells. Proc. Natl.Acad. Sci. USA 96:4494–99

8. Vogelstein B, Fearon ER, Kern SE, Hamilton SR, Preisinger AC, et al. 1988. Genetic alterations duringcolorectal-tumor development. N. Engl. J. Med. 319:525–32


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


9. Albertini RJ, Nicklas JA, O’Neill JP, Robison SH. 1990. In vivo somatic mutations in humans: measure-ment and analysis. Annu. Rev. Genet. 24:305–26

10. Bielas JH, Heddle JA. 2003. Elevated mutagenesis and decreased DNA repair at a transgene are associatedwith proliferation but not apoptosis in p53-deficient cells. Proc. Natl. Acad. Sci. USA 100:12853–58

11. Chu EHY, Boehnke M, Hanash SM, Kuick RD, Lamb BJ, et al. 1988. Estimation of mutation rates basedon the analysis of polypeptide constituents of cultured human lymphoblastoid cells. Genetics 119:693–703

12. DeMars R, Held KR. 1972. The spontaneous azaguanine-resistant mutants of diploid human fibroblasts.Humangenetik 16:87–110

13. Cervantes RB, Stringer JR, Shao C, Tischfield JA, Stambrook PJ. 2002. Embryonic stem cells and somaticcells differ in mutation frequency and type. Proc. Natl. Acad. Sci. USA 99:3586–90

14. Jackson AL, Loeb LA. 1998. The mutation rate and cancer. Genetics 148:1483–9015. Ferrin LJ, Beckman RA, Loeb LA, Mildvan AS. 1986. Kinetic and magnetic resonance studies of the

interaction of Mn2+, substrates and templates with DNA polymerases. In Manganese in Metabolism andEnzyme Function, ed. VL Schramm, FC Wedler, pp. 259–73. New York: Academic

16. Petruska J, Sowers LC, Goodman MF. 1986. Comparison of nucleotide interactions in water, proteinsand vacuum: model for DNA polymerase fidelity. Proc. Natl. Acad. Sci. USA 83:1559–62

17. Loeb LA, Monnat R. 2008. DNA polymerases and human disease. Nat. Rev. Genet. 9:594–60418. Uyemura D, Lehman IR. 1976. Biochemical characterization of mutant forms of DNA polymerase I

from Escherichia coli. I. The polA12 mutation. J. Biol. Chem. 251:4078–8419. Modrich P, Lahue R. 1996. Mismatch repair in replication fidelity, genetic recombination and cancer

biology. Annu. Rev. Biochem. 65:101–3320. Kunkel TA, Schaaper RM, Loeb LA. 1983. Depurination-induced infidelity of DNA synthesis with

purified DNA replication proteins in vitro. Biochemistry 22:2378–8421. Kiel MJ, He S, Ashkenazi R, Gentry SN, Teta M, et al. 2007. Haematopoietic stem cells do not asym-

metrically segregate chromosomes or retain BrdU. Nature 449:238–4222. Nishimura S. 2006. 8-hydroxyguanine: from its discovery in 1983 to the present status. Proc. Jpn. Acad.

82:127–4123. Floyd RA. 1995. Measurement of oxidative stress in vivo. In The Oxygen Paradox, ed. KJA Davies, F

Ursini, pp. 89–103. Padova, Italy: CLEUP Univ. Press24. Lindahl T, Wood RD. 1999. Quality control by DNA repair. Science 286:1897–90525. Barrett MT, Sanchez CA, Prevo LJ, Wong DJ, Galipeau PC, et al. 1999. Evolution of neoplastic cell

lineages in Barrett esophagus. Nat. Genet. 22:106–926. Bronner MP, O’Sullivan JN, Rabinovitch PS, Crispin DA, Chen L, et al. 2008. Genomic biomarkers to

improve ulcerative colitis neoplasia surveillance. Am. J. Pathol. 173:1853–6027. Baik SC, Youn HS, Chung MH, Lee WK, Cho MJ, et al. 1996. Increased oxidative DNA damage in

Heliocobacter pylori–infected human gastric mucosa. Cancer Res. 56:1279–8228. Fausto N, Webber EM. 1993. Mechanisms of growth regulation in liver regeneration and hepatic carcino-

genesis. In Progress in Liver Diseases, ed. JL Boyer, RK Okner, 11:115–37. Philadelphia: W.B. Saunders29. Feig DI, Loeb LA. 1994. Oxygen radical induced mutagenesis is DNA polymerase specific. J. Mol. Biol.

