ABSTRACT
The replication-error positive (RER+) phenotype characterizes tumour cells with microsatellite instability. This `mutator phenotype' is thought to induce spread mutations throughout the genome, thus increasing the risk of tumour development. Here we analyse spontaneously arising mutations at the tetranucleotide CCGG (MspI recognition site), at positions 14 067-14 070 of the p53 gene sequence, in three colon cancer cell lines, two with microsatellite instability and one without this characteristic. This restriction site covers hot-spot codon 248, which is often mutated in colon carcinomas. Using the MspI RFLP-PCR assay we found that the mean mutation frequency at this site was not different among the cell lines considered. Taking the substitutions separately, none of the mutations involving codon 248 arose with significantly higher frequency in each of the RER+ cell lines (HCT116 and DLD1) compared with the RER- one (SW480). Only the CG transversion at nt 14 067 (codon 247) occurred with a slightly higher, but biologically insignificant, frequency in one of the RER+ cell lines (HCT116). Our in vitro data support the previously reported lack of correlation between microsatellite instability and p53 mutations in RER+ tumour specimens.
Spontaneously arising mutations are thought to be responsible for tumour development in cells with the replication-error positive (RER+) phenotype. Although this characteristic was originally discovered because it provoked microsatellite instability in hereditary non-polyposis colorectal cancer (HNPCC) patients (1 ,2 ), it is somehow extended to other types of neoplasms with or without hereditary features (3 -6 ). The reason for microsatellite instability in these cells is inactivation of proteins involved in DNA mismatch repair. These proteins and the genes encoding them have come to the forefront of cancer research because of their potential involvement in tumourigenesis and resistance to DNA-interacting antineoplastic drugs (7 -9 ). Seven mammalian genes encoding DNA mismatch repair-related proteins have been identified thus far (hMLH1, hMSH2, PMS1, PMS2, GTBP, MRP1 and REP-3) and their specific activities in binding and promoting the repair of different types of mismatches are under study (10 -14 ).
Although mutations in these genes are known to play a major role in induction of microsatellite instability, DNA regions not involving repeated sequences can also be affected by replication errors, depending on the specific mismatch repair deficiency and the gene sequence taken into consideration. Mutations at the selectable locus encoding the purine salvage enzyme hypoxanthine phosphoribosyl transferase (HPRT), for instance, have been previously studied in cultured RER+ and RER- cells (15 ,16 ). These studies provide further insight into the mutational spectrum specificity of each alteration affecting the molecular mechanisms of mismatch repair. The HPRT locus has been chosen in these studies because of the ease with which its mutant clones are selected. When looking for spontaneous mutations at specific loci a strong background of cells that are phenotypically or genotypically wild-type is present, thus introducing the need for a sensitive method to select mutant cells. If spontaneous mutations at oncogenes or oncosuppressor genes are the subject of study, then it is very difficult to avoid a massive background of cells with wild-type features. A partial notion of mutations induced in cancer-related genes in RER+ cells can be obtained through the analysis of the genes of interest in tumour specimens. In this case, natural selection has already occurred during tumour development and there is no need for particularly sensitive methods to detect mutations along the sequence of interest. The prevalence of mutations at the transforming growth factor [beta] type II receptor gene (TGF[beta]IIR) and at the BAX gene is strongly correlated with microsatellite instability in colorectal tumours (17 -19 ). In contrast, APC mutation prevalence is not dramatically different when comparing RER+ and RER- tumours, even if a substantial excess of frameshift mutations at mononucleotide repeats has been reported in the former versus the latter cases (20 ).
