Mechanisms of SpontaneousHuman CancersStanley VenittSection of Molecular Carcinogenesis, Institute of Cancer Research:Royal Cancer Hospital, Sutton, Surrey, United Kingdom

The causes of much of human cancer remain obscure. The fraction that is spontaneous isunknown and cannot be calculated until all known external causes have been accounted for. Thisis not a feasible proposition. However, there is substantial evidence that about 80% of humancancer could be avoided by eliminating tobacco consumption; by dietary changes; by reducinginfection with certain viruses, bacteria, and parasitic worms; and, in white populations, byavoiding sunburn. Alcohol, occupational and medical carcinogens, and certain patterns ofreproductive behavior also contribute to the cancer burden. Cancers that cannot be attributed tothese causes, and for which no other causes can be found, could be considered spontaneousand to arise from endogenous processes. Epidemiological evidence suggests that spontaneousand induced cancers share the same mechanism. Cancer is a genetic disorder of somatic cells.An accumulation of mutant genes that control the cell cycle, maintain genomic stability, andmediate apoptosis is central to carcinogenesis. Spontaneous mutation may cause spontaneouscancer. Endogenous causes of mutation include depurination and depyrimidation of DNA;proofreading and mismatch errors during DNA replication; deamination of 5-methylcytosine toproduce C to T base pair substitutions; and damage to DNA and its replication imposed byproducts of metabolism (notably oxidative damage caused by oxygen free radicals). Deficienciesin cellular defense mechanisms may also provoke spontaneous mutation. These include defectiveDNA excision-repair; low levels of antioxidants, antioxidant enzymes, and nucleophiles that trapDNA-reactive electrophiles; and enzymes that conjugate nucleophiles with DNA-damagingelectrophiles. Mechanisms underlying many of these cellular defenses are under genetic control.Thus, germ line mutations or polymorphisms of genes that govern them may also contribute tospontaneous cancer. Environ Health Perspect 1 04(Suppl 3):633-637 (1996)

Key words: cancer, avoidable, spontaneous, mutation, DNA damage, p53, deamination, CpGdinucleotides

What Is Spontaneous Cancer?Before addressing the mechanisms thatmight underlie spontaneous cancer, itwould be useful to examine what such aterm might mean. Spontaneous can bedefined in several ways. For example, theShorter Oxford Dictionary (1) provides sev-eral definitions, including "Of naturalprocesses: having a self-contained cause ororigin"; "growing or produced naturally

This paper was presented at the 2nd InternationalConference on Environmental Mutagens in HumanPopulations held 20-25 August 1995 in Prague,Czech Republic. Manuscript received 22 November1995; manuscript accepted 28 November 1995.

The author thanks the Cancer Research Campaignand the Medical Research Council of the UnitedKingdom for financial support.

Address correspondence to Dr. Stanley Venitt,Section of Molecular Carcinogenesis, Institute ofCancer Research: Royal Cancer Hospital, CotswoldRoad, Sutton, Surrey SM2 5NG, U.K. Telephone: +44181 643 8901. Fax: +44 181 770 7290. E-mail:s.venitt@icr.ac.uk

without cultivation or labour"; and "pro-duced, developed, or coming into existenceby natural processes or changes." It isbeyond dispute that much of humancancer is not spontaneous as defined abovebut is caused by exogenous agents or is self-inflicted by habits such as smoking anddrinking alcoholic beverages or resultsfrom germ-line mutations in critical genes.(I shall not attempt to grapple with thequestion as to whether such mutations arethemselves spontaneous, and the cancersthat result are also spontaneous.) Suchknowledge has been gained by studyingdifferences in cancer incidence and mortal-ity among different settled communitiesand between migrants and those who staybehind. Variation with time in cancer inci-dence and mortality within communitiesand the identification of specific causes orpreventive factors have also contributed toour present knowledge of the causes of

cancer (2). Moving from cause to mecha-nism has been made possible by remark-able advances in the molecular biology andgenetics of cancer, and it is probable thatthe key events in carcinogenesis will beunderstood at the molecular level withinthe lifetime of a graduate student justembarking on a career in cancer research.Nevertheless, as things stand, the causes ofmuch of human cancer remain obscure.The proportion that is truly spontaneous(and therefore unavoidable) is unknownand cannot be calculated unless and untilall known external and genetic causes havebeen accounted for.

