Conditional Mutator Phenotypes in hMSH2-Deficient Tumor Cell Lines

See allHide authors and affiliations

Science  05 Sep 1997:
Vol. 277, Issue 5331, pp. 1523-1526
DOI: 10.1126/science.277.5331.1523


Two human tumor cell lines that are deficient in the mismatch repair protein hMSH2 show little or no increase in mutation rate relative to that of a mismatch repair–proficient cell line when the cells are maintained in culture conditions allowing rapid growth. However, mutations accumulate at a high rate in these cells when they are maintained at high density. Thus the mutator phenotype of some mismatch repair–deficient cell lines is conditional and strongly depends on growth conditions. These observations have implications for tumor development because they suggest that mutations may accumulate in tumor cells when growth is limited.

The autosomal dominant syndrome of hereditary nonpolyposis colon cancer (HNPCC) is characterized by early onset of colon tumors as well as cancers of the endometrium, stomach, upper urinary tract, small intestine, and ovary. Mutations in two human homologs of the Escherichia colimismatch repair genes MutS and MutL(hMSH2 and hMLH1) are found in the great majority of HNPCC patients (1). Less frequent germline mutations of another MutL homolog (hPMS2) are also associated with this disease (2). HNPCC patients inherit a mutant allele of a mismatch repair gene (3), and the second wild-type allele is mutated or lost as an early event in tumor development (3, 4). This second event renders cells mismatch repair–deficient (5, 6), presumably leading to a mutator phenotype that drives the accumulation of mutations required for tumor development (7). Two other genes encoding homologs of the E. coli MutS gene (hMSH3 and hMSH6) appear to be involved in the repair of some types of DNA replication errors or damage (8). A variety of in vitro assays indicate that heterodimers formed between these proteins and hMSH2 exhibit considerable specificity in the types of errors they recognize and bind. Thus, hMSH2 may play a central role in recognition of DNA replication errors while hMSH3 and hMSH6 modify the specificity of this recognition (8).

To examine the consequences of hMSH2 deficiency in human tumor cells, we measured mutation rates in two tumor cell lines withhMSH2 mutations: SK-UT-1, which was derived from a uterine tumor; and 2774, which originated from an ovarian tumor (6,9-11). SK-UT-1 has a 2–base pair (bp) deletion in exon 10 of the hMSH2 coding sequence that results in a truncation (9); and 2774 has a base substitution in exon 14, resulting in a missense mutation (Arg → Pro) (10). SK-UT-1 has no detectable hMSH2 protein, whereas 2774 retains a full-length mutant protein (12). The levels of hMLH1 and hPMS2 in these cell lines are similar to those in repair-proficient cell lines. Surprisingly, mutation rates of growing 2774 and SK-UT-1 cells at the X-linked locus encoding the purine salvage enzyme hypoxanthine guanine phosphoribosyl transferase (HPRT) (Table1) were lower than that measured for the mismatch repair–proficient SV40-transformed fibroblast line MRC-5 (rate = 1.4 × 10 7 mutations per cell per generation). This contrasts with the 130- to 190-fold increase in mutation rate found in tumor cell lines deficient in other mismatch repair components (Table 2). Other laboratories have reported difficulties in isolating HPRTmutants from hMSH2-deficient cell lines (13), although increased mutation rates have been measured in the hMSH2-deficient LoVo cell line (14, 15). To test the possibility that the low HPRT mutation rate in the hMSH2-deficient cells is due to the presence of multiple active X chromosomes (16), we measured the rate of mutation to ouabain resistance (OuaR ). Because these mutations act dominantly, they are not obscured in polyploid cells (17). However, even at this locus there was no change in mutation rate for cell line 2774 relative to that in mismatch repair–proficient MRC-5, whereas that in SK-UT-1 was elevated 7.1-fold (18). Thus, these data indicate that the two hMSH2-deficient tumor cell lines develop a weaker mutator phenotype than lines deficient in other mismatch repair genes.

Table 1

Accumulation of mutations in tumor cell lines maintained at high density.

View this table:
Table 2

HPRT mutation rates in growing and high-density cultures of mismatch repair–deficient tumor cells.

View this table:

To determine whether mutation rates increased in these hMSH2-deficient cells under less optimal culture conditions, we examinedHPRT and OuaR mutant frequencies in 2774 and SK-UT-1 cells that had been maintained for 2 weeks in a high-density growth-limited state. For 2774 cells, the HPRTmutant frequency of replicas maintained at high density was 7900-fold higher and the frequency of OuaR mutants was >67-fold higher relative to those of replica cultures kept under optimal growth conditions (Table 1). SK-UT-1 cells maintained at high density showed a 34-fold increase in frequency of HPRTmutants.

To determine the nature of the mutations in the cells maintained in restrictive conditions, we picked independent colonies from each of the replica cultures and generated HPRT cDNAs by reverse transcription (19). The mutant cDNAs were amplified and sequenced. Over half of the mutations (7 of 13) identified in 2774 were frame shifts at mono- or dinucleotide runs (Table3). The remaining HPRTmutations were transitions, transversions, and a frame shift outside the runs. In SK-UT-1 cells, six of seven HPRT mutations were frame shifts. This distinctive pattern of frame shifts is also evident in other log-growing mismatch repair–deficient cell lines (14, 20).