235:33–4130. The test describedwithin is the standard ofreference for detectionof mutagenic chemicals.

30. Ames BN, Durston WE, Yamasaki E, Lee FD. 1973. Carcinogens are mutagens: a simple testsystem combining liver homogenates for activation and bacteria for detection. Proc. Natl. Acad.

Sci. USA 70:2281–8531. Auerbach C, Robson JM. 1946. Chemical production of mutations. Nature 154:30232. Loeb L, Harris CC. 2008. Advances in chemical carcinogenesis: a historical review and prospective.

Cancer Res. 68:6863–7233. Hecht SS. 1999. Tobacco smoke carcinogens and lung cancer. J. Natl. Cancer Inst. 91:1194–21034. Wynder EL, Graham EA. 1950. Tobacco smoking as a possible etiologic factor in bronchogenic carci-

noma. JAMA 143:329–36 35. Position paper of theAmerican Association ofCancer Research on theassociation betweensmoking and cancer.

35. Loeb LA, Ernster VL, Warner KE, Abbotts J, Laszlo J. 1984. Smoking and lung cancer: anoverview. Cancer Res. 44:5940–58

36. Friedberg EC, Aguilera A, Gellert M, Hanawalt PC, Hays JB, et al. 2006. DNA repair: from molecularmechanism to human disease. DNA Repair 5:986–96


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


37. Hoeijmakers JH. 2001. Genome maintenance mechanisms for preventing cancer. Nature 411:366–7438. Harris CC. 1993. p53: at the crossroads of molecular carcinogenesis and cancer risk assessment. Science

262:1980–8139. Kolodner RD, Marsischky GT. 1999. Eukaryotic DNA mismatch repair. Curr. Opin. Genet. Dev. 9:89–9640. Al-Tassan N, Chmiel NH, Maynard J, Fleming N, Livingston AL, et al. 2002. Inherited variants of MYH

associated with somatic G:C–>T:A mutations in colorectal tumors. Nat. Genet. 30:227–3241. King MC, Marks JH, Mandell JB. 2003. Breast and ovarian cancer risks due to inherited mutations in

BRCA1 and BRCA2. Science 302:643–4642. Riballo E, Critchlow SE, Teo SH, Doherty AJ, Priestley A, et al. 1999. Identification of a defect in DNA

ligase IV in a radiosensitive leukaemia patient. Curr. Biol. 9:699–70243. Masutani C, Araki M, Yamada A, Kusumoto R, Nogimori T, et al. 1999. Xeroderma pigmentosum variant

(XP-V) correcting protein from HeLa cells has a thymine dimer bypass DNA polymerase activity. EMBOJ. 18:3491–501

44. Ellis NA, Groden J, Ye T-Z, Straughen J, Lennon DJ, et al. 1995. The Bloom’s syndrome gene productis homologous to RecQ helicases. Cell 83:655–66

45. Kamath-Loeb AS, Lan L, Nakajima S, Yasui A, Loeb LA. 2007. The Werner syndrome protein interactsfunctionally with translesion DNA polymerases. Proc. Natl. Acad. Sci. USA 104:10394–99

46. Levine AJ. 1997. p53, the cellular gatekeeper for growth and division. Cell 88:323–3147. Goode EL, Ulrich CM, Potter JD. 2002. Polymorphisms in DNA repair genes and associations with

cancer risk. Cancer Epidemiol. Biomark. Prev. 11:1513–3048. Webber EM, Wu JC, Wang L, Merlino G, Fausto N. 1994. Overexpression of transforming growth

factor-α causes liver enlargement and increased hepatocyte proliferation in transgenic mice. Am. J.Pathol. 145:398–408

49. Hecker E. 1967. Phorbol esters from croton oil. Chemical nature and biological activities. Naturwis-senschaften 54:282–84

50. Rous P. 1911. Transmission of a malignant new growth by means of a cell-free filtrate. JAMA 56:19851. Varmus HE, Shank PR, Hughes SE, Kung HJ, Heasley S, et al. 1979. Synthesis, structure, and integration

of the DNA of RNA tumor viruses. Cold Spring Harb. Symp. Quant. Biol. 43(Part 2):851–6452. Presents a detailedanalysis of the requiredphenotypes fortumorigenesis.

