Hypermutability in Carcinogenesis

Corresponding author: Bernard S. Strauss, Department of Molecular Genetics and Cell Biology, The University of Chicago, 920 E. 58th St., Chicago, IL 60637, bs19{at}midway.uchicago.edu (E-mail).

	ABSTRACT

TOP ABSTRACT Are tumors hypermutable? TP53 mutation Silent mutation Multiple TP53 mutations Mutations in other genes Other examples of... LITERATURE CITED

The presence of numerous chromosomal changes and point mutations in tumors is well established. At least some of these changes play a role in the development of the tumors. It has been suggested that the number of these genetic changes requires that tumorigenesis involves an increase in mutation rate. However, the presence of numerous changes can also be accounted for by efficient selection. What is required to settle the issue is some measure of nonselected mutations in tumors. In order to determine whether the tumor suppressor TP53 (coding for the protein p53) is hypermutable at some stage of carcinogenesis, the frequency of silent and multiple mutations in this gene has been examined. Silent mutations make up ~3% of the total recorded but constitute 9.5% of the mutations found in tumors with multiple mutations. Multiple closely linked mutations are also observed. Such multiple mutations suggest the operation of an error-prone replication process in a subclass of cells. The published data indicate that TP53 is hypermutable at some stage of tumor development. It is not yet clear whether TP53 is unique or whether other genes display a similar pattern of silent and multiple mutations.

	Are tumors hypermutable?

TOP ABSTRACT Are tumors hypermutable? TP53 mutation Silent mutation Multiple TP53 mutations Mutations in other genes Other examples of... LITERATURE CITED

Tumors accumulate genetic alterations as they progress, and it seems very likely that at least a portion of these genetic alterations play a role in the etiology of the disease (VOGELSTEIN et al. 1988 ; GOYETTE et al. 1992 ). This accumulation of changes is so striking that many investigators suppose that a condition of genetic instability is a necessary concomitant of tumorigenesis (LOEB et al. 1974 ; LOEB 1991 ; CHENG and LOEB 1993 ). The genetic changes observed are of different kinds and include both point mutations and qualitative and quantitative chromosome alterations. Chromosome instability might occur without affecting the fidelity of the replication process, that is, without changing the rate of gene mutation. The question to be addressed here is whether a general increase in the point mutation rate occurs at some stage in carcinogenesis.

It has been argued that the normal mutation rate is insufficient to generate the changes required in a single cell and that there must, therefore, be a special process at work in cells destined to engender tumors (LOEB 1991 ). Some years ago it would have been supposed that exogenous environmental agents acting as mutagens were in large part responsible for this increase (AMES 1973), but such arguments are no longer as compelling (AMES et al. 1995 ). Many of the changes observed in tumors are associated with G:CA:T transitions characteristic of spontaneous change rather than the G:CT:A transversions characteristic of chemical mutagenesis (HOLLSTEIN et al. 1991 ). Current hypotheses suggest that the developing tumor cells carry "mutator" genes that diminish normal cellular surveillance mechanisms and increase the spontaneous mutation rate (LOEB 1994 ). Such surveillance mechanisms do exist (MODRICH 1995 ), and certain types of tumors occur more frequently in individuals with defects in genes responsible for the coding of the necessary surveillance proteins. The most spectacular recent finding is the association of certain colon carcinomas with defects in the mismatch repair genes (FISHEL et al. 1993 ; MODRICH 1995 ), but the earlier association of a defect in excision repair with skin carcinomas is equally striking (CLEAVER 1968 ). These correlations are persuasive arguments for the tumor hypermutability hypothesis.

The mutation frequency for point mutations is very significantly increased in the cells of individuals deficient in mismatch repair (BHATTACHARYYA et al. 1995 ; MALKHOSYAN et al. 1996 ). However, PARSONS et al. 1995 found a subset of patients whose somatic cells were devoid of mismatch repair activity and were hypermutable, but such individuals did not have the number of early tumors that might be expected. In addition, the tumors that do occur in patients with hereditary nonpoyposis colorectal cancer represent a limited set, that is, certain kinds of tumors occur but not others. It may be that the importance of mismatch repair deficiency is not primarily its effect on the generalized mutation rate but rather on the mutation rate of genes containing repeated nucleotide sequences within the coding region of critical genes, for example, the observed frameshift in a run of 10 (A's) in the TGF-ßII receptor gene (LU et al. 1995 ; MARKOWITZ et al. 1995 ; AKIYAMA et al. 1996 ; RAMPINO et al. 1997 ; SOUZA et al. 1996 ). Protection against single base changes that occur at the growing point during DNA synthesis is provided by both the mismatch repair and proofreading systems (FRIEDBERG et al. 1995 ). Frameshifts in long-repeated sequences are to some extent ignored by the proofreading system but are efficiently recognized by the mismatch repair system (STRAUSS et al. 1997 ; TRAN et al. 1997 ). Mismatch repair deficiency may therefore affect genes with such sequences to a disproportionate extent. If this explanation is correct, mismatch repair deficiency makes certain critical changes more likely, but the general increase in mutation need not result in generalized tumorigenesis. Demonstration that hypermutability is a general property of cells undergoing neoplastic transformation requires additional kinds of evidence.

The example of xeroderma pigmentosum has long been used as an argument for the role of DNA repair in removing tumorigenic lesions (CLEAVER 1968 ) and, indirectly, for generalized mutagenesis as a causative factor in cancer induction. One of the signs of xeroderma pigmentosum is an increased frequency of sunlight-induced skin carcinoma. Such tumors are associated with dinucleotide "signature" mutations (BRASH et al. 1991 ; DUMAZ et al. 1993 ; MATSUMURA et al. 1996 ), and the lack of repair activity ensures the persistence of sunlight-induced lesions long enough to result in replication errors. These characteristics of xeroderma patients link mutation and cancer. Some years ago, CAIRNS 1981 proposed that the xeroderma defect resulted only in skin tumors, and his paper provoked the collection of data showing that internal tumors are also increased in xeroderma patients (KRAEMER et al. 1984 ; KRAEMER et al. 1994 ). "Knockout" xeroderma mice (XPA deficient) do develop spontaneous liver tumors and an increased frequency of benzpyrene-induced lymphomas as compared to normal mice (DE VRIES et al. 1997 ). Deficiency in repair synthesis, however, is not necessarily the cause of the tumorigenesis because individuals with trichothiodystropy and some cases of Cockaynes' syndrome have excision repair activities as low as some xeroderma patients but do not have an increased frequency of skin tumors, notwithstanding their sensitivity to light. Despite several attempts, this paradox has not been explained (CHU and MAYNE 1996 ; MARIONNET et al. 1996 ; STARY and SARASIN 1996 ; TAKAYAMA et al. 1996 ; WEEDA et al. 1997 ). The genes involved are complex, with both repair and transcription functions. These examples suggest that heightened mutability is not necessarily accompanied by a general increase in tumorigenesis and also that increased mutability can be associated with only specific sorts of tumors.

