mutations in the p53 tumor suppressor gene: clues to ... · the p53 tumor suppressor gene has come...

[CANCERRESEARCH54, 4855-4878, Septemberi5, 1994J

Perspectives in Cancer Research

Mutations in the p53 Tumor Suppressor Gene: Clues to Cancer Etiology

and Molecular Pathogenesist

M. S. Greenblatt, W. P. Bennett, M. Hollstein, and C. C. Harris2Laboratory of Human Carcinogenesis, Division of Cancer Etiology, National Cancer institute, Bethesda, Maryland 20892 (M. S. G., W. P. B., C. C. HI, and Division ofToxicology, German Cancer Research Center, D-69120 Heidelburg, Germany [M. H.]

I. Introduction

The p53 tumor suppressor gene has come to the forefront of cancerresearch because it is commonly mutated in human cancer and thespectrum of p53 mutations in these cancers is providing clues to theetiology and molecular pathogenesis of neoplasia (1â€”3).Detection ofp53 abnormalities may have diagnostic, prognostic, and therapeuticimplications (4).

The 15-year history of p53 investigations is a paradigm in cancerresearch, illustrating the convergence of previously parallel lines ofbasic, clinical, and epidemiological investigation and the rapid transfer of research findings from the laboratory to the clinic. p53 is clearlya component in biochemical pathways central to human carcinogenesis; p53 protein alterations due to missense mutations and loss of p53

protein by nonsense or frameshift mutations provide a selective advantage for clonal expansion of preneoplastic and neoplastic cells (5).The potential for a missense mutation to cause loss of tumor suppressor function and gain of oncogenic activity, i.e., to transform cells bytwo mechanisms, is one explanation for the commonality of p53mutations in human cancer. Recent studies investigating the mechanisms underlying the biological activity of p53 indicate that theprotein is involved in gene transcription, DNA synthesis and repair,genomic plasticity, and programmed cell death (1â€”6).These complexbiochemical processes are performed by multicomponent protein machines; therefore, it is not surprising that the p53 protein formscomplexes with other cellular proteins (Fig. 1) and that some viraloncoproteins alter the functions of these machines by binding to p53and perturbing its interaction with other cellular protein components.

In this Perspective, we will focus on the origin of p.53 mutations,the mutational spectrum of p.53 in human cancers, and the hypothesesgenerated by the analysis of p53 mutations in premalignant andmalignant cells. The interpretation ofp53 mutations in human cancersis based on observations of the patterns of DNA damage induced bychemical and physical mutagens in model systems. In this Introduction, we will review these data, which provide the background formany of the inferences drawn from p53 mutational analysis.

A. Origins of Mutation

Both exogenous carcinogens and endogenous biological processesare known to cause mutations (7â€”9).Classes of DNA damage includedeletion, insertion, and base substitution, either transition (change ofa pyrimidine to another pyrimidine or a purine to another purine) ortransversion (change of a pynmidine to a purine or vice versa;

Received 4/27/94; accepted 7/20/94.Thecostsof publicationof thisarticleweredefrayedin partby the paymentof page

charges. This article must therefore be hereby marked advertisement in accordance with18 U.S.C. Section 1734 solely to indicate this fact.

i Due to space limitations and the breadth of this topic, comprehensive reviews which

are the source of additional citations will be cited frequently.2 To whom requests for reprints should be addressed, at Laboratory of Human Carcino

genesis, Building 37, Room 20)5, National Cancer Institute, Bethesda, MD 20892.

Ref. 10). Important sources of the spontaneous generation of pointmutations in human cells include deamination of cytosine and 5-methylcytosine (1 1), DNA polymerase infidelity (12), depurination (13),and oxidative damage from free radicals generated by biologicalprocesses (14, 15).

The most well-studied endogenous mechanism of DNA damage atthis time is the phenomenon of deamination of 5-methylcytosine.Methylation of DNA is one epigenetic mechanism involved in regulating gene expression. Cytosine and 5-methylcytosine residues canspontaneously deaminate to uracil and thymine, respectively, which ifnot repaired will result in G:Câ€”>A:T transitions. These mutationsoccur most frequently at CpG dinucleotides (a cytosine followed by aguanine), which are frequently methylated (9, 11, 16).

DNA replication in humans is performed by multicomponent protein

machines that include several DNA polymerases, which vary in characteristics such as exonuclease activity and fidelity of replication. Errorrates vary from 112,000 for DNA polymerase-@3,which repairs only shortgaps, to 1/50,000- 1/1,000,000 for other polymerases which contain

exonuclease activity (17). The most common base substitutions observedare G:Câ€”*A:T and A:Tâ€”@G:C transitions and G:Câ€”*T:A transversions.Imbalances in deoxynucleoside triphosphate pools (18), mutations in

DNA polymerase-a (19), and slippage of DNA polymerase at nucleotide

repeats (Ref. 8; section II.B.2) are examples of mechanisms contributingto infidelity of DNA synthesis. Considering the multiplicity of proteinsinvolved in DNA synthesis, DNA repair, mitosis, and the cell cycle, thenumber of potential gene targets which could interfere with replicationfidelity in somatic cells is large.

Abnormalities in some of these processes could result in the mutator phenotype (20, 21), which could increase the probabilities ofboth neoplastic transformation and the generation of increasinglymalignant subclones during tumor progression (22, 23). If suchmutations occur in germ cells and are nonlethal, they could lead toheritable syndromes of increased cancer risk. The hypothesis thatinherited genomic instability may be associated with increased cancerrisk is supported by studies of two diseases. Children with Bloom'ssyndrome have an increased risk of childhood cancers, and their cellsshow an increase in cytogenetic abnormalities (24). Germline mutations in two DNA mismatch repair genes, hMSH2 (25, 26) andhMLHJ (27, 28), recently have been identified as the genetic abnor

malities responsible for HNPCC.3 Mutations of these genes are

3 The abbreviations used are: HNPCC, hereditary nonpolyposis colorectal carcinoma;

aprt, adenine phosphoribosyl transferase; dhfr, dihydrofolate reductase; hprt, hypoxanthine-guanine phosphoribosyl transferase; CHO, Chinese hamster ovary; BPDE, 75,8R-dthydroxy-9R,1OS-epoxy-7,8,9,10-tetrahydrobenzo(a)pyrene; BP, benzo(a)pyrene; IHC,immunohistochemistry; SSCP, single-stranded conformation polymorphism; PCR, polymersac chain reaction; XP-C, xeroderma pigmentosum, complementation group C; PAH, polycyclic aromatic hydrocarbon; AFB, aflatoxin B@;HCC, hepatocellular carcinoma; HBV,hepatitisB virus; SCL@ small cell lung carcinoma;NSCLC, ron-small cell lung carcinomaNPC, nasopharyngeal carcinoma EBV, Epstein-Barr virus; HPV, human papilloma virus;CAT, chloramphemcol acetyltransferase; LFS, Li-Fraumeni Syndrome; MP, mean pairwise.

4855

Research. on January 13, 2020. © 1994 American Association for Cancercancerres.aacrjournals.org Downloaded from

http://cancerres.aacrjournals.org/

U G:CtoC:Gâ€¢G:CtoT:Aâ€¢G:CtoA:Tâ€¢A:TtoT:Aâ€¢A:TtoG:CU A:TtoC:G@@ U-Non-CpGâ€¢CC:TTtoThAA@ Othir

p53 MUTATIONS AND CANCER

flG.1 FIG. 2ATr.lnsdc:' @atic@n[ Mutations and [email protected]$on and

4.â€• Dvrnairi Sequence Specific NucI@r Localization

@ DNA Binding Domain@ Domains110 27314O@120100 1 24â€¢

@ I 235 282@ I

4O@

20 .kli .a

2% 8%

17%

as

17%

BASAL CELL@ MURINE (UV-INDUCED)

2% 7%1O%@ @j1% 13%

2E 2F 2GORAL

10/0 10%

PHARYNX/LARYNX

29%

Fig. 1. Schematic of the p53 molecule. Verticallines, distribution of mutations; horizontal bars, regions of protein binding. The p53 protein consists of 393 amino acids with functionaldomains, evolutionarily conserved domains, and regions designated as mutational hotspots. Functional domains include the transactivation region (amino acids 20-42, orange area),sequence-specific DNA binding region (amino acids 100â€”293),nuclear localization sequence (amino acids 316â€”325,dark green area), and oligomenzation region (amino acids 319â€”360,lightgreen area). Cellular or oncoviral proteins bind to specific areas of the p53 protein. Evolutionarily conserved domains (amino acids 17â€”29,97â€”292,and 324â€”352;rose-colored area) weredetermined using the MACAW program as described in the text. Seven mutational hotspot regions within the large conserved domain are identified: amino acids 130â€”142,151â€”164,171â€”181,193-200, 213-223, 234-258, and 270-286. Functional domains and protein binding sites were compiled from references in text and from X. Wang and C. Harris(unpublished data). Verticallines above the schematic, missense mutations;lines below schematic, nonmissense mutations. The majority of missense mutations are in the conserved hydrophobicmidregion, whilenonmissensemutations(nonsense,frameshift,splicing,and silent mutations)are distributedthroughoutthe protein,determinedprimarilyby sequencecontext, as describedin the text.

Fig. 2. A-G. spectrum of p53 mutations in various tumor types. (n) the numberof mutations for which exact base change or deletion was confirmed by DNA sequencing. A, the mutationalspectrumofp53 in all reportedtumorsand cell lines(n = 2567) in the database.B, squamous(n 32 induding Bowen's disease)and (C)basal cell (n 52)carcinomas ofthe skin in humansand (D) UV light-inducedmurine skin tumors (n = 30) exhibit frequent G:Câ€”Ã·A:Ttransitionsand occasional tandem mutations at dipynmidine sites (usually CC:GGâ€”@1T:AA),bothcharacteristicof UV-induceddamage.Squamouscell carcinomashave been linkedwith occupationalexposureto PAHa,and one-fourthcontainG:C-@T:A transversions,which are consistentwith the mutations caused by these carcinogens in model systems. The mutational spectra differ by site of origin of head and neck squamous carcinomas [(E) nasopharynz n 17; (F) oral,n 69; (G) pharvnxilar,nx. n 42J. 4856

MUTATIONALSPECTRUMOFALLHUMANCANCER

.4.4@V @P $V@ LARGE T-Ag ______

ISBIS H$P70 HSX 5431@ 1.4 EISMK Â£[email protected] IIYNPA NPA@

15@41 MOM2

SKINCANCERS

2B SQUAMOUS CELL

3%

HEADANDNECKCANCERS

NASOPHARYNX6%

LEGENDFORFiGURE2 A-O



LARGECELL2K

p53 MIJFATIONS AND CANCER

FIG.2H-0LUNGCANCERSBYCELLTYPE

LUNGCANCERS

2H ADENOCARCINOMA 21 SQUAMOUS

14%

45%

8%

2J SMALL CELL4%

5%

13%

2M NON-SMOKERS6% 6%

6%/0

11%I@ 3%

8%

COLONSOMATICMUTATIONS LI-FRAUMENIGERMUNEMUTATiONS

2N

8%

5%

15%

Fig. 2. (continued) H-O, spectrumof p53 mutationsin various tumortype&The mutationalspectrain lung cancers differ among histological types [(H) adenocarcinonia n 61; (1) sqisimousn = 75; (J) 5771011cell (SfLC), n = 92; (K) large @dLn 27]. Menocarcinomasdisplaya lowermutationprevalence,a kwer percentageof G:Câ€”@TAtranavetsions,aixi a higherpercentageofG:C-Ã·Altaiultians.The mutational specus in lungeancervaiy byanddnghistccy[(L)nnokers n = 169;(M)noiwnoke,@ n = 161.Themostcommon mutations in(N)coloncarcinoma(n = 324)and (0) LFS (germline p53 mutations, n = 47) are GC-*A@T transitions at @pGdinucIeOtideS,which may be the result of spontaneous deamination of 5-inethylcytosine.

4857

a SMOKERS



Table ISpontaneous and induced mutational spectra: in vitro and anima!modeli'Gene

and Mutagen (n)â€•[Ref.]G:C-@A:TG:C-'T:AG:Câ€”@C:GA:Tâ€”*G:CA:T-@T:AA:Tâ€”@C:GCC:GGâ€”'TT:AADe1C@dCHO

aprt geneSpontaneous (30) [281]UV light (31) [282]Ionizing radiation (16) [39]BPDE (21) [283]71

5219513

66

623

66

140

3607

101390

61900

6006

331

50

30

5Human

aprt geneSpontaneous (74) [43]UV light (55) [43]BPDE (102) [43]MNNG (103)(43]ENU (81) [43]28

4714633112

13691398

7132520

152

10178

917

258

7159NR

NRNRNRNR15

21050

001

0Animal

p53 gene@Mouse skin

UV light (33) [44, 451BP (7) [46]Butadiene (10)eDMBA (18) [46, 48]MCA (7) [284]42

1410222918

714017570

141011140

03011018

00

1700

006021

00000

0101100

006

0Rat

esophagusNMBA (10)[285]90000000100Rat

livercholine-deficient diet (9)

[286]550111 101 101 10


associated with deletions and insertions in microsatellite sequences ofsomatic cells and a genetic predisposition to tumors of the colon andother tissues (29). Somatic mutations in !ZMSH2 have also beenobserved. Therefore, the mutator phenotype can be both a predisposing inherited factor initiating carcinogenesis or a secondary somatic

change produced during this multistage process.A well-studied mechanism by which chemical carcinogens and

their ultimate carcinogenic metabolites cause mutations is by formingcovalent adducts with the nucleotides in DNA, increasing the probability of errors during DNA replication (30). Some small carcinogenDNA adducts, such as 06-methylguanine resulting from alkylatingagents, may cause DNA polymerase to misread the base pairing dueto the altered hydrogen bonding properties of a base which contains anadditional methyl or ethyl group. Laboratory studies have establishedthat the most common mutations caused by alkylating agents areG:Câ€”+A:T transitions, consistent with 06-methylguanine mispairing

with thymine (31). Bulky carcinogen-DNA adducts may render thebases unreadable and stall the replication machinery. DNA polymerases may fill these noninstructive sites preferentially with adenine.If this occurs opposite to a guanine (a common target for bulkycarcinogens), a thymine will pair with the adenine in the next roundof DNA synthesis, resulting in a G:Câ€”*T:A transversion (32). Studiesusing prokaryotic and eukaryotic cells and site-specific mutagenesisassays (13, 33) have shown that some carcinogenic agents produce a

â€œfingerprintâ€•(1, 34, 35); specific types and locations of DNA adductshave been linked with the mutational spectrum of an agent. Thesepatterns have been studied in eukaryotes using both exogenous genesintroduced into the appropriate host cell by a shuttle vector (33, 36)and endogenous genetic loci (34) such as apri, dhfr, and hprt.

B. Mutations in Model Systems

The carcinogen-induced mutational spectra in prokaryotes and intarget genes in eukaryotes can be compared to spontaneous mutationalspectra, although uncertainty exists in the extrapolation of in vitro datato human carcinogenesis. Confounding variables which could affect

experimental systems include differences in the genes and cell typesstudied, interspecies variability of DNA repair, and the potentialmutagenicity of transfection and shuttle vectors. The interpretation ofmutational spectra is complicated, and observed mutations may notreflect all events at the DNA level. The DNA sequence can influenceboth carcinogen-DNA adduct formation and DNA repair, and somemutations may be either more readily detected due to clonal expansionor less readily detected due to inhibition of the mutant cell population.

1. Mutational Spectra in Housekeeping Genes

The majority of spontaneous mutations in the model systems usedare single base substitutions. Studies in prokaryotes indicate thatcarcinogen-induced mutations occur at nucleotides which are targetsfor carcinogen-DNA interactions and alterations (37). Differenceshave been observed between the spontaneous mutational spectraof the endogenous aprt gene of CHO cells and both bacterial genesand exogenous genes carried by shuttle vectors in mammalian cells(38). This may reflect either cell-specific or gene-specific biologicalproperties.

A comparison of the spontaneous mutational spectrum at the aprt

locus in CHO cells with the spectra induced by UV radiation, ionizingradiation, or BPDE illustrates carcinogen-specific damage (Table 1).The mutational spectrum of UV light is consistent with the mutagenesis models of the promutagenic cyclobutane and pyrimidine-pyrimidone (6â€”4)photoproducts. Ionizing radiation often causes chromosomal abnormalities, deletions, and losses (39, 40). BPDE, whichprimarily binds to the 2-amino group of deoxyguanine, producespredominantly the G:Câ€”*T:A transversions predicted of a bulky carcinogen-DNA adduct; a similar mutational spectrum for BPDE hasbeen observed in human cell assays (41). The hprt locus allowscomparison of mutational spectra obtained by exposing cultured cellsto carcinogens with those observed in human donor lymphocytesexposed in vivo to environmental carcinogens (42). A recent comprehensive database has been developed which contains more than

a Values, percentagesof each class of mutation;MNNG, N-methyl-N'-nitro-N-nitrosoguanidine;ENU, N-ethyl-N'-nitrosourea;DMBA, dimethylbenzanthracene;MCA, methylcholanthrene; NMBA, nitrosomethylbenzylamine; CHO< Chinese Hamster Ovary; Ref., reference no; NR, not reported.

h Number of mutations sequenced.

C Deletions.

d Insertions.

e R. Wiseman, personalcommunication.I Examples of mutation prevalence may be found in Refs. 287 and 288.

4858




1000 human hprt mutations (43). The mutational spectra found atthe aprt locus of CHO cells and the hprt locus of human cells aredetailed in Table 1. The most striking similarities are the highpercentages of G:Câ€”'T:A transversions at both loci in cells exposed to BPDE and high percentages of G:Câ€”*A:T transitions inUV-exposed cells, consistent with models predicted for theseagents. The largest difference between the two systems is in thespontaneous mutational spectrum. In the aprt locus, G:Câ€”A:Ttransitions account for 71% of the 30 spontaneous alterations,whereas in the hprt locus, only 28% of the 74 mutations areG:Câ€”t@A:T (P = 0.001, Fisher's exact test). Thus, cell-specific or

gene-specific features may affect spontaneous mutation patterns.

2. Mutational Spectra in Animal Models

Animal models provide opportunities for mutational spectrum studies that cannot be performed with human tumors. For example, doseresponse experiments can define the mutational spectra of carcinogensin vertebrate neoplasms. The spectrum of carcinogen-induced mutations in cancer-related and housekeeping genes can be compared inrodent and human cells, assessing the relevance of these modelsystems to human carcinogenesis. Furthermore, transgenic animals

can provide mechanistic insights into the interplay between environmental carcinogens and host susceptibility factors such as carcinogenmetabolism and alterations of oncogenes and tumor suppressor genes.

The frequency of p53 alterations in murine primary tumors isgenerally lower than in human tumors. p53 mutations have been foundin tumors of the breast (42%), lung (8%), skin (30%), and other tissues(Table 1). Alterations generally have not been found in colon or livertumors. The low prevalences in lung and colon tumors are surprisingfor several reasons: (a) the gene is highly conserved between miceand humans; (b) p.53 mutation frequency is high in human lung andcolon cancers; (c) the morphological features of lung and colonadenocarcinomas are similar in humans and mice; and (d) lung andcolon tumors from both species frequently contain ras mutations. Themutation database in laboratory animals is still limited (Table 1), andmore studies of different carcinogens and dose regimens are needed.The most well-studied system is skin carcinogenesis in the mouse.Several carcinogens have been examined in rat, hamster, and nonhuman primate models, but mutation rates are generally too low togenerate meaningful spectra.

Uv exposure efficiently induces squamous skin carcinomas inseveral strains of inbred mice (44, 45). In these studies, as in humantumors, there is a striking preference for single base transitions andtandem mutations at dipyrimidine sites (30 of 33; 91% of mutations).BP, a carcinogen found in tobacco smoke and burning fossil fuels,

induces papillomas and squamous carcinomas in Sencar mice whenapplied topically. p53 mutations were identified in 1 of 8 papillomas and 6 of 12 carcinomas (46) with 5 G:Câ€”@T:A transversions,1 G:Câ€”t@C:G transversion, and 1 G:Câ€”@.A:T transition (Table 1).These data provide the first characterization of the p53 mutationspectrum of BP-induced tumors in vertebrates. Skin tumors induced by 7,12-dimethylbenz(a)anthracene provide a contrastingmutation spectrum. Eighteen p53 mutations have been reported(46â€”48), and the mutation pattern is notable for a high frequency(40%) of mutations at A:T base pairs and a low frequency ofG:Câ€”â€•T:A transversions. This is consistent with the mutagenic

specificity of 7,12-dimethylbenz(a)anthracene (48).These mutational spectra of aprt, dhfr, and hprt genes and animal

models serve as a foundation with which to compare the spectraobserved in cancer-related genes in humans. Differences in spontaneous mutation patterns may reflect differences in inherent characteristics of various cell types (e.g., DNA repair capacities), whereas

induced mutation patterns that override the spontaneous profiles aremore likely to reflect mutagen-DNA interactions.

II. p53 Mutational Spectrum Analysis in Human Cancers

One of the goals of carcinogenesis research is to identify the precisemolecular alterations responsible for neoplastic transformation, abnormal differentiation, and growth control. Molecular epidemiologyhas developed to attempt to integrate traditional epidemiologicalinvestigation of cancer risk factors with the substantial expansion ofknowledge of the molecular mechanisms of cellular processes (49).

Mutational spectrum analysis, the study of the types and locationsof DNA alterations, describes the often characteristic patterns of DNAchanges caused by endogenous and exogenous mutagens. Alterationsof cancer-related genes found in tumors not only represent the interactions of carcinogens with DNA and cellular DNA repair processesbut also reflect the selection of those mutations which provide premalignant and malignant cells with a growth advantage. Study of thefrequency, timing, and mutational spectra of abnormalities of p53 andother cancer-related genes can thus generate and test a variety ofhypotheses in carcinogenesis. These include questions regarding carcinogen-DNA interactions, functions of the affected gene products,mechanisms of carcinogenesis in specific organs or tissues, and features of general cell biology such as DNA replication and repair.

The information derived from this type of mutational spectrumanalysis is most relevant when it addresses alterations of genes vital tocarcinogenesis. A burgeoning literature (reviewed in Refs. 1, 2, 5, and6) attests to the important role of p53 protein in normal cell functionand neoplastic transformation. Mutations of p53 are the most commongenetic abnormality yet found in human cancers. The prevalence ofp53 mutations varies among tumor types, ranging from 0 to 60% in

major cancers (Table 2), and is over 80% in some histological subtypes. All classes of mutations occur in the p53 gene (Fig. 2A);three-fourths of substitutions occur at G:C base pairs.

The p53 gene is well-suited to mutational spectrum analysis forseveral reasons. First, since p53 mutations are common in manyhuman cancers, a sizable database has accrued whose analysis canyield statistically valid conclusions (50). Its modest size (1 1 exons,393 amino acids) permits study of the entire coding region, and it ishighly conserved in vertebrates, allowing extrapolation of data fromanimal models (51). Point mutations which alter p53 function aredistributed over a large region of the molecule, especially in thehydrophobic midportion (1), where many base substitutions alter p53conformation and sequence-specific transactivation activity (sec.II.C.1.); thus, correlations between distinct mutants and functional

changes are possible. Frameshift and nonsense mutations which truncate the protein can be located outside of these regions, so evaluationof the entire DNA sequence yields relevant data. This situation differsfrom that of the ras oncogenes. Transforming mutations of ras occurprimarily in three codons and are thus limited to a few sequencemotifs which identify a critical functional domain (52). The diversityof p53 mutational events permits more extensive inferences regardingmechanisms of DNA damage and mutation.