52. Hanahan D, Weinberg RA. 2000. The hallmarks of cancer. Cell 100:57–7053. Reya T, Morrison SJ, Clarke MF, Weissman IL. 2001. Stem cells, cancer, and cancer stem cells. Nature

414:105–1153a. Pepper JW, Sprouffske K, Maley CC. 2007. Animal cell differentiation patterns suppress somatic evo-

lution. PLoS Comput. Biol. 3(12):e5054. Humphries A, Wright NA. 2008. Colonic crypt organization and tumorigenesis. Nat. Rev. Cancer 8:415–

2455. Cairns J. 2006. Cancer and the immortal strand hypothesis. Genetics 174:1069–7256. Prindull GA, Fibach E. 2007. Are postnatal hemangioblasts generated by dedifferentiation from com-

mitted hematopoietic stem cells? Exp. Hematol. 35:691–70157. Seminal paperdescribing cancer an asevolutionary process.

57. Nowell PC. 1976. The clonal evolution of tumor cell populations. Science 194:23–2858. Merlo LM, Pepper JW, Reid BJ, Maley CC. 2006. Cancer as an evolutionary and ecological process.

Nat. Rev. Cancer 6:924–3559. Cairns J. 1975. Mutation selection and the natural history of cancer. Nature 255:197–20060. Armitage P, Doll R. 1954. The age distribution of cancer and a multi-stage theory of carcinogenesis.

Br. J. Cancer 8:1–1261. Knudson AG. 1996. Hereditary cancer: two hits revisited. J. Cancer Res. Clin. Oncol. 122:135–4062. Renan MJ. 1993. How many mutations are required for tumorigenesis? Implications from human cancer

data. Mol. Carcinog. 7:139–4663. Rangarajan A, Hong SJ, Gifford A, Weinberg RA. 2004. Species- and cell type–specific requirements for

cellular transformation. Cancer Cell 6:171–8364. Mahale AM, Khan ZA, Igarashi M, Nanjangud GJ, Qiao RF, et al. 2008. Clonal selection in malignant

transformation of human fibroblasts transduced with defined cellular oncogenes. Cancer Res. 68:1417–2665. Bodmer W. 2008. Genetic instability is not a requirement for tumor development. Cancer Res. 68:3558–61


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


66. Loeb LA, Bielas JH, Beckman RA. 2008. Cancers exhibit a mutator phenotype: clinical implications.Cancer Res. 68:3551–57

67. Lengauer C, Kinzler KW, Vogelstein B. 1998. Genetic instabilities in human cancers. Nature 396:643–4968. Moolgavkar SH, Luebeck EG. 2003. Multistage carcinogenesis and the incidence of human cancer. Genes

Chromosomes Cancer 38:302–669. First paper topropose that cancersexhibit a mutatorphenotype.

69. Loeb LA, Springgate CF, Battula N. 1974. Errors in DNA replication as a basis of malignantchange. Cancer Res. 34:2311–21

70. Loeb LA, Loeb KR, Anderson JP. 2003. Multiple mutations and cancer. Proc. Natl. Acad. Sci. USA100:776–81

71. Beerenwinkel N, Antal T, Dingli D, Traulsen A, Kinzler KW, et al. 2007. Genetic progression and thewaiting time to cancer. PLoS Comput. Biol. 3:e225

72. Tomlinson IP, Novelli MR, Bodmer WF. 1996. The mutation rate and cancer. Proc. Natl. Acad. Sci. USA93:14800–3

73. Beckman RA, Loeb LA. 2006. Efficiency of carcinogenesis with and without a mutator mutation.Proc. Natl. Acad. Sci. USA 103:14140–15

73a. Beckman RA. 2009. Mutator mutations enhance tumorigenic efficiency across fitness landscapes.PLoS One 4:e5860

74. Beckman RA, Loeb LA. 2005. Negative clonal selection in tumor evolution. Genetics 171:2123–3175. Robins H, Alexe G, Harris SL, Levine AJ. 2005. The first twenty-five years of p53 research. In Twenty-Five

Years of p53 Research, ed. P Hainaut, KG Wiman, pp. 1–26. Dordrecht: Springer76. Wells RD, Sinden RR. 1993. Defined ordered sequence DNA, DNA structure, and DNA-directed

mutation. In Genome Analysis, Genome Rearrangement and Stability, 7:107–38. Plainview, NY: Cold SpringHarb. Lab. Press