Avoidable CancerThere is substantial evidence that a majorproportion (perhaps about 80%) of humancancer could be avoided by reducing oreliminating tobacco consumption; bychanges in diet and nutrition; by reducinginfection with certain viruses, bacteria, andparasitic worms; and, especially for white-skinned populations, by avoiding sunburn.Consumption of alcoholic beverages, cer-tain aspects of sexual and reproductivebehavior, and occupational and medicalexposure to carcinogens also contribute tothe cancer burden (Figure 1) (2). Povertyand wealth must also be taken into accountwhen discussing the causes of cancerbecause there is impressive evidence thatsome cancers occur more frequently inpoor communities and that others aremore frequent among the rich (3) (seeFigure 2 for some examples). Those can-cers that cannot be attributed to thesecauses and for which no other causes canbe found could be considered spontaneousand to arise from endogenous processes.

TobaccoAlcohol

DietFood additives

Sex, reproductionOccupation

PollutionIndustrial products

MedicinesUV, ionizing radiation

InfectionsUnknown

.

t

0 20 40 60 80

Percentage of deaths

Figure 1. Proportions of cancer deaths attributed tovarious causes. Data from Doll and Peto (2). The verti-cal line on each bar represents the best point estimateand the bar itself represents the range of acceptableestimates.

Environmental Health Perspectives * Vol 104, Supplement 3 * May 1996 633

S. VENITT

160 -

140 -

120 -

100 -

80 -

60 -

An _

Esophagus160 -

140 -

120 -

100 -

80 -

60 -

40-

20 -

Il

- 200-

*020

.w_

co

c 150-

100-

200 -

150 -

100 -

50 -

0-

50 -

0-

Stomach200 -

150 -

100 -

50 -

0-

Lung Breast120 -i

100-

80 -

60 -

40 -

20-

0-

Melanoma Colon1201

100 -

80

40

0

Rectum120

100

80 -

60-

40 -

20-

0-

Deprivation category

Affluent * Median Deprived

Figure 2. The relationship between deprivation category and incidence of eight different cancers in Scotland from 1979 to 1982. Data from Carstairs and Morris (3).

Nature, Nurture and LuckWith rare exceptions, development ofcancer in an individual depends on a com-plex interplay among genetics, environ-ment, and the play of chance-"nature,nurture and luck" and "what people do tothemselves and what they have done tothem" (2). It is therefore impossible, atpresent, to disentangle the relative contri-butions that each of these factors makes tothe risk of cancer in an individual or withina given population. The problem has beenaddressed by Knudson (4). He suggestedthat the population can be divided intofour oncodemes (an oncodeme is definedas a demographic unit with a peculiar sen-sitivity to a particular cancer), dependingon the relative contributions of environ-ment and genetics to the risk of cancer.These oncodemes are background (due torandom mutations in normal people);environmental (due to environmental car-cinogens [chemicals, radiation, or viruses,or combinations of these] in normalpeople); environmental/genetic (due to

genetic susceptibility to environmentalcarcinogens); and genetic (genetic suscepti-bility is more important than spontaneousor environmentally induced events). Mosthuman cancer probably occurs in the sec-ond and third oncodemes because exposureto environmental carcinogens is difficult orimpossible to avoid and cancers that aredue entirely to an inherited predispositionare known to be rare (5).

Do Spontaneous andInduced Cancers Sharethe Same Mechanism?There is circumstantial evidence that themechanisms that provoke what are thoughtof as spontaneous cancers appear to be thesame as or similar to those that underliecancers caused by exogenous agents. Figure3 depicts one such piece of evidence, whichis based on the analysis of the age distribu-tion of lung cancer in smokers and non-smokers by Doll (6). He showed that theexcess rate of annual lung cancer incidence

,

c, /100 10

=CO

1-M10o100

Years

Figure 3. Lung cancer death rates in nonsmokers as afunction of age and in regular cigarette smokers as afunction of duration of smoking. Data from Doll (6).

among regular smokers, which dependsstrongly on the duration of smoking, canbe distinguished from the background ratederived from the incidence of lung cancer