For this Perspective, we compiled 2567 mutations in human carcinomas or cell lines from over 300 papers published or in press throughJanuary 1, 1994, as identified by searches of Medline and CurrentContents in January, 1994. Only mutations confirmed by DNAsequencing were entered into the database. Selected additional reportswere added to illuminate specific discussions. An updated version ofthe database is available from the European Molecular Biology Laboratory Data Library through e-mail; details of its compilationhave been reported separately (50). Here we analyze these data withattention to: (a) questions regarding carcinogenesis which have been

4859



Table 2 Preyalence and spectra ofp53 mutations in humancancers'@Tumor

type

(n)â€•Prey.p.53

Mut.c %G:C-a A:T %G:Câ€”aT:A %G:C-+ C:G %A:Tâ€”'G:C %A:Tâ€”*T:A %A:T-@ C:G %Del/Ins/ OtherCpG W HOTSPOTS1

37 41 17 8 11 6 4 13

4860

p53 MLflA11ONS AND CANCER

All tumors 24 â€¢175lb H(2567) â€¢245G> S,D,C,v

â€¢248lb W,Q249 R> S,Mâ€¢273lb H,C

Lung 56 24 40 9 7 5 3 11 9 157V> F(897) â€¢24@lb W,L,Q

249 R> M,S*273 R> L,H,C

Colon 50 63 9 3 11 4 1 8 47 â€˜175@>H(960) â€¢245G> S,D,V

â€¢248lb W,Qp273 lb H,Cp282 lb W

Esophagus 45 38 16 3 15 14 3 12 18 None(279)

Ovary 44 42 9 12 18 3 6 9 23 @273@>C,H(386)

Pancreas 44 41 13 6 17 8 5 10 21 @273@>H(170)

Skin 44 41 12 6 4 4 2 31 17 @248R>W(220) 278 P> S,F

Gastric 41 47 6 3 17 5 6 16 35 None(314)

Headandneck 37 31 18 11 14 5 3 19 13 @@4flbQWJ@(524)

Bladder 34 37 14 21 9 3 2 14 16 280R>T(308)

Sarcoma 31 47 14 6 7 6 1 19 21 None(339)

Prostate 30 44 13 6 16 0 6 16 28 None

(87)Hepatocellular 29 20 37 8 11 12 3 9 9 249 lb S,M,W

(716)Brain 25 50 11 5 14 1 3 15 30 *175@>H

(456) â€¢248R> Q,WAdrenal 23

(31)Breast 22 36 13 8 11 7 6 16 23 @175R>H

(1536) â€¢248R> Q,W*273 R> H,C

Endometrium 22 51 10 4 14 0 6 14 37 â€¢24@B.> W,Q(224)

Mesothelioma 22

(23)Renal 19

(102)Thyroid 13 48 9 13 9 0 4 17 28 *248R>Q,G,W(299) 273 lb.H

Hematological 12 49 7 7 12 4 5 15 31 â€¢175P.> H(1916) â€¢24@lb. Q,W

Carcinoid 11(61)

Melanoma 9(70)

Parathyroid 8

(13)Cervix 7 35 35 10 0 0 10 10 25 â€˜273@>C(350)

Neuroblastoma 1(212)

Wilms 0

(41)Testes 0(40)

Pituitary 0(27)

Pheochromocytoma 0

(47)

a Mutationalspectrum of a tumor type was calculatedonly if 20 mutationswere reported.b n, the number of tumors of each cell type evaluated for p53 mutation by PCR-based techniques as compiled in the p53 mutation database (50).

C Prevalence of p53 mutations in cancers from each organ/tissue, as detected by PCR-based techniques, screening at least exons 5â€”8. Premalignant lesions such as adenomas or

dysplastic lesions are not included in prevalence data. Prevalence in all cancers is a crude estimate derived from statistics on worldwide cancer incidence and the prevalence of p53mutations for each type as calculated from the reports in the mutation database (see text).

d Del, deletion; Ins, insertions.

e@ of mutationswhich are G:Câ€”@A:T transitionsat CpG dinucleotides.I Hotspots for the entire database are defined as codons at which at least 50 mutations have been reported. Hotspots for individual tumor types include codons at which at least 10

mutations, or 10% of all mutations for that tumor, have been reported. Letters refer to amino acid change, using standard one letter code. Amino acid substitutions which occur morethan once are listed in descending order of frequency at each codon.

*, CpG site.



p53 MIThATIONS AND CANCER

addressed by mutation analysis, and conclusions reached to date; (b)how mutational spectrum analysis and in vitro experiments providecomplementary information on carcinogenesis; and (c) new hypotheses generated from the accumulated information to be specificallyaddressed by future studies of p53 mutations.

A. Analytical/Technical Considerations

1. Methods ofAnalysis for p53 Abnormalities

The methods of analysis for mutations can bias conclusions regarding the prevalence and nature of p53 mutations. A few caveatsconcerning techniques and experimental design are warranted beforeexamining the data in detail.

Solid tumors are heterogeneous, consisting of tumor cells mixedwith nonneoplastic stromal cells and necrotic debris. Since tumorsundergo clonal evolution, different sites within a malignancy may alsovary in genotype and phenotype (21); a false negative result might

arise from sampling error. Microscopic assessment is important toensure that tissue samples contain adequate amounts of the desiredhistological features and to avoid analysis of necrotic debris. Samplesmay be enriched for tumor content by several methods, such asmanual microdissection of slides (removing only the tumor tissue forstudy) and cell sorting using DNA flow cytometry (53, 54).

DNA sequencing is required for precise identification of mutations,

but an altered p53 gene or protein can be detected by other immunological or chemical tests (55â€”58).Because DNA sequencing is timeand labor intensive, studies evaluating human cancers for p53 mutations often use screening tests, performing DNA sequencing only ontumors which are likely to contain mutations. The sensitivity andspecificity of these techniques are critical factors which could introduce bias into p53 mutation prevalence data.

The wild-type p53 protein usually resides in the cell nucleus, buthas a very short half-life and is present in such small quantities that itcannot be detected by routine IHC. Missense mutations often increasethe half-life and quantity of the p53 protein, allowing its detection byIHC (58). Many investigators have used IHC as a screening test priorto DNA sequencing, and staining is often considered a surrogatemarker for gene mutation even in the absence of confirmatory DNAstudies. In the 84 studies we identified which report IHC and sequencing in the same tumor sets, positive staining was found in 44% oftumors by IHC, while 36% contained mutations. However, the sensitivity of IHC in these studies was only 75% (range, 36â€”100%),andthe positive predictive value was only 63% (range, 8â€”100%), withconsiderable variation among tumor types. Thus, the status of the p53protein by IHC should not be equated with wild-type or mutantgenotype.

Discrepancies between IHC and sequencing results may be due tofalse negative and positive results from technical aspects of samplepreparation and analysis or the result of biological phenomena responsible for the discordance (reviewed in Ref. 58). Mutations whichresult in deletion or truncation of the protein (nonsense and frameshift) do not cause protein accumulation. Therefore, IHC will be lesssensitive in detecting mutations in tumor types with high proportionsof these nonmissense mutations. Some tumors, including melanoma(59, 60) and testicular carcinoma (61, 62), are frequently immunopositive for p53 in the absence of mutation. Potential mechanisms forpositive IHC due to accumulation and staining of nonmutant p53protein include stabilization of p53 protein by binding to viral orcellular proteins and DNA damage from chemical or physical genotoxic agents (58). Another hypothesis is that mutations are occurringin a gene downstream of p53, such as WAF1/Cipi, a Mr 21,000protein which is transactivated by the p53 protein and is an inhibitorof cyclin-dependent kinases (63, 64). Interruption of the p53 pathway

could interfere with a feedback loop which regulates expression ofwild-type p53. This can be tested by searching for mutations in genesdownstream in the p53 biochemical pathway in the cancers in whichstaining is frequently detected but which rarely contain p53 mutations.

Other methods of screening tumors to identify those likely tocontain mutations involve detection of altered electrophoretic mobility patterns of DNA sequences containing point mutations. The mostcommon techniques, SSCP (65â€”67),denaturant gradient gel electrophoresis, and constant denaturant gel electrophoresis (68, 69), have asensitivity and specificity of approximately 90% in detecting mutations confirmed by DNA sequencing (67, 69, 70). Sequencing shouldbe done to confirm that positive results represent mutations and nottechnical artifacts or germline polymorphisms, which have beendescribed at codons 72 and 213 (71, 72).

PCR-based techniques, including SSCP, constant denaturant gelelectrophoresis, and DNA sequencing, are subject to false negativeand false positive results. The polymerases used in PCR to amplifyspecific gene sequences are subject to error rates from 1/10,000 tomore than 1/500 base pairs, depending on reaction conditions (73).Large deletions might not be detected if the missing segment includesprimer sites, and the wild-type gene of stromal tissue is amplified.Contamination of a PCR reaction with other airborne PCR productsmay also lead to false positive or negative sequencing results (74).Therefore, it is imperative that all mutations found by direct sequencing of PCR products be confirmed by sequencing of a second, mdcpendent PCR product. This practice has not been followed uniformly;hence, some mutations in the database may represent false positives.There also is considerable inconsistency among papers in the extent ofanalysis. Mutation prevalence data are included in this report only ifa minimum of exons 5â€”8was sequenced. Due to underdetection,mutation prevalence rates are likely underestimated in the database,although reported rates for some tumors were probably inflated due topreselection of samples by screening tests. Thus, absolute mutationprevalence rates are difficult to assess, given the heterogeneity oftechniques in cited reports.

2. Extent ofp53 Gene Analysis

The p53 gene consists of 11 exons; exons 2â€”11 code for the proteinof 393 amino acids. Early studies noted that most p53 mutationsoccurred in the regions of the gene which are highly conservedthrough evolution and presumably of functional importance, primarilyin exons 5â€”8(codons 126â€”306) (55). Based on these preliminarydata, most investigators have confined their analyses to exons 5â€”8(and sometimes exon 4). Partly because of this bias in searching, 95%of the reported mutations have been found in exons 5â€”8and theirintervening introns (mutations in introns which affect splicing of thenext exon have been classified as mutations of the deleted exon). Wehave identified 50 studies in which sequencing of the entire codingregion of p53 was reported. Of the 560 mutations reported in thesepapers, 87% were in exons 5â€”8,and most of the others were in exons4 (8%) and 10 (4%). The distribution varied among tumor types.Mutations outside exons 5â€”8were most frequent in bladder (28%)and hepatocellular (24%) carcinomas and least frequent in head andneck and hematopoietic malignancies (2% each). Over 70% of thesemutations generated stop codons or frameshifts; therefore, they likelywould be missed by IHC screening.

Thus, evaluation of only exons 5â€”8is likely to underestimate theprevalence of p53 mutations, especially nonmissense mutations, byover 20% in some tumors. Analysis of exons 4 and 10 is thereforewarranted when mutations are not found in exons 5â€”8.The tissuespecific differences in the prevalence of mutations in various exons

4861




are clues that detailed p.53 mutational spectrum analysis can reveal

subtle patterns of genetic damage.Another potential source of underestimating the prevalence of p.53

mutations might be the presence of mutations in unevaluated DNAsequences, such as its 2 promoters or its 10 introns. Although theintron bases adjacent to exons, which are responsible for splicing, areoften examined, complete evaluation of introns is not common. Reverse transcription-PCR, a technique using reverse transcriptase toamplify mRNA and sequence the resulting complementary DNA,cannot evaluate introns, which have been removed by splicing. Mutations in intronic regulatory regions may be prevalent and undetected byscreening with IHC (if protein does not accumulate) and PCR-basedtechniques (if primers do not encompass the relevant intron sequence).

3. Epidemiokgkal Considerations

Specific features of populations evaluated for p53 mutations mayalso create bias in the conclusions. For some cancer types, differencesin frequency or spectrum have been detected between groups which

differ in ethnicity or nationality, possibly due to specific carcinogenexposures or inherited features in these populations (sec. II.D.3.). Ifatypical subgroups are overrepresented, features of the mutationalspectrum of other groups may be obscured.

In order to yield valid conclusions, molecular epidemiology mustadhere to standard epidemiological principles of statistical analysis:i.e., selection of appropriate control and study groups, assessment forpotential confounding variables (which may be different for eachtumor type), and adequate sample sizes to minimize type I errors (e.g.,the premature assignment of relationships between putative carcinogens and cancers) and provide the statistical power to detect importantdifferences and avoid type II errors (the erroneous conclusion that nodifference exists). Although these admonitions may seem elemental,the literature contains many examples of preliminary conclusionsbased on small sample sizes, which often pass into general opinionwithout sufficient critique. The database provides a powerful tool tohighlight the areas in which data are inadequate and focus future endeavors.

B. Patterns of Mutation: Insights into DNA Replicationand Repair

1. DNA Repair and Strand Bias of Mutations

Since all organisms are constantly exposed to exogenous and endogenous mutagens, sophisticated systems of DNA repair to maintainthe integrity of the genome have evolved. DNA repair may be definedas a cellular response which restores the normal nucleotide sequenceand stereochemistry of DNA following damage. DNA damage oftencauses a physical block to transcription; therefore, repair is generallyfaster in actively transcribed genes and is closely linked to othercellular pathways which control gene expression (75). Several recently described genes code for enzymes exhibiting both DNA repairand gene transcription activities, suggesting a potential molecularmechanism for control of cellular proliferation until damage is corrected (76, 77). In addition, the p53 gene is repaired faster than thedhfr gene, and much more efficiently than the inactive genomicregions (78). In vitro studies and observations in repair-deficient organisms indicate that repair is important in determining mutation patterns.

Importantly, repair is faster on the transcribed (template) strand ofDNA than on the nontranscribed (coding) strand (75). Experiments inbacteria, yeast, and mammalian cells and in vivo mouse skin carcinogenesis models have shown that mutations induced by exogenouscarcinogens preferentially occur on the nontranscribed strand of genessuch as dhfr, hprt (79), and p53 (46, 75). That is, the mutations

induced on the nontranscribed (coding) strand by UV light or BP aremuch more likely to be CC-+ U than GGâ€”AA and Gâ€”@T thanCâ€”@A, respectively. This difference is thought to be an evolutionary

adaptation which ensures that structure and function of essentialproteins with short half-lives are maintained, increasing the likelihoodof cell survival. Repair of the nontranscribed strand can occur laterand still preserve the fidelity ofthe gene in the daughter cells. If repairof the nontranscribed strand does not take place, then daughter cellsmight arise with mutations that are lethal or that confer a survivaladvantage on the mutant clone.

The hypothesis that deficient DNA repair can affect mutationpatterns in humans was tested by measuring the repair of UV-inducedcyclobutane pyrimidine dimers in the individual DNA strands of p53in a normal, repair-proficient human fibroblast strain and in fibroblasts from a patient with the repair-deficient disorder XP-C (78).Selective repair of the transcribed DNA strand of p53 is observed inboth normal and XP-C fibroblasts, and the strand bias of repair ismore pronounced in XP-C cells. Since mutations may occur due toreplication errors at the sites of unrepaired DNA damage, these resultspredict that the majority of p.53 mutations in skin cancers, especiallythose from patients with XP-C, would occur on the nontranscribedstrand. Indeed, 100% of tandem mutations in skin cancers from XP-Cpatients and four of seven from non-XP patients exhibit such a strandbias (80, 81). More data in non-XP cancers are needed. Tumors fromXP-Cpatientsare muchmorelikelythan sporadiccancersto containtandem pyrimidine transitions (58 versus 11%), suggesting a differential ability to repair this type of DNA damage. This sequence of studiesis an example of testing in the laboratory of a hypothesis generated by

analysis of the p53 mutation spectrum and then predicting the result offurther molecular epidemiological investigation of human tumors.

Some tumor types have been linked by epidemiological and laboratory evidence to specific carcinogens with known mutational spectra, such as lung cancer and exposure to cigarette smoke (sectionII.D.5.C). Ninety-one percent of G:Câ€”+T:A transversions in lung

cancer occur on the nontranscribed DNA strand (i.e., are Gâ€”+T ratherthan Câ€”+A), supporting the hypothesis that DNA repair is an important determinant of mutation specificity of cancer-related genes inhuman cancer. The hypothesis may be extended to postulate that anontranscribed strand bias in the mutation pattern of a specific basesubstitution or tumor type implies that an exogenous carcinogen maybe responsible. Evaluating the entire p53 database for strand bias maytherefore provide clues to suspected and unrecognized carcinogens.

Four patterns of base substitution exhibit strand bias: Gâ€”+T (87%on the nontranscribed strand) and Tâ€”*G (67%) transversions, andAâ€”pG (70%) and nonCpG site G-+ A (65%) transitions (Table 3).G:Câ€”*A:T transitions occurring at CpG dinucleotides are reportedseparately from other G:Câ€”@A:T transitions because of differences inmechanism discussed in section I. Bulky carcinogens such as BP,other PAHs (82), and AFB (82â€”84)have been associated with transitions and transversions at both G:C and A:T base pairs (85). Therelative contributions of these and other agents to human somaticmutations are unknown.

In support of the hypothesis that tumors which are associated withenvironmental carcinogens should contain a high proportion of basesubstitutions with bias toward the nontranscribed strand, Table 3 listsall tumor types in descending frequency of the prevalence of the basesubstitution types which exhibit strand bias (Gâ€”*T, non-CpG siteGâ€”+A, A-t' G, and Tâ€”t'G). Among the tumors which most frequentlycontain these types of mutations are tobacco-associated cancers of thelung (most frequent), cervix, esophagus, head and neck (fourth, fifth,and sixth most frequent, respectively), and hepatocellular carcinoma(second most frequent), which is associated with dietary AFB exposure in certain geographic regions (section II.D.5.B). Most of these

4862



Table 3DNA strandbias by tumor%

G-+ATumorTotal% of% Gâ€”TonNon-CpG on% A-@G%Tâ€”Gtype

(n)Â°Muts

at strandbias siteâ€•% Gâ€”'D%

Gâ€”aAnon-CpG% Aâ€”@G%Tâ€”*Gnontran strandâ€•nontran strandon

nontranstrandon

nontranstrandAlltumors50171711487657067(2567)Lung6640157395529164(328)HCC61381011389777571(224)Pancreas56132117510069730(63)Cervix55351001086100NAe50(20)Esophagus531619153895576100(115)Headand52181814385787550neck(148)Brain50122014350715033(104)Ovary5091717686747980(159)Prostate50131616621004040100(32)Sarcoma48142671677383100(85)Breast46131611687667277(223)Bladder4514219018681780(99)Endometrial4310121461001786100(49)Hematological4371912688866962(236)

Skin401222527742'@20100(110)Gastric4061117671693857(121)Thyroid39917941003810050(46)Colon36915II268646650(324)


a ,@ number of mutations detected in benign and malignant tumors at each site. These numbers differ from values obtained from Table 2 by multiplying the total numbertested X prevalence rates because: (a) some tumors contain multiple mutations; (b) mutations in nonmalignant precursors are included here but not in Table 2 prevalence data; and(c) only sequenced mutations are included here, whereas mutations determined by other techniques may be included in prevalence statistics.

b Percentage of p53 mutations that are one of the indicated base substitutions associated with strand bias.

C Percentage of p53 mutations of indicated base substitution on both strands.

d Percentage of indicated base substitution on nontranscribed strand.

e NA, not applicable.I@ light-induced transitions at tandem pyrimidine sites are a cause of Câ€”*T transitions with nontranscnbed strand bias in experimental models. Therefore, in skin cancer, Câ€”'T

transitions are considered to be carcinogen-induced nontranscribed strand mutations.

substitutions exhibit marked strand bias. When strand bias is notapparent, it may reflect a mechanism of mutation which does notexhibit strand bias [such as non-bulky DNA adducts of alkylatingagents, which cause mainly G:Câ€”A:T transitions due to mispairing(31, 86)]. Alternatively, multiple carcinogens could be involved, andthe bias patterns could negate each other (e.g., in esophageal and lungcancer, one tobacco-associated carcinogen might cause Gâ€”A andanother, Câ€”pT transitions). In vitro studies which establish the mutational spectrum of a mutagen can be complementary to analyses ofclinical specimens for strand bias in base substitutions and can implicate potential tumor-specific carcinogens (section II.D.5). In thelower half of Table 3 (i.e., tumors which exhibit few mutations withstrand bias) are cancers which have not been strongly associated withenvironmental carcinogens (87), supporting the nontranscribed strandbias-exogenous carcinogen hypothesis.

2. Deletions and Insertions: DNA Replication and the Mispairing

Hypothesis

Previous analyses of short deletions and insertions in p53 (88) andother genes associated with genetic disease (89) have concluded that

the DNA sequence context is the most important factor determiningthese events. Almost all short deletions and insertions occur at monotonic runs of two or more identical bases or at repeats of 2- to 8-basepair DNA motifs, either in tandem or separated by a short interveningsequence. Several mechanisms are probably involved (90). The mostwell-studied is called slipped mispairing, a misalignment of the template DNA strands during replication that leads to either deletion, ifthe nucleotides excluded from pairing are on the template strand, orinsertion, if they are on the primer strand. When direct repeat sequences mispair with a complementary motif nearby, the interveningoligonucleotide sequence may form a loop between the two repeatmotifs and be deleted (88, 89). More lengthy runs and sequencerepeats are more likely to generate frameshift mutations. In vitroresearch which detects errors in replication of reporter genes hashelped quantify this phenomenon (reviewed in Ref. 17).

The database contains 285 deletions or insertions of 30 base pairs,including 98 single base pair deletions, 40 single base pair insertions,and 217 total events of @6base pairs. Almost all (265; 90%) are atmonotonic runs and short direct repeats, supporting the prior observations that mispairing is the primary mechanism of short deletions

4863



p53 MUTATIONSAND CANCER

and insertions in human genes. Both deletions and insertions increasein frequency as the number of repeated bases in a run increases. Theregion of p53 which most frequently contains deletions or insertionsis codons 151â€”159.Twenty-seven (9% of all deletions and insertionsand 1% of all p53 mutations) have been reported at this G:C-richsequence with multiple runs and direct repeats, confirming in vitroobservations regarding sequence motifs which predispose to mispairing and other mechanisms of frameshift mutations.

Much remains unknown regarding the cellular phenomena whichinfluence the risk of mispairing. Factors which affect DNA polymerase function, e.g., binding to polymerase accessory proteins, are areasof interest for mechanistic study (89). Some bulky chemical mutagensinduce frameshift mutations in prokaryotic models through mechanisms related to adduct-induced deformation of the DNA helix. Theseadduct-induced mutations occur more frequently in monotonic ortandem base runs and at G:C-rich sequences (82, 90, 91). The contribution of chemical carcinogens to frameshift mutations of p53 ineukaryotic cells is potentially significant. Tumor-related differences indeletion prevalence may reflect organ-specific carcinogens or tissuespecific properties that regulate replication fidelity.

It is uncertain whether the recently described human mismatch

repair genes hMSH2 and hMLHJ (25â€”28),which are responsible forcausing mutations at tandem repeat DNA sequences (microsateilites)in families with FINPCC (section I.A; Refs. 92 and 93), play a role in

p53 deletions and insertions. Most events in colon and other cancersoccur at consecutive base runs, but 40% are at short tandem repeats.None of the seven deletions at the most mutable tandem repeatsequence in the p53 gene (GTG GTG GTG at codons 216â€”218)arefound in colon cancers. Future studies might include evaluation ofHNPCC tumors for slippage mutations at repeat sequences, assessmont of the frequency of microsatellite changes in cancers whichexhibit frequent deletions and insertions, and evaluation of mutationsin hMLHJ, hMSH2, and other mismatch repair genes.