The prevalence of p53 and RAS mutations in different kinds of tumours has not been correlated with RER+ phenotype (4 ,5 ,17 ,21 ). Alterations of the p53 gene are sometimes found in RER+ tumour samples, but no information has been reported about the sequence specificity (if it exists) of RER+ phenotype-induced p53 mutations. A knowledge of the p53 mutational spectrum in RER+ tumour samples would give us an idea of those mutations favourably selected during the process of carcinogenesis in HNPCC patients. The mutational spectrum of the p53 gene at specific loci has been studied using the RFLP-PCR assay (22 ,23 ). This genotypic mutation analysis, unlike other systems, measures the frequencies of in vitro and in vivo mutations without previous selection and without the need for a selectable phenotype (24 -27 ). We used the MspI RFLP-PCR assay to study the induction of mutations at positions 14 067-14 070 of the p53 gene sequence (5'-CCGG-3') in RER+ (HCT116 and DLD1) and RER- (SW480) colon cancer cell lines. This site spans the CpG dinucleotide of codon 248 (CGG), which is the strongest hot-spot of p53 in human tumours and is very frequently found mutated in colon cancers considered as a whole (28 ). In addition, the CpG dinucleotide of this site is prone to deamination of the 5-methylcytosine residue; this spontaneous alteration leads to a G:T mismatch that might be unrepaired in mismatch repair-deficient cell lines. We considered all the possible base substitutions at the MspI site (positions 14 067-14 070 of the p53 gene) and compared the mean mutation frequency measured for each cell line. We also considered the base substitutions separately and evaluated the differences in their frequencies among the three cell lines.
DLD1, HCT116 and SW480 human colon carcinoma cell lines were obtained from the American Type Culture Collection (ATCC). For these cell lines the ATCC catalog (WWW site http://www.atcc.org/catalogs.html) reports a minimum of at least 20 previous passages (DLD1) and a maximum of ~100 previous passages (SW480). In our laboratory, cells were passaged in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum for 8-10 additional cell divisions and then frozen for DNA preparation. We did not clone the cell lines in order to maximize the number of replications they had undergone and, consequently, accumulation of a substantial number of mutations.
In the entire HPRT gene mutation rates (per cell, per generation) of 1.5 * 10-5 and 1.9 * 10-5 have been reported for the HCT116 and DLD1 lines respectively (15 ); for the RER- line SW480 the mutation rate in HPRT was <4 * 10-9 (16 ). Bhattacharyya et al. (15 ) analysed the mutational spectrum of the HPRT gene in the HCT116 and DLD1 cell lines (RER+) and identified 39 major and minor hot-spots. Consequently, the average mutation rate per hot-spot can be estimated as ~3.8 * 10-7 for HCT116, ~4.9 * 10-7 for DLD1 and <1.0 * 10-10 for SW480 cells. Since we can confidently assume that our three cell lines had undergone at least 30 cell replications, the expected average mutation frequency for a major or minor hot-spot in the HPRT gene becomes ~1.1 * 10-5 (HCT116), ~1.5 * 10-5 (DLD1) and <3.0 * 10-9 (SW480). Because there is no known reason for a negative selective effect on in vitro growth for a cell carrying a mutation at codon 248 of p53 and assuming that the hot-spot was linked to 5-methylcytosine deamination at the CpG site (as suggested by Greenblatt et al.; 28 ), we would have expected a noticeable accumulation of mutations in the two RER+ lines.
Genomic DNA from the three cell lines was isolated as previously described (29 ). The wild-type status of p53 codon 248 in the bulk lines has been verified through PCR amplification and sequence analysis of p53 exon 7 (data not shown). The RFLP-PCR protocol included restriction of the DNA aliquots with MspI: 132 [mu]g genomic DNA were overdigested with 3 U enzyme/[mu]g DNA at 37oC overnight and the restriction reaction was repeated two additional times. Twenty copies of mutant standard (MS) were added to each sample. MS was prepared as previously described (23 ; see also Fig. 1 ). The MS is identical to the amplified p53 sequence except for a CT substitution at the MspI site (position 14 067) and a double substitution outside this site. These mutations allow its distinction from mutants `naturally' present in cellular DNA. The preparations were separated by electrophoresis on agarose gel (Sigma, St Louis, MO) and the fragments migrating between 400 and 500 bp (containing unrestricted copies of p53 exon 7 plus MS) were recovered using the Qiaex II Gel Extraction Kit (Qiagen, Santa Clarita, CA) following the manufacturer's instructions. The samples were again digested with MspI and then used as templates for PCR amplification. All subsequent RFLP-PCR steps have already been described elsewhere (23 ). High fidelity amplification of p53 exon 7 from residue 13 999 to 14 114 was performed using primers 1 and 2 of Figure 1 . Aliquots of these PCR products served as templates in a second PCR in which primers 3 and 4 of Figure 1 were used. These primers contained 5'-tails of 12 nt with EcoRI recognition sequences (5'-AGT GAA TTC TAT-3') in order to facilitate subsequent steps of cloning.