Environmental Health Perspectives * Vol 104, Supplement 3 * May 1996634

SPONTANEOUS CANCER

1000 -l Female breast cancer

100 -

10-

2(

1000 -

100 -

10 -

1-

0.1 -

a)CLa)

co-j

12000

100 -

10 -

1 -

0.1 -21

100-

10 -

1 -

0.1 -

10 -

1 -

0.1

England(Birmingham)

XBmIndia0/(Bombay)

Male lung cancer

England(Birmingham)

Male colon cancer

England(Birmingham) }

Male oral cancer

India(Bombay)

*- England(Birminghar

20

Testis cancer

England(Birmingham) ,,*W/,/

Japan(Miyagi)

of slopes could also be construed as indi-rect evidence that exogenous carcinogens-in this case tobacco smoke-enhance orintensify aberrant biological processes thatoccur spontaneously. Examination of therelationship between incidence and age in

I populations at high and low risk of partic-40 60 80 ular cancers adds support for this con-

tention, if we assume that the cancers inlow-risk populations are more likely to bespontaneous than those in the high-riskpopulation. Some examples of thisapproach are shown in Figure 4. In eachcase, the age-incidence curve in the low-

)_______________ incidence population runs parallel with40 60 80 that of the high-incidence population but

is shifted to the right. This holds true forthe four examples of epithelial cancers(breast, lung, colon, and oral cancer) andfor testicular cancer in which the incidencedeclines after age 30. These comparisonsdo not prove that spontaneous cancershares the same mechanism as cancers

40 60 80 caused by tobacco smoke or oral tobacco orare related to reproductive history or diet;but these data do suggest that such exoge-nous factors enhance or accelerate a biolog-ical process that would occur anyway butat a much lower incidence and at olderages. Thus, if exogenous carcinogens do

m) accelerate spontaneous carcinogenesis, it islikely that when all the preventable cancers

40 60 80 have actually been prevented (an unlikelyeventuality), the remaining cancers willappear at later ages and at much lower rates.

10 20 30 40 50 60 70

Log age, 5-year intervals

Figure 4. Log cancer incidence as a function of log agein high- and low-incidence countries. Data from theInternational Agency for Research on Cancer (7).

in nonsmokers. After adjustment for theeffects of dose (numbers of cigarettessmoked), the slope (4.22) of the lineobtained by plotting log lung cancer inci-dence for smokers as a function of logduration of smoking is remarkably similarto that obtained (4.17) by plotting logincidence of lung cancer as a function oflog age in nonsmokers. Doll interpretedthis as suggesting that "...non-smokers areexposed from birth to an agent that iscapable of causing bronchial carcinoma(albeit a very weak one)...." This similarity

Cancer Is a Genetic DisorderofSomatic CellsIt is now generally accepted that cancer is agenetic disorder of somatic cells and thatmutation of genes, whose products mediatesignal transduction, control the cell cycle,maintain genomic stability, and mediateapoptosis and cellular senescence, is centralto carcinogenesis. The number of muta-tions necessary to establish the full cancerphenotype is still under investigation andmay vary from one type of cancer toanother; this number of mutations is prob-ably at least 2 and, in many types of cancer,not less than 5 or 6, as judged by interpre-tation of the age dependence of varioushuman cancers [reviewed by Lawley (8)].Direct evidence for the serial accumulationof independent mutations during carcino-genesis has been obtained from studies ofthe occurrence of mutant protooncogenes,tumor-suppressor genes, and mismatch-repair genes at successive histopathologicalstages that mark progression from normaltissue to a fully malignant tumor. The

most well-documented and widely quotedexample is colorectal cancer (9,10).

Spontaneous Cancer Arisesfrom Spontaneous MutationAccepting that an accumulation of somaticmutations, however provoked, can initiateand sustain carcinogenesis, there willalways be a background incidence for anygiven cancer because spontaneous muta-tion is inescapable, given the inherentchemical instability of DNA, the prodi-giously large target that it presents tomutagens, the number of times it has toreplicate, and the less-than-perfect defensesthat have evolved to protect its integrity asa self-replicating carrier of information. Itcould be argued that cancer is indeed aninevitable by-product of evolution becausemutation is the driving force of evolution,and without evolution there would be nolife-forms for cancer to happen to. It is rea-sonable to propose, therefore, that sponta-neous cancer arises from spontaneousmutation (oncodeme 1). Other factorscould contribute; for example, a sponta-neous increase in the rate of cell divisionmight increase the probability of conver-sion of a endogenously generated promuta-genic lesion into a mutation. The risk ofsustaining a crucial mutation at a criticaltime in a crucial cell, in the absence of anyexogenous cause or predisposing geneticbackground, will depend to some extent onchance (i.e., bad luck).