Table 3

HPRT mutations in 2774 and SK-UT-1 cells maintained at high density.

View this table:

We next examined microsatellites in single cell clones isolated from line 2774 to determine whether they exhibited a similar conditional instability (21). Four loci were amplified with DNA purified from 20 independent subclones isolated from growing 2774 cells. No novel microsatellite alleles were detected at these loci (frequency <1.7%). In contrast, novel alleles were detected in 36% of the subclones isolated from 2774 cultures maintained at high density. In some subclones there were losses or gains of multiple repeat units (Fig. 1). This pattern is consistent with a previous report of increased microsatellite instability in subclones of 2774 maintained for a prolonged period (10). We also examined microsatellite stability in HPRT mutants isolated from cultures maintained at high density and again observed dramatic alterations in a high proportion (46%) of subclones (Fig. 1).

Figure 1

Microsatellite instability in single cell clones of 2774 cultures maintained at high density without selection (A) and in HPRT mutant strains selected from high-density replicas (B). The dinucleotide repeat locusD10S197 was assayed. The left lane of each panel contains microsatellites amplified from DNA purified from uncloned, growing 2774 cells. Single cell clones obtained from growing cultures of 2774 had no alterations at this locus (21).

Our attempts to determine whether other mismatch repair–deficient cell lines exhibit a similar response were complicated by the high frequency of mutants accumulating in replica cultures before they reached high density. As a result, replica cultures of the hPMS2-deficient endometrial carcinoma cell line HEC-1-A maintained at high density showed only a small (2.3-fold) increase in HPRT mutant frequency relative to that of growing replicas (Table 2). More interesting was the loss of variation in the number of mutants in the high-density replica cultures (Table 2). The change in variation observed in high-density cultures was highly significant (P < 0.0001), and the variation in mutant number per high-density replica was similar to that found when a single culture of HEC-1-A cells was repeatedly sampled (Table 2). High-density cultures of the colon cancer tumor cell line DLD-1, which is deficient in hMSH6 and has a missense mutation of DNA polymerase δ (22), did not show a similar effect (Table 2). Because mutation is a random process, the number of mutants is normally highly variable, depending on when the mutation occurs during the outgrowth of a replica culture (11, 17). The loss of variability in HEC-1-A replicas maintained at high density indicates that an inductive process leads to the accumulation of mutants under such conditions. Alternatively, the much lengthened time for the expression of the HPRT deficiency in the high-density cultures may have improved the recovery of mutants. However, the fact that DLD-1 did not accumulate mutants at high density over the same length of time argues against the latter hypothesis.

Although we have clearly seen the conditional mutator phenotype in two hMSH2-deficient lines, it is not known whether this property is restricted to a subset of mismatch repair–deficient cells or occurs in a wide variety of tumor cells. The extraordinary increase in mutant frequency in 2774 cells could be the result of the stable expression of the mutant hMSH2 protein. This protein may retain some residual activity (6) that suppresses the mutation rate under growing conditions. Under conditions of stress, this activity may be saturated by an elevated level of DNA replication errors or damage. The increase in mutant frequency in SK-UT-1 cells maintained at high density relative to that in growing cells was not as dramatic. In part this can be attributed to the higher (nearly 50-fold) baseline mutant frequency of the growing cells. HEC-1-A, which is deficient in hPMS2, also continues to accumulate HPRT mutants when maintained under restrictive conditions. However, the mutator phenotype in these cells is not strictly conditional because HEC-1-A has a high rate of mutation under optimal growth conditions.

The frame shifts in hMSH2-deficient cells at high density indicate that the elevated mutant frequency is the consequence of DNA replication occurring in the absence of mismatch repair. This DNA synthesis could occur in a small subpopulation of cells that continue to grow in spite of the restrictive culture conditions. Alternatively, the accumulation of mutations may be the product of an error-prone DNA repair pathway induced by the suboptimal culture conditions. If the first hypothesis is correct, the mutation rate must be very high, as only a small proportion of cells in these cultures (7.5%) have an S-phase DNA content. The high level of microsatellite instability in subclones isolated from high-density cultures and the HPRT mutant strains appears to be consistent with the idea that widespread catastrophic DNA synthesis occurs in these cells. Thus, the conditional mutator phenotype may reflect the loss of a checkpoint that prevents cells from entering the S phase when environmental conditions are not optimal, which is not unlike the checkpoint that arrests cells exposed to hypoxia (23). With respect to the possibility of error-prone repair in cells maintained at high density, it was recently reported that transient exposure of a mouse tumor cell line to hypoxia modestly increased the mutant frequency of a target gene (24). Mismatch repair has thus far been primarily associated with correction of DNA replication errors in growing cells, but may have another function or functions outside the S phase (25).

Our observations raise the possibility that mutations in some cells may accumulate in a time-dependent manner in the absence of growth, as proposed by Strauss (26). Furthermore, because the majority of cells in a tumor may not be in a microenvironment conducive to the rapid growth that occurs in cell culture, the conditions in high-density cultures described here may more closely resemble conditions in the tumor.

  • * To whom correspondence should be addressed. E-mail: mark.meuth{at}


Stay Connected to Science

Navigate This Article