As previously described (23 ), we cloned the RFLP-PCR products into [lambda]gt10 vectors, infected Escherichia coli C600Hfl cells and transferred the lysis plaques onto colony/plaque screens (NEF-978; New England Nuclear, Boston, MA). Specific probes were prepared according to standard conditions (30 ) by end-labelling of oligonucleotide 19mers (residues 14 061-14 079) corresponding to the wild-type sequence of p53 and to oligomers containing all 12 possible single base changes in the tetranucleotide CCGG of MspI recognition site 14 067-14 070. The MS oligonucleotide was a 20mer encompassing residues 14 058-14 077. The colony/plaque screens were hybridized overnight with the radioactive probes at 65oC (except for MS probe hybridization, carried out at 58oC) in 0.75 M NaCl, 50 mM NaH2PO4, 5 mM EDTA, pH 7.4, 0.3% SDS. Selective washing temperatures between 60oC and 70oC were used to exclude non-specific hybridization. In order to obtain statistically meaningful data, at least 103 plaques on four different Petri dishes were analysed for each RFLP-PCR product with the different probes. After the content of MS in the RFLP-PCR product was determined by [lambda] plaque oligonucleotide hybridization, a relationship was established to the known number of MS copies which had been added to the cellular DNA at the outset of the experiment. This relationship was used to calibrate the frequency per base pair of each single base pair change from its content in the RFLP-PCR product (31 ,32 ).
A simple factorial analysis of variance (ANOVA) for repeated measures and fixed effects design was performed to examine the effect of two independent factors (cell line and mutation type) on a dependent variable (mutation frequency) (33 ). The F tests related to mutation type and their interaction with the cell lines were statistically significant (P < 0.001), i.e. the mutation frequencies varied across the 12 mutation types.
Consequently, to obtain a more detailed analysis we analysed the frequencies of mutations among the cell lines for each single mutation type (one-way ANOVA). Two out of the 12 F statistics performed were significant (P < 0.01). To detect the pairs of cell lines showing a significant difference in mutation frequencies we used Tukey's method for multiple comparisons. This method produces tests of significance between every pair of means, taking into account that when the number of comparisons increases, so does the likelihood of defining the differences as `significant' (34 ).
Moreover, we used a combined significance test (the inverse [chi]2 method of Fisher; 35 ) to summarize the results of the 12 studies, each of which examined the mutation frequencies for a different mutation type. Such a procedure was applied to test whether any of the P values obtained in the 12 studies could be significant only by chance. This hypothesis could be rejected at P < 0.01.
In order to shed light on the mechanisms of mutation induction by the RER+ phenotype in colon cancer, we analysed spontaneously arising mutations at the tetranucleotide CCGG (positions 14 067-14 070), covering the last base of codon 247 and the three bases of codon 248 of the p53 coding region, in two RER+ and one RER- colon cancer cell lines. The MspI RFLP-PCR assay allowed us to detect mutations arising at this site with very low frequencies (10-7-10-6). Briefly, this assay amplifies enriched mutated sequences of the region of interest using a proof-reading polymerase enzyme [Pyrococcus furiosus (Pfu) DNA polymerase] to minimize background mutations (see Fig. 1 ). The fragments bearing mutations at codons 247 (last base) and 248 are enriched by virtue of their resistance to the endonuclease activity of the restriction enzyme MspI. The PCR products are cloned into [lambda] bacteriophages and the frequencies of clones with specific single base substitutions in the MspI site are revealed through [lambda] plaque hybridization with mutant-specific labelled oligonucleotides. These values, compared with the frequency of a MS, added at the beginning of the experiment, serve to establish the frequency of each mutation before PCR (see Materials and Methods). The mutation frequencies obtained should be considered relative rather than absolute frequencies, since a precise proportionality between bona fide mutants and the MS cannot be guaranteed throughout the RFLP-PCR protocol (23 ).