C. Determinationof Functionsand Propertiesofp53

Mutational spectrum analysis complements in vitro studies of therelationship between p53 structure and function and suggests manyexperiments which can test the hypothesis that some mutants of p53alter its function and confer upon cells a growth advantage. In othertumor suppressor genes (e.g., RB and APC), the most common abnormalities found are deletions or nonsense mutations which eliminate protein expression, indicating that loss of their function promotestumorigenesis (94, 95). The presence of mutated p53 protein in tumorssuggests that cells acquire a selective growth advantage from not onlyloss but also missense alteration of p53, frequently accompanied byloss or mutation of the second allele. Analysis of all p53 mutations inhuman cancers reveals a nonrandom distribution within the p53 molecule, implying that specific mutant proteins possess different growthstimulatory properties (Fig. 1). Identification of these locations hasfocused research of p53 structure and function toward specific regionsof the protein, such as exons 5â€”8.What is the functional importance ofthese regions? Are mutations more frequent here because of the gain andloss of specific functions? Are the types of p53 mutations seen in otherregions of the molecule different than those in the conserved regions?

1. Structure-Function Relationships ofps3 Protein

Homologies have been noted between p53 domains and otherproteins, providing clues to both structure and function. The conformations of wild-type and some mutant p53 proteins and their relationship to the critical functions of sequence-specific DNA bindingand transactivation have been examined using full-length and trun

cated proteins. Some of these inferences have been corroborated byrecent evidence describing the nuclear magnetic resonance analysisand the crystal structure of several p53 domains (96, 97).

A schematic of the p53 molecule, indicating the functional regionsmost frequently evaluated, is shown in Fig. 1. The NH2-terminal ofp53 contains an acidic domain similar to those found in other franscription factors (5). Chimeric proteins containing amino acids 20â€”42fused to the DNA binding site of Gal4 will transactivate reportergenes linked to Gal4-binding DNA sequences (98). Interestingly, thisregion also is responsible for binding to cellular protein Mdm2, anddirect inhibition of transactivation is the likely mechanism by whichMdni2 interferes with p53 function (99â€”102).A proline-rich linkingregion between amino acids 60 and 92 begins the hydrophobic midsection of the protein. Between amino acids 100 and 295 lies thecritical region (delineated by studies of truncated and chimeric proteins) which determines p53 conformation, specific binding to DNAconsensus sequences, and sequence-specific transcriptional activityand which may be essential for growth suppression (103). This regionalso binds to SV4O T-antigen (104). The recently described crystalstructure of the DNA-binding domain identifies several motifs whichdirectly contact DNA, including nine amino acids which abut the p53consensus binding sequence (97). The positively charged carboxyterminus contains the basic nuclear localization sequence (aminoacids 316â€”325)and oligomerization domain (amino acids 319â€”360;Refs. 5, 96, 105, and 106). Oligomerization into dimers or tetramersappears to be required for DNA binding and transactivation function(97, 107, 108). The observation that the nonsense mutationCGCâ€”@TGA at codon 342 has been reported 10 times in humantumors supports the experimental evidence that truncation of the last50 amino acids does lead to a selection advantage in vivo.

Many mutations in the hydrophobic midregion change the conformation of the molecule, perhaps by partial denaturation. This exposesan epitope between amino acids 213â€”217,which is the target formAb24O, a monoclonal antibody that binds specifically to conformationally altered or denatured p53 (97). This antibody has been used asa marker of altered p53 conformation in immunoprecipitation andimmunoblotting studies. The importance of conformation in p53 function has been demonstrated using truncated and mutant p53 molecules(including temperature-sensitive mutants whose conformationchanges from wild-type at 32Â°Cto mutant at 37Â°C)and by modulatingwild-type conformation by oxidation. Molecules whose conformationis altered by mutation or oxidation (109) lose sequence-specific DNAbinding and transactivation activity; temperature-sensitive mutantsexhibit this loss only at 37Â°C,when they exist in mutant conformation(110, 111). Chimeric proteins, with the central p53 conformationregion linked to foreign nuclear localization and transactivation regions, maintain sufficient p53 transactivating properties to suppress growth of lung and colon carcinoma cells (103, 112), indicatingthat the amino and carboxy termini contribute generic functions but notspecificity.

Humans and laboratory animals which develop tumors carrying p53mutations may also develop antibodies against the protein, and thisphenomenon is now being investigated as a possible marker fordiagnosis and prognosis. Anti-p53 antibodies occur in patients whosetumors contain missense mutations, but they are usually directedagainst epitopes in the amino and carboxy termini and not theconformation-determining midregion, suggesting that the native protein can be immunogenic. Antibodies have been detected in the serumof patients with carcinomas of the bladder (28%), lung (15%), breast(12%), esophagus (16%), liver (25%), pancreas (19%), and lymphomas (5%) and at least seven other types of cancer, as well as inpatients with precursor lesions such as Barrett's esophagus (5%; Refs.113â€”118). In several anecdotal cases, the antibodies have been de

4864



p53 MUTATION5AND CANCER

tected in serum samples in high-risk patients (Barrett's esophagus,tobacco smokers, and occupational vinyl chloride exposure) severalyears before the clinical detection of cancer.4 Hypotheses regardinganti-p53 antibodies include: (a) does the immune response affectcarcinogenesis, e.g., does presence of anti-p53 antibody affect rates ofprogression to malignancy of preneoplastic lesions? (b) can the detection of anti-p53 antibodies be used as a screening test in patients athigh risk for specific cancers (e.g., smokers, Barrett's esophagus, andoccupational carcinogens)? and (c) can the presence of serum anti-p53antibodies identify individuals most responsive to immunotherapy?

2. Distribution of Mutations and the Relationship to p53

Conservation through Evolution

The clonal selection theory of carcinogenesis posits a growth advantage for clones containing mutant p53; therefore, sites wheremissense mutations frequently occur presumably identify amino acidswhose replacement leads to loss of p53 suppressor function andperhaps gain of oncogenic function. Previous reviews have noted thatmajor mutational â€œhotspotsâ€•occur in four of five domains labeled Iâ€”V(I, amino acids 13â€”19; II, amino acids 117â€”142; III, amino acids

171â€”181;IV, amino acids 234â€”258;and V, amino acids 270â€”286),which are highly conserved in evolution and are presumed to containamino acids essential for p53 function (Fig. 1; Ref. 51). Of the 393amino acids in human p53, 96 (24%) are conserved in 8 species forwhich the p53 sequence is reported. By comparing all hotspot sites tothe pattern of AA conservation in all species, we can address thehypothesis that common mutations affect conserved amino acidsimportant to p53 function.

Fig. 1 demonstrates the distribution of missense and nonmissensemutations in the database. Missense mutations are the most commontype (79% of mutations) in the conserved midregion but are uncommon (23%) in the amino and carboxy termini, where nonmissensemutations (nonsense, frameshift, splicing, and silent mutations) predominate. The rarity of missense mutations in the amino and carboxytermini suggests that in these regions the integrity of a single aminoacid residue is not essential to function or conformation. Of note, nomutations have been reported in conserved domain I. Nonconservedresidues within exons 5â€”8also show a low proportion of missensemutations, supporting the hypothesis that substitutions of nonconserved amino acids do not provide cells with a positive selectionadvantage. The most commonly mutated amino acids in the DNAbinding region are close to the protein-DNA interface (97). Thedistribution of deletions and insertions in p53 appears to be relatedprimarily to DNA sequence repeats (Fig. 1; section II.B.2) and not tocharacteristics of the protein; as predicted, they occur frequently atboth conserved and nonconserved codons.

We analyzed patterns of evolutionary conservation using the Multiple Alignment Construction & Analysis Workbench (MACAW)program (119). This method identifies the most highly conservedportion as amino acids 97â€”292(MP-scores, 359.4 and 471.2 for twocontiguous conserved regions). This area corresponds almost exactlyto the region which determines sequence-specific DNA binding andtransactivation (section II.C.1). When lower stringency for sequencesimilarity is allowed, secondary conserved blocks are noted betweenamino acids 17â€”29(MP-score, 40.4), corresponding to the transactivation region identified through Gal4 binding experiments, and aminoacids 324â€”352(MP-score 114.3), encompassing the oligomerizationdomain. Higher stringencies fail to identify subregions within aminoacids 97â€”292,including domains IIâ€”V,that are significantly morehomologous than the sequence as a whole. Detailed analysis of the

4 G. Trivers and T. Soussi, personal communication.

patterns of mutation clusters in conserved domains Ilâ€”Vand elsewhere adds to the evolutionary data and provides clues to amino acidsof greater and lesser importance.

The predominance of missense over nonmissense mutations beginsand ends abruptly at codons 130 and 286 (Fig. 1), suggesting thatthese are the boundaries of the conformation-determining region. Thisis shorter than the span identified by in vitro and evolutionary studiesas important for DNA binding specificity (amino acids 93â€”293)andsuggests that changes to amino acids 93â€”130may not alter p53conformation significantly. This region contains three short (3-sheetsand an unstructured loop, of which only one amino acid (Lys 120)interacts with DNA (97). Forty-four codons are relative hotspots formissense mutation, arbitrarily defined as more than 12 mutationsreported in the database (>0.5% of all mutations); 40 of the 44 hotspotamino acids (91%) are conserved in all 8 species. This represents only42% of the 96 conserved AM, suggesting that mutations at manyother conserved sites confer little or no growth advantage; this hypothesis can be tested by functional assays of mutants created bysite-directed mutagenesis.

Some amino acids in the previously identified hotspot domains areaffected much more frequently than others, and within each domainare some highly conserved amino acids which are rarely mutated. Forexample, only 3 of the 25 amino acids in domain II (132 Lys, 135 Cys,and 141 Cys) are relative hotspots. All three of these residues begin orend (3-strands, although other amino acids which flank (3-strands arenot hotspots (97). Of the six hotspot codons which contain 23% ofreported p53 mutations (Fig. 1), five contain CpG dinucleotides(codons 175, 245, 248, 273, and 282 but not codon 249), suggestinga shared mechanism (deamination of 5-methylcytosine) for the preyalent G:Câ€”@A:T transitions reported at these sites. The resultingmutants at some of these codons have been shown to alter p53 transcriptional activity and growth suppression (section II.D.5.F). Only two ofthese amino acids (Arg 248 and 273) directly contact DNA, but all appearto be Critical in stabilizing the protein-DNA interface (97).

Some relative hotspots for missense mutations occur outside thepreviously recognized conserved domains, marking three unrecognized regions of potential functional significance (Fig. 1). Betweendomains II and III lies a stretch of 14 amino acids, 151â€”164,of whichone-half are conserved and at which 167 missense mutations (and 33nonmissense mutations) have been reported. This region forms a(3-strand which is not adjacent to DNA (97). Its DNA sequence is richin G:C base pairs, often repetitive. No codons in this span are mutatedas frequently as the six hotspots noted above. Despite the fact that fiveCpG sites occur here (section II.C.2), G:Câ€”t'A:T transitions are rare.

This suggests that these sites may not be methylated or that the AAsubstitutions resulting from deamination (three of which are silent)offer little growth advantage, perhaps because they are not adjacent toDNA (97). These hypotheses also are testable using in vitro mutagen

esis, evaluating the mutants for transactivation, growth suppression,and other properties, as has been done for other CpG site mutations(section II.D.5.F).

Two other highly mutated areas lie between domains III and IV.Amino acids 193â€”200are almost completely conserved; missensemutations are common at the first three codons (His-Leu-Ile). Codons213â€”223also define a region of high conservation, with missensemutational hotspots at codons 216 (Val) and 220 (Tyr). Two otherconserved moieties which are frequently altered are 205 Tyr and 266Gly. Of these mutational hotspots, five contain a single favored

mutant: 157 G:Câ€”@T:A, Val>Phe; 194 G:Câ€”*T:A, Lys>Arg; 195T:Aâ€”*C:G, Ile>Thr; 205 A:G-+ G:C, Tyr>Cys; and 220 A:T-+ G:C,

Tyr>Cys. Codon 157 mutations, especially G:Câ€”*T:A transversion,

are especially frequent in lung cancer, perhaps indicating a tobacco4865




specific carcinogen (section II.D.5.C) or a growth advantage peculiar

to bronchial cells.Taken together, these specificities do not exclude that certain re

gions of DNA may be susceptible to specific mutagenic events, with

only favored mutants providing a selection advantage. The findingthat tyrosine residues are affected at several hotspots suggests thatthey are important to p53 function. Potential reasons include contribution to p53 conformation or modulation of phosphorylation status,

although a mechanism has not yet been demonstrated.Some other specific residues with potential importance are serines

9, 15, 37, 315, and 392, targets for phosphorylation by DNA-activatedprotein kinase, p34cdc kinase, and casein kinase II. In vitro mutagenesis studies creating alanine residues which are not phosphorylatedhave implicated serine 15 in human p53 as playing a role in suppressor function; conflicting results have been found by altering serine392 (120â€”122). That no mutations of these amino acids have beenreported is inconsistent with a strong selective advantage and arguesagainst the hypothesis that these residues contribute significantly tosuppressor function.

Our conclusions regarding evolutionary aspects of p53 are: (a) themost highly conserved region corresponds to the area of p53 responsible for sequence-specific DNA binding functions (amino acids 97â€”292); (b) the transactivation, nuclear localization, and oligomerizationdomains are only moderately conserved, provide generic functionswhich are not p53-specific, and are not significantly functionallyaltered by single amino acid substitutions; and (c) based on mutationalspectrum analysis, essential conserved residues and favored mutantsoccur at previously described domains Ilâ€”Vand three additionalregions (1) codons 151â€”164,especially codons 151â€”152(Pro-Pro),157â€”158(Val-Arg), and 163 (Tyr); (2) codons 193â€”195,especiallythe codon 194 Lys>Arg and 195 Ile>Thr mutants; and (3) codons213â€”220,especially the Tyr>Cys mutant at codon 220. Two otherresidues which are both conserved and frequently mutated are codon205, another tyrosine which often mutates to a cysteine, and glycine atcodon266.A searchof the SwissProt,PIRInternational,andtranslatedGenbank files found no other reported sequences homologous to thesethree regions; therefore, their specific properties remain obscure.

D. Clues to Carcinogenesis

1. p53 in the Context of Multistep Carcinogenesis

Is alteration of function of the p53 pathway a prerequisite forneoplastic transformation? The frequency of p53 mutations argues fora critical role, but for most tumor types mutations have been found inonly 20â€”50%of cases, with a maximum of 81% in oral mucosasquamous cell, 70% in small cell lung cancers, and 68% in anaplastic/undifferentiated thyroid cancer. Current techniques are unlikely tomiss more than 15â€”20% of coding sequence abnormalities. How

frequent and important to functional abrogation are mutations in thep53 promoter or unevaluated introns, epigenetic inactivation of p53,

and/or alteration of genes or proteins downstream in the p53 pathway?Is p53 only one of several pathways whose disruption is sufficient forneoplastic transformation? Support for this concept comes from theobservations that both p53 and ras mutations are common events butoften occur independently (67, 123â€”127),and either may be associated with aggressive tumor behavior (4, 128, 129). Does neoplastictransformation require a critical mass of genetic injury, in which p53is a frequent but not essential casualty? In this scenario, loss of p53genomic stabilization properties would predispose a cell to an acceleration in the rate of genetic damage and greatly increase the likelihood of neoplastic transformation and/or malignant progression. Research which attempts to unify molecular and cellular theories ofcarcinogenesis should address these and other questions. Evaluation

of mutation patterns of p53 and other genes in clinical tumors can bean important part of this process.

Some cancers are thought to arise from a single cell, while forothers a â€œfielddefectâ€•of the entire mucosal surface is thought to resultin multiple clones of transformed cells and multiple primary neo

plasms (130). Field cancerization is thought to occur in epithelia of theupper aerodigestive tract, bladder, and ulcerative colitis-associatedcolon (131â€”133).Careful identification of the stages of this processand molecular analysis at each stage can help illuminate the processesof clonal evolution in specific tumors. The multiplicity of p53 mutational events indicates that identical mutations are unlikely to occurindependently, and specific mutants may be used as clonal markers.For example, the field cancerization hypothesis for head and necksquamous carcinoma is supported by molecular analyses which havefound discordant mutations among nonneoplastic respiratory mucosa,primary cancers, and second upper aerodigestive primary lesions,suggesting that p53 is an important early target for mutations whichcan occur at multiple mucosal sites (132, 133).

2. Timing ofps3 Mutations in Human Cancer

Several human cancers develop through multiple, morphologicallydefined stages, such as the adenoma-carcinoma sequence in the colonand the dysplasia progression in the oral and bronchial mucosae. Theanatomically defined stages provide milestones for marking the timing of p53 mutation, and there is evidence that some tumor typesundergo early and others late p53 mutation. The issue has bothmechanistic and practical implications, including the use of p53 as apotential marker for early diagnosis and as an intermediate biomarkerfor chemoprevention studies. This section reviews current data assessing early and late alterations in human tumors. Early is defined hereas more than one morphological stage before invasion of the basementmembrane, which is the traditional hallmark of malignancy. Latemeans just before, or after, crossing the basement membrane. Alterations in the p53 pathway (mutation or epigenetic inactivation) arethought to be generally early events in cancers of the lung, esophagus,head and neck, breast, cervix, and stomach and generally late eventsin cancers of the brain, thyroid, liver, and ovary (134â€”139).Cancersof the colon, bladder, liver, and perhaps other organs may developthrough multiple pathways; p53 alterations may be early events insome pathways and late events in others for these tumor types. Someof the better characterized examples are discussed here.

Squamous cell carcinomas account for 25% of lung cancers and thelarge majority of cancers of the upper aerodigestive tract. Thesetumors usually progress through multiple, histologically recognizablestages of dysplasia (usually subdivided as mild, moderate, and severe), carcinoma in situ, and transgression of the basement membrane,which marks the beginning of invasive cancer. Despite the small sizeof preneoplastic lesions, which frequently measure less than 100 p@m,some genetic studies have been done, including allelic deletion, SSCP,and sequence analysis. However, the bulk of the data come from IHCusing p53 protein accumulation as a surrogate marker for missensemutation (section II.A.1).

Genetic and cytogenetic analyses of preinvasive bronchial lesionshave shown that p53 mutations can precede invasion. Three reports(140â€”142) document missense mutations, p53 allelic deletion, andchromosome l'7p deletion in dysplastic mucosa associated with carcinoma. However, composite IHC data show p53 protein accumulation in 0% of normal bronchial mucosae, 7% of squamous metaplasias, 25% of mild dysplasias, 32% of moderate dysplasias, 69% ofsevere dysplasias, 57% of carcinomas in situ, 70% of microinvasivecarcinomas, and 76% of fully invasive lung carcinomas (140â€”144).These results support a multistage model for squamous lung carcinoma in

4866




which p53 mutation occurs in 25% of the earliest neoplastic lesions, andthe large majority of mutations occur before invasion develops.

The timing of p53 alterations in the oral cavity, oropharynx, andlarynx is addressed by at least 10 published reports, usually using IHCas a marker for putative missense mutation in small progenitor lesions.Several groups have reported p53 alterations (jositive IHC or mutation) at or before the stage of severe dysplasia, including somehistologically normal mucosa (132, 145â€”149).Future studies couldcontribute to this area by providing additional histological detail toIHC studies, further molecular data, or correlating additional factors

(e.g., DNA content and other genetic changes) with p53 alterations.The timing of p53 mutation in squamous neoplasms in the esophagusappears similar to those in the oropharynx (150â€”152).

Sporadic colorectal cancers generally develop through an adenomacarcinoma sequence which has been extensively analyzed for alterations in oncogenes and tumor suppressor genes (reviewed in Ref.153). Allelic deletion and p53 point mutation rarely occur until thelate adenoma/carcinoma boundary (154; reviewed in Ref. 153). Incontrast, in ulcerative colitis, an inflammatory condition predisposingto colorectal cancer in non-polypoid, non-adenomatous, dysplasticmucosa, p53 mutation and allelic deletion are found in the earliestrecognized dysplastic lesions (131, 155, 156). It is notable that severalIHC studies find protein accumulation at earlier stages of sporadic

colon cancer, suggesting a nonmutational mechanism of proteinaccumulation during incipient neoplasia (157, 158).

3. p53 Genotype and Cancer Phenotype

The conventional method for describing the phenotype and predicting the clinical behavior of a cancer has been histological grading asassessed by light microscopy of tumor cells. IHC, which identifiesspecific cellular antigens, and other specialized staining techniquescan yield more detailed phenotypes and sometimes clinically usefulinformation, but these traditional measures are merely crude descriptions of essential tumor behaviors. Correlation of phenotypes, such ashistological grade or proliferative rate, with precise molecular events,such as alterations of p53 or other genes (e.g., metastasis-related genenm23, RB, and ras), promises to expand our understanding of tumorbehavior and perhaps foster new paradigms of risk assessment, prevention, early detection, determination of prognosis, and antineoplastic therapies.

The relationship of p53 mutations to tumor grade and stage hasbeen evaluated in many tumor types. Associations of p53 mutation orpositive IHC staining with higher grade and more advanced stage arefrequent, although sample sizes are often too small to draw statistically valid conclusions from individual studies. A relationship between p53 mutation and advanced tumor stage has been noted for

cancers of the endometrium (159), cervix (159), ovary (138, 139,160), liver (136, 137), prostate (161), and bladder (162, 163), indicating that for these tumors p53 mutation may be a late event contributing to tumor progression, but not for lung (67, 164, 165), headand neck (132, 149, 166), and breast (134, 135), in which mutationsare known to occur in early stage tumors and precursor lesions.

Studies of brain neoplasms are archetypes for establishing the roleof p53 in tumor progression and clonal dedifferentiation. Low gradegliomas, which are indolent and often curable by resection, maytransform to high grade, aggressive, rapidly lethal glioblastomas. Twostudies evaluating both low grade tumors and recurrent high gradetumors document the acquisition of p53 mutations in five of sixtumors after transformation from low to high grade, although lowgrade tumors may also contain p53 mutations (167, 168). The p53mutational spectra of low grade astrocytomas versus high gradeglioblastomas are similar, supporting the concept that the high grade

tumors evolve from the low grade precursors and are subject to thesame mutagens. As 45% of mutations are G:Câ€”@A:T transitions atCpG dinucleotides plus short deletions, the mutagens responsible for

progressive genetic damage in these tumors may be endogenouscellular processes (section II.D.5.F).

Recent studies have correlated p53 mutation with metastasis andother important but less well-studied tumor phenotypes, such asangiogenesis (169). The regulation of specific genes which contributeto these behaviors [such as MDR-1 (170) and thrombospondin (171)]is one area in which the known transcriptional activating function ofp53 could contribute to acquisition of the metastatic phenotype. Sincep53 regulates genomic stability, and prevents cell cycle entry inresponse to DNA damage, loss of p53 function would be expected tocorrelate with aneuploidy and increased proliferative rates, two features often associated with tumor progression and poor prognosis.Preliminary studies have noted a statistically significant associationbetween aneuploidy and p53 mutation for ovarian (172), colorectal(173), and gastric (54) cancers and a positive trend in esophagealcancer (53). Other properties which are being studied for a relationship to p53 function include radiosensitivity and hormone responsiveness (161, 174â€”176).