Arguments against the hypothesis that tumors are necessarily hypermutable have a long history (ARMITAGE and DOLL 1957 ; FISHER 1958 ). Two recent discussions have been published by TOMLINSON et al. 1996 and by SIMPSON 1997 . The development of each tumor is an evolutionary event and, as in macroevolution, mutations are subject to selection, and some mutations are neutral and subject to genetic drift. The presence of any number of mutations in a tumor is the result of these different forces. Both very early and very recent calculations indicate that, given the power of selection and the biology of tumor development, selection is capable of yielding clones with multiple mutations without an increase in the normal mutation rate. Such calculations do not settle the issue, but they certainly indicate that the presence of multiple mutations in tumors is not sufficient to establish the case for general tumor hypermutability.

Some of the evidence that tumor cells are not necessarily hypermutable has already been summarized. As an example, HARWOOD et al. 1991 reported that two lines derived from a human colorectal carcinoma had mutation rates of 5 x 10^-8 and 3 x 10^-8 per cell per generation for the apr t gene, rates "as low or lower" than for diploid human cell lines. WITTENKELLER et al. 1997 compared the mutation rates of hypoxanthine-guanine phosphoribosyltransferase in an immortalized human bronchial epithelial cell line and its tumorigenic derivative and found that the tumor line had slightly lower rates. Of course, the control "line" had been immortalized so that even though not a tumor, it was hardly "normal." BUETTNER et al. 1996 found an increased mutation frequency (for a transgenic lac gene) in only one of four tumors arising in p53 knockout mice, and even in that tumor the increase was only 2.3 fold. The p53 knockout mutation rate in normal tissue was not different from the normal (NISHINO et al. 1995 ). The rate of point mutation is not permanently higher than the normal somatic mutation rate for some genes in (some) tumor cells.

	TP53 mutation

TOP ABSTRACT Are tumors hypermutable? TP53 mutation Silent mutation Multiple TP53 mutations Mutations in other genes Other examples of... LITERATURE CITED

The TP53 tumor suppressor gene, whose product, the p53 protein, plays an important role in regulation of the cell cycle (LEVINE 1997 ), is much studied: numerous mutations have been reported, and there are several data bases available that summarize the information (DEVRIES et al. 1996 ; CARRIELLO et al. 1997; HAINUT et al. 1997 ). The analysis below utilizes the data base first prepared by HOLLSTEIN et al. 1996 which is available as an Excel spreadsheet (HAINUT et al. 1997 ). This data base includes all reported mutations in tumors and attempts to exclude mutations at known polymorphic sites. Up to the end of 1996 (HAINUT et al. 1997 ), there were 6005 total p53 mutations recorded. It seemed possible that an analysis of the mutational data base, and in particular the analysis of the frequency with which silent mutations occur, would permit an estimate of the nonselected mutation frequency for the TP53 gene. Although knowledge of the mutation frequency does not permit estimation of the mutation rate, a comparison of the (nonselected) mutation frequencies of genes in normal and tumor tissue might give some indication of whether tumors are hypermutable.

	Silent mutation

TOP ABSTRACT Are tumors hypermutable? TP53 mutation Silent mutation Multiple TP53 mutations Mutations in other genes Other examples of... LITERATURE CITED

Silent mutations are defined as nucleotide substitutions that do not result in amino acid changes, owing to the redundancy of the genetic code. There were 202 silent mutations in the tumors (HAINUT et al. 1997 ). However, nine of the silent mutations recorded occur at codons reported as polymorphic. Consequently, there are () x 100 = 3.2% silent mutations in the sample. Their exact positions and the nature of the substitutions are readily retrieved from the data base (HAINUT et al. 1997 ). The percentage of silent mutations is about the same in tumors and in cell lines (STRAUSS 1997 ).

A certain proportion of the recorded mutations are probably due to PCR errors, a problem that has been discussed previously (STRAUSS 1997 ). Although errors probably do occur, many of the papers reporting mutations do amplify several samples independently and take other precautions to reduce the possibility of error. The percentage of silent mutations reported in the 1996 update (42/915) is slightly higher than the overall value, making it more likely that ~3% represents an accurate estimate of the frequency of silent mutations among all TP53 mutations. The methodology for the detection of mutations is necessarily becoming more reliable as the utility of mutational analysis as a diagnostic tool becomes evident. The question of whether the changes represent preexisting polymorphisms is harder to address because many of the investigations utilize paraffin-embedded archival samples as their source of tissue and do not have normal tissue controls.

If mutation were random and neutral, 23.5% of all mutations in TP53 would be silent. This figure is based on the actual usage of codons in p53 and the probability of a mutation in that codon being silent. The percentage of observed silent mutations is much lower. However, inactivation of p53 function plays a role in the etiology of tumors and so frameshifts, missense, and nonsense mutations would be expected to have a selective advantage. If we suppose that a first mutation in p53 inactivates its function and contributes to tumorigenesis, then it is possible that additional mutations can accumulate in a nonselective way. A tabulation of the occurrence of silent mutations in tumors with multiple p53 mutations might therefore make possible a nonbiased estimate of their frequency. In the most recent data, there are 305 tumors with multiple mutations, and these tumors had a total of 693 mutations. After removal of four possibly polymorphic silent changes, 66 mutations, or 9.5%, were silent. This figure is still an underestimate of the total percentage of silent mutations because, as indicated above, a nonsilent mutation is (presumably) required for entry into the data set. If one assumes that among the tumors with multiple p53 mutations, silent mutations are only found in tumors with a p53-inactivating missense, nonsense, or frameshift mutation, then a corrected silent mutation frequency can be obtained by dividing the total number of silent mutations (66) among the multiple mutations (693) by the total number of mutations in that sample minus the total number of tumors with multiple mutations (305) as a correction for the ascertainment bias. The corrected frequency is 17% ([66/693-305] x 100). Because not all silent mutations need be nonselective (see below), the figure for silent mutations is probably somewhere between 9.5% and 17%. What is critical is that considerable numbers of silent mutations occur. An essentially similar argument, in which the second mutations in doublets (including the silent mutations) are considered as "hitchhikers," has been made by RODIN et al. 1998 .