The Background Mutation RateofHuman Cells Is Very LowEach cell division of a somatic diploidmammalian cell requires the accurate andtimely distribution of 6 x 109 base pairs ofDNA (2.2 m in length) to each daughtercell. In long-lived species like Homo sapi-ens, this process must operate accuratelyand unremittingly over many decades.Errors can arise if the replication machineselects the wrong nucleotides during poly-merization of a new daughter strand usingthe template provided by the parentalstrand. Such misincorporation can result inmismatches. Several editing mechanismsprevent accumulation of coding errors dur-ing DNA replication, which is remarkablyaccurate, with error frequencies that rangebetween 10-9 and 10-11 per base pair repli-cated (11). The background mutation rateis estimated to be about 1.4 x 10-10 muta-tions per base pair per cell generation, basedon measurement of forward mutations inthe hypoxanthine guanine phosphoribosyl-transferase (hprt) gene in cultured human

Environmental Health Perspectives * Vol 104, Supplement 3 * May 1996 635

S. VENITT

lymphoid cells or electrophoretic proteinvariants at unselected loci (12,13). Loeb(12,13), in considering the role of muta-tion in carcinogenesis, has argued that thisbackground mutation rate cannot accountfor more than two or three mutations(bearing in mind that cancer is a clonal dis-ease originating from a single mutant cell)and cannot explain the numbers of mutantgenes already known to occur in cancers, letalone the larger numbers that may be foundin the future, even if clonal evolution isassumed to increase the selective pressureon mutations that confer a large growthadvantage. He therefore suggested that amutator phenotype that increases the rateof mutation is necessary to explain thenumber of serial mutations required forcarcinogenesis. Such a phenotype wouldresult from loss or impairment of mecha-nisms that prevent mutations arising duringand after DNA replication. If this pheno-type were to arise early in carcinogenesis, itwould explain the genomic instability thatis characteristic of human cancers and thenumber of mutant genes observed in them.Mutator phenotypes and their correspond-ing genotypes have been discovered in thehuman germ line and in human cancers,suggesting indeed that genomic instabilityprovides a powerful drive to carcinogenesis.Examples of such germ-line mutationsinclude those in genes underlying rarerecessively inherited disorders such asataxia telangiectasia (14) and more com-mon, dominantly inherited genes such asMSH2, hMLHI, hPMSJ and hPMS2mismatch-repair genes that predispose tohereditary nonpolyposis colorectal cancer(HNPCC) (15). The role of mutator phe-notypes in spontaneous cancer is unknown,but there is no reason to exclude them fromconsideration in that context.

Endogenous Processes thatLead to MutationLindahl (16) has reviewed the variouschemical and enzymatic processes that may

account for spontaneous mutation. Theseinclude spontaneous depurination anddepyrimidation of DNA in the aqueousmilieu of the cell, mismatch and proofread-ing errors during DNA replication, anddeamination of 5-methylcytosine at CpGdinucleotides to produce C to T base pairsubstitutions (see Table 1).

Deamination of 5-MethylcytosineAcs as an Endogenous MutagenIn about 99% of the mammalian genome,CpG dinucleotides are underrepresentedby about one-quarter of the proportionexpected from the observed frequency of Cand G in the DNA. There is a correspond-ing overabundance of TpG and CpAdinucleotides. This depletion ofCpG dinu-cleotides in mammalian genomes is thoughtto result from the readiness of 5-methylcy-tosine (5mG) to undergo oxidative deami-nation at the exocyclic amino group toform thymine, which, being a normal base,is not readily recognized by repair enzymes;this results in mutations, in particular, the5mC - T transition (CpG -> TpG).Despite the rarity of CpG dinucleotidesand the existence of a G-T mismatch repairsystem, this transition occurs about 10times as frequently as other transitions andis frequently seen in inherited diseases-over 35% of point mutations leading tohuman genetic disease are CpG -> TpGtransitions (11). Thus, 5mCpG sites areparticularly prone to loss by mutation, and5mC can be thought of as an endogenousmutagen. CpG -÷ TpG transitions are alsocommon in the p53 gene in human can-cers. It is estimated that about half of allcancers that occur in the United Kingdomand the United States contain p53 muta-tions (17-19). Mutations, mainly of themissense type (transitions and transver-sions), are scattered over a wide area of thegene but occur most frequently in thoseregions most highly conserved during evo-lution, particularly exons 5 to 8. At leastfour hotspots for mutation can be identified,