77. Farber RA, Petes TD, Dominska M, Hudgens SS, Liskay RM. 1994. Instability of simple sequencerepeats in a mammalian cell line. Hum. Mol. Genet. 3:253–56

78. Hanawalt PC, Spivak G. 2008. Transcription-coupled DNA repair: two decades of progress and surprises.Nat. Rev. Mol. Cell Biol. 9:958–70

79. Stamatoyannopoulos JA, Adzhubei I, Thurman RE, Kryukov GV, Mirkin SM, Sunyaev SR. 2009. Humanmutation rate associated with DNA replication timing. Nat. Genet. 41:393–95

80. Balaban GB, Herlyn M, Clark WH Jr, Nowell PC. 1986. Karyotypic evolution in human malignantmelanoma. Cancer Genet. Cytogenet. 19:113–22

81. Smith G, Carey FA, Beattie J, Wilkie MJV, Lightfoot TJ, et al. 2002. Mutations in APC, Kirsten-ras,and p53—alternative genetic pathways to colorectal cancer. Proc. Natl. Acad. Sci. USA 99:9433–38

82. Provided the firstcomplete whole-genome sequence of atumor and matchednormal and identifiednearly 1000nonsynonymoussomatic mutations.

82. Ley TJ, Mardis E, Ding L, Fulton B, McLellan M, et al. 2008. DNA sequencing of a cytogeneticallynormal acute myeloid leukaemia genome. Nature 456:66–72

83. Druker BJ. 2004. Imatinib: paradigm or anomaly? Cell Cycle 3:833–3584. Wang TL, Rago C, Silliman N, Ptak J, Markowitz S, et al. 2002. Prevalence of somatic alterations in

the colorectal cancer cell genome. Proc. Natl. Acad. Sci. USA 99:3076–8085. Wang Z, Shen D, Parsons DW, Bardelli A, Sager J, et al. 2004. Mutational analysis of the tyrosine

phosphatome in colorectal cancers. Science 304:1164–6686. Bardelli A, Parsons DW, Silliman N, Ptak J, Szabo S, et al. 2003. Mutational analysis of the tyrosine

kinome in colorectal cancers. Science 300:94987–88. First cancergenome studies tosequence mutations inall known coding exonsof any cancer.

87. Sjoblom T, Jones S, Wood LD, Parsons DW, Lin J, et al. 2006. The concensus coding sequencesof human breast and colorectal cancers. Science 314:268–74

88. Wood LD, Parsons DW, Jones S, Lin J, Sjoblom T, et al. 2007. The genomic landscapes of humanbreast and colorectal cancers. Science 318:1108–13

89. Kudla G, Murray AW, Tollervey D, Plotkin JB. 2009. Coding-sequence determinants of gene expressionin Escherichia coli. Science 324:255–58

90. Getz G, Hofling H, Mesirov JP, Golub TR, Meyerson M, et al. 2007. Comment on “The ConsensusCoding Sequences of Human Breast and Colorectal Cancers.” Science 317:1500

91. Rubin AF, Green P. 2007. Comment on “The Consensus Coding Sequences of Human Breast andColorectal Cancers.” Science 317:1500


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


92. Forrest WF, Cavet G. 2007. Comment on “The Consensus Coding Sequences of Human Breast andColorectal Cancers.” Science 317:1500

93. Parmigiani G, Lin J, Boca S, Sjoblom T, Jones S, et al. 2007. Response to comments on “The ConsensusCoding Sequences of Human Breast and Colorectal Cancers.” Science 317:1500

94. The most extensivecharacterization to dateof mutations in a singlegene family acrossdifferent cancer types.

94. Greenman C, Stephens P, Smith R, Dalgliesh GL, Hunter C, et al. 2007. Patterns of somaticmutation in human cancer genomes. Nature 446:153–58

95. Stephens P, Edkins S, Davies H, Greenman C, Cox C, et al. 2005. A screen of the complete proteinkinase gene family identifies diverse patterns of somatic mutations in human breast cancer. Nat. Genet.37:590–92

96. Davies H, Hunter C, Smith R, Stephens P, Greenman C, et al. 2005. Somatic mutations of the proteinkinase gene family in human lung cancer. Cancer Res. 65:7591–95

97. Bignell G, Smith R, Hunter C, Stephens P, Davies H, et al. 2006. Sequence analysis of the protein kinasegene family in human testicular germ-cell tumors of adolescents and adults. Genes Chromosomes Cancer45:42–46