These data support a critical role for p53 in tumor progression. Theresults to date of p53 mutation analysis indicate that the details of thisprocess are complex and probably mutant- and tissue-specific. Specific p53 functions potentially involved include the pathways of genetransactivation, cell cycle checkpoint control, and programmed celldeath and suggest unbounded opportunities for in vitro testing. Futureresearch might address p53 mutation and mechanisms of tumor invasion, metastatic properties such as angiogenesis, and drug resistance.Such studies could clarify questions generated by clinicopathologicalobservations, which are often inconclusive due to unavoidableconfounding variables or sample size limitations.

4. Immortalization and Neoplastic Transformation:In Vitro Features

Much in vitro carcinogenesis research uses cultured human cellsderived from normal or malignant tissues (177). Tumor cell lines oftendiffer from the corresponding in vivo normal tissues in properties suchas protein or marker expression, response to growth modulators, andoncogene abnormalities. p53 mutations are found twice as frequentlyin all cell lines compared with all clinical tumor specimens (53% vs26%); 15 of 20 tumor types have an excess of mutations in cell lines.Potential explanations for this observation include: (a) mutationsmight be underdetected in clinical specimens due to technical reasons;(b) mutations could arise frequently during the process of establishinga cell line and, therefore, reflect in vitro mutagenic events which arenot relevant to in vivo carcinogenesis; (c) clones with mutated p53which comprise a minority of the in situ tumor could have an in vitrogrowth advantage over clones without such mutations. According tothis hypothesis, the mutations would reflect in vivo mutagens, but p53mutation would not be an important early event. The genetic instability triggered by loss of p53 function would lead eventually to clonalevolution and tumor progression.

Mutational spectrum analysis supports the hypothesis that cell linemutations do reflect in vivo mutagens. Distribution of mutation sitesand percentages of each type of base substitution, frameshift mutations, and CpG site mutations are similar in clinical tumors and celllines. Experience in the establishment of cell lines from cervical(178â€”180), thyroid (181, 182), nasopharyngeal (183, 184), and othercancers strongly suggests that cells with abnormal p53 function havea selective growth advantage in vitro. This advantage also may existin athymic nude mice, as suggested by the higher prevalence of p53

4867




mutations in xenografts than clinical specimens of head and neckcarcinomas (183, 185). This hypothesis warrants further investigationinto the features of in vitro cell growth and what properties ofwild-type p53 may be inhibitory.

5. Carcinogens and Specific Organ Carcinogenesis

The narrow mutational spectra exhibited by some mutagens haspopularized the idea that each agent might leave a specific identifyingâ€œfingerprintâ€•of site and type of DNA damage (35). It is probablymore realistic to expect that carcinogens will produce mutation patterns which are characteristic and instructive but not as unique asfingerprints. Table 2 contains the mutational spectra of all cancers forwhich more than 20 mutations were reported at the time of analysis ofthe database, in descending order of mutation prevalence.

A) Skin Carcinoma and UV Light. A clear example of the valueof mutational spectrum analysis in identifying carcinogen-specificmutations is seen in skin carcinomas, the most common type of humancancer, in which the role of UV light as the major carcinogen isunquestioned. Exposure to UV light increases the risk of both basalcell and squamous cell carcinoma, as well as the less common butmore lethal melanoma (186, 187). This physical mutagen producesdistinctive pyrimidine dimers that, if unrepaired, can produce tandemmutations, most characteristically CCâ€”tTT transitions. Tandemdipyrimidine mutations are infrequently caused by mutagens otherthan UV light (188, 189) and have rarely been observed in noncutaneous malignancies, i.e., 10 in over 2400 tumors with p53 mutation.Thus, the observation that these tandem mutations are common insquamous cell (9%) and basal cell (12%) carcinomas of the skin (Fig.2, B-D) directly incriminates both exposure to UV light as the causeof damage to the p53 gene and the loss of its tumor-suppressorfunction in the development ofthe cancers (190). The detection of rarecells with p53 mutations at dipynmidine sites in sun-exposed nonmalignant skin indicates that UV light initiated these mutations at theearliest stage of skin carcinogenesis (191).

Mutations at dipyrimidine sites in skin carcinomas show a nonrandom distribution among sites within the p.53 gene. Of the 30 CC:GGdinucleotides (18 on the nontranscribed strand) in conserved domains,transitions are common (5 or more) at only five (codons 151â€”152,247â€”248,178â€”179,278, and 281â€”282);frequency of mutations alsovaries at other dipyrimidine sites (CT:GA, TT:AA, and TC:AG).

Recent work shows that rates of cyclobutane dimer repair vary amongcodons within p53 (192). Comparison of the frequency of transitionmutations to rates of repair shows that 94% of C:Gâ€”@T:A or CC:GGâ€”TT:AA transitions occur at sites where more than 10% ofcyclobutane pyrimidine dimers remain 24 h after UV treatment,suggesting that slow repair of these dimers is responsible for some ofthe site specificity of UV-induced mutations. The dipyrimidine siteswhich are frequently mutated in XP-C patients were not evaluated inthis study. It remains unknown whether variations in site-specificDNA repair in XP patients follow the same patterns as normal

individuals and whether these patterns explain the different dipyrimidine mutation site specificity seen in XP-C. Mutant selection due tofunctional changes of specific amino acid substitutions also is likely toplay a role in determining this specificity. To further define themechanisms responsible for site specificity of mutations, study of thefunctional consequences (e.g., transactivation) of the commonlyfound mutants in skin and other cell types would be of interest.

Whereas p53 mutations are common in basal cell carcinoma andsquamous cell carcinoma, they are infrequent in melanomas (Table 2).Since sunlight also is a risk factor for melanomas (reviewed in Ref.193), the rarity of p53 mutations is surprising, although the frequentaccumulation of p53 in the nuclei of these tumor cells (59, 186, 194)

supports a role for the p53 biological pathway in tumorigenesis.Evaluation in melanomas of other genes and proteins which interactwith and stabilize p53 (e.g., Mdm2) or act downstream in the p53pathway [e.g., WAF1/Cipi (63)] would be informative. The p1611â€•â€•gene, which like WAF1/Cipi encodes an inhibitor of cyclin-dependent kinases, is another candidate tumor suppressor gene involved inmelanomas and other cancers (122, 195, 196).

B) Hepatocellular Carcinoma, Aflatoxin, and Hepatitis. Theviral-chemical etiology and multistage pathogenesis of HCC are archetypal of human carcinogenesis. It is estimated that 75â€”90%ofHCC cases are attributable to HBV (197). The relative risk of HCC iselevated in viral hepatitis carriers with chronic active hepatitis (198).These findings suggest that cell proliferation and/or inflammatoryresponse associated with chronic active hepatitis are the critical fac

tors responsible for the increased probability of neoplastic transformation of the precursor cells of HCC.

HBVDNAintegratesintoHCCcellsat randomsitesin thegenome.It contains the X gene, which codes for a protein (HBX) that modu

lates the transactivation of many cellular genes and is a candidate viraloncoprotein (199). Transgenic mice containing HBX gene in theirgermline have an increased frequency of HCC (200). HBX proteinbinds with p53 in vitro and in vivo (201), inhibits p53 sequencespecific DNA binding and transactivation activities, partially disruptsp53 oligomerization, and prevents p53 binding to transcription-repaircoupling factor ERCC3 (202). These mechanisms of epigenetic p53inactivation could account for the transactivation properties of HBX.Correlation of p53 staining and mutation data with HBX proteinexpression in HCCS could help assess the interactions between HBVand p53.

AFB also is considered to be a significant etiological factor in

certaingeographicareas (e.g., southernAfricaand Asia),wherethismycotoxin is consumed in food contaminated by Aspergillus flavis(reviewed in Ref. 203). Epidemiological studies conducted in the1970s and 1980s provided statistical evidence that dietary consumption of AFB was positively correlated with incidence of HCC andsuggested a synergistic interaction with alcoholic beverage consumption (204) and chronic active viral hepatitis (205â€”207).

Many investigators have reported data on the presence of chronicHBV infection and p53 mutations in the same tumors (208â€”216).

Their results indicate that HBV infection alone does not influence therate of p53 mutation and that AFB exposure is the most importantinfluence on mutation prevalence (28 versus 29% mutation prevalencein HBV-positive versus HBV-negative cases in low AFB exposure

areas; 54% prevalence in high AFB exposure areas, where almost allcases were HBV positive). The combination of these exposures maybe synergistic in causing mutation, but data are insufficient for defmitive conclusions (28 of 49 AFB positivefHBV-positive cases versus 1of S AFB positive/HBV-negative cases contain mutations; Refs. 208,211, and 216).

The results from a recent prospective cohort study of 18,244 peopleprovide convincing evidence that AFB has an etiological role inhepatocellular carcinogenesis and indicate a synergy between HBVand AFB (217, 218). This nested case-control analysis shows statistically significant associations among the presence of AFB and itsmetabolites in urine specimens, serum HBV surface antigen positivity, and HCC risk. The finding of the promutagenic AFB-N7-guanineadduct in the urine is further evidence that AFB has been activated toits electrophilic ultimate carcinogenic metabolite, AFB 8,9-oxide.Human hepatocytes in vitro can enzymatically activate AFB to its8,9-oxide (219); the interindividual variation in forming the AFB-N7-guanine adduct may be 10-fold or greater (220). Since several isoforms of cytochrome P450 enzymes can activate AFB (221), theinterpretation of pharmacokinetics data and its use in risk assessment

4868




will be complicated. The mutational spectrum of AFB (222, 223) hasbeen established in experimental systems; G:Câ€”t@T:A transversionsare the most common base substitutions (224â€”226). The high frequency of AGGâ€”'AGT transversions on the nontranscribed strand atp53 codon 249 in HCCs from areas of China and Mozambique witha high HCC incidence could be due to the high mutability of the thirdbase of codon 249 by AFB and/or a selective growth advantage of

hepatocyte clones carrying this specific 249@ mutant in liverchronically infected by HBV.

The preferential mutability hypothesis has been tested by Aguilar eta!. (227), who found that in a human liver cell line exposed to AFB,the third base in codon 249 is preferentially but not exclusivelymutated compared to immediately adjacent codons, suggesting thatboth preferential mutability and clonal selection are involved in human hepatocellular carcinogenesis. This highly sensitive and specificgenotypic mutation assay also can be used to determine the p53mutation load in nontumorous liver tissue from donors living ingeographic areas of high HCC incidence in which urinary AFB and/ormacromolecular-AFB adducts have been detected. A pilot study usinga genotypic mutation assay has compared the load of codon 249scr(AGG->AGT) mutant cells in nontumorous liver tissues from HCCpatients from Qidong, People's Republic of China, urban Thailand,and the United States (228). The loads of codon 249@Tmutations wereelevated in specimens from Qidong, slightly elevated in one biopsyfrom Thailand, but did not exceed background levels in specimensfrom the United States. These results indicate a positive associationbetween loads of codon 249ser mutation and dietary exposure to AFB.They also suggest that, in some populations, p53 mutation occursearly in hepatocarcinogenesis, and clonal expansion of these mutantcells occurs in the nonmalignant tissue. A number of hypotheses couldaccount for a selection advantage for these mutant cells, includingaltered responses to growth factors or resistance to programmed celldeath induced by hepatitis viral infection. These and other hypotheses canbe tested by in vitro models using human hepatocytes (219) and clinicopathological studies, such as correlation of aflatoxin-DNA adducts withcodon249G:Câ€”*T:Atransversionsin individualpatients.Thegenotypicmutation assay also might be applied to other tumor-specific hotspots.

C) Tobacco:Lungand OtherCarcinomas.Tobaccosmokingisthe leading preventable cause of cancer mortality, accounting for 40%of cancer deaths in men and 20% in women. Cigarette smoke is acomplex mixture with thousands of constituents including tumorinitiators, promoters, complete carcinogens, and cocarcinogens (229).The contributions of many of these agents to carcinogenesis in thelung and other organs are not known. Some of the compounds foundin tobacco smoke, most notably N-nitrosamines and PAHs, have beenpurified, tested in model systems, and found to be in vitro carcinogenswhich induce characteristic DNA abnormalities (section I.B). Thisfinding has encouraged mutational spectrum analysis in tobaccoassociated cancers to test the hypothesis that patterns of mutation canimplicate specific agents in tobacco responsible for inducing humancancers. Compounds of interest include N-nitrosamines, PAHs (e.g.,BP), active oxygen species, aldehydes, and metals. The p53 muta

tional spectra in tobacco-associated cancers show surprising intertumor variation, suggesting that many different tobacco-related carcinogens may be germane to human cancers.

1) Lung Carcinoma. Cigarette smoking is now thought to beresponsible for 90% of lung carcinomas in men and 78% in women(230). A large body of epidemiological evidence has confirmed adose-response relationship proportional to duration and amount ofsmoking (231, 232). How does mutational spectrum analysis furtherdefine this association?

p53 mutations are common in lung cancer, with the highest preyalence (70%) in SCLC and the lowest (33%) in adenocarcinomas. The

prevalent mutation is G:Câ€”@'T:A transversion with a predominance ofguanine residues on the nontranscribed DNA strand (Fig. 2, H-K), andthe frequency of transitions at CpG sites (9%) is lower than in almostall other cancers. This spectrum is consistent with data for severaldifferent types of chemical carcinogens found in tobacco smoke. Thehighly mutagenic metabolites of PAHs such as BP preferentiallyattack deoxyguanines and lead to mutations (Table 1), and one of thequantitatively minor adducts from the tobacco-specific N-nitrosamine4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone also leads toG:Câ€”@T:A transversions (233).

Analysis of p53 mutations in the database in relation to smokinghistory yields statistically significant conclusions consistent with theexperimental data. The frequency of p53 mutation and of G:Câ€”@T:Atransversion on the nontranscribed DNA strand are positively conelated with lifetime cigarette consumption (234). Compared with nonsmokers, smokers have significantly lower frequencies of all transitions (35 versus 69%; P < 0.05), G:Câ€”*A:T transitions (21 versus69%; P < 0.001), and higher rates of G:Câ€”*T:A transversions with a

DNA nontranscribed strand bias (29 versus 0%; P < 0.01; Fig. 2, L-M;Ref. 235). These observations are consistent with the model of preferential repair of carcinogen-induced damage on the transcribed DNAstrand (75, 236). Cigarette smoke and some of its components havebeen shown in vitro (229, 237) to increase mutation frequency andcause defects in DNA repair, which may contribute to the p53 mutation spectrum. These results also have implications in the currentcontroversies concerning the relative risk of lung cancer in nonsmokers exposed to environmental (passive) tobacco smoke. Molecular

analysis would contribute important data to the debate by determiningwhether critical mutations in lung cancers from these individualsshow a spectrum similar to smokers (a dose-response increase inG:Câ€” T:A transversions with exposure to environmental tobacco

smoke), although a large sample set may be required to detect changesin the spectrum.

Evidence exists of potential differences in pathways of carcinogenesisamong histological types of lung cancer. In vitro studies of lung cancerdifferentiation support a pluripotent cell of origin common to all types, ormultiple cells which can differentiate by variable, often overlappingpathways (238â€”240).Sufficient data exist to analyze the mutationalspectra of squamous cell, large cell, adenocarcinoma, and SCLC (Fig. 2,H-K). p53 mutations are found in 70% of SCLC and 47% of NSCLC,including 65% of squamous and 60% of large cell, but only 33% ofadenocarcinomas (140, 234, 241â€”246),a prevalence which more closelyapproximates that of other adenocarcinomas than of other lung cancerhistologies. The mutational spectra are notable for an excess ofA:Tâ€”@G:C mutations in SCLC (11 versus 6%) and an excess of deletions

and insertions in NSCLC (14 versus 4%). Subdividing NSCLC typesreveals that the prevalence of G:Câ€”@T:A transversions in large cell,squamous, and small cell carcinomas is similar (43â€”49%),but they areless common in adenocarcinoma. This finding is consistent with theweaker association of adenocarcinoma with cigarette smoking (247â€”249), since G:Câ€”Ã·T:A transversions are thought to result from bulkymolecules found in tobacco smoke. In adenocarcinomas, higher proportions of G:Câ€”@A:T (non-CpG site) and G:Câ€”@C:G mutations are seen,suggesting the presence of other carcinogens. Since adenocarcinomas arethe most common cell type in women and squamous cell is less frequent,analysis of mutational spectrum by gender (controlling for cell type andsmoking history) may provide clues to different carcinogenesis pathways,hormonal influences, or differing susceptibilities to tobacco-associatedcarcinogens.

2) Other Tobacco-associated Cancers. Epidemiological studies

have linked cigarette smoking to other cancers with attributable risksranging from 20 to 90% (250). Mutational spectrum analysis of someof these tumors should reflect the effects of relevant tobacco-associ

4869




ated carcinogens. Despite the similarity of risk factors, the mutationalspectra of these tumors differ from each other and differ significantly

from the lung cancer spectrum. This may be due to different constituents of tobacco or nontobacco-related carcinogens.

Head and Neck and Esophageal Carcinomas: The combination oftobacco use and drinking alcoholic beverages is responsible for approximately 75% of esophageal and 90% of head and neck cancers,most of them squamous cell (250). Two lines of evidence derivedfrom the database implicate environmental carcinogens in head and

neck carcinogenesis: (a) over one-half of all mutations in head andneck cancers are substitutions at base pairs which usually demonstratenontranscribed strand bias (Table 3; section II.B.1), and 80% of theseoccur on the nontranscribed strand; and (b) CpG transitions (suggestive of endogenous events; section II.D.5.F) are uncommon (13%).Compared with lung cancers, G:Câ€”*T:A transversions are less common (18 versus 40%), while A:Tâ€”*G:C (14%) and non-CpGG:Câ€”*A:T (18%) transitions are prominent, as are deletions andinsertions (19%). Of 18 deletions and insertions, 12 can be localizedto repetitive DNA sequences, consistent with the hypothesis thatmispairing during replication causes frameshifts (section II.B.2), and

two are large deletions at splice site codons, suggesting that the actual

DNA alteration is an intronic point mutation. Chemical carcinogenssuch as PAHs in tobacco smoke may contribute to these frameshiftmutations (section II.B.2).

Subdivision of head and neck carcinomas by anatomic region oforigin shows differing patterns of p53 mutations among the various

primary mucosal sites (Fig. 2, E-G). In laryngeal and pharyngeal

primaries, the mutation prevalence is relatively low (34%), but thespectrum resembles that of lung cancer. G:Câ€”@'T:A transversions,primarily on the nontranscribed strand, are the most common mutation(34%), and non-CpG G:Câ€”@A:T transitions also are common. Primary tumors in the oral cavity have a high mutation prevalence (81%)

and exhibit a spectrum in which non-CpG site G:Câ€”*A:T transitions,A:Tâ€”*G:C transitions, deletions, and insertions predominate. Thus, inthis class of cancers, subgroups which are often grouped togetherwhen addressing carcinogenesis and therapy demonstrate mutationaldifferences, which could be due to differences in carcinogen exposure

(e.g.,solubletobacco-specificN-nitrosaminesinoralcancersversustobacco combustion products in laryngeal primaries) or variation incarcinogen activation or DNA repair between distinct mucosal sites.

These hypotheses can be tested in the laboratory and by molecularepidemiological studies of mutational spectra of well-characterized

tumor types in populations with well-defined exposures.

The epidemiology of primary NPC differs from other head andneck sites; these tumors are associated with dietary N-nitrosamines,EBV, and perhaps genetic factors (183). Clinical NPCS rarely containmutated p53 (4%). One hypothesis for this low frequency of mutationcould be the epigenetic inactivation of p53 by EBV proteins (251).The spectrum of the few p53 mutations in NPC tumors and lines isunusual, reflecting a lack of mutations known to be associated withtobacco-related carcinogens or N-nitrosamines (G:Câ€”*T:A, non-CpGG:Câ€”*A:T, and A:Tâ€”@G:C substitutions). G:Câ€”*C:G transversions,which comprise only 3% of p53 mutations at other head and neck sites

and 7% in all cancers, are common (5 of 17; 29% of mutations;

P@ 0.00001 versus other head and neck sites), and four of five are on

the nontranscribed strand, suggesting the action of a chemical carcinogen. Further molecular epidemiological studies are indicated to generate a statistically meaningful database.

The spectrum in esophageal squamous cell carcinomas also showsless frequent G:Câ€”*T:A transversions than in lung carcinomas (16

versus 40%), with corresponding increases in G:Câ€”@'A:T (non-CpGsite 22 versus 16%) and A:Tâ€”@G:C (14 versus 7%) transitions, as wellas A:Tâ€”*T:A transversions (15 versus 5%). As discussed (section I.B;

Table 1), PAHs, N-nitrosamines, and alkylating agents are among thecompounds which cause these substitutions in experimental systems.Risk factors for squamous cell carcinoma in addition to tobacco smokingare alcoholic beverage consumption, and in China, dietary N-nitrosamineingestion (which could account for non-CpG site G:Câ€”*A:T transitions;Ref. 252). Future studies could address the carcinogenicity and mutational spectra of these suspected carcinogens in cultured human esopha

geal and upper aerodigestive cells or in animal models.As in the lung, adenocarcinoma of the esophagus is increasing in

frequency relative to squamous cell; the last decade has seen a 3-foldincrease in the incidence of adenocarcinoma and a 23% decrease in

squamous cell carcinoma (253, 254). Adenocarcinoma of the esophagus resembles carcinoma of the gastric cardia and usually ariseswithin the intestinal metaplasia of Barrett's esophagus. Chronic gastric reflux is considered a risk factor for this condition, but risk factorsfor neoplastic transformation are unknown. Few studies of the mutations in esophageal adenocarcinomas or Barrett's esophagus havebeen performed, but they suggest differences in mutagens from squamous carcinoma. Of the mutations reported, 6 of 9 (67%) have beenG:Câ€”*A:T transitions at CpG sites, compared with a 14% incidenceof CpG site mutations in squamous cell esophageal cancer. Evaluationof more adenocarcinomas will be important to generate statisticallyvalid comparisons with squamous carcinomas of the esophagus and

with gastric cardia cancers.Bladder Carcinoma: One-half of bladder carcinomas in men and

one-third in women are attributable to smoking (255). A significantassociation has been found in superficial bladder cancers between p53nuclear staining by IHC and the number of cigarettes smoked (256).The p53 mutational spectrum in bladder cancer is unusual with a highfrequency of G:Câ€”@C:G transversions (21 versus 7% for all tumors)with a nontranscribed DNA strand bias. G:Câ€”@A:T transitions at bothCpG and non-CpG sites are the most common mutation, with nontranscribed DNA strand bias at non-CpG sites. One study has found ahotspot G:Câ€”@C:G (Tyr>Lys) mutation at codon 280, but this has notbeen confirmed. Two reports have compared the spectra in a total of 34smokers and 23 nonsmokers (163, 257) with similar spectra found in thetwo groups. Further epidemiological studies, including quantitative

smoking histories, are warranted to evaluate the data for more subtlepatterns.

D) Radiation. 222Radon is a colorless, odorless, inert gas, a naturally occurring decay product of @8uraniumwhose decay emits a-partide radiation. Conclusive epidemiological evidence implicates occupational exposure to radon in uranium miners as a cause of lungcancer (reviewed in Refs. 258 and 259). Early epidemiological studiesof domestic radon and lung cancer risk exposure suggest that thedose-response and relative risk for residential radon exposure is thesame as for occupational exposure (260â€”262). Radon exposure and

cigarette smoking increase risk synergistically, and analysis of humantumors may reflect interactions between radon and tobacco carcinogens (258, 259, 262).