Table 1. DNA instability, damage, and defense mechanisms involved in spontaneous mutation.

Event Example of damage Defense

DNA depurination Apurinic site AP endonuclease + error-free repairDeamination of cytosine Replacement of C by U Uracil glycosylase + error-free repairDeamination of 5-methylcytosine Replacement of C by T Mismatch repairto thymine Evolution to deplete genome of CpGs

Alkylation of guanine and adenine 7-Methylguanine, 3MeA-glycosylase + error-free repair3-methyladenine Spontaneous loss of 7MeG

Oxidative damage 8-Hydroxyguanine Specific glycosylase + error-free repair;Antioxidants; catalase, superoxide dismutase

the frequency and distribution of whichvary among different kinds of cancer.Transitions at the CpG dinucleotide occurfrequently in p53 in many cancers andprobably reflect endogenous mutation. Areduction in the proportion of CpG muta-tions at the expense, for example, of anincrease in G -> T transversions may signalthe effect of an exogenous mutagen. Anexample is shown in Figure 5. In the germline and in colorectal cancer, the majorityof p53 mutations are G--A and C ->Ttransitions that occur predominantly atCpG dinucleotides (17,19). This spectrumis very similar to that seen in the germ linein an unrelated gene, hemophilia B, and is

40-

30-20 -

10-

* Transitions * CpG Mutations* Transversions i Others

Hemophilia B

it-

O - -

40 -

30 -

20 -

10 -

0

Eco0

0

cmCDa,a)0I

40-

30 -

20 -

10-

4

31

21

Germ line p53

Breast (femalel

LEo .

o0 Lung

:0 U

10-<< cD 1 *F c u<C. CD > CD E S Xt t t t t t t t t t t t ==- rCD e< C- <- eCc:-< s 1CDI- :t--

Type of mutation

Figure 5. Spectrum of somatic mutations in exons 5 to8 of p53 in the colorectum, female breast, lung, germline, and hemophilia B gene. Abbreviations: dm, doublemutant; del, deletion; ins, insertion; int sp, intronsplice. Redrawn from Biggs (17).

Environmental Health Perspectives * Vol 104, Supplement 3 - May 1996

_- CM

636

SPONTANEOUS CANCER

thought to represent mutations arisingspontaneously. In lung cancer, however,there is a marked reduction in the propor-tion of such mutations, concomitant withan increase in the proportion of G -> Ttransversions, such that these predominate.This is consistent with the action of exoge-nous carcinogens present in tobacco smoke,which cause G -* T transversions. Breastcancer is an interesting case: the spectrumof mutations is intermediate between thatfor colorectal cancer and that for lungcancer. This implies that an environmentalagent or agents could be responsible for aproportion of human breast cancer (17).Given that deamination of 5-methylcyto-sine to thymine appears to be a powerfulsource of endogenous mutation, it is

interesting to note that all 46 CpG sites onboth strands in exons 5 to 8 of the p53gene are methylated and that methylation isindependent of tissue type (20).

Oxidative Damage,Defects in Detoxification,and DNA RepairOther events that could produce sponta-neous mutations include insults to DNAand its replication imposed by products ofmetabolism, the most notable being oxida-tive DNA damage caused by oxygen freeradicals produced by normal aerobicmetabolism (16,21). Deficiencies in thosemechanisms that have evolved to protectcells against genetic accidents may also

contribute to spontaneous mutation. Theseinclude defective DNA excision-repair,low levels of antioxidants and antioxidantenzymes (e.g., superoxide dismutase), lowlevels of nucleophiles (e.g., glutathione)that trap DNA-reactive electrophiles, anddefects in enzymes (e.g., glutathione S-transferase) that conjugate nucleophileswith DNA-damaging electrophiles (16).Suboptimal concentrations of substrates(e.g., S-adenosylmethionine) that areinvolved in biomethylation of DNA mightalso contribute to spontaneous mutation(22). The mechanisms that operate inthese three categories are all under geneticcontrol. Thus, germ-line mutations orpolymorphisms of genes that govern themmay also contribute to spontaneous cancer.