Individual investigations from the published data base in which substantial numbers of p53 mutations have been recorded do give results nearer to the expected value for silent mutations. In one of the cases in which silent mutations were found and in which the surrounding tissue was also sequenced, 11 out of 29 single nucleotide changes were silent, and of these, only two were found in the surrounding tissue at a known polymorphic site (HONGYO et al. 1995 ). In another investigation, a total of 64 bladder tumors yielded 45 p53 mutations of which 11 (24%) were silent (TAYLOR et al. 1996 ). Two of the tumors had two silent mutations each, and for these polymorphism was eliminated by sequencing blood DNA. None of the other silent mutations were at the sites of known polymorphisms (WESTON and GODBOLD 1997 ).

Silent mutations need not be selectively neutral and, in fact, are not in many organisms (SHARP et al. 1995 ). Codon usage differs, and the availability of tRNAs is different for the different codons. In Drosophila, in yeast, and in E. coli it has been shown that codons are selected. This does not seem to be true for humans in which codon usage for the same amino acid in different genes expressed at the same time may differ widely (SHARP et al. 1993 ; see Table 6 in STRAUSS 1997 ). It has also been pointed out that silent mutation might produce alternative splice sites (HONGYO et al. 1995 ). The production of a 5' GT, for example, could result in a new splice site and an altered p53 protein, but no such altered protein has been isolated. Notwithstanding these caveats, the hypothesis that most of the silent mutations are selectively neutral for tumor formation in humans remains a reasonable one.

If silent mutations are effectively neutral, then their frequency in tumors, as compared to the frequency of newly occurring (i.e., nonpolymorphic) mutations in normal tissue, should provide a clue to the question of the mutation rate in tumors. There is only the most limited data that would permit an estimate of what the expected frequency of a newly arising mutation in a tumor might be, considering that a sufficient number of generations would have had to occur between the mutation and the assay to result in a clone of cells, most of which contained the mutation, so that it would be detected by current PCR methodology. AGUILAR et al. 1994 estimated the frequency for mutation at one nucleotide of the frequently mutated p53 codon 249 in nonmalignant samples of liver from hepatocellularcarcinoma patients. The detection methodology depends on the resistance of mutated sites to restriction-enzyme digestion. Values of <2 to 13 x 10^-7 per base were obtained for mutation at the third position of codon 249 in three nonmalignant liver samples. Assuming that all positions in the TP53 gene have the same mutability, but that only one-fourth of the changes would give silent mutations, the highest frequency of silent mutations would be x 1179 coding bases x 1.3 x 10^-6 = 3.8 x 10^-4, or 0.04%. This is about 1/100 of the overall frequency of silent mutations in the p53 set and 1/400 of the corrected frequency in multiple mutants. Another estimate can be obtained from values for transgenic lacI mutations in the mouse, where NISHINO et al. 1995 report a frequency of 3.7 x 10^-5 mutations per gene. The effective mutational size of the lacI gene is ~210 bp, or ~one-fifth the p53 size. If this frequency applied to humans, one might expect 3.7 x 10^-5 x 5 x 1/4 (to correct for silent mutations), or 4.6 x 10^-5 = 0.005%. Once again, the measured p53 silent mutation frequency in human tumors is much higher.

Values for mutation frequencies in blood cells of normal young adults are available for four genes (ALBERTINI et al. 1993 ): hemoglobin (4 x 10^-8; red blood cells), hprt (5 x 10^-6; T cells ), glycophorin A (10 x 10^-6; red blood cells), and HLA (3 x 10^-5; T cells). The hprt frequency has been described in human kidney epithelium as age-dependent, ranging from 5 x 10^-5 (children) to 2.5 x 10^-4 (>80 yr) (MARTIN et al. 1996 ). Measurement of the somatic mutation frequency of a gene involved in O-acetylation of sialic acid in normal colonic mucosa from patients with either Crohn's disease, hereditary nonpolyposis colorectal cancer, or sporadic colorectal cancer gave values of 4.2 x 10^-4, 4.3 x 10^-4, and 6.6 x 10^-4, respectively, with no significant differences between the sets, although there was a significant age dependence seen (WILLIAMS et al. 1995 ). These frequencies are about two orders of magnitude lower than the frequency of silent mutations (even the uncorrected frequency) observed in TP53. The data on silent mutations in TP53 show that this gene is hypermutable in tumors.

	Multiple TP53 mutations

TOP ABSTRACT Are tumors hypermutable? TP53 mutation Silent mutation Multiple TP53 mutations Mutations in other genes Other examples of... LITERATURE CITED

Are there any other indications of a special nature for p53 mutations in tumors? To answer this question, it is useful to look at the distribution of mutations in tumors with multiple p53 mutations (Table 1). A problem with the tabulated data (HOLLSTEIN et al. 1996 ; HAINUT et al. 1997 ) is that it is not easy to find the number of tumors scored with no mutations, but it can be estimated (M. HOLLSTEIN, personal communication) as ~24,000 or 48,000 TP53 genes. There were 6005 mutations recorded. This gives a P(0) of 0.87 and a value of 0.13 mutations per p53 gene based on the Poisson expectation, P(0) = e^-m.