Two studies report the molecular analysis of the p.53 gene in lungcancers from uranium miners (140, 263). Missense p53 mutations

were found frequently in both studies. A mutational hotspot (16 of 30mutations) at codon 249, AGGâ€”@A@ Argâ€”+Met, was observed inminers from the Colorado Plateau who were exposed to high amountsof radon. Three of the codon 249met mutations occurred in lungcancers from never smokers, implicating a non-tobacco-associatedcarcinogen. Since these miners also were heavily exposed to othercarcinogens found in the mines, the hypothesis that radon was responsible for the specific base substitution requires further testing.

Other sources of ionizing radiation which have been associatedwith the development of hematological and solid malignancies indude therapeutic radiation, atomic bomb explosions, and the radio

4870




logical contrast agent Thorotrast. Mutational spectrum analysis ofcancersin thesepatientscouldindicateradiation-inducedevents.p53mutations in lung cancers from nine nonsmoking atomic bomb survivors from Hiroshima, Japan, were compared to eight nonexposedcontrols; no differences in prevalence or mutational spectrum werenoted in this small study (235). The study of larger numbers and agreater variety of radiation-related cancers would contribute morestatistically meaningful data to the issue of radiation-induced mutagenesis in human cancer.

E) Cervical Carcinoma and Human Papifioma Virus. Epidemiological studies established the link between cervical carcinoma andHPV types 16 and 18, and the molecularbasis for this associationemerging from laboratory research features the p53 pathway. HPVtypes 16 and 18 both produce E6 protein, which complexes with

cellular protein AP6 and degrades p53 in the cytoplasm (264, 265), anexample of epigenetic p53 inactivation. Mutation data would supportthis hypothesis if p53 mutations were found in HPV-negative morefrequently than HPV-positive cancers, and IHC would be expected todemonstrate no p53 staining in HPV-positive cases and nuclear staining in HPV-negative cases with missense mutations. Early reportsfound that mutation frequency was 100% in HPV-negative and 0% inHPV-positive tumors and cell lines (179, 266, 267), but this dramaticdifference has not been validated in further studies of clinical tumors(159, 178, 268â€”271).These later studies confirmed the rarity ofmutations in HPV-positive tumors (4%) and lines (7%); mutationshave been found in 7 of 39 of HPV-negative carcinomas now reported(18%; P 0.0007 versus HPV-positive tumors).

These data indicate that dysfunction of the p53 pathway contributesto cervical carcinogenesis, but unless mutations or other alterations inthe p53 pathway are demonstrated in all HPV-negative tumors, itcannot be considered essential. Future studies might address the lowrate of mutations in HPV-negative tumors by searching for other p53pathway abnormalities, such as occult mutations, downstream geneabnormalities, or other epigenetic mechanisms of inactivation.

F) Endogenous Sources of Mutation. As discussed in section I.A,not all mutations are caused by exogenous carcinogens. Normalcellular events which are mutagenic include deamination of 5-methylcytosine at CpG dinucleotides, DNA polymerase infidelity due tomispairing at repeat sequences or other mechanisms, and oxygenradical damage (reviewed in section l.A. and Refs. 14 and 272).Studies of constitutive cellular genes to determine baseline mutationrates in human and other mammalian cells have found rates ofbetween 10 â€˜Â°and i0@ mutations/gene/cell generation (10, 85).However, these may underestimate the true mutation rate because ofinsensitivity of the assays to some classes of mutations (e.g., only

hemoglobin variants with point mutations were detectable) or failureof an assay's cell culture conditions to select clones with mutationsbut no growth advantage (10, 85). This background level of mutationsimplies that populations will always exhibit a baseline rate of cancer,even in the absence of carcinogen exposure. The rates of endogenousmutations may also be affected by exogenous events; future studies onthe effects of carcinogens and cocarcinogens on deamination anddeletion are warranted. Mutational spectrum analysis may be able todefine these baseline mutational frequencies by identifying the proportions of mutations in cancer-related genes which are caused byendogenous processes.

1) CpG Deamination. It cannot be proven that an individualG:Câ€”@A:T transition found at a CpG dinucleotide in a tumor resultsfrom deamination of 5-methylcytosine (as opposed to mispairing of anOÂ°-methylguanine adduct with thymine, for example). However, thein vitro rates of spontaneous deamination, the clustering of spontaneous G:Câ€”+A:T mutations in experimental systems at sites wherecytosine is methylated, and the large database of observed mutations

suggests that this is the usual mechanism by which G:Câ€”'A:T transitions arise at CpG sites in human tumors. Examination of the p53database for associated features such as amino acid selection andstrand bias of CpG transitions may yield further insights into thisprocess in tumor cells.

The p53 coding region contains 39 CpG dinucleotides (i.e., 78 basepairs) which are potential sites for G:Câ€”@A:T transition from deamination. Transitions at these sites account for 24% of all p53 mutationsdescribed, with large variations among tumors. Seven sites (codons 175,196, 213, 245, 248, 273, and 282) account for 90% of CpG transitions, or21% of all p53 mutations. These mutations occur equally on both DNAstrands; strand bias in recognition and repair of G:T mismatches is thusunlikely to contribute to the selection of mutants.

Some mechanisms by which cells with some CpG transitions acquire a selective growth advantage have been studied. At codons 248and 273, both C:Gâ€”@T:A and G:Câ€”*A:T transitions commonly occur,implying that both amino acid changes (Arg>Trp and >Gln at 248,Arg>Cys and >His at 273) result in a selection advantage. At codons175 and 282, only one of the two mutations is selected (175, Arg>Hisbut not >Cys; 282, Arg>Trp but not >Gln). Soussi and colleagueshave constructed all potential transition mutants at CpG hotspotcodons 175, 248, and 273 and assessed conformation and in vitrotranscriptional activation in the sarcoma cell line SAOS, which contains no endogenous p53 protein.5 Transactivation, which is thoughtto contribute to p53 tumor suppressor function, was assessed bytranscriptional assays using the reporter gene CAT linked to an upstream p53 binding region. Wild-type and mutant p53 constructscotransfected into target cells with the reporter constructs were compared in their transcriptional activating abilities. The favored Arg>Hismutant at codon 175 assumes the mutant conformation and transactivation properties, whereas the rare Arg>Cys mutant retains wildtype conformation and activity. Interestingly, both mutants at codons248 (Arg>Trp and >Gln) and 273 (Arg>Cys and >His) maintainapparent wild-type conformation by antibody analysis but acquiremutant transactivation function. These two amino acids are in directcontact with DNA (97). This evidence supports the studies (sectionII.C.1) which connect p53 transactivation activity with growth suppression, but they imply that conformational changes as detected bythese methods are not necessary for modulation of this function (atleast in SAOS sarcoma cells). Future studies could address otherfeatures of CpG site mutants which could eliminate p53 suppressoractivity. It is possible that the more infrequent mutants are functionally silent, and their presence in cancer cells does not contribute to thecancer phenotype.

Thus, observational and experimental analysis supports the hypothesis that selection advantage due to altered transactivation and perhapsother functions is responsible for the hotspot mutations at CpGdinucleotides. Evidence argues against strand bias in mismatch repairof CpG transitions.

2) Cancers with High Rates of CpG Mutations, Deletions, andInsertions. The reasons discussed above argue strongly that deamination of 5-methylcytosine, an endogenous mechanism of mutation, isprimarily responsible for G:Câ€”@A:T transitions at CpG dinucleotides(273). Similarly, deletions and insertions may result from misalignment due to polymerase infidelity during replication, another endogenous mechanism. Transitions and frameshifts also may be induced bycarcinogen-DNA adducts and other mechanisms (section II.B.1.A).We hypothesize, however, that the sum of CpG transitions, deletions,and insertions, and potentially of other mutations which may eventually be linked specifically to endogenous processes, may be used as a

5 K. Ory and T. Soussi, personal communication.

4871



p53 MUTATIONSAND CANCER

Fig. 3. The mutational spectrum in hepatocellularcarcinomas varies by geographic origin in Asia (Qidong, n = 34; other mainland China, n = 11; Taiwan, n = 26; and Japan, n = 123).

Legend:

crude indicator of the proportion of mutations which are an inevitableconsequence of biological processes.

Transitions at CpG sites, deletions, and insertions comprise 37% of allreported p53 mutations but 45% of mutations of brain, colon (Fig. 2N),stomach, endometrial, thyroid, and hematological cancers. This suggeststhat endogenous cellular processes may contribute to a significant fraction of these tumors. The mutational spectrum of LFS implicates 5-methylcytosine deamination as the primary event in the formation of p53germline mutations; 53% of germline mutations are G:Câ€”+A:T transitions at CpG sites (Fig. 20). The second most frequent class of mutations(13%) are deletions and insertions, mostly of a single base pair at repeatsequences, a pattern which is consistent with the slipped mispairingmechanism. Thus, two-thirds of LFS cases can be attributed to endogenous-type mutational events, compared with slightly more than one-thirdof p53 somatic mutations found in human cancers. Mutations known orsuspected to be associated with carcinogen exposure (A:Tâ€”t'G:C andnon-CpG site G:Câ€”*A:T transitions, G:Câ€”@T:A transversions, and tandem mutations) represent only 23% of LFS mutations versus 46% ofsomatic mutations (Fig. 2, A and 0).

Comparison to germline mutations of other genes associated withcancer or other diseases supports the concept that these putativemechanisms of endogenous mutations are important sources of germline mutation, but mutation patterns are not consistent and probablyreflect specific functional characteristics of the gene and protein. Forexample, germline mutations in the APC gene cause familial adenomatous polyposis, a syndrome in which multiple colon polyps, adenomas, and eventually carcinomas develop in all affected persons.

Germline and somatic mutations in APC result in truncation or complete loss of the protein in 96% of cases and in an altered protein inonly 4% (274). Thus, frameshift mutations are prominent (48% ofmutations), and CpG site G:Câ€”+A:T transitions, although the mostfrequent base substitution, comprise only 14% of all mutations. Similar to p53 germline mutations, however, three-quarters ofAPC germline mutations are consistent with the endogenous processes of deamination and slipped mispairing. The spectrum of mutations in the factorIX gene, which causes hemophilia B, contains similarities and differences from the spectra in the tumor suppressor genes (273, 275).Approximately two-thirds are missense substitutions, less than 10%are deletions and insertions, and the remaining one-quarter are splicesite alterations and promoter mutations. Transitions at CpG sites are

common (approximately 36%), but other base substitutions are more

prevalent in factor IX mutations than in p53 germline mutations.Expansion of the data sets on germline mutations in other genes suchas HPRT, APC, DCC, hMLHJ, and hMSH2 may suggest patternswhich are now inapparent.

3) Mutational Spectrum Differences among Populations. Although the focus of mutational spectrum analysis in HCC has been onthe role of codon 249@ mutations and AFB exposure, deeper scrutinyof the data produces additional hypotheses. Since the spectrum inHCC is particularly dependent on a population's environmental cx

posure, subclassification of the data by geographic region may revealpatterns hidden by the AFB-related tumors. Fig. 3 depicts the mutational spectra of tumors from Qidong, other regions of mainlandChina, Taiwan, and Japan. While tumors from Qidong contain anoverwhelming preponderance of G:Câ€”@T:A transversions (95%) due

to association with AFB, Japanese and Taiwanese tumors harbor thistype in only 26% of cases each. That both of these latter groupsexhibit some codon 249@1mutations suggests that AFB may be a riskfactor even in populations not normally considered to be at high riskfor dietary contamination. The remainder of the spectrum differsbetween the groups, however, and suggests other potential carcinogens. In Taiwanese patients, the most frequent base change for allHCCs is A:Tâ€”@T:A transversion (41%); the strand bias (all 11 mutations are Aâ€”@T) suggests that an exogenous carcinogen is responsible. For Japanese patients, G:Câ€”*A:T (at CpG and non-CpG sites)and A:Tâ€”'G:C transitions (also with nontranscribed strand bias) areboth common, suggesting environmental carcinogens; deletions andinsertions, linked to both endogenous and exogenous mutationalmechanisms, also are frequent. The identification of relevant carcinogens with these spectra will be important. Only nine mutations havebeen reported in HCC from Western countries; five are G:Câ€”+A:Ttransitions, three of which are at CpG sites. Thus, no pattern suggestingcarcinogen exposure is yet discernible.

Environmental exposures or inherited characteristics which vary

among populations may be important determinants of mutationalspectra in other tumors as well. Sufficient data are available toevaluate lung, gastric, and breast cancers for differences in spectrumbetween Japanese and Western patients. Compared to Western patients, Japanese lung tumors show a higher frequency of G:Câ€”@A:Ttransitions (29 versus 20%, P 0.13; most are at non-CpG sites) anda lower frequency of G:Câ€”+T:A transversions (30 versus 43%,P = 0.01), both with a nontranscribed strand bias. These differences

4872

@,J-JAPAN




are not found in SCLC but are seen in both adenocarcinomas(G:Câ€”sA:T transitions: 48 versus 20%, P = 0.02; G:Câ€”+T:A transversions: 14 versus 27%, P = 0.21) and non-adeno NSCLC (24 versus20%, P = 0.004; 29 versus 55%, P 0.0003, respectively). Deletionsand insertions are more frequent in Japanese tumors (18 versus 10%,P = 0.22). These results suggest differences between Western and

Japanese individuals, either in exogenous carcinogens or in cancersusceptibility factors such as germline polymorphisms in genesinvolved in carcinogen metabolism (49).

Most of the gastric tumors analyzed have been from Japan, whereincidence rates are much higher than in Western countries. Epidemiological studies in Japan have found an association between gastriccancer and dietary salt and N-nitrosamines (276). Comparison of themutational spectra reveals that endogenous-type mutations are seen in72% of Western and 55% of Japanese tumors. More Western tumorsmust be analyzed to provide a more statistically valid spectrum. Themajor type of p53 mutation seen in Japanese but not Western tumorsis A:Tâ€”@G:C transition (20 of 69 versus 1 of 18, P = 0.03), withstrand bias favoring Tâ€”@C on the nontranscribed strand, the oppositeof the usual A:Tâ€”*G:C strand bias. Perhaps this represents the consequences of alkylthymine adducts from N-nitrosamines (277â€”279)inthe Japanese diet.

No specific exogenous carcinogens have been conclusively linkedwith development of human breast carcinoma. Although only 25% ofbreast cancers contain p53 mutations, 15% of these are G:Câ€”@'T:Atransversions, 16% are G:Câ€”'A:T transitions at non-CpG sites, and11% are A:Tâ€”@'G:C transitions, with nontranscribed strand bias of 89,63, and 75%, respectively, suggesting that exogenous bulky carcinogens may play a role. The responsible carcinogens may vary amongpopulations. Two reports of breast cancers from Japanese women(135, 280) indicate a mutational spectrum different from tumors inWestern women. G:Câ€”*T:A transversions are uncommon in Japanesecases (1 of 48 versus 25 of 175, P = 0.02), while A:Tâ€”@G:Ctransitions are more common (8 of 48 versus 10 of 175, P = 0.01).Thus, A:Tâ€”*G:C transitions with nontranscribed strand bias areprominent in the spectra of three different types of tumors in Japanesepatients (HCC, gastric, and breast). Whether this represents the effectof a regional environmental agent(s) or population differences incarcinogen metabolism or DNA repair is an area for future investigation.Transitions at CpG sites, deletions, and insertions are common in bothpopulations, suggesting a similar contribution of endogenous mutations.

These analyses suggest that molecular studies with careful identification of geographic and ethnic features of populations can make asignificant contribution to carcinogenesis research. Future researchshould evaluate statistically valid sample sizes in order to addressspecific hypotheses of differences in carcinogenesis among populations. For example, studies of Japanese populations in different geographic regions (Japan, Hawaii, and mainland United States) havedemonstrated differences in cancer rates. Molecular studies on thesecancers should be done to search for changes in mutational spectra,which would suggest that environmental carcinogens differ amonggeographic regions for several tumor types.

4)p53 Mutation and Prognosis. Since tumors with p53 mutationsoften behave more aggressively, the value of p53 as a prognostic factorhas been addressed. Studies have implicated p53 protein expression as anindependent prognostic factor in carcinomas of the breast, stomach,colon/rectum, bladder, and NSCLC (reviewed in Ref. 129).

Future studies might integrate hypotheses of carcinogenesis intostudies evaluating clinical outcomes. Since p53 mutation may correlate with other indicators of poor prognosis, such as aneuploidy, geneamplification, and high proliferative fraction (section II.D.3), it willbe important to adhere to careful epidemiological methodologies toassure that p53 status adds information to established prognostic

markers. This can be accomplished by performing prospective studiesin conjunction with large clinical trials to ensure statistical power,facilitate multivariate analyses to control for other established prognostic variables, and assess the responses to therapies in relation top53 status. The ability to perform IHC and PCR-based diagnosticstudies from archived tissue means that some of these studies could beaccomplished on materials from completed clinical trials which mayalready have reached their end points, expediting the answers to somequestions. Comparison of multiple molecular markers in the sametumors can help to determine the relative importance to carcinogenesis of abnormalities in various cellular pathways.

Such studies can address both specific and general hypotheses regarding carcinogenesis, for example: that tumors with p53 mutations might bemore resistant to therapy-induced apoptosis; that tumors with mutationsare more likely to be associated with amplification ofother cancer-relatedgenes which influence therapy; whether all cancers have abnormalities of

the p53 pathway, or of a specific combination of pathways; the relationship of response to therapies to molecularly determined tumor phenotypes and genotypes; identification of subsets of patients with favorableand unfavorable prognoses that should be analyzed separately and cxcluded from standard treatment protocols; and whether innovative therapies (e.g., antisense oligonucleotides and gene therapy) can be tailoredto specific molecular features of a tumor to improve the probability ofsuccessful outcome.

ifi. Conclusions

p53 mutational spectrum analysis contributes to a variety of mech

anistic questions in carcinogenesis, including integration of molecularand cellular hypotheses. Examples which are of ongoing interestinclude: what is the timing and sequence of the genetic changescontributing to neoplastic transformation, progression, and metastasis? Does cancer always arise from a single transformed clone, or can

multiple transforming events occur in some tissues? How do carcinogen-host interactions, such as individual differences in carcinogenmetabolism and DNA repair, influence an individual's risk of developing cancer? Do different histological types of cancer contain characteristic mutations reflecting a pattern of mutagen exposure, and docancers of similar histological types in different organs share thesepatterns? Are abnormalities of the p53 pathway critical for all tumors?In cancers with low rates of p53 mutation, is p53 inactivated byepigenetic events, or is its function interrupted by changes in othergenes which participate in p53-mediated pathways?

The emerging field of molecular epidemiology promises importantcontributions to cancer research, but full realization of its potentialwill require careful adherence to the tenets of both laboratory scienceand epidemiology. This will require increased interaction betweenlaboratory researchers, clinicians, and epidemiologists in order todesign studies which can adequately address important hypotheses.

For example, published reports of the molecular features of cancers(and other diseases) should include important epidemiological datasuch as gender, age, ethnicity, geographic origin, and, when appropriate, associated medical conditions, tobacco smoking, and occupational or other exposures. Large scale clinical trials which include dataon the relevant variables can provide tumor tissue for molecular study.This may be accomplished using archived materials from past studies;future trials could collect and preserve tissue specifically for molecular analyses. Attention to specimen handling at all times wouldminimize tissue degradation and maximize the reliability of laboratoryfindings.

Mutational spectrum analysis and molecular epidemiology can beapplied in innovative ways to both generate and address basic andpractical hypotheses including the areas of chemical, physical, and bio

4873




carcinogenesis. Nature (Land.), 363: 558â€”561, 1993.30. Yuspa, S. H., and Poirier, M. C. Chemical carcinogenesis: from animal models to

molecular models in one decade. Adv. Cancer Res., 50: 25â€”70,1988.31. Horsfall, M. J., Gordon, A. J., Bums, P. A., Zielenska, M., van der Vliet, G. M., and

Glickman, B. W. Mutational specificity of alkylating agents and the influence ofDNA repair. Environ. Mol. Mutagen., 15: 107â€”122,1990.

32. Strauss, B. S. The â€œAruleâ€•of mutagen specificity: a consequence of DNA polymerase bypass of non-instructional lesions? Bioessays, 13: 79â€”84, 1991.

33. Basu, A. K., and Essigmann, J. M. Site specifically modified oligodeoxynucleotidesas probes for the structural and biological effects of DNA-damaging agents. Chem.Rca. Toxicol., 1: 1â€”18,1988.

34. Skandalis, A., and Glickman, B. W. Endogenous gene systems for the study ofmutational specificity in mammalian cells. Cancer Cells, 2: 79â€”83, 1990.

35. Vogelstein, B., and Kinzler, K. W. Carcinogens leave fingerprints. Nature (Land.),355: 209â€”210,1992.

36. Sarasin, A. Use of shuttle vectors for the analysis of DNA damage processing inmammalian cells. Mutat. Res., 220: 101â€”106,1989.

37. Miller, J. H. Carcinogens induce targeted mutations in Escherichia co!i. Cell, 31: 5â€”7,1982.

38. Mawr, M., Gordon, A. J. F., Bernelot-Moens, C., and Glickman, B. W. Comparisonof the mutational specificities exhibited by BPDE in Escherichia co!i and CHO cells.in: M. W. Lambert and J. Laval (eds.), DNA Repair Mechanisms and their BiologicalImplications in Mammalian Cells, pp 205â€”231.New York: Plenum Publishing Corp.,1989.

39. Grosovsky, A. J., de Boer, J. G., de Jong, P. J., Drobetsky, E. A., and Glickman, B. W.Base substitutions, frameshifts, and small deletions constitute ionizing radiationinduced point mutations in mammalian cells. Proc. NatI. Acad. Sci. USA, 85:185â€”188,1988.

40. Liber, H. L, Call, K. M., and Little, J. B. Molecular and biochemical analyses ofspontaneous and X-ray-induced mutants in human lymphoblastoid cells. Mutat. Res.,178: 143â€”153,1987.

41. Yang, J. L., Chen, R. H., Maher, V. M., and McCormick, J. J. Kinds and location ofmutations induced by (Â±)-7@,8a-dihydroxy-9a,10a-epoxy-7,8,9,10-tetrahydro-benzo[a]pyrene in the coding region of the hypoxanthine (guanine) phosphoribosyltransferase gene in diploid human fibroblasts. Carcinogenesis (Land.), 12: 71â€”75,1991.

42. Albertini, R. J., Sullivan, L. M., Berman, J. K., Greene, C. J., Stewart, J. A., Silveira,J. M., and O'Neill, J. P. Mutagenicity monitoring in humans by autoradiographicassay for mutant T-lymphocytes. Mutat. Res., 204: 481â€”492,1988.

43. Cariello, N. F., Craft, T. R., Vrieling, H., Van Zeeland, A. A., Adams, T., and Skopek,T. R. Human HPRT mutant database: software for data entry and retrieval. Environ.Mol. Mutagen., 20: 81â€”83,1992.

44. Kress, S., Sutter, C., Strickland, P. T., Mukhtar, H., Schweizer, J., and Schwarz, M.Carcinogen-specific mutational pattern in the p53 gene in ultraviolet B radiationinduced squamous cell carcinomas of mouse skin. Cancer Res., 52: 6400â€”6403,1992.

45. Kanjilal, S., Pierceall, W. E., Cummings, K. K., Kripke, M. L., and Ananthaswamy,H. N. High frequency of p53 mutations in ultraviolet radiation-induced murineskin tumors: evidence for strand bias and tumor heterogeneity. Cancer Res., 53:2961â€”2964,1993.