REFERENCES

1. The Shorter Oxford Dictionary. Oxford: Clarendon Press, 1973.2. Doll R, Peto R. The causes of cancer. Quantitative estimates of

avoidable risks of cancer in the United States today. J NatlCancer Inst 66:1191-1308 (1981).

3. Carstairs V, Morris R. Deprivation and Health in Scotland.Aberdeen: Aberdeen University Press, 1991.

4. Knudson AG Jr. Genetic oncodemes and antioncogenes. In:Biochemical and Molecular Epidemiology of Cancer (HarrisCC, ed). New York:Alan R. Liss, 1986:127-134.

5. Digweed M. Human genetic instability syndromes: single genedefects with increased risk of cancer. Toxicol Lett 67:259-281(1993).

6. Doll R. The age distribution of cancer: implications for modelsof carcinogenesis. J Roy Stat Soc A 134:133-166 (1971).

7. IARC. Cancer Incidence in Five Continents, Vol III (WaterhouseJ, Muir C, Correa P, Powell J, eds). Lyon:International Agencyfor Research on Cancer, 1976.

8. Lawley PD. Historical origins of current concepts of carcino-genesis. Adv Cancer Res 65:17-111 (1994).

9. Fearon ER,Vogelstein B. A genetic model for colorectal tumori-genesis. Cell 61:759-767 (1990).

10. Cavanee WK, White RL. The genetic basis of cancer. Sci AmMarch:50-57 (1995).

11. Cooper DN, Krawczak M. Human Gene Mutation. Oxford:Bios Scientific Publishers, 1993.

12. Loeb LA. Mutator phenotype may be required for multistagecarcinogenesis. Cancer Res 51:3075-3079 (1991).

13. Loeb LA. Microsatellite instability: marker of a mutator pheno-type in cancer. Cancer Res 54:5059-5063 (1994).

14. Savitsky K, Bar-Shira A, Gilad S, Rotman G, Ziv Y, VanagaiteL, Tagle DA, Smith S, Uziel Y, Sfez S, Ashkenazi M. A singleataxia telangiectasia gene with a product similar to PI-3 kinase.Science 268:1749-1753 (1995).

15. Eshleman JR, Markowitz SD. Microsatellite instability ininherited and sporadic neoplasms. Curr Opin Oncol 7:83-89(1995).

16. Lindahl T. Instability and decay of the primary structure ofDNA. Nature 362:709-715 (1993).

17. Biggs PJ, Warren W, Venitt S, Stratton MR. Does a genotoxiccarcinogen contribute to human breast cancer? The value ofmutational spectra in unravelling the aetiology of cancer.Mutagenesis 8:275-283 (1993).

18. Levine AJ, Perry ME, Chang A, Silver A, Dittmer D, Wu M,Welsh D. The 1993 Walter Hubert Lecture: the role of the p53tumour-suppressor gene in tumorigenesis. Br J Cancer69:409-416 (1994).

19. Greenblatt MS, Bennett WP, Hollstein M, Harris CC.Mutations in the p53 tumor suppressor gene: clues to cancer eti-ology and molecular pathogenesis. Cancer Res 54:4855-4878(1994).

20. Tornaletti S, Pfeifer GP. Complete and tissue-independentmethylation of CpG sites in the p53 gene: implications formutations in human cancers. Oncogene 10: 1493-1499 (1995).

21. Ames BN, Gold LS, Willett WC. The causes and prevention ofcancer. Proc Natl Acad Sci USA 92:5258-5265 (1995).

22. Shen JC, Rideout WM III, Jones PA. High frequencymutagenesis by a DNA methyltransferase. Cell 71:1073-1080(1992).

Environmental Health Perspectives - Vol 104, Supplement 3 - May 1996 637