There is a deficiency of double mutations but an excess of single and multiple (>2) mutations (Table 1). There are several interpretations possible for the deviations of the observed from expected value. A trivial explanation is that because only ~16% of the publications report sequencing all exons, many second mutations were just not observed. A second possibility (suggested by J. CAIRNS in a personal communication) is that there is a fraction of hypermutable tumors. Yet a third, not necessarily exclusive, possibility is suggested by an examination of the position of the mutations, some of which are closely linked. A tabulation of tumors with two closely linked mutations, and including six tumors with three mutations located no more than 10 codons from each other, has recently been presented (STRAUSS 1997 ). The current data set provides additional cases. A particularly interesting example is provided by SHIPMAN et al. 1996 who carefully examined TP53 mutations in 24 paired samples of normal and non-small-cell lung carcinomas from the same individuals. Three of these tumors had multiple TP53 mutations, including one with four mutations, shown by cloning to be the result of three mutations in one allele and one in the other. SHIPMAN et al. 1996 eliminated germinal polymorphism as an explanation for the findings. Counting the site of the first change, a deletion at position (1), a second deletion occurred at base 36 followed by a base substitution at position 46. The authors suggest that it is the first change that creates a stop codon 23 bp downstream, and the following mutations are presumably of no physiological importance. TAYLOR et al. 1996 studied 64 bladder tumors. Thirty-two had p53 mutations, and of these, nine had multiple p53 mutations. One of these tumors had five mutations at four codons (nos. 89, 101, 132 and 214). Although this tumor came from an individual exposed to arylamines, multiple mutations occurred about equally in exposed and control groups, and there were no differences between exposed and control groups in the codons mutated or in the pattern of changes—most were transitions. There is no evidence that such multiple changes occur as part of a single (replicative) event, but the proximity of such changes and their frequency certainly suggests that hypothesis. This possibility was first raised by SEIDMAN et al. 1987 and by HARWOOD et al. 1991 . SEIDMAN et al. 1987 found multiple point mutations in a shuttle vector introduced into repair-proficient lymphoblastoid cells, especially after introduction of a nick. The tumor cell lines studied by HARWOOD et al. 1991 were peculiar in that although the spontaneous mutation rate was low, those cells that did have mutations tended to have multiple mutations.

	Mutations in other genes

TOP ABSTRACT Are tumors hypermutable? TP53 mutation Silent mutation Multiple TP53 mutations Mutations in other genes Other examples of... LITERATURE CITED

The frequency of silent mutations and the observation of closely linked multiple mutations are arguments for the hypothesis that at least at some stage of tumorigenesis, the TP53 gene is hypermutable and/or mutates by a special mechanism. The question is then whether genes other than TP53 are hypermutable at some stage of tumorigenesis. Useful data are available for only a few genes. These include the APC gene (SMITH et al. 1994 ; BEROUD and SOUSSI 1996 , BEROUD and SOUSSI 1997 ; DEVRIES et al. 1996 ), the BRCA1 (FUTREAL et al. 1994 ; MERAJVER et al. 1995 ) and BRCA2 cancer susceptibility genes (LANCASTER et al. 1996 ; MIKI et al. 1996 ; TENG et al. 1996 ), CDKN2 (cyclin-dependent kinase inhibitor, p16) (POLLOCK et al. 1996 ; SMITH-SORENSEN and HOVIG 1996 ), WAF1 (SHIOHARA et al. 1994 ; KOOPMAN et al. 1995 ), and HPRT (CARIELLO et al. 1997 ). Although numerous germinal mutations have been reported in BRCA1 and BRCA2, there is a paucity of somatic point mutations, probably because of the presence of numerous deletions that occur at a high-enough frequency to mask the point mutations (SMITH et al. 1996 ). POLLOCK et al. 1996 have compiled a list of 124 somatic mutations in the CDKN2 (p16) tumor suppressor. Of these, four were silent but one silent mutation occurred at the site of a dipyrimidine change and a second at the site of a known polymorphism. This small sample therefore gives a frequency of 2/124, or ~1.6%. Data bases are available for the APC gene in tumors (BEROUD and SOUSSI 1996 , BEROUD and SOUSSI 1997 ) and for HPRT in cell cultures. The APC data base includes 339 germ line mutations and 486 from tumors. Most of the mutations are either frameshift or stop mutations. Only 6 of 339 germ line mutations and 20 of 486 mutations in tumors were missense. (The difference is barely significant; P < 0.05.) No silent mutations have been recorded. Double or triple mutations were reported in 18 of the tumors. In one report (ODA et al. 1996 ), eight of 13 sporadic hepatoblastomas were reported to have APC point mutations, and of these, two cases had double mutations. One of these had a CCT to ACT change at codon 1319 and a AGC to AAC change at codon 1321! The other double had mutations in codon 1395 and 1534. Tests for loss of heterozygosity were inconclusive.

WAF1 mutations were not found in a study of 315 samples from 14 different malignancies, although two polymorphisms were detected, validating the sensitivity of the methodology (SHIOHARA et al. 1994 ). A second study of WAF1 in a different set of tumors also found several polymorphisms in 158 brain tumors but no mutations, although p53 mutations were found in 22 of the tumors (KOOPMAN et al. 1995 ). If hypermutability was widespread, and mutation to silent alleles was at least as frequent as in TP53, ~13 silent mutations might have been expected.

The data are inconclusive, but it certainly remains possible that mutations in TP53 are particularly easy to detect or that the gene is uniquely mutable and/or that hypermutation of p53 is a likely (but not necessary) part of tumorigenesis. The particular structure of a functional p53 protein might make dominant negative effects common so that the first mutation inactivates function and that many second mutations would have no effect (this suggestion is from M. HOLLSTEIN). As has been suggested by several authors, inactivation of p53 might be important in inactivating the cell death mechanisms, thereby permitting the accumulation of deleterious changes. However, in contrast to the APC gene (POWELL et al. 1992 ), mutation in p53 is often a late event in carcinogenesis (BAKER et al. 1990 ; NERI et al. 1993 ; OHGAKI et al. 1993 ; BERCHUCK and BOYD 1995 ; CRAANEN et al. 1995 ; ILYAS et al. 1996 ; KAHLENBERG et al. 1996 ). For that reason, it is unlikely that inhibition of cell death due to p53 mutation permits p53 mutations to accumulate. It is also unlikely that TP53 mutation is the initiating event in carcinogenesis.