97a. Jones S, Chen WD, Parmigiani G, Diehl F, Beerenwinkel N, et al. 2008. Comparative lesion sequencingprovides insights into tumor evolution. Proc. Natl. Acad. Sci. USA 105:4283–88

98. Jones S, Zhang X, Parsons D, Lin J, Leary R, et al. 2008. Core signaling pathways in human pancreaticcancers revealed by global genomic analyses. Science 321:1801–6

99. Prehn RT, Main JM. 1957. Immunity to methylcholanthrene-induced sarcomas. J. Natl. Cancer Inst.18:769–78

100. Segal N, Parsons D, Peggs K, Velculescu V, Kinzler K, et al. 2008. Epitope landscape in breast andcolorectal cancer. Cancer Res. 68:889–92

101. Leary RJ, Lin JC, Cummins J, Boca S, Wood LD, et al. 2008. Integrated analysis of homozygous deletions,focal amplifications, and sequence alterations in breast and colorectal cancers. Proc. Natl. Acad. Sci. USA105:16224–29

102. Parsons DW, Jones S, Zhang X, Lin JC, Leary RJ, et al. 2008. An integrated genomic analysis of humanglioblastoma multiforme. Science 321:1807–12

103. Cancer Genome Atlas Network. 2008. Comprehensive genomic characterization defines human glioblas-toma genes and core pathways. Nature 455:1061–68

104. Balss J, Meyer J, Mueller W, Korshunov A, Hartmann C, von Deimling A. 2008. Analysis of the IDH1codon 132 mutation in brain tumors. Acta Neuropathol. 116:597–602

105. Yan H, Parsons DW, Jin G, McLendon R, Rasheed BA, et al. 2009. IDH1 and IDH2 mutations ingliomas. N. Engl. J. Med. 360:765–73

106. Zhao S, Lin Y, Xu W, Jiang W, Zha Z, et al. 2009. Glioma-derived mutations in IDH1 dominantlyinhibit IDH1 catalytic activity and induce HIF-1α. Science 324:261–61

107. Weir BA, Woo MS, Getz G, Perner S, Ding L, et al. 2007. Characterizing the cancer genome in lungadenocarcinoma. Nature 450:893–98

108. Ding L, Getz G, Wheeler DA, Mardis ER, McLellan MD, et al. 2008. Somatic mutations affect keypathways in lung adenocarcinoma. Nature 455:1069–75

109. Tomasson MH, Xiang Z, Walgren R, Zhao Y, Kasai Y, et al. 2008. Somatic mutations and germlinesequence variants in the expressed tyrosine kinase genes of patients with de novo acute myeloid leukemia.Blood 111:4797–808

110. Loriaux MM, Levine RL, Tyner JW, Frohling S, Scholl C, et al. 2008. High-throughput sequenceanalysis of the tyrosine kinome in acute myeloid leukemia. Blood 111:4788–96

111. Campbell PJ, Stephens PJ, Pleasance ED, O’Meara S, Li H, et al. 2008. Identification of somati-cally acquired rearrangements in cancer using genome-wide massively parallel paired-end sequencing.Nat. Genet. 40:722–29

112. Lai LA, Paulson TG, Li X, Sanchez CA, Maley C, et al. 2007. Increasing genomic instability duringpremalignant neoplastic progression revealed through high resolution array-CGH. Genes ChromosomesCancer 46:532–42

113. Shames DS, Girard L, Gao B, Sato M, Lewis CM, et al. 2006. A genome-wide screen for promotermethylation in lung cancer identifies novel methylation markers for multiple malignancies. PLoS Med.3:e486


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.


114. Stoecklein NH, Hosch SB, Bezler M, Stern F, Hartmann CH, et al. 2008. Direct genetic analysis ofsingle disseminated cancer cells for prediction of outcome and therapy selection in esophageal cancer.Cancer Cell 13:441–53

115. Michor F, Hughes TP, Iwasa Y, Branford S, Shah NP, et al. 2005. Dynamics of chronic myeloidleukaemia. Nature 435:1267–70

116. Radich JP, Dai H, Mao M, Oehler V, Schelter J, et al. 2006. Gene expression changes associated withprogression and response in chronic myeloid leukemia. Proc. Natl. Acad. Sci. USA 103:2794–99

117. Talpaz M, Shah NP, Kantarjian H, Donato N, Nicoll J, et al. 2006. Dasatinib in imatinib-resistantPhiladelphia chromosome–positive leukemias. N. Engl. J. Med. 354:2531–41

118. Golub TR. 2003. Mining the genome for combination therapies. Nat. Med. 9:510–11119. Demonstrates theextent of subclonalmutationalheterogeneityidentifiable by currentsequencingtechnologies.