46. Ruggeri, B., DiRado, M., Zhang, S. Y., Bauer, B., Goodrow, T., and Klein-Szanto,A. J. P. Benzo(a)pyrene-induced murine skin tumors exhibit frequent and characteristic G to T mutations in the p53 gene. Proc. Natl. Acad. Sci. USA, 90: 1013â€”1017,1993.

47. Ruggeri, B., Caamano, J., Goodrow, T., DiRado, M., Bianchi, A., Trono, D., Conti,C. J., and Klein-Szanto, A. J. Alterations of the p53 tumor suppressor gene duringmouse skin tumor progression. Cancer Res., 51: 6615â€”6621, 1991.

48. Burns, P. A., Kemp, C. J., Gannon, J. V., Lane, D. P., Bremner, R., and Balmain, A.Loss of heterozygosity and mutational alterations of the p53 gene in skin tumours ofinterspecific hybrid mice. Oncogene, 6: 2363â€”2369,1991.

49. Shields, P. G., and Harris, C. C. Molecular epidemiology and the genetics ofenvironmental cancer. JAMA, 266: 681â€”687, 1991.

50. Hollstein, M., Rice, K., Greenblatt, M. S., Soussi, T., Fuchs, R., Sortie, T., Hovig, E.,Smith-Sorensen, B., Montesano, R., and Harris, C. C. Database of p53 gene somaticmutations in human tumors and cell lines. Nucleic Acids Res., 22(Supp.): in press,1994.

51. Soussi, T., Caron de Fromentel, C., and May, P. Structural aspects of the p53 proteinin relation to gene evolution. Oncogene, 5: 945â€”952,1990.

52. Park, M., and Vande Woude, G. F. Principles of molecular cell biology of cancer:oncogenes. In: V. T. DeVita, Jr., S. Heilman, and S. A. Rosenberg (ads.), Cancer.Principles and Practice of Oncology, Ed. 3, pp. 45â€”66.New York: J. B. LippincottCo., 1989.

53. Macsaws, C., Tamura, G., Suzuki, Y., Ishida, K., Saito, K., and Satodate, R. Sensitivedetection of p53 gene mutations in esophageal endoscopic biopsy specimens by cellsorting combined with polymerase chain reaction single-strand conformation polymorphism analysis. Jpn. J. Cancer Res., 83: 1253â€”1256,1992.

54. Tamura, G., Kihana, T., Nomura, K., Terada, M., Sugimura, T., and Hirohashi, S.Detection of frequent p53 gene mutations in primary gastric cancer by cell sorting andpolymerase chain reaction single-strand conformation polymorphism analysis. CancerRes., 51: 3056â€”3058, 1991.

55. Nigro, J. M., Baker, S. J., Preisinger, A. C., et a!. Mutations in the p53 gene occur indiverse human tumor types. Nature (Land.), 342: 705â€”708,1989.

56. Hsu, 1. C., Metcalf, R. A., Sun, T., Welsh, J., Wang, N. J., and Harris, C. C. p53 genemutational hotspot in human hepatocellular carcinomas from Qidong, China. Nature(Land.), 350: 427â€”428,1991.

57. Hinds, P. W., Finlay, C. A., Quartin, R. S., Baker, S. J., Fearon, E. R., Vogelstein, B.,and Levine, A. J. Mutant p53 DNA clones from human colon carcinomas cooperate

4874

logical carcinogens, risk assessment, host factors, control of cell andtissue behaviors, mechanisms of subcellular communication, proteinfunction, DNA replication, transcription, translation, and evolution.

Acknowledgments

We appreciate discussion and comments from our colleagues, includingEttore Appella, Mark Boguski, Kathleen Forrester, Stephen Friend, S. PerwezHussain, David Lane, W. Edward Mercer, Ruggero Montesano, MosheOren, Jennifer Pietenpol, Varda Rotter, Thierry Soussi, Glen Trivers, BertVogelstein, Xin Wang, and Stuart Yuspa. The expert graphic and editorial

assistance of Dorothea Dudek and the assistance of Peter Shields withstatistical analyses are also appreciated.

References

1. Hollstein, M., Sidransky, D., Vogelstein, B., and Harris, C. C. p53 mutations inhuman cancers. Science (Washington DC), 253: 49â€”53,1991.

2. Levine, A. J., Momand, J., and Finlay, C. A. The p53 tumour suppressor gene. Nature(Land.), 351: 453â€”456, 1991.

3. Harris, C. C. The p53 tumor suppressor gene: at the crossroads of molecularcarcinogenesis, molecular epidemiology and cancer risk assessment. Science (Washington DC), 262: 1980â€”1981, 1993.

4. Harris, C. C., and Hollstein, M. Clinical implications of the p53 tumor-suppressorgene. N. EngI. J. Med., 329: 1318â€”1327,1993.

5. Vogelstein, B., and Kinzler, K. W. p53 function and dysfunction. Cell, 70: 523â€”526,1992.

6. Montenarh, M. Functional implications of the growth-suppressor/oncoprotein p53.Int. J. Oncol., 1: 37â€”45,1992.

7. Lindahl, T. Instability and decay of the primary structure of DNA. Nature (Land.),362:709â€”715,1993.

8. Kunkel, T. A. Nucleotide repeats. Slippery DNA and diseases. Nature (Land.), 365:207â€”208,1993.

9. Holliday, R., and Grigg, G. W. DNA methylation and mutation. Mutat. Res., 285:61â€”67,1993.

10. Strauss, B. The origin of point mutations in human tumor cells. Cancer Res., 52:249â€”253,1992.

11. Ehrlich, M., Zhang, X. Y., and Inamdar, N. M. Spontaneous deamination of cytosineand 5-methylcytosine residues in DNA and replacement of 5-methylcytosine residueswith cytosine residues. Mutat. Res., 238: 277â€”286,1990.

12. Loeb, L. A., and Cheng, K. C. Errors in DNA synthesis: a source of spontaneousmutations. Mutat. Res., 238: 297â€”304,1990.

13. Loeb, L. A., and Preston, B. D. Mutagenesis by apurinic/apyrimidinic sites. Annu.Rev. Genet., 20: 201â€”230,1986.

14. Lutz, W. K. Endogenous genotoxic agents and processes as a basis of spontaneouscarcinogenesis. Mutat. Res., 238: 287â€”295,1990.

15. Breimer, L. H. Molecular mechanisms of oxygen radical carcinogenesis and mutagenesis: the role of DNA base damage. Mol. Carcinog., 3: 188â€”197,1990.

16. Rideout, W. M., III, Coetzee, G. A., Olumi, A. F., and Jones, P. A. 5-Methylcytosineas an endogenous mutagen in the human LDL receptor and p53 genes. Science(Washington DC), 249: 1288â€”1290, 1990.

17. Kunkel, T. A. Misalignment-mediated DNA synthesis errors. Biochemistry, 29:8003-8011, 1990.

18. Phear, G., Nalbantoglu, J., and Meuth, M. Next-nucleotide effects in mutations drivenby DNA precursor pool imbalances at the apri locus of Chinese hamster ovary cells.Proc. Nati. Acad. Sci. USA, 84: 4450â€”4454,1987.

19. Liu, P. K., Chang, C. C., Trosko, J. E., Dube, D. K., Martin, G. M., and Loeb, L A.Mammalian mutator mutant with an aphidicolin-resistant DNA polymerase a. Proc.Natl. Acad. Sci. USA, 80: 797â€”801,1983.

20. Loeb, L. A., Springgate, C. F., and Battula, N. Errors in DNA replication as a basisof malignant changes. Cancer Res., 34: 2311â€”2321,1974.

21. Nowell, P. C. The clonal evolution of tumor cell populations. Science (WashingtonDC), 194: 23â€”28,1976.

22. Foulds, L. The experimental study of tumor progression: a review. Cancer Res., 14:317â€”339,1954.

23. Chorazy, M. Sequence rearrangements and genome instability. A possible step incarcinogenesis. J. Cancer Res. Clin. Oncol., 109: 159â€”172,1985.

24. German, J. Bloom's syndrome. X. The cancer proneness points to chromosomemutation as a crucial event in human neoplasia. in: J. German (ed), ChromosomeMutation and Neoplasia, pp. 347â€”357.New York: Alan R. lAss, 1983.

25. Fishel, R., Lescoe, M. K., Rao, M. R., Copeland, N. G., Jenkins, N. A., Gather, J.,Kane, M., and Kolodner, R. The human mutator gene homolog MSH2 and itsassociation with hereditary nonpolyposis colon cancer. Cell, 75: 1027â€”1038,1993.

26. Leach, F. S., Nicolaides, N. C., Papadopoulos, N., et a!. Mutations of a mutS homologin hereditary nonpolyposis colorectal cancer. Cell, 75: 1215â€”1235,1993.

27. Bronner, C. E., Baker, S. M., Morrison, P. T., et a!. Mutation in the DNA mismatchrepair gene homologue hMLH1 is associated with hereditary non-polyposis coloncancer. Nature (Land.), 368: 258â€”261,1994.

28. Papadopoulos, N., Nicolaides, N. C., Wei, Y. F., et a!. Mutation of a mutL homologin hereditary colon cancer. Science (Washington DC), 263: 1625â€”1629,1994.

29. Ionov, Y., Peinado, M. A., Malkhosyan, S., Shibata, D., and Perucho, M. Ubiquitoussomatic mutations in simple repeated sequences reveal a new mechanism for colonic




with ras in transforming primary rat cells: a comparison of the â€œhotspot: mutantphenotypesâ€•. Cell Growth & Differ., I: 571â€”580,1990.

58. Hall, P. A., and Lane, D. P. p53 in tumour pathology: can we trustimmunohistochemistry?â€”revisited. J. Pathol., 172: 1â€”4,1994.

59. Castresana, J. S., Rubio, M. P., Vazquez, J. J., Idoate, M., Sober, A. J., Seizinger,B. R., and Barnhill, R. L. Lack of allelic deletion and point mutation as mechanismsof p53 activation in human malignant melanoma. tnt. J. Cancer, 55: 562â€”565,1993.

60. Weiss, J., Schwechheimer, K., Cavenee, W. K., Herlyn, M., and Arden, K. C.Mutation and expression of the p53 gene in malignant melanoma cell lines. tnt. J.Cancer, 54: 693â€”699,1993.

61. Heimdal, K., Lathe, R. A., Lystad, S., HoIm, R., Fossa, S. D., and Borresen, A. L Nogermline TP53 mutations detected in familial and bilateral testicular cancer. GenesChromosom. Cancer (Phila.), 6: 92â€”97,1993.

62. Peng, H. Q., Hogg, D., Malkin, D., Bailey, D., Gallie, B. L., Bulbul, M., Jewett, M.,Buchanan, J., and Goss, P. E. Mutations of the p53 gene do not occur in testis cancer.Cancer Res., 53: 3574â€”3578, 1993.

63. El-Deity, W. S., Tokino, T., Velculescu, V. E., Levy, D. B., Parsons, R., Trent, J. M.,Lin, D., Mercer, W. E., Kinzler, K. W., and Vogelstein, B. WAF1, a potentialmediator of p53 tumor suppression. Cell, 75: 817â€”825, 1993.

64. Harper, J. W., Adami, G. R., Wei, N., Keyomarsi, K., and Elledge, S. J. The p21talk-interacting protein Cipi is a potent inhibitor of G, cyclin-dependent kinases. Cell,75: 805â€”816,1993.

65. Orita, M., Iwahana, H., Kanazawa, H., Hayashi, K., and Sekiya, T. Detection ofpolymorphisms of human DNA by gel electrophoresis as single-strand conformationpolymorphisms. Proc. Natl. Acad. Sci USA, 86: 2766â€”2770, 1989.

66. Murakami, Y., Hayashi, K., and Sekiya, T. Detection of aberrations of the p53 allelesand the gene transcript in human tumor cell lines by single-strand conformationpolymorphism analysis. Cancer Res., 51: 3356â€”3361, 1991.

67. Mitsudomi, T., Steinberg, S. M., Nau, M. M., et a!. p53 gene mutations in non-smallcell lung cancer cell lines and their correlation with the presence of ras mutations andclinical features. Oncogene, 7: 171â€”180,1992.

68. Borresen, A. L., Hovig, E., Smith-Sorensen, B., Malkin, D., Lystad, S., Andersen,T. I., Nesland, J. M., Isselbacher, K. J., and Friend, S. H. Constant denaturant gelelectrophoresis as a rapid screening technique for p53 mutations. Proc. Nail. Acad.Sci. USA, 88: 8405â€”8409, 1991.

69. Smith-Sorensen, B., Gebhardt, M. C., Kloen, P., McIntyre, J., Aguilar, F., Cerutti, P.,and Borresen, A. L. Screening for TPS3 mutations in osteosarcomas using constantdenaturant gel electrophoresis (CDGE). Hum. Mutat., 2: 274â€”285,1993.

70. Condie, A., Eeles, R., Borresen, A. L., Coles, C., Cooper, C., and Prosser, J. Detectionof point mutations in the p53 gene: comparison of single-strand conformation polymorphism, constant denaturant gel electrophoresis, and hydroxylamine and osmiumtetroxide techniques. Hum. Mutat., 2: 58â€”66,1993.

71. Matlashewski, G. J., Tuck, S., Pim, D., Lamb, P., Schneider, J., and Crawford, L. V.Primary structure polymorphism at amino acid residue 72 of human p53. Mol. Cell.Biol., 7: 961â€”963,1987.

72. Carbone, D., Chiba, I., and Mitsudomi, T. Polymorphism at codon 213 within the p53gene. Oncogene, 6: 1691â€”1692,1991.

73. Eckert, K. A., and Kunkel, T. A. The fidelity of DNA polymerases used in thepolymerase chain reactions. in: M. J. McPherson, P. Quirke, and G. R. Taylor (eds.),PCR: A Practical Approach, pp. 225â€”245.New York: Oxford University Press, 1991.

74. Jackson, D. P., Hayden, J. D., and Quirke, P. Extraction of nucleic acid from fresh andarchival material. in: M. J. McPherson, P. Quirke, and G. R. Taylor (eds.), PCR: APractical Approach, pp. 29â€”50.New York: Oxford University Press, 1991.

75. Bohr, V. A., Phillips, D. H., and Hanawalt, P. C. Heterogeneous DNA damage andrepair in the mammalian genome. Cancer Res., 47: 6426â€”6436, 1987.

76. Xanthoudakis, S., Miao, G., Wang, F., Pan, Y. C., and Curran, T. Redox activation ofFos-Jun DNA binding activity is mediated by a DNA repair enzyme. EMBO J., Ii:3323â€”3335, 1992.

77. Schaeffer, L., Roy, R., Humbert, S., Moncollin, V., Vermeulen, W., Hoeijmakers,J. H., Chambon, P., and Egly, J. M. DNA repair helicase: a component of BTF2(TFIIH) basic transcription factor. Science (Washington DC), 260: 58â€”63, 1993.

78. Evans, M. K., Taffe, B. G., Harris, C. C., and Bohr, V. A. DNA strand bias in therepair of the p53 gene in normal human and xeroderma pigmentosum group Cfibroblasts. Cancer Res., 53: 5377â€”5381,1993.

79. Bohr, V. A. Gene specific DNA repair. Carcinogenesis (Land.), 12: 1983â€”1992,1991.

80. Sato, M., Nishigori, C., Zghal, M., Yagi, T., and Takebe, H. Ultraviolet-specificmutations in p53 gene in skin tumors in xeroderma pigmentosum patients. CancerRes., 53: 2944â€”2946, 1993.

81. Dumaz, N., Drougard, C., Sarasin, A., and Daya-Grosjean, L. Specific UV-inducedmutation spectrum in the p53 gene of skin tumors from DNA-repair-deficient xeroderma pigmentosum patients. Proc. Natl. Acad. Sci. USA, 90: 10529â€”10533,1993.

82. Gupta, P. K., Johnson, D. L., Reid, T. M., Lee, M. S., Romano, L. J., and King, C. M.Mutagenesis by single site-specific arylamine-DNA adducts. Induction of mutationsat multiple sites. J. Biol. Chem., 264: 20120â€”20130, 1989.

83. Klein, J. C., Bleeker, M. J., Lutgerink, J. T., et a!. Use of shuttle vectors to study themolecular processing of defined carcinogen-induced DNA damage: mutagenicity ofsingle Oâ€•-ethylthymine adducts in HeLa cells. Nucleic Acids Res., 18: 4131â€”4137,1990.

84. Moriya, M. Single-stranded shuttle phagemid for mutagenesis studies in mammaliancells: 8-oxoguanine in DNA induces targeted G. Câ€”pT.A transversions in simiankidney cells. Proc. NatI. Acad. Sci USA, 90: 1122â€”1126, 1993.

85. Harris, C. C. Chemical and physical carcinogenesis: advances and perspectives.Cancer Res., 51: 5023sâ€”5044s,1991.

86. Hussain, S. P., Aguilar, F., and Cerutti, P. Mutagenesis of codon 248 of the humanp53 tumor suppressor gene by N-ethyl-N-nitrosourea. Oncogene, 9: 13â€”18,1994.

87. Fraumeni, J. F., Hoover, R. N., Devesa, S. S., and Kinlen, L. J. Epidemiology ofcancer. in: V. T. DeVita, S. Hetlman, and S. A. Rosenberg (eds.), Cancer: Principleand Practice of Oncology, pp. 150â€”181.Philadelphia: J. B. Lippincott, 1993.

88. Jego, N., Thomas, G., and Hamelin, R. Short direct repeats flanking deletions, andduplicating insertions in p53 gene in human cancers. Oncogene, 8: 209â€”213, 1993.

89. Krawczak, M., and Cooper, D. N. Gene deletions causing human genetic disease:mechanisms of mutagenesis and the role of the local DNA sequence environment.Hum. Genet., 86: 425â€”441,1991.

90. Ripley, L. S. Frameshift mutation: determinants of specificity. Annu. Rev. Genet., 24:189â€”213,1990.

91. Lambert, I. B., Napolitano, R. L., and Fuchs, R. P. Carcinogen-induced frameshiftmutagenesis in repetitive sequences. Proc. Natl. Acad. Sci. USA, 89: 1310â€”1314,1992.

92. Aaltonen, L. A., Peltomaki, P., Leach, F. S., et a!. Clues to the pathogenesis offamilial colorectal cancer. Science (Washington DC), 260: 812â€”815, 1993.

93. Thibodeau, S. N., Bren, G., and Schaid, D. Microsatellite instability in cancer of theproximal colon. Science (Washington DC), 260: 816â€”819, 1993.

94. Dunn, J. M., Phillips, R. A., Becker, A. J., and Gallie, B. L. Identification of germlineand somatic mutations affecting the retinoblastoma gene. Science (Washington DC),241:1797â€”1800,1988.

95. Harbour, J. W., Lxi, S. L., Whang-Peng, J., Gazdar, A. F., Minna, J. D., and Kaye,F. J. Abnormalities in structure and expression of the human retinoblastoma gene inSCLC. Science (Washington DC), 241: 353â€”357,1988.

96. Clore, G. M., Omichinski, J. G., Sakaguchi, K., Zambrano, N., Sakamoto, H.,Appella, E., and Gronenborn, A. M. High-resolution solution structure of the ohgomerization domain of p53 by multi-dimensional NMR. Science (Washington DC),265:386â€”391,1994.

97. Cho, Y., Gorina, S., Jeffrey, P., and Pavletich, N. P. Crystal structure of a p53 tumorsuppressor-DNA complex: a framework for understanding how mutations inactivatep53. Science (Washington DC), 265: 346â€”355,1994.

98. Unger, T., Nau, M. M., Segal, S., and Minna, J. D. p53: a transdominant regulator oftranscription whose function is ablated by mutations occurring in human cancer.EMBO J., 11: 1383â€”1390,1992.

99. Momand, J., Zambetti, G. P., Olson, D. C., George, D., and Levine, A. J. The mdm-2oncogene product forms a complex with the p53 protein and inhibits p53-mediatedtransactivation. Cell, 69: 1237â€”1245,1992.

100. Ohiner, J. D., Pietenpol, J. A., Thiagahingam, S., Gyuris, J., Kinzler, K. W., andVogelstein, B. Oncoprotein MDM2 conceals the activation domain of tumoursuppressor p53. Nature (Land.), 362: 857â€”862, 1993.

101. Chen, J., Marechal, V., and Levine, A. J. Mapping of the p53 and mdm-2 interactiondomains. Mol. Cell. Biol., 13: 4107â€”4114, 1993.

102. Brown, D. R., Deb, S., Munoz, R. M., Subler, M. A., and Deb, S. P. The tumorsuppressor p53 and the oncoprotein simian virus 40 T-antigen bind to overlappingdomains on the MDM2 protein. Mol. Cell. Biol., 13: 6849â€”6857, 1993.

103. Srinivasan, R., Roth, J. A., and Maxwell, S. A. Sequence-specific interaction of aconformational domain of p53 with DNA. Cancer Res., 53: 5361â€”5364,1993.

104. Ruppert, J. M., and Stillman, B. Analysis of a protein-binding domain of p53. Mol.Cell. Biol., 13: 3811â€”3820,1993.

105. Sturzbecher, H. W., Brain, R., Addison, C., Rudge, K., Remm, M., Grimahdi, M.,Keenan, E., and Jenkins, J. R. A C-terminal a-helix plus basic region motif is themajor structural determinant ofp53 tetramerization. Oncogene, 7: 1513â€”1523,1992.

106. Milner, J., and Medcalf, E. A. Cotranslation of activated mutant p53 with wild typedrives the wild-type p53 protein into the mutant conformation. Cell, 65: 765â€”774,1991.

107. Tarunina, M., and Jenkins, J. R. Human p53 binds DNA as a protein homodimer butmonomeric variants retain full transcription transactivation activity. Oncogene, 8:3165â€”3173,1993.

108. Pietenpol, J. A., Tokino, T., Thiagalingam, S., El-Deity, W. S., Kinzler, K. W., andVogelstein, B. Sequence-specific transcriptional activation is essential for growthsuppression by p53. Proc. Nath. Acad. Sci. USA, 91: 1998â€”2002, 1994.

109. Hainaut, P., and Milner, J. Redox modulation of p53 conformation and sequencespecific DNA binding in vitro. Cancer Res., 53: 4469â€”4473, 1993.

110. Martinez, J., Georgoff, I., and Levine, A. J. Cellular localization and cell cycleregulation by a temperature-sensitive p53 protein. Genes 0ev., 5: 151â€”159,1991.

111. Michalovitz, D., Halevy, 0., and Oren, M. Conditional inhibition of transformationand of cell proliferation by a temperature-sensitive mutant of p53. Cell, 62: 671â€”680, 1990.

112. Reed, M., Wang, Y., Mayr, G., Anderson, M. E., Schwedes, J. F., and Tegtmeyer,P. p53 domains: suppression, transformation, and transactivation. Gene Exp., 3:95â€”107,1993.

113. Davidoff, A. M., Iglehart, J. D., and Marks, J. R. Immune response to p53 isdependent upon p53IHSP7O complexes in breast cancers. Proc. NatI. Acad. Sci.USA, 89: 3439â€”3442, 1992.

114. Schhichtholz, B., Legros, Y., Gillet, D., Gaillard, C., Marty, M., Lane, D., Calvo, F.,and Soussi, T. The immune response to p53 in breast cancer patients is directedagainst immunodominant epitopes unrelated to the mutational hot spot. Cancer Res.,52: 6380â€”6384, 1992.

I 15. Crawford, L. V., Pim, D. C., and Bulbrook, R. D. Detection of antibodies against thecellular protein p53 in sera from patients with breast cancer. Int. J. Cancer, 30:403â€”408, 1982.