	Other examples of hypermutability

TOP ABSTRACT Are tumors hypermutable? TP53 mutation Silent mutation Multiple TP53 mutations Mutations in other genes Other examples of... LITERATURE CITED

The data indicate that the mutation process in (some) tumor cells and in (some) genes is different from mutation in normal cells. What occurs in tumorigenesis is reminiscent of the events involved in the generation of antibodies in which a limited portion of the genome becomes hypermutable for a limited time (NEUBERGER and MILSTEIN 1995 ). Antibody production is an example of microevolution in much the same way as tumor formation is because a large number of random mutations are generated, and those mutations that produce a more efficient antibody molecule are selected. The mutations occur at random. Silent mutations are found in 17% of nonproductively rearranged V_H chains and in 29% of productively rearranged chains (DORNER et al. 1997 ), close to the expected value. Antibody molecules may contain multiple changes, some possibly occurring in the same replication cycle (STORB 1996 ). This special mutational process occurs exclusively in a sequence defined by precise boundaries (LEBECQUE and GEARHART 1990 ) and is limited to a particular time in development. The nature of the error-generating mechanism in that process remains a mystery, although it appears to be linked to transcription (PETERS and STORB 1996 ).

The hypermutability that occurs in TP53 during tumorigenesis is reminiscent of the hypermutability of starving bacterial cells (STRAUSS 1992 ), a comparison that has not eluded the attention of workers in that field (SNIEGOWSKI et al. 1997 ; TADDEI et al. 1997 ; TORKELSON et al. 1997 ). A subpopulation of E. coli cells kept under starvation conditions become hypermutable. Although hypermutability is attained by modulation of the mismatch repair system, a special mode of DNA replication may also be required, possibly set off by double-strand breaks in the DNA (TORKELSON et al. 1997 ). The mutational process affects many genes, and selection gives the appearance of adaptation (FOSTER 1997 ). The factors that result in a hypermutable state remain to be determined.

Both experimental and modeling studies suggest that clonal populations may increase their rate of evolution by transient changes in their mutation rates (SNIEGOWSKI et al. 1997 ; TADDEI et al. 1997 ). In bacterial populations, it would appear that selection is not the unique determinant of adaptation, as has been generally accepted (DRAKE 1991 , DRAKE 1992 ). The demonstration of a surprisingly high frequency of mutator strains among bacterial pathogens appears to illustrate this paradigm (LECLERC et al. 1996 ), an example of the advantage of high mutability in populations themselves subject to selection (MAO et al. 1997 ). Recent experiments with certain of the mismatch repair-deficient tumor cells indicate that mutations accumulate only when the cells are maintained at high density, rather than under conditions of rapid growth (RICHARDS et al. 1997 ). The data suggest that in tumors, as in starved bacteria, mutation rates are transiently increased under conditions of stress and that mutation rates in general may be influenced by the conditions of growth (BOESEN et al. 1994 ). It may be that the initial events in carcinogenesis result in a condition of transient hypermutability, perhaps as a result of a peculiar form of DNA synthesis related to the functional mismatch repair deficiency (RICHARDS et al. 1997 ). The processivity factor for DNA synthesis, proliferating cell nuclear antigen has been reported to form a physical complex with the mismatch repair proteins, and mutations in proliferating cell nuclear antigen have a mutator effect ( JOHNSON et al. 1996 ; UMAR et al. 1996 ). Some of the genes mutated as a result of this hypermutable state may then play a role in the rapid progression/evolution of the tumor.

	FOOTNOTES

This paper is dedicated to DR. JOHN DRAKE on the occasion of his retirement as Editor of GENETICS, in the hope that he may find it amusing.

	ACKNOWLEDGMENTS

I am grateful to DR. MONICA HOLLSTEIN for her help in providing access to the updated p53 data base and for suggesting that looking at multiple mutations would be instructive. The suggestion that p53 might be hypermutable comes in the first instance from an e-mail conversation with DR. JOHN CAIRNS. DR. CAIRNS also pointed out the advantages of using the Poisson distribution for the analysis of the mutation data. I am grateful to DR. BRIAN CHARLESWORTH for emphasizing the possible occurrence of polymorphisms among the silent mutations. Financial support for this work was provided by grant CA-32436 from the National Cancer Institute, National Institutes of Health.

	LITERATURE CITED

TOP ABSTRACT Are tumors hypermutable? TP53 mutation Silent mutation Multiple TP53 mutations Mutations in other genes Other examples of... LITERATURE CITED

AGUILAR, F., C. C. HARRIS, T. SUN, M. HOLLSTEIN, and P. CERUTTI, 1994  Geographic variation of p53 mutational profile in nonmalignant human liver. Science 264:1317-1319[Abstract/Free Full Text].

AKIYAMA, Y., R. IWANAGA, T. ISHIKAWA, K. SAKAMOTO, and N. NISHI et al., 1996  Mutations of the transforming growth factor-beta type II receptor gene are strongly related to sporadic proximal colon carcinomas with microsatellite instability. Cancer 78:2478-2484[Medline].

ALBERTINI, R. J., J. A. NICKLAS, J. C. FUSCOE, T. R. SKOPEK, and R. F. BRANDA et al., 1993  In vivo mutations in human blood cells: biomarkers for molecular epidemiology. Env. Health Perspectives 99:135-141.

AMES, B. N., W. DURSTON, E. YAMASAKI, and F. LEE, 1973  Carcinogens are mutagens: a simple test system combining liver homogenates for activation and bacteria for detection. Proc. Natl. Acad. Sci. USA 70:2281-2285[Abstract/Free Full Text].

AMES, B. N., L. S. GOLD, and W. C. WILLETT, 1995  The causes and prevention of cancer. Proc. Natl. Acad. Sci. USA 92:5258-5265[Abstract/Free Full Text].

ARMITAGE, P. and R. DOLL, 1957  A two-stage theory of carcinogenesis in relation to the age distribution of human cancer. Brit. J. Cancer 11:161-169[Medline].

BAKER, S. J., A. C. PREISINGER, J. M. JESSUP, C. PARASKEVA, and S. MARKOWITZ et al., 1990  53 gene mutations occur in combination with 17p allelic deletions as late events in colorectal tumorigenesis. Cancer Res. 50:7717-7722[Abstract/Free Full Text].

BERCHUCK, A. and J. BOYD, 1995  Molecular basis of endometrial cancer. Cancer 76(10 Suppl.):2034-2040[Medline].

BEROUD, C. and T. SOUSSI, 1996  APC gene: database of germline and somatic mutations in human tumors and cell lines. Nucleic Acid Res. 24:121-124[Abstract/Free Full Text].