119. Campbell PJ, Pleasance ED, Stephens PJ, Dicks E, Rance R, et al. 2008. Subclonal phylogeneticstructures in cancer revealed by ultradeep sequencing. Proc. Natl. Acad. Sci. USA 105:13081–86

120. Bielas J, Loeb LA. 2005. Quantification of random genomic mutations. Nat. Methods 2:285–90121. Bielas JH, Loeb KR, Rubin BP, True LD, Loeb LA. 2006. Human cancers express a mutator phenotype.

Proc. Natl. Acad. Sci. USA 103:18238–42122. Zheng L, Dai H, Zhou M, Li M, Singh P, et al. 2007. Fen1 mutations result in autoimmunity, chronic

inflammation and cancers. Nat. Med. 13:812–19


Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.

AR403-FM ARI 9 December 2009 12:46

Annual Review ofPathology:Mechanisms ofDisease

Volume 5, 2010Contents

Molecular Pathogenesis of Necrotizing FasciitisRandall J. Olsen and James M. Musser � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 1

The Pathobiology of Glioma TumorsCandece L. Gladson, Richard A. Prayson, and Wei Michael Liu � � � � � � � � � � � � � � � � � � � � � � � � � � �33

Mutational Heterogeneity in Human Cancers: Origin andConsequencesJesse J. Salk, Edward J. Fox, and Lawrence A. Loeb � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �51

Fibrogenic Reactions in Lung DiseaseJun Araya and Stephen L. Nishimura � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �77

The Senescence-Associated Secretory Phenotype: The Dark Sideof Tumor SuppressionJean-Philippe Coppe, Pierre-Yves Desprez, Ana Krtolica, and Judith Campisi � � � � � � � � � � �99

Epithelial Barriers in Homeostasis and DiseaseAmanda M. Marchiando, W. Vallen Graham, and Jerrold R. Turner � � � � � � � � � � � � � � � � � � 119

Nonalcoholic Fatty Liver Disease: Pathology and PathogenesisDina G. Tiniakos, Miriam B. Vos, and Elizabeth M. Brunt � � � � � � � � � � � � � � � � � � � � � � � � � � � � 145

Pathogenesis of PreeclampsiaBrett C. Young, Richard J. Levine, and S. Ananth Karumanchi � � � � � � � � � � � � � � � � � � � � � � � � 173

Preinvasive Breast CancerDennis C. Sgroi � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 193

The Pathogenesis of Acute Pulmonary Viral and Bacterial Infections:Investigations in Animal ModelsMary F. Lipscomb, Julie Hutt, Julie Lovchik, Terry Wu, and C. Rick Lyons � � � � � � � � � � � 223

Mammalian Sirtuins: Biological Insights and Disease RelevanceMarcia C. Haigis and David A. Sinclair � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 253

Mitochondrial Energetics and TherapeuticsDouglas C. Wallace, Weiwei Fan, and Vincent Procaccio � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 297

v

Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.

AR403-FM ARI 9 December 2009 12:46

p63 in Epithelial Survival, Germ Cell Surveillance, and NeoplasiaChristopher P. Crum and Frank D. McKeon � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 349

Indexes

Cumulative Index of Contributing Authors, Volumes 1–5 � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 373

Cumulative Index of Chapter Titles, Volumes 1–5 � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 376

Errata

An online log of corrections to Annual Review of Pathology, Mechanisms of Disease articlesmay be found at http://pathol.annualreviews.org

vi Contents

Ann

u. R

ev. P

atho

l. M

ech.

Dis

. 201

0.5:

51-7

5. D

ownl

oade

d fr

om a

rjou

rnal

s.an

nual

revi

ews.

org

by U

NIV

ER

SIT

Y O

F W

ASH

ING

TO

N o

n 01

/19/

10. F

or p

erso

nal u

se o

nly.

Documents

Mutational Heterogeneity in Human Cancers: Origin and ...depts.washington.edu/loeblabs/pdf/2009_Salk.pdf · University of Washington, Seattle, Washington 98195; email: [email protected]