Distinct Signatures for Mutator Sensitivity of lacZ Reversions and for the Spectrum of lacI/lacO Forward Mutations on the Chromosome of Nondividing Escherichia coli

Corresponding author: J. Gowrishankar, Centre for DNA Fingerprinting and Diagnostics, ECIL Rd., Nacharam, Hyderabad 500 076, India., shankar{at}cdfd.org.in (E-mail)

Communicating editor: S. T. LOVETT

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

A conditional lethal galE(Ts)-based strategy was employed in Escherichia coli, first to eliminate all growth-associated chromosomal reversions in lacZ or forward mutations in lacI/lacO by incubation at the restrictive temperature and subsequently to recover (as papillae) spontaneous mutations that had arisen in the population of nondividing cells after shift to the permissive temperature. Data from lacZ reversion studies in mutator strains indicated that the products of all genes for mismatch repair (mutHLS, dam, uvrD), of some for oxidative damage repair (mutMT), and of that for polymerase proofreading (dnaQ) are required in dividing cells; some others for oxidative damage repair (mutY, nth nei) are required in both dividing and nondividing cells; and those for alkylation damage repair (ada ogt) are required in nondividing cells. The spectrum of lacI/lacO mutations in nondividing cells was distinguished both by lower frequencies of deletions and IS1 insertions and by the unique occurrence of GC-to-AT transitions at lacO +5. In the second approach to study mutations that had occurred in nondividing cells, lacI/lacO mutants were selected as late-arising papillae from the lawn of a galE⁺ strain; once again, transitions at lacO +5 were detected among the mutants that had been obtained from populations initially grown on poor carbon sources such as acetate, palmitate, or succinate. Our results indicate that the lacO +5 site is mutable only in nondividing cells, one possible mechanism for which might be that random endogenous alkylation (or oxidative) damage to DNA in these cells is efficiently corrected by the Ada Ogt (or Nth Nei) repair enzymes at most sites but not at lacO +5. Furthermore, the late-arising papillae from the second approach were composed almost exclusively of dominant lacI/lacO mutants. This finding lends support to "instantaneous gratification" models in which a spontaneous lesion, occurring at a random site in DNA of a nondividing cell, is most likely to be fixed as a mutation if it allows the cell to immediately exit the nondividing state.

SPONTANEOUS mutations are believed to arise randomly in dividing cells in a two-step process in which an incorrect nucleotide that is first incorporated in one DNA strand during replication then templates the incorporation of its complement in the other. Mutational avoidance mechanisms have accordingly evolved either (i) to reduce the probability of occurrence of the first step or (ii) to correct the mismatch before it is fixed as a mutation by the second one (MAKI 2002 ). In Escherichia coli, for example, the first step probability is modulated by dnaQ encoding the proofreading component of DNA polymerase III, as well as by the products of genes such as mutT, mutM, ung, mug, nth, nei, ada, and ogt, which act to remove nucleotides that are damaged (by oxidation, deamination, or alkylation) and that have an increased miscoding potential. The products of genes such as mutS, mutH, mutL, dam, and uvrD (of the postreplication mismatch repair pathway) act immediately after the first step to remove misincorporated nucleotides, and the mutY gene product similarly acts to remove a miscoded A opposite an oxidatively damaged G nucleotide. The need and the efficiency of each of these error-avoidance and error-correction mechanisms is demonstrated by the magnitude of the mutator phenotypes exhibited by the appropriate mutant strains (reviewed in DRAKE 1991 ; ECHOLS and GOODMAN 1991 ; MILLER 1992 , MILLER 1996 ; FRIEDBERG et al. 1995 ; HUTCHINSON 1996 ; RUPP 1996 ; MAKI 2002 ).

Spontaneous mutations have also been described in nondividing or very slowly dividing cells (that is, in the absence of overt DNA replication), but the subject is more controversial. Typically, the mutants have been identified as late-arising colonies or papillae following a nonlethal selection (for growth on a carbon source or for reversion of an auxotrophy), and for this reason they have also been called adaptive mutations (BRIDGES 1998 ; FOSTER 1999 ; ROSENBERG 2001 ). The best-studied example is the reversion to Lac⁺ of a frameshift lacI-Z33 mutation on F'128 in strain FC40. In brief, the results obtained so far indicate that the high frequency of adaptive Lac⁺ reversions in FC40 is a unique phenomenon that is dependent on RecA and other SOS- and homologous recombination-related reactions, as well as on the episomal location of lacI-Z33 and on F'-related conjugational transfer functions (FOSTER 1999 ; GODOY and FOX 2000 ; ROSENBERG 2001 ). Adaptive mutation in the episomal strain is also apparently associated with large arrays of tandemly amplified copies of the lac genes (ANDERSSON et al. 1998 ; HENDRICKSON et al. 2002 ) and with an increased occurrence of mutations at unselected loci on both F and the chromosome (FOSTER 1999 ; BRIDGES 2001 ; ROSENBERG 2001 ). Two distinct classes of models have been proposed to explain this phenomenon: (i) the amplification-mutagenesis model, which invokes lac amplification as an essential intermediate in the pathway to adaptive mutagenesis, with SOS induction and generalized mutagenesis being unavoidable side effects of the process (ANDERSSON et al. 1998 ; HENDRICKSON et al. 2002 ; SLECHTA et al. 2002 ); and (ii) the hypermutable state models, in which SOS-induced mutagenesis, and not amplification, is causally related to adaptive reversion (ROSENBERG 2001 ).

Far fewer studies have been done on adaptive mutations in chromosomal genes, and even so diverse targets and assay procedures have been employed (although most of them have scored for reversions, that is, for dominant mutations; FOSTER 1999 ; GODOY et al. 2000 ; SLECHTA et al. 2002 ). Some studies have also employed lethal selection (e.g., for rifampicin resistance) on cells from aging colonies as a means to obtain mutations in presumptively nondividing cells (TADDEI et al. 1995 , TADDEI et