Table 1 shows the percentage of mutations of all base pair changes in codons 247 and 248 at the p53 gene in RER+ and RER- cell lines, reported as mean percentage (+- SD) over the total number of plaques obtained from each PCR product. Even at first glance it is evident that an excess of mutations in the two RER+ lines is lacking (in contrast to the previously reported increases of >300 times for the HPRT gene). By calibrating these data with the content of MS we obtained the relative mutation frequencies of each single base substitution at the MspI site. These values are graphically reported in Figure 2 . The global frequency of mutation was not significantly different in the multiple comparison between the three cell lines (HCT116 versus DLD1, HCT116 versus SW480, DLD1 versus SW480). Although we considered a single cell line as a representative of the RER- phenotype, mutation frequencies are rather low in the RER+ cells, which makes it unlikely that RER- cells other than SW480 can have a significantly lower frequency. When considering the mutations by type, regardless of cell line, some of them showed frequencies differing markedly from one another (Fig. 2 ). This effect might be determined by preferential errors produced by the DNA polymerase used in PCR (31 ,36 ). The relative frequencies of particular base pair changes also depend on the sequence under study. For the sequence we analysed (tetranucleotide 5'-CCGG-3' between positions 14 067 and 14 070 of the p53 gene sequence) the relative frequencies of base substitutions introduced by Pfu DNA polymerase have not been previously reported.
Table 1
When considering the single base substitutions separately, some mutations found in the RER+ cells seemed to be slightly different in frequency from those found in the RER- ones and vice versa. A statistical analysis (one-way ANOVA; see Materials and Methods) was performed on these data and the comparison of each mutation frequency between RER+ and RER- cells revealed only two statistically meaningful differences: (i) the tetranucleotide CCGG mutated into gCGG more frequently in HCT116 than in SW480 cells (P < 0.001); (ii) SW480 cells had a significantly higher frequency of CCGG CCtG mutations when compared with DLD1 cells (P < 0.01) (Fig. 2 ). Experimental fluctuations were accountable for every other difference found in the mutation frequencies among the three cell lines considered. We also applied a combined significance test to determine whether statistically meaningful differences could be found by chance among the 36 mutation frequencies obtained for the 12 mutation types analysed. The inverse [chi]2 method of Fisher was used to combine and summarize the results of the 12 one-way ANOVAs separately performed on each mutation type. The hypothesis of statistical significance obtained by chance could be rejected at P < 0.01.
*To whom correspondence should be addressed at: Department of Experimental Oncology, National Institute for Cancer Research, Largo Rosanna Benzi 10, I-16132 Genoa, Italy. Tel: +39 10 5600 211; Fax: +39 10 5600 210; Email: parodis@hp380.ist.unige.it
DLD1
HCT116
SW480
Mutant standard
9.1 +- 1.4
9.5 +- 1.1
8.9 +- 1.3
(A)(A)C CGG (wild-type)
3.0 +- 0.5
2.8 +- 0.7
3.2 +- 1.8
(A)(A)C CGc
32.1 +- 3.3
25.1 +- 4.5
23.8 +- 4.8
(A)(A)C CGt
8.4 +- 1.1
7.5 +- 2.7
5.7 +- 2.5
(A)(A)C CGa
6.1 +- 0.7
11.0 +- 2.6
9.2 +- 3.3
(A)(A)C CcG
8.4 +- 1.8
6.7 +- 2.8
8.3 +- 2.1
(A)(A)C CtG
3.3 +- 1.4
7.0 +- 3.1
11.8 +- 3.0
(A)(A)C CaG
7.9 +- 2.0
7.8 +- 1.7
5.6 +- 1.2
(A)(A)C gGG
0.90.8
1.2 +- 0.9
0.8 +- 0.5
(A)(A)C tGG
7.72.6
5.0 +- 1.9
6.5 +- 1.2
(A)(A)C aGG
5.41.3
3.6 +- 1.6
2.9 +- 0.5
(A)(A)g CGG
3.40.3
5.5 +- 0.1
3.2 +- 0.6
(A)(A)t CGG
0.40.3
0.4 +- 0.3
0.4 +- 0.3
(A)(A)a CGG
0.80.6
1.1 +- 0.8
0.7 +- 0.4
REFERENCES