116. Caron de Fromentel, C., May-Levin, F., Mouriesse, H., Lemerle, J., Chandrasekaran,K., and May, P. Presence of circulating antibodies against cellular protein p53 in anotable proportion of children with B-cell lymphoma. Int. J. Cancer, 39: 185â€”189,1987.

117. Lubin, R., Schlichtholz, B., Bengoufa, D., et a!. Analysis of p53 antibodies inpatients with various cancers define B-cell epitopes of human p53: distribution on

4875




primary structure and exposure on protein surface. Cancer Res., 53: 5872â€”5876,I 993.

118. Vohkmann, M., Muller, M., Hofmann, W. J., Meyer, M., Hagelstein, J., Rath, U.,Kommerell, B., Zentgraf, H., and Galle, P. R. The humorahimmune response to p53in patients with hepatocellular carcinoma is specific for malignancy and independentof the a-fetoprotein status. Hepatology, 18: 559â€”565, 1993.

119. Schuler, G. D., Altschul, S. F., and Lipman, D. J. A workbench for multiplealignment construction and analysis. Proteins, 9: 180â€”190,1991.

120. Milne, D. M., Palmer, R. H., and Meek, D. W. Mutation of the casein kinase IIphosphorylation site abolishes the anti-proliferative activity of p53. Nucleic AcidsRes., 20: 5565â€”5570,1992.

121. Fiscella, M., Ullrich, S. J., Zambrano, N., Shields, M. T., Lin, D., Lees-Miller, S. P.,Anderson, C. W., Mercer, W. E., and Appella, E. Mutation of the serine 15phosphoryhation site of human p53 reduces the ability of p53 to inhibit cell cycleprogression. Oncogene, 8: 1519â€”1528,1993.

122. Fiscella, M., Zambrano, N., Uhlrich, S. J., Unger, T., Lin, D., Coo, B., Mercer, W. E.,Anderson, C. W., and Appella, E. The carboxy-terminal serine 392 phosphorylationsite of human p53 is not required for wild-type activities. Oncogene, in press, 1994.

123. Lehman, T. A., Bennett, W. P., Metcalf, R. A., Reddel, R., Welsh, J. A., Ecker, J.,Modahi,R. V., Ullrich, S., Romano, J. W., Appella, E., Testa, J. R., Gerwin, B. I.,and Harris, C. C. p53 mutations, ras mutations and p53-heat shock 70 proteincomplexes in human hung cell lines. Cancer Res., Si: 4090â€”4096, 1991.

124. Enomoto, T., Fujita, M., Inoue, M., Rice, J. M., Nakajima, R., Tanizawa, 0., andNomura, T. Alterations of the p53 tumor suppressor gene and its association withactivation of the c-K-ras-2 protooncogene in premahignant and malignant lesions ofthe human uterine endometrium. Cancer Res., 53: 1883â€”1888,1993.

125. Shaw, P., Tardy, S., Benito, E., Obrador, A., and Costa, J. Occurrence of Ki-ras andp53 mutations in primary colorectal tumors. Oncogene, 6: 2121â€”2128,1991.

126. Kalthoff, H., Schmiegeh, W., Roeder, C., et a!. p53 and K-RAS alterations inpancreatic epithehial cell lesions. Oncogene, 8: 289â€”298, 1993.

127. Nikolaidou, A., Liloglou, T., Malliri, A., Ergazaki, M., Tiniakos, G., Tiniakos, D.,and Spandidos, D. A. Detection of hepatitis B virus DNA and mutations in K-rasand p53 genes in human hepatocellular carcinomas. Int. J. Oncol., 3: 593â€”601,1993.

128. Rodenhuis, S., and Shebos, R. J. Clinical significance of ras oncogene activation inhuman lung cancer. Cancer Res., 52: 2665sâ€”2669s, 1992.

129. Dowell, S. P., and Halt, P. A. The clinical relevance of the p53 tumour suppressorgene. Cytopathol., 5: 133â€”145,1994.

130. Slaughter, T. P., Southwick, H. W., and Smejkel, W. Field cancerization in oralstratified epithehium. Cancer (Phila.), 6: 963-968, 1953.

131. Yin, J., Harpaz, N., Tong, Y., Huang, Y., Laurin, J., Greenwald, B. D., Hontanosas,M., Newkirk, C., and Mehtzer,S. J. p53 point mutations in dysplastic and cancerousulcerative colitis lesions. Gastroenterohogy, 104: 1633â€”1639,1993.

132. Nees, M., Homann, N., Discher, H., Andl, T., Enders, C., Herold-Mende, C.,Schuhmann, A., and Bosch, F. X. Expression of mutated p53 occurs in tumor-distantepithehia of head and neck cancer patients: a possible molecular basis for thedevelopment of multiple tumors. Cancer Res., 53: 4189â€”4196, 1993.

133. Chung, K. Y., Mukhopadhyay, T., Kim, J., Casson, A., Ro, J. Y., Goepfert, H.,Hong, W. K., and Roth, J. A. Discordant p53 gene mutations in primary head andneck cancers and corresponding second primary cancers of the upper aerodigestivetract. Cancer Rca, 53: 1676â€”1683,1993.

134. Thor, A. D., Moore, D. H., II, Edgerton, S. M., et a!. Accumulation of p53 tumorsuppressor gene protein: an independent marker of prognosis in breast cancers. J.NatI. Cancer Inst., 84: 845â€”855, 1992.

135. Tsuda, H., Iwaya, K., Fukutomi, T., and Hirohashi, S. p53 mutations and c-erbB-2amplification in intraductal and invasive breast carcinomas of high histologic grade.Jpn. J Cancer Res., 84: 394â€”401, 1993.

136. Murakami, Y., Hayashi, K., Hirohashi, S., and Sekiya, T. Aberrations of the tumorsuppressor p53 and retinoblastoma genes in human hepatocellular carcinomas.Cancer Rca., 51: 5520â€”5525, 1991.

137. Nishida, N., Fukuda, Y., Kokuryu, H., Toguchida, J., Yandell, D. W., Ikenega, M.,Imura, H., and Ishizaki, K. Role and mutational heterogeneity of the p53 gene inhepatocellular carcinoma. Cancer Res., 53: 368â€”372, 1993.

138. Kohler, M. F., Kerns, B. J., Humphrey, P. A., Marks, J. R., Brat, R. C., Jr., andBerchuck, A. Mutation and overexpression of p53 in early-stage epithehial ovariancancer. Obstet. Gynecol., 81: 643â€”650,1993.

139. Teneriello, M. G., Ebina, M., Linnoila, R. I., Henry, M., Nash, J. D., Park, R. C., andBirrer, M. J. p53 and Ki-ras gene mutations in epithehial ovarian neoplasms. CancerRes., 53: 3103-3108, 1993.

140. Vahakangas, K. H., Samet, J. M., Metcalf, R. A., Welsh, J. A., Bennett, W. P., Lane,D. P., and Harris, C. C. Mutations of p53 and ras genes in radon-associated lungcancer from uranium miners. Lancet, 339: 576â€”580, 1992.

141. Sundaresan, V., Ganly, P., Hasleton, P., Rudd, R., Sinha, G., Bleehen, N. M., andRabbitts, P. p53 and chromosome 3 abnormalities, characteristics of malignantlung tumours, are detectable in preinvasive lesions of the bronchus. Oncogene, 7:1989â€”1997,1992.

142. Sozzi, G., Miozzo, M., Donghi, R., Pilotti, S., Cariani, C. T., Pastorino, U., DellaPorta, G., and Pierotti, M. A. Deletions of l7p and p53 mutations in preneoplasticlesions of the hung. Cancer Rca., 52: 6079â€”6082, 1992.

143. Nuorva, K., Soini, Y., Kamel, D., Autio-Harmainen, H., Ristehi, L., Ristehi, J.,Vahakangas, K., and Paakko, P. Concurrent p53 expression in bronchial dysphasiasand squamous cell lung carcinomas. Am. J. Pathol., 142: 725â€”732,1993.

144. Bennett, W. P., Colby, T. V., Vahakangas, K. H., et a!. p53 protein accumulatesfrequently in early bronchial neophasia. Cancer Res., 53: 4817â€”4822, 1993.

145. Gusterson, B. A., Anbazhagan, R., Warren, W., Midgehy, C., Lane, D. P., O'Hare,M., Stamps, A., Carter, R., and Jayatilake, H. Expression of p53 in premalignant andmalignant squamous epithehium.Oncogene, 6: 1785â€”1789,1991.

146. Dolcetti, R., Doglioni, C., Maestro, R., Gasparouo, D., Barzan, L, Pastore, A.,Romanelli, M., and Boiocchi, M. p53 over-expression is an early event in thedevelopment of human squamous-cell carcinoma of the larynx: genetic and prognostic implications. Int. J. Cancer, 52: 178â€”182,1992.

147. Coltrera, M. D., Zarbo, R. J., Sakr, W. A., and Gown, A. M. Markers for dysphasiaof the upper aerodigestive tract. Suprabasal expression of PCNA, p53, and CK19 inalcohol-fixed, embedded tissue. Am. J Pathol., 141: 817â€”825, 1992.

148. Ogden, G. R., Lane, D. P., and Chishohm, D. M. p53 expression in dyskeratosiscongenita: a marker for oral premahignancy? J. Chin. Pathol., 46: 169â€”170,1993.

149. Boyle, J. 0., Hakim, J., Koch, W., van der Riet, P., Hruban, R. H., Roa, R. A.,Correo, R., Eby, Y. J., Ruppert, J. M., and Sidransky, D. The incidence of p53mutations increases with progression of head and neck cancer. Cancer Res., 53:4477â€”4480, 1993.

150. Bennett, W. P., Hohlstein, M. C., Metcalf, R. A., Welsh, J. A., He, A., Thu. S.,Kusters, I., Resau, J. H., Trump, B. F., Lane, D. P., and Harris, C. C. p53 mutationand protein accumulation during multistage human esophageal carcinogenesis. Cancer Res., 52: 6092â€”6097, 1992.

151. Sasano, H., Goukon, Y., Nishihira, T., and Nagura, H. In situ hybridization andimmunohistochemistry of p53 tumor suppressor gene in human esophageal cardnoma. Am. J. Pathol., 141: 545â€”550,1992.

152. Wang, L. D., Hong, J. Y., Qiu, S. L., Gao, H., and Yang, C. S. Accumulation of p53protein in human esophageal precancerous lesions: a possible early biomarker forcarcinogenesis. Cancer Res., 53: 1783â€”1787,1993.

153. Fearon, E. R., and Vogehstein, B. A genetic model for cohorectal tumorigenesis. Cell,61: 759â€”767,1990.

154. Kikuchi-Yanoshita, R., Konishi, M., Ito, S., a' a!. Genetic changes of both p53alleles associated with the conversion from cohorectal adenoma to early carcinomain familial adenomatous polyposis and non-familial adenomatous pohyposis patients.Cancer Res., 52: 3965â€”3971,1992.

155. Burmer, G. C., Crispin, D. A., Kolli, V. R., Haggitt, R. C., Kulander, B. G., Rubin,C. E., and Rabinovitch, P. S. Frequent loss of a p53 allele in carcinomas and theirprecursors in ulcerative colitis. Cancer Commun., 3: 167â€”172,1991.

156. Burmer, G. C., Rabinovitch, P. S., Haggitt, R. C., Crispin, D. A., Brentnahh, T. A.,Kohhi, V. R., Stevens, A. C., and Rubin, C. E. Neophastic progression in ulcerativecolitis: histology, DNA content, and loss of a p53 allele. Gastroenterology, 103:1602â€”1610,1992.

157. Pignatelhi, M., Stamp, G. W., Kafiri, G., Lane, D., and Bodmer, W. F. Overexpression of p53 nuclear oncoprotein in colorectal adenomas. Int. J. Cancer, 50:683â€”688, 1992.

158. van den Berg, F. M., Tigges, A. J., Schipper, M. E., den Hartog-Jager, F. C., Kroes,w. G., and Walboomers,J. M. Expressionof the nuclearoncogenep53 in colontumours. J. Pathol., 157: 193â€”199,1989.

159. Tsuda, H., and Hirohashi, S. Frequent occurrence of p53 gene mutations in uterinecancers at advanced clinical stage and with aggressive histological phenotypes. Jpn.J. Cancer Res., 83: 1184â€”1191,1992.

160. Kohler, M. F., Marks, J. R., Wiseman, R. W., Jacobs, I. J., Davidoff, A. M.,Clarke-Pearson, D. L, Soper, J. T., Bast, R. C., Jr., and Berchuck, A. Spectrum ofmutation and frequency of allehic deletion ofthe p53 gene in ovarian cancer. J. Natl.Cancer Inst., 85: 1513â€”1519,1993.

161. Navone, N. M., Troncoso, P., Pisters, L L., Goodrow, T. L., Palmer, J. L., Nichols,W. W., Von Eschenbach, A. C., and Conti, C. J. p53 protein accumulation and genemutation in the progression of human prostate carcinoma. J. Nath. Cancer Inst., 85:1657â€”1669,1993.

162. Fujimoto, K., Yamada, Y., Okijima, E., Kakizoe, T., Sasaki, H., Sugimura, T., andTerada, M. Frequent association of p53 gene mutation in invasive bladder cancer.Cancer Res., 52: 1393â€”1398,1992.

163. Habuchi, T., Takahashi, R., Yamada, H., et a!. Influence of cigarette smokingand schistosomiasis on p53 gene mutation in urothehial cancer. Cancer Res., 53:3795â€”3799,1993.

164. Chiba, I., Takahashi, T., Nau, M. M., D'Amico, D., Curiel, D. T., Mitsudomi, T.,Buchhagen, D. L., Carbone, D., Piantadosi, S., Koga, H., Reissman, P. T.,Slamon, D. J., Holmes, E. C., and Minna, J. D. Mutations in the p53 gene arefrequent in primary, resected non-small cell hung cancer. Oncogene, 5:1603â€”1610, 1990.

165. Lahmann, D., Puts, B., Reich, U., Bohm, J., PraOer, H., and HÃ¶fler,H. Mutationalspectrum of the p53 gene in human small-cell hung cancer and relationship tochinicopathohogicaldata. Am. J Pathol., 142: 907â€”915,1993.

166. Lee, N. K., Ye, Y. W., Chen, J., Li, X., Waber, P. G., and Nisen, P. D. p53,retinoblastoma, and human papihlomavirus in squamous cell carcinoma and adjacentnormal mucosa of the upper aerodigestive tract. Arch. Otoharyngol. Head NeckSurg., 119: 1125â€”1131,1993.

167. Sidransky, D., Mikkelsen, T., Schwechheimer, K., Rosenblum, M. L., Cavanee, W.,and Vogehstein, B. Clonal expansion of p53 mutant cells is associated with braintumour progression. Nature (Land.), 355: 846â€”847, 1992.

168. Hayashi, Y., Yamashita, J., and Yamaguchi, K. Timing and role of p53 genemutation in the recurrence of ghioma. Biochem. Biophys. Res. Commun., 180:1145â€”1150,1991.

169. Gasparini, G., Weidner, N., Mahuta, S., Pozza, F., Boracchi, P., Mezzetti, M.,Testohin, A., and Bevihacqua, P. Intratumoral microvessel density and p53 protein:correlation with metastasis in head-and-neck squamous-cell carcinoma. Int. J.Cancer, 55: 739â€”744,1993.

170. Chin, K. V., Ueda, K., Pastan, 1., and Gottesman, M. M. Modulation of activity ofthe promoter of the human MDRJ gene by Ras and p53. Science (Washington DC),255: 459â€”462,1992.

171. Dameron, K. M., Volpert, 0. V., and Bouck, N. Regulation of an angiogenesisinhibitor in human fibrobhasts by the p53 tumor suppressor gene. J. Cell Biochem.,

4876




18C: 187, 1994.172. Kihana, T., Tsuda, H., Teshima, S., Okada, S., Matsuura, S., and Hirohashi, S. High

incidence of pS3 gene mutation in human ovarian cancer and its association withnuclear accumulation of p53 protein and tumor DNA aneuploidy. Jpn. J. CancerRes., 83: 978â€”984,1992.

173. Remvikos, Y., Tominaga, 0., Hammel, P., Laurent-Puig, P., Salmon, R. J.,Dutrihhaux,B., and Thomas, G. Increased p53 protein content of colorectal tumourscorrelates with poor survival. Br. J. Cancer, 66: 758â€”764, 1992.

174. Jung, M., Notario, V., and Dritschiho, A. Mutations in the p.5.3 gene in radiationsensitive and -resistant human squamous carcinoma cells. Cancer Res., 52:6390â€”6393, 1992.

175. Brachman, D. G., Beckett, M., Graves, D., Haraf, D., Vokes, E., and Weichsehbaum,R. R. p53 mutation does not correlate with radiosensitivity in 24 head and neckcancer cell lines. Cancer Res., 53: 3667â€”3669,1993.

176. O'Connor, P. M., Jackman, J., Jondle, D., Bhatia, K., Magrath, I., and Kohn, K. W.Role of the p53 tumor suppressor gene in cell cycle arrest and radiosensitivity ofBurkitt's lymphoma cell lines. Cancer Rca., 53: 4776â€”4780, 1993.

177. Harris, C. C. Human tissues and cells in carcinogenesis research. Cancer Rca., 47:1â€”10,1987.

178. Chen, T. M., Chen, C. A., Hsieh, C. Y., Chang, D. Y., Chen, Y. H., and Defendi,V. The state of p53 in primary human cervical carcinomas and its effects inhuman papillomavirus-immortahized human cervical cells. Oncogene, 8: 1511â€”1518,1993.

179. Crook, T., Wrede, D., and Vousden, K. H. p53 point mutation in HPV negativehuman cervical carcinoma cell lines. Oncogene, 6: 873â€”875,1991.

180. Scheffner, M., Munger, K., Byrne, J. C., and Howley, P. M. The state ofthe p53 andretinoblastoma genes in human cervical carcinoma cell lines. Proc. Nath. Acad. Sci.USA, 88: 5523â€”5527,1991.

181. Fagin, J. A., Matsuo, K., Karmakar, A., Chen, D. L., Tang, S. H., and Koeffler, H. P.High prevalence of mutations of the p53 gene in poorly differentiated human thyroidcarcinomas. J. Chin. Invest., 91: 179â€”184,1993.

182. Wright, P. A., Lemoine, N. R., Goretzki, P. E., Wylhie, F. S., Bond, J., Hughes, C.,Roher, H. D., Williams, E. D., and Wynford-Thomas, D. Mutation of the p53 genein a differentiated human thyroid carcinoma cell line, but not in primary thyroidtumours. Oncogene, 6: 1693â€”1697,1991.

183. Effert, P., McCoy, R., Abdeh-Hamid, M., Flynn, K., Zhang, Q., Busson, P., Tursz,T., Liu, E., and Raab-Traub, N. Alterations of the p53 gene in nasopharyngealcarcinoma. J. Virol., 66: 3768â€”3775, 1992.

184. Spruck, C. H., Thai, Y. C., Huang@D. P., Yan& A. S., RideOut,W. M., GonzalezZulueta, M., Choi, P., La, K. W., Yu, M. C., and Jones, P. A. Absence of p53 genemutations in primary nasopharyngeal carcinomas. Cancer Res@52: 4787-4790, 1992.

185. Somers, K. D., Merrick, M. A., Lopez, M. E., Incognito, L. S., Schechter, G. L., andCasey, G. Frequent p53 mutations in head and neck cancer. Cancer Res., 52:5997â€”6000,1992.

186. Stretch, J. R., Gatter, K. C., Rahfkiaer,E., Lane, D. P., and Harris, A. L. Expressionof mutant p53 in melanoma. Cancer Rca., 51: 5976â€”5979,1991.

187. Council on Scientific Affairs, Harmful effects of ultraviolet radiation. JAMA, 262:380â€”384, 1989.

188. Reid, T. M., and Laeb, L. A. Mutagenic specificity of oxygen radicals produced byhuman leukemia cells. Cancer Rca., 52: 1082â€”1086,1992.

189. Reid, T. M., and Loeb, L. A. Tandem double CCâ€”.iT mutations are produced byreactive oxygen species. Proc. NatI. Acad. Sci. USA, 90: 3904â€”3907, 1993.

190. Brash, D. E., Rudolph, J. A., Simon, J. A., Lin, A., McKenna, G. J., Baden, H. P.,Halperin, A. J., and Ponten, J. A role for sunlight in skin cancer: UV-inducedp53 mutations in squamous cell carcinoma. Proc. NatI. Acad. Sci. USA, 88:10124â€”10128, 1991.

191. Nakazawa, H., English, D., Randell, P. L, Nakazawa, K., Martel, N., Armstrong,B. K., and Yamasaki, H. UV and skin cancer: specific p53 gene mutation in normalskin as a biologically relevant exposure measurement. Proc. Nath. Acad. Sci. USA,91: 360â€”364,1994.

192. Tornaletti, S., and Pfeifer, G. P. Slow repair of pyrimidine dimers at p53 mutationhotspots in skin cancer. Science (Washington DC), 263: 1436â€”1439, 1994.

193. Balch, C. M., Houghton, A. N., and Peters, L. J. Cutaneous melanoma. in: V. T.DeVita, S. Hehhman, and S. A. Rosenberg (eds.), Cancer: Principles and Practice ofOncology, pp. 1612â€”1662.Philadelphia: J. B. Lippincott Co., 1993.

194. McGregor, J. M., Yu, C. C., Dublin, E. A., Barnes, D. M., Levison, D. A., andMacDonald, D. M. p53 immunoreactivity in human malignant melanoma anddysplastic naevi. Br. J. Dermatol., 128: 606â€”611, 1993.

195. Kamb, A., Gruis, N. A., Weaver-Fehdhaus, J., Liu, Q., Harshman, K., Tavtigian,S. V., Stockert, E., Day, R. S., Johnson, B. E., and Skolnick, M. H. A cell cycleregulator potentially involved in genesis of many tumor types. Science (WashingtonDC), 264: 436â€”440, 1994.

196. Nebert, T., Miura, K., Wu, D. J., Lais, A., Takabayashi, K., and Carson, D. A.Deletions of the cychin-dependent kinase-4 inhibitor gene in multiple human cancers. Nature (Land.), 368: 753â€”756,1994.

197. Beashey, R. P. Hepatitis B virus. The major etiology of hepatocellular carcinoma.Cancer (Phila.), 61: 1942â€”1956,1988.

198. Beasley, R. P., Hwang, L. Y., Un, C. C., and Chien, C. S. Hepatocellular carcinomaand hepatitis B virus. A prospective study of 22,707 men in Taiwan. Lancet, 2:1129â€”1133,1981.

199. Koshy, R., and Hofschneider, P. H. Transactivation by hepatitis B virus maycontribute to hepatocarcinogenesis. Curr. Top. Microbiol. Immunoh., 144: 265â€”281,1989.

200. Kim, C. M., Koike, K., Saito, I., Miyamura, T., and Jay, G. HBx gene of hepatitisB virus induces liver cancer in transgenic mice. Nature (Land.), 351: 317â€”320,1991.

201. Feitelson, M. A., Zhu, M., Duan, L-X., and Landon, W. T. Hepatitis B X antigen and

p53 are associated in vitro and in liver tissues from patients with primary hepatocellular carcinoma. Oncogene, 8: 1109â€”1117, 1993.