BEROUD, C. and T. SOUSSI, 1997  53 and APC gene mutations: software and databases. Nucleic Acids Res. 25:138[Abstract/Free Full Text].

BHATTACHARYYA, N. P., A. GANESH, R. G. PHEA, B. RICHARDS, and A. SKANDALIS et al., 1995  Molecular analysis of mutations in mutator colorectal carcinoma cell lines. Human Molecular Genetics 4:2057-2064[Abstract/Free Full Text].

BOESEN, J. J., M. J. NIERICKER, N. DIETEREN, and J. W. SIMONS, 1994  How variable is a spontaneous mutation rate in cultured mammalian cells? Mutat. Res. 307:121-129[Medline].

BRASH, D. E., J. A. RUDOLPH, J. A. SIMON, A. LIN, and G. J. MCKENNA et al., 1991  A role for sunlight in skin cancer: UV-induced p53 mutations in squamous cell carcinoma. Proc. Natl. Acad. Sci. USA 88:10124-10128[Abstract/Free Full Text].

BUETTNER, V. L., K. A. HILL, H. NISHINO, D. J. SCHAID, and C. S. FRISK et al., 1996  Increased mutation frequency and altered spectrum in one of four thymic lymphomas derived from tumor prone p53/Big Blue double transgenic mice. Oncogene 13:2407-2413[Medline].

CAIRNS, J., 1981  The origin of human cancers. Nature 289:353-357[Medline].

CARIELLO, N. F., G. R. DOUGLAS, M. J. DYCAICO, G. S. PROVOST, and T. SOUSSI, 1997  Databases and software for the analysis of mutations in the human p53gene, the human hprt gene and both the lacI and lacZ gene in transgenic rodents. Nucleic Acids Res. 25:136-137[Abstract/Free Full Text].

CHENG, K. C. and L. A. LOEB, 1993  Genomic instability and tumor progression: mechanistic considerations. Advances in Cancer Research 60:121-56[Medline].

CHU, G. and L. MAYNE, 1996  Xeroderma pigmentosum, Cockayne syndrome and trichothiodystrophy: do the genes explain the diseases? Trends in Genetics 12:187-192[Medline].

CLEAVER, J. E., 1968  Defective repair replication of DNA in xeroderma pigmentosum. Nature 216:652-656.

CRAANEN, M. E., P. BLOK, W. DEKKER, G. J. OFFERHAUS, and G. N. TYTGAT, 1995  Chronology of p53 protein accumulation in gastric carcinogenesis. Gut 36:848-852[Abstract/Free Full Text].

DE VRIES, A., C. T. VAN OOSTROM, P. M. DORTANT, R. B. BEEMS, and C. F. VAN KREIJL et al., 1997  Spontaneous liver tumors and benzo[a]pyrene-induced lymphomas in XPA-deficient mice. Mol. Carcinog. 19:46-53[Medline].

DEVRIES, E. M., D. O. RICKE, T. N. DEVRIES, A. HARTMANN, and H. BLASZYK et al., 1996  Database of mutations in the p53 and APC tumor suppressor genes designed to facilitate molecular epidemiological analyses. Human Mutation 7:202-213[Medline].

DORNER, T., H. P. BREZINSCHEK, R. I. BREZINSCHEK, S. J. FOSTER, and R. DOMIATI-SAAD et al., 1997  Analysis of the frequency and pattern of somatic mutations within nonproductively rearranged human variable heavy chain genes. J. Immunol. 158:2779-2789[Abstract].

DRAKE, J. W., 1991  A constant rate of spontaneous mutation in DNA-based microbes. Proc. Natl. Acad. Sci. USA 88:7160-7164[Abstract/Free Full Text].

DRAKE, J. W., 1992  Mutation rates. BioEssays 14:137-140[Medline].

DUMAZ, N., C. DROUGARD, A. SARASIN, and L. DAYA-GROSJEAN, 1993  Specific UV-induced mutation spectrum in the p53 gene of skin tumors from DNA-repair-deficient xeroderma pigmentosum patients. Proc. Natl. Acad. Sci. USA 90:10529-10533[Abstract/Free Full Text].

FISHEL, R., M. LESCOE, M. RAO, N. COPELAND, and N. JENKINS et al., 1993  The human mutator gene homolog MSH2 and its association with hereditary nonpolyposis colon cancer. Cell 75:1027-1038[Medline].

FISHER, J. C., 1958  Multiple-mutation theory of carcinogenesis. Nature 181:651-652[Medline].

FOSTER, P. L., 1997  Nonadaptive mutations occur on the F' episome during adaptive mutation conditions in Escherichia coli. J. Bacteriol. 179:1550-1554[Abstract/Free Full Text].

FRIEDBERG, E. C., G. C. WALKER and W. SIEDE, 1995 DNA Repair and Mutagenesis. ASM Press, Washington, DC.

FUTREAL, P. A., Q. LIU, D. SHATTUCK-EIDENS, C. COCHRAN, and K. HARSHMAN et al., 1994  BRCA1 mutations in primary breast and ovarian carcinomas. Science 266:120-122[Abstract/Free Full Text].

GOYETTE, M. C., K. CHO, C. L. FASCHING, D. B. LEVY, and K. W. KINZLER et al., 1992  Progression of colorectal cancer is associated with multiple tumor suppressor gene defects but inhibition of tumorigenicity is accomplished by correction of any single defect via chromosome transfer. Mol. Cell. Biol. 12:1387-1395[Abstract/Free Full Text].

HAINUT, P., T. SOUSSI, B. SHOMER, M. HOLLSTEIN, and M. GREENBLATT et al., 1997  Database of p53 gene somatic mutations in human tumors and cell lines: updated compilation and future prospects. Nucleic Acids Res. 25:151-157[Abstract/Free Full Text].

HARWOOD, J., A. TACHIBANA, and M. MEUTH, 1991  Multiple dispersed spontaneous mutations: a novel pathway of mutation in a malignant human cell line. Mol. Cell. Biol. 11:3163-3170[Abstract/Free Full Text].

HOLLSTEIN, M., D. SIDRANSKY, B. VOGELSTEIN, and C. C. HARRIS, 1991  53 mutations in human cancers. Science 253:49-53[Abstract/Free Full Text].

HOLLSTEIN, M., B. SHOMER, M. GREENBLATT, T. SOUSSI, and E. HOVIG et al., 1996  Somatic point mutations in the p53 gene of human tumors and cell lines: updated compilation. Nucleic Acids Res. 24:141-146[Abstract/Free Full Text].

HONGYO, T., G. S. BUZARD, D. PALLI, C. M. WEGHORST, and A. AMOROSI et al., 1995  Mutations of the K-ras and p53 genes in gastric adenocarcinomas from a high-incidence region around Florence, Italy. Cancer Res. 55:2665-2672[Abstract/Free Full Text].

ILYAS, M., M. KENDALL, H. JALAL, C. LINTON, and N. ROONEY, 1996  Changes in Bcl-2 and p53 expression in recurrent B-cell lymphomas. J. Pathol. 180:249-253[Medline].

JOHNSON, R. E., G. K. KOVVALI, S. N. GUZDER, N. S. AMIN, and C. HOLM et al., 1996  Evidence for involvement of yeast proliferating cell nuclear antigen in DNA mismatch repair. J. Biol. Chem. 271:27987-27990[Abstract/Free Full Text].

KAHLENBERG, M. S., D. L. STOLER, M. BASIK, N. J. PETRELLI, and M. RODRIGUEZ-BIGAS et al., 1996  53 tumor suppressor gene status and the degree of genomic instability in sporadic colorectal cancers. J. Natl. Cancer Inst. 88:1665-1670[Abstract/Free Full Text].

KOOPMAN, J., D. MAINTZ, S. SCHILD, J. SCHRAMM, and D. N. LOUIS et al., 1995  Multiple polymorphisms, but no mutations, in the WAF1/CIP1 gene in human brain tumours. Brit. J. Cancer 72:1230-1233[Medline].

KRAEMER, K. H., M. M. LEE, and J. SCOTTO, 1984  DNA repair protects against cutaneous and internal neoplasia: evidence from xeroderma pigmentosum. Carcinogenesis 5:511-514[Abstract/Free Full Text].

KRAEMER, K. H., D. D. LEVY, C. N. PARRIS, E. M. GOZUKARA, and S. MORIWAKI et al., 1994  Xeroderma pigmentosum and related disorders: examining the linkage between defective DNA repair and cancer. Journal of Investigative Dermatology 103(5 Suppl.):96S-101S[Medline].

LANCASTER, J. M., R. WOOSTER, J. MANGION, C. M. PHELAN, and C. COCHRAN et al., 1996  BRCA2 mutations in primary breast and ovarian cancers. Nature Genetics 13:238-240[Medline].

LEBECQUE, S. G. and P. J. GEARHART, 1990  Boundaries of somatic mutation in rearranged immunoglobulin genes: 5' boundary is near the promoter and 3' boundary is ~1 kb from V(D)J gene. J. Exp. Med. 172:1717-1727[Abstract/Free Full Text].

LECLERC, J. E., B. LI, W. L. PAYNE, and T. A. CEBULA, 1996  High mutation frequencies among Escherichia coli and Salmonella pathogens. Science 274:1208-1211[Abstract/Free Full Text].

LEVINE, A. J., 1997  53, the cellular gatekeeper for growth and division. Cell 88:323-331[Medline].

LOEB, L. A., 1991  Mutator phenotype may be required for multistage carcinogenesis. Cancer Res. 51:3075-3079[Free Full Text].

LOEB, L. A., 1994  Microsatellite instability: marker of a mutator phenotype in cancer. Cancer Res. 54:5059-5063[Free Full Text].

LOEB, L., C. SPRINGGATE, and N. BATTULA, 1974  Errors in DNA replication as a basis of malignant change. Cancer Res. 34:2311-2321[Abstract/Free Full Text].

LU, S. L., Y. AKIYAMA, H. NAGASAKI, K. SAITOH, and Y. YUASA, 1995  Mutations of the transforming growth factor-beta type II receptor genes and genomic instability in hereditary nonpolyposis colorectal cancer. Biochem. Biophys. Res. Commun. 216:452-457[Medline].

MALKHOSYAN, S., A. MCCARTY, H. SAWAI, and M. PERUCHO, 1996  Differences in the spectrum of spontaneous mutations in the hprt gene between tumor cells of the microsatellite mutator phenotype. Mutat. Res. 316:249-259[Medline].

MAO, E. F., L. LANE, J. LEE, and J. H. MILLER, 1997  Proliferation of mutators in a cell population. J. Bacteriol. 179:417-422[Abstract/Free Full Text].

MARIONNET, C., X. QUILLIET, A. BENOIT, J. ARMIER, and A. SARASIN et al., 1996  Recovery of normal DNA repair and mutagenesis in trichothiodystrophy cells after transduction of the XPD human gene. Cancer Res. 56:5450-5456[Abstract/Free Full Text].

MARKOWITZ, S., J. WANG, L. MYEROFF, R. PARSONS, and L. SUN et al., 1995  Inactivation of the type II TGF-ß receptor in colon cancer cells with microsatellite instability. Science 268:1336-1338[Abstract/Free Full Text].

MARTIN, G. M., C. E. OGBURN, L. M. LOLGIN, A. M. GOWN, and S. D. EDLAND et al., 1996  Somatic mutations are frequent and increase with age in human kidney epithelial cells. Hum. Mol. Genet. 5:215-221[Abstract/Free Full Text].

MATSUMURA, Y., C. NISHIGORI, T. YAGI, S. IMAMURA, and H. TAKEBE, 1996  Characterization of p53 gene mutations in basal-cell carcinomas: comparison between sun-exposed and less-exposed skin areas. Int. J. Cancer 65:778-780[Medline].

MERAJVER, S. D., T. M. PHAM, R. F. CADUFF, M. CHEN, and E. L. POY et al., 1995  Somatic mutations in the BRCA1 gene in sporadic ovarian tumours. Nature Genetics 9:439-443[Medline].

MIKI, Y., T. KATAGIRI, F. KASUMI, T. YOSHIMOTO, and Y. NAKAMURA, 1996  Mutation analysis in the BRCA2 gene in primary breast cancers. Nature Genetics 13:245-247[Medline].

MODRICH, P., 1995  Mismatch repair, genetic stability and tumour avoidance. Phil. Trans. R. Soc. Lond. B 347:89-95[Medline].

NERI, A., L. BALDINI, D. TRECCA, L. CRO, and E. POLLI et al., 1993  53 gene mutations in multiple myeloma are associated with advanced forms of malignancy. Blood 81:128-135[Abstract/Free Full Text].

NEUBERGER, M. and C. MILSTEIN, 1995  Somatic hypermutation. Curr. Opin. Immunol. 7:248-254[Medline].

NISHINO, H., A. KNOLL, V. L. BUETTNER, C. S. FRISK, and Y. MARUTA et al., 1995  53 wild-type and p53 nullizygous Big Blue transgenic mice have similar frequencies and patterns of observed mutation in liver, spleen and brain. Oncogene 11:263-270[Medline].

ODA, H., Y. IMAI, Y. NAKATSURU, J. HATA, and T. ISHIKAWA, 1996  Somatic mutations of the APC gene in sporadic hepatoblastomas. Cancer Res. 56:3320-3323[Abstract/Free Full Text].

OHGAKI, H., P. KLEIHUES, and P. U. HEITZ, 1993  53 mutations in sporadic adrenocortical tumors. International Journal of Cancer 54:408-410.

PARSONS, R., G. M. LI, M. LONGLEY, P. MODRICH, and B. LIU et al., 1995  Mismatch repair deficiency in phenotypically normal human cells. Science 268:738-740[Abstract/Free Full Text].

PETERS, A. and U. STORB, 1996  Somatic hypermutation of immunoglobulin genes is linked to transcription initiation. Immunity 4:57-65[Medline].

POLLOCK, P. M., J. V. PEARSON, and N. K. HAYWARD, 1996  Compilation of somatic mutations of the CDKN2 gene in human cancers: non-random distribution of base substitutions. Genes, Chromosomes & Cancer 15:77-88[Medline].

POWELL, S. M., N. ZILZ, Y. BEAZER-BARCLAY, T. BRYAN, and S. HAMILTON et al., 1992  APC mutations occur early during colorectal tumorigenesis. Nature 359:235-237[Medline].

RAMPINO, N., H. YAMAMOTO, Y. IONOV, Y. LI, and H. SAWAI et al., 1997  Somatic frameshift mutations in the BAX gene in colon cancers of the microsatellite mutator phenotype. Science 275:967-969[Abstract/Free Full Text].

RICHARDS, B., H. ZHANG, G. PHEAR, and M. MEUTH, 1997  Conditional mutator phenotypes in hMSH2-deficient tumor cell lines. Science 277:1523-1526[Abstract/Free Full Text].

RODIN, S. N., G. P. HOLMQUIST, and A. S. RODIN, 1998  CpG transition strand asymmetry and hitch-hiking mutations as measures of tumorigenic selection in shaping the p53 mutation spectrum. Int. J. Molec. Med. in press.

SEIDMAN, M. M., A. BREDBERG, S. SEETHARAM, and K. H. JRAEMER, 1987  Multiple point mutations in a shuttle vector propagated in human cells: evidence for an error-prone DNA polymerase activity. Proc. Natl. Acad. Sci. USA 84:4944-4948[Abstract/Free Full Text].

SHARP, P. M., M. AVEROF, A. T. LLOYD, M. G. AND J, and F. PEDEN, 1995  DNA sequence evolution: the sounds of silence. Phil. Trans. Roy. Soc. Lond. B 349:241-247.

SHARP, P. M., M. STENICO, J. F. PEDEN, and A. T. LLOYD, 1993  Codon usage: mutational bias, translational selection, or both? Biochemical Society Transactions 21:835-841[Medline].

SHIOHARA, M., W. S. EL-DEIRY, M. WADA, T. NAKAMAKI, and S. TAKEUCHI et al., 1994  Absence of WAF1 mutations in a variety of human malignancies. Blood 84:3781-3784[Abstract/Free Full Text].

SHIPMAN, R., P. SCHRAML, M. COLOMBI, G. RAEFLE, and P. DALQUEN et al., 1996  Frequent TP53 gene alterations (mutation, allelic loss, nuclear accumulation) in primary non-small cell lung cancer. Europ. J. Cancer 32A:335-341.

SIMPSON, A. J. G., 1997  The natural somatic mutation frequency and human carcinogenesis. Adv. Cancer Res. 71:209-240[Medline].

SMITH, A. J., H. S. STERN, M. PENNER, K. HAY, and A. MITRI et al., 1994  Somatic APC and K-ras codon 12 mutations in aberrant crypt foci from human colons. Cancer Res. 54:5527-5530[Abstract/Free Full Text].

SMITH, T. M., M. K. LEE, C. I. SZABO, N. JEROME, and M. MCEUEN et al., 1996  Complete genomic sequence and analysis of 117 kb of human DNA containing the gene BRCA1.. Genome Research 6:1029-1049[Abstract/Free Full Text].

SMITH-SORENSEN, B. and E. HOVIG, 1996  CDKN2A (p16^INK4A) somatic and germline mutations. Human Mutation 7:294-303[Medline].

SNIEGOWSKI, P. D., P. J. GERRISH, and R. E. LENSKI, 1997  Evolution of high mutation rates in experimental populations of E. coli. Nature 387:703-705[Medline].

SOUZA, R. F., R. APPEL, J. YIN, S. WANG, and K. N. SMOLINSKI et al., 1996  Microsatellite instability in the insulin-like growth factor II receptor gene in gastrointestinal tumours. Nature Genetics 14:255-257[Medline].

STARY, A. and A. SARASIN, 1996  The genetic basis of xeroderma pigmentosum and trichothiodystrophy syndromes. Cancer Surveys 26:155-171[Medline].

STORB, U., 1996  The molecular basis of somatic hypermutation of immunoglobulin genes. Curr. Opin. Immunol. 8:206-214[Medline].

STRAUSS, B., 1992  The origin of point mutations in human tumor cells. Cancer Res. 52:249-253[Free Full Text].

STRAUSS, B. S., 1997  Silent and multiple mutations in p53 and the question of the hypermutability of tumors. Carcinogenesis 18:1445-1452[Abstract/Free Full Text].

STRAUSS, B. S., D. SAGHER, and S. ACHARYA, 1997  Role of proofreading and mi