202. Wang, X. W., Forrester, K., Yeh, H., Feitelson, M. A., Gu, J. R., and Harris, C. C.Hepatitis B virus X protein inhibits p53 sequence-specific DNA binding, transcriptional activity, and association with transcription factor ERCC3. Proc. Nail. Acad.Sci. USA, 91: 2230â€”2234,1994.

203. Wogan, G. N. Aflatoxins as risk factors for hepatocellular carcinoma in humans.Cancer Res., 52: 21 14sâ€”2118s, 1992.

204. Buhatao-Jayme, J., Almero, E. M., Castro, M. C., Jardeheza, M. T., and Sahamat,L. A. A case-control dietary study of primary liver cancer risk from aflatoxinexposure. Int. J. Epidemiol., 11: 112â€”119,1982.

205. Peers, F., Bosch, X., Kaldor, J., Linsell, A., and Phuijmen, M. Aflatoxin exposure,hepatitis B virus infection andhiver cancer in Swaziland. Int. J. Cancer, 39:545-553,1987.

206. Van Rensburg, S. J., Cook-Mozaffari, P., Van Schalkwyk, D. J., Van der Watt, J. J.,Vincent, T. J., and Purchase, I. F. Hepatocelluhar carcinoma and dietary aflatoxin inMozambique and Transkei. Br. J. Cancer, 51: 713â€”726,1985.

207. Yeh, F. S., Yu, M. C., Mo, C. C., Lao, S., Tong, M. J., and Henderson, B. E.Hepatitis B virus, aflatoxins, and hepatocehiular carcinoma in southern Guangxi,China.CancerRes.,49: 2506-2509,1989.

208. Bressac, B., Kew, M., Wands, J., and Ozturk, M. Selective G to T mutations ofp53 gene in hepatocelluhar carcinoma from southern Africa. Nature (Land.), 350:429â€”431, 1991.

209. Chahlen, C., Lunec, J., Warren, W., Collier, J., and Bassendine, M. F. Analysis of thep53 tumor-suppressor gene in hepatocelluhar carcinomas from Britain. Hepatohogy,16: 1362â€”1366,1992.

210. Hohhstein, M. C., Wild, C. P., Bleicher, F., Chutimataewin, S., Harris, C. C.,Srivatanakuh, P., and Montesano, R. p53 mutations and aflatoxin B, exposure inhepatocehlular carcinoma patients from Thailand. Int. J. Cancer, 53: 51â€”55,1993.

21 1. Li, D., Ceo, Y., He, L., Wang, N. J., and Gu, J. Aberrations of p53 gene in humanhepatocellular carcinoma from China. Carcinogenesis (Land.), 14: 169â€”173,1993.

212. Oda, T., Tsuda, H., Scarpa, A., Sakamoto, M., and Hirohashi, S. p53 gene mutationspectrum in hepatocelhular carcinoma. Cancer Res., 52: 6358â€”6364, 1992.

213. Debuire, B., Paterhini, P., Pontisso, P., Basso, G., and May, E. Analysis of the p.53gene in European hepatocellular carcinomas and hepatoblastomas. Oncogene, 8:2303â€”2306, 1993.

214. Buetow, K. H., Sheffield, V. C., Zhu, M., Thou, T., Shen, F., Hino, 0., Smith, M.,McMahon, B. J., Lanier, A. P., Landon, W. T., Redeker, A. G., and Govindarajan,S. Law frequencyof p53 mutationsobservedin a diversecollectionof primaryhepatocehiuharcarcinomas. Proc. Nail. Acad. Sci. USA, 89: 9622â€”9626,1992.

215. Teramoto, T., Satonaka, K., Kitazawa, S., Fujimori, T., Hayashi, K., and Maeda, S.p53 gene abnormalities are closely related to hepatoviral infections and occur at alate stage of hepatocarcinogenesis. Cancer Res., 54: 231â€”235,1994.

216. Fujimoto, Y., Hampton, L. L., Wirth, P. J., Wang, N. J., Xie, J. P., and Thorgeirsson,S. S. Alterations of tumor suppressor genes and allehiclosses in human hepatocelhular carcinomas in China. Cancer Res., 54: 281-285, 1994.

217. Ross, R. K., Yuan, J. M., Yu, M. C., Wogan, G. N., Qian, G. S., Tu, J. T., Groopman,J. D., Gao, Y. T., and Henderson, B. E. Urinary aflatoxin biomarkers and risk ofhepatocellular carcinoma. Lancet, 339: 943â€”946,1992.

218. Qian, G. S., Ross, R. K., Yu, M. C., Yuan, J. M., Gao, Y. T., Henderson, B. E.,Wogan, G. N., and Groopman, J. D. A follow-up study of urinary markers ofaflatoxin exposure and liver cancer risk in Shanghai, People's Republic of China.Cancer Epidemiol. Biomarkers. Prey., 3: 3â€”10,1994.

219. Pfeifer, A. M. A., Cole, K. E., Smoot, D. T., Weston, A., Groopman, J. D., Shields,P. G., Vignaud, J-M., Juillerat, M., Lipsky, M. M., Trump, B. F., Lechner, J. F., andHarris, C. C. SV4O T-antigen immortalized normal human liver epithehial cellsexpress hepatocyte characteristics and metabolize chemical carcinogens. Proc. Nath.Acad. Sci. USA, 90: 5123â€”5127,1993.

220. Autrup, H., Harris, C. C., Wu, S. M., Bao, L Y., Pci, X. F., Lu, S., Sun, T. T., andHsia, C. C. Activation of chemical carcinogens by cultured human fetal liver,esophagus and stomach. Chem. Bioh. Interact., 50: 15â€”25,1984.

221. Aoyama, T., Yamano, S., Guzehian, P. 5., Gelboin, H. V., and Gonzalez, F. J. Fiveof 12 forms of vaccinia virus-expressed human hepatic cytochrome P450 metabolically activate aflatoxin B,. Proc. Nath. Acad. Sci. USA, 87: 4790â€”4793, 1990.

222. Lin, J. K., Miller, I. A., and Miller, E. C. 2,3-Dihydro-2-(guan-7-yl)-3-hydroxyaflatoxin B,, a major acid hydrolysis product of aflatoxin B,-DNA or -ribosomalRNA adducts formed in hepatic microsome-mediated reactions and in rat liver invivo. Cancer Res., 37: 4430â€”4438, 1977.

223. Essigmann, J. M., Croy, R. G., Nadzan, A. M., Busby, W. F., Jr., Reinhold, V. N.,Buchi, G., and Wogan, G. N. Structural identification of the major DNA adductformed by aflatoxin B, in vitro. Proc. Nath. Acad. Sci. USA, 74: 1870â€”1874,1977.

224. McMahon, G., Davis, E. F., Huber, L J., Kim, Y., and Wogan, G. N. Characterization of c-Ki-ras and N-ms oncogenes in aflatoxin B,-induced rat liver tumors.Proc. NaIl. Acad. Sci. USA, 87: 1104-1108, 1990.

225. Levy, D. D., Groopman, J. D., Lim, S. E., Seidman, M. M., and Kraemer, K. H.Sequence specificity of aflatoxin B,-induced mutations in a phasmid replicated inxeroderma pigmentosum and DNA repair proficient human cells. Cancer Res., 52:5668â€”5673, 1992.

226. Trottier, Y., Waithe, W. I., and Anderson, A. Kinds of mutations induced byaflatoxin B, in a shuttle vector replicating in human cells transiently expressingcytochrome P450IA2 cDNA. Mol. Carcinog., 6: 140â€”147,1992.

227. Aguilar, F., Hussain, S. P., and Cerutti, P. Aflatoxin B, induces the transversion ofG to T in codon 249 of the p53 tumor suppressor gene in human hepatocytes. Proc.NatI. Acad. Sci. USA, 90: 8586â€”8590, 1993.

228. Aguilar, F., Harris, C. C., Sun, T., Hollstein, M., and Cerutti, P. Geographicvariation of p53 mutational profile in nonmalignant human liver. Science (Wash.

4877




ingtonDC),264:1317â€”1319,1994.229. DeMarini, D. M. Genotoxicity of tobacco smoke and tobacco smoke condensate.

Mutat. Res., 114: 59â€”89, 1983.230. Shophand, D. R., Eyre, H. J., and Pechacek, T. F. Smoking-attributable cancer

mortality in 1991 : is lung cancer now the leading cause of death among smokers inthe United States? J. Nath. Cancer Inst., 83: 1142â€”1148, 1991.

231 . Doll, R., and Peto, R. Cigarette smoking and bronchial carcinoma: dose and timerelationships among regular smokers and lifelong non-smokers. J. Epidemiol. Cammun. Health, 32: 303â€”313,1978.

232. U.S. Department of Health, Education and Welfare Smoking and Health. in: AReport of the Surgeon General. DHEW Publication No. (PHS) 79â€”50066,pp. 5-11,1988.

233. Ronai, Z. A., Gradia, S., Peterson, L. A., and Hecht, S. S. G to A transitions and Gto T transversions in codon 12 of the Ki-ras oncogene isolated from mouse lungtumors induced by 4-(methylnitrosamino)-l-(3-pyridyh)-l-butanone (NNK) and rehated DNA methyhating and pyridyloxobutyhating agents. Carcinogenesis (Land.),14:2419â€”2422,1993.

234. Suzuki, H., Takahashi, T., Kuroishi,T., Suyama, M., Ariyoshi, Y., and Ueda, R.p53mutations in non-small cell hung cancer in Japan: association between mutations andsmoking. Cancer Res., 52: 734â€”736, 1992.

235. Takeshima, Y., Scyama, T., Bennett, W. P., Metcalf, R. A., Welsh, J. A., Akiba, S.,Fujihara, M., Hayashi, Y., Yonehara, S., Ito, T., Mizuno, T., hnai, K., Yamakido, M.,Nakamura, N., Akiyama, M., Tokuoka, S., Mabuchi, K., Land, C. E., and Harris,C. C. p53 gene mutations in lung cancers from Japanese non-smokers and atomicbomb survivors. Lancet, 342: 1520â€”1521,1993.

236. Hanawalt, P., and Mellon, I. Stranded in an active gene. Curr. Biol., 3: 67â€”69,1993.237. Grafstrom, R. C., Fornace, A. J., Jr., Autrup, H., Lechner, J. F., and Harris, C. C.

Formaldehyde damage to DNA and inhibition of DNA repair in human bronchialcells. Science (Washington DC), 220: 216â€”218, 1983.

238. McDowell, E. M., and Trump, B. F. Histogenesis of preneoplastic and neophasticlesions in tracheobronchial epithehium. Sun'. Synth. Pathol. Res., 2: 235â€”279,1983.

239. Berendsen, H. H., de Leij, L., Poppema, S., Postmus, P. E., Boes, A., Sluiter, H. J.,and The, H. Clinical characterization of non-small-cell hung cancer tumors showingneurocndocrine differentiation features. J. Chin. Oncoh., 7: 1614â€”1620,1989.

240. Graziano, S. L, Mazid, R., Newman, N., Tatum, A., 01cr, A., Mortimer, J. A.,Gulho, J. J., DiFino, S. M., and Scahzo, A. J. The use of neuroendocrine immunoperoxidase markers to predict chemotherapy response in patients with non-small-cellhung cancer. J. Chin. Oncol., 7: 1398â€”1406,1989.

241. Takahashi, T., Suzuki, H., Hida, T., Sekido, Y., Ariyoshi, Y., and Ueda, R. The p53gene is very frequently mutated in small-cell hung cancer with a distinct nucleotidesubstitution pattem. Oncogene, 6: 1775â€”1778,1991.

242. Miller, C. W., Simon, K., Asho, A., Kok, K., Yokota, J., Buys, C. H. C. M., Terada,M., and Koeffler, H. P. p.53 mutations in human hung tumors. Cancer Rca., 52:1695â€”1698,1992.

243. D'Amico, D., Carbone, D., and Mitsudomi, 1. High frequency of somaticallyacquired p53 mutations in small cell lung cancer cell lines and tumors. Oncogene,7: 339â€”346,1992.

244. Sameshima, Y., Matsuno, Y., Hirohashi, S., Shimosato, Y., Mizoguchi, H., Sugimura, T., Terada, M., and Yokota, J. Alterations of the p53 gene are common andcritical events for the maintenance of malignant phenotypes in small-cell lungcarcinoma. Oncogene, 7: 451â€”457, 1992.

245. Kishimoto, Y., Murakami, Y., Shiraishi, M., Hayashi, K., and Sekiya, T. Aberrationsof the p53 tumor suppressor gene in human non-small cell carcinomas of the hung.Cancer Res., 52: 4799â€”4804,1992.

246. Lohmann, D., Putz, B., Reich, U., Bohm, J., Prauer, H., and Hofler, H. Mutationalspectrum of the p53 gene in human small-cell hung cancer and relationship tochinico-pathohogicah data. Am. J. Pathol., 142: 907â€”915,1993.

247. Osann, K. E., Anton-Culver, H., Kurosaki, T., and Taylor, 1. Sex differences inhung-cancer risk associated with cigarette smoking. Int. J. Cancer, 54: 44â€”48,1993.

248. Jedrychowski, W., Becher, H., Wahrendorf, J., Basa-Cierpialek, Z., and Gomola, K.Effect of tobacco smoking on various histological types of lung cancer. J. CancerRes. Chin. Oncol., 118: 276â€”282,1992.

249. Stockwell, H. G., Goldman. A. L, Lyman, G. H., Noss, C. I., Armstrong, A. W.,Pinkham,P. A., Candelora,E. C., and Brusa, M. R. Environmentaltobaccosmoke andhung cancer risk in nonsmoking women. J Nati. Cancer Inst., 84: 1417â€”1422,1992.

250. Blum, A. Curtailing the tobacco pandemic. in: V. T. DeVita, S. Hehhman,and S. A.Rosenberg (eds.), Cancer: Principle and Practice of Oncology, pp. 480â€”491.Philadehphia: J. B. Lippincott, 1993.

251. Szekehy, L., Sehivanova, G., Magnusson, K. P., Klein, G., and Wiman, K. G.EBNA-5, an Epstein-Barr virus-encoded nuclear antigen, binds to the retinobhastomaand p53 proteins. Proc. Nath.Acad. Sci. USA, 90: 5455â€”5459,1993.

252. Nishihira, T., Hashimoto, Y., Katayama, M., Mon. S., and Kuroki, T. Molecular andcellular features of esophageal cancer cells. J. Cancer Res. Clin. Oncol., 119:441â€”449,1993.

253. Blot, W. J., Devesa,S. S., Kneller,R. W., and Fraumeni,J. F., Jr. Rising incidenceofadenocarcinomaof the esophagusand gastric cardia. JAMA, 265: 1287â€”1289,1991.

254. Blot, W. J., Devesa, S. S., and Fraumeni, J. F., Jr. Continuing climb in rates ofesophageal adenocarcinoma: an update. JAMA, 270: 1320, 1993.

255. Hartge, P., Silverman, D., Hoover, R., el a!. Changing cigarette habits and bladdercancer risk: a case-control study. J. Nath.Cancer Inst., 78: 1119â€”1125, 1987.

256. Zhang, Z. F., Sarkis, A. S., Cordon-Cardo, C., Dalbagni, G., Mehamed,J., Aprikian,A., Pohiack,D., Sheinfeld, J., Herr, H. W., Fair, W. R., Reuter, V. E., and Begg, C.Tobacco smoking, occupation, and p53 nuclear overexpression in early stage bladdercancer. Cancer Epidemiol. Biomarkers Prey., 3: 19â€”24,1994.

257. Spruck, C. H., III, Rideout, W. M., III, Olumi, A. F., Ohneseit, P. F., Yang, A. S.,Tsai, Y. C., Nichols, P. W., Horn, T., Hermann, G. G., Steven, K., Ross, R. K., Yu,

M. C., and Jones, P. A. Distinct pattem of p53 mutations in bladder cancer:relationship to tobacco usage. Cancer Res., 53: 1162â€”1166, 1993.

258. Samet, J. M. Radon and lung cancer. J. Nail. Cancer Inst., 81: 745â€”757,1989.259. Moolgavkar, S. H., Luebeck, E. G., Krewski, D., and Ziehinski, J. M. Radon,

cigarette smoke, and lung cancer: a re-analysis of the colorado plateau uraniumminers' data. Epidemiology, 4: 204â€”217,1993.

260. Neuberger, J. S. Residential radon exposure and hung cancer: an overview ofongoing studies. Health Phys., 63: 503â€”509,1992.

261. Samet, J. M., and Homung, R. W. Review of radon and lung cancer risk. Risk Anal.,10: 65â€”75,1990.

262. Pershagen, G., Akerbiom, G., Axelson, 0., et a!. Residential radon exposure andlung cancer in Sweden. N. Engl. J. Med., 330: 159â€”164,1994.

263. Taylor, J. A., Watson, M. A., Devereux, T. R., Michels, R. Y., Saccomanno, G., andAnderson, M. Mutational hotspot in the p53 gene in radon-associated lung tumorsfrom uranium miners. Lancet, 343: 86â€”87,1993.

264. Scheffner, M., Wemess, B. A., Huibregtse, J. M., Levine, A. J., and Howley, P. M.The E6 oncoprotein encoded by human papillomavirus types 16 and 18 promotes thedegradation of p53. Cell, 63: 1129â€”1136, 1990.

265. Liang, X. H., Volkmann, M., Klein, R., Herman, B., and Lackett, S. J. Calocalization of the tumor-suppressor protein p53 and human papillomavirus E6protein in human cervical carcinoma cell lines. Oncogene, 8: 2645â€”2652,1993.

266. Srivastava, S., Tong, Y. A., Devadas, K., Thu. Z. Q., Chen, Y., Pirollo, K. F., andChang, E. H. The status of the p53 gene in human papilloma virus positive or negativecervicalcarcinomacell lines. Carcinogenesis(Land.), 13: 1273â€”1275,1992.

267. Crook, T., Wrede, D., Tidy, J. A., Mason, W. P., Evans, D. J., and Vousden, K. H.Chonahp53 mutation in primary cervical cancer: association with human-papillomavirus-negative tumours. Lancet, 339: 1070â€”1073, 1992.

268. Borresen, A. L, Helland, A., Nesland, J., HoIm, R., Trope, C., and Kaern, J.Papillomaviruses, p53, and cervical cancer. Lancet, 339: 1350â€”1351,1992.

269. Fujita, M., Inoue, M., Tanizawa, 0., Iwamoto, S., and Enomoto, T. Alterations of thep53 gene in human primary cervical carcinoma with and without human papilhomavirus infection. Cancer Res., 52: 5323â€”5328,1992.

270. Busby-Earle, R. M., Steel, C. M., Williams, A. R., Cohen, B., and Bird, C. C.Papilhomaviruses, PS3, and cervical cancer. Lancet, 339: 1350, 1992.

271. Naito, M., Satake, M., Sakai, E., Hirano, Y., Tsuchida, N., Kanzaki, H., Ito, Y., andMori, T. Detection ofp53 gene mutations in human ovarian and endometrial cancersby pohymerase chain reaction-single strand conformation polymorphism analysis.Jpn. J. Cancer Rca., 83: 1030â€”1036,1992.

272. Willey, J. C., and Harris, C. C. Cellular and molecular biological aspects of humanbronchogemc carcinogenesis. CRC Crit. Rev. Hematol. Oncol., 10: 181â€”209,1990.

273. Jones, P. A., Buckley, J. D., Henderson, B. E., Ross, R. K., and Pike, M. C. Fromgene to carcinogen: a rapidly evolving field in molecular epidemiology. Cancer Res.,51: 3617â€”3620,1991.

274. Nagase, H., and Nakamura, Y. Mutations of the APC gene. Hum. Mutat., 2:425-434,1993.

275. Giannehhi,F., Green, P. M., High, K. A., et aL HaemophihiaB: database of pointmutations and short additions and deletionsâ€”fourthedition, 1993. Nucleic ACIdSRes., 21: 3075â€”3087,1993.

276. Sugimura, T. Studies on environmental chemical carcinogenesis in Japan. Science(Washington DC), 233: 312â€”318,1986.

277. Ziehenska, M., Beranek, D., and Guttenphan, J. B. Different mutational profilesinduced by N-nitroso-N-ethylurea: effects of dose and error-prone DNA repair andcorrelations with DNA adducts. Environ. Mol. Mutagen., 11: 473â€”485, 1988.

278. Singer, B. in vivo formation and persistence of modified nucheosides resulting fromalkylating agents. Environ. Health Perspect., 62: 41â€”48,1985.

279. Pe@, A. E. Methyhation of the O@position of guanine in DNA is the most likelyinitiating event in carcinogenesis by methyhating agents. Cancer Invest., 2: 223â€”231,1984.

280. Sasa, M., Kondo, K., Komaki, K., Uyama, T., Morimoto, T., and Monden, Y.Frequency of spontaneous p53 mutations (CpG site) in breast cancer in Japan.Breast. Cancer Rca. Treat., 27: 247â€”252,1993.

281. de Jong, P. J., Grosovsky, A. J., and Ghickman,B. W. Spectrum of spontaneousmutation at the APRT locus of Chinese hamster ovary cells: an analysis at the DNAsequence level. Proc. Nati. Acad. Sci. USA, 85: 3499â€”3503, 1988.

282. Drobetsky, E. A., Grosovsky, A. J., and Ghickman, B. W. The specificity ofUV-induced mutations at an endogenous locus in mammalian cells. Proc. NatI.Acad. Sci. USA, 84: 9103â€”9107,1987.

283. Mazur, M., and Ghickman, B. W. Sequence specificity of mutations induced bybenzo[a]pyrene-7,8-dioh-9,10-epoxide at endogenous aprt gene in CHO cells.Somat. Cell Mol. Genet., 14: 393â€”400,1988.

284. Halevy, 0., Rodeh, J., Pehed, A., and Oren, M. Frequentp53 mutations in chemicallyinduced murine fibrosarcoma. Oncogene, 6: 1593â€”1600,1991.

285. Lozano, J. C., Nakazawa, H., Cros, M. P., Cabral, R., and Yamasaki, H. Gâ€”@Amutations in p53 and Ha-ras genes in esophageal papillomas induced by N-nitrosomethyhbenzyhamine in two strains of rats. Mol. Carcinog., 9: 33â€”39,1994.

286. Smith, M. L, Yeheswarapu, L, Scalamogna, P., Locker, J., and Lambardi, B. p53mutations in hepatocelluharcarcinomas induced by a choline-devoid diet in maleFischer-344 rats. Carcinogenesis (Land.), 14: 503â€”510,1993.

287. Hegi, M. E., Soderkuist, P., Foley, J. F., et a!. Characterization of p53 mutations inmethyhene chloride-induced lung tumors from B6C3F1 mice. Carcinogenesis(Land.), 14: 803â€”810,1993.

288. Goodrow, T. L., Storer, R. D., Leander, K. R., et a!. Murine p53 intron sequence 5â€”8and their use in pohymerase chain reaction/direct sequencing analysis of p53 mutations in CD-i mouse liver and hung tumors. Mol. Carcinog., 5: 9â€”15,1992.

4878



1994;54:4855-4878. Cancer Res M. S. Greenblatt, W. P. Bennett, M. Hollstein, et al. Etiology and Molecular Pathogenesis

Tumor Suppressor Gene: Clues to Cancerp53Mutations in the

Updated version

http://cancerres.aacrjournals.org/content/54/18/4855.citation

Access the most recent version of this article at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected] at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cancerres.aacrjournals.org/content/54/18/4855.citationTo request permission to re-use all or part of this article, use this link



http://cancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]



mutations in the p53 tumor suppressor gene: clues to ... · the p53 tumor suppressor gene has come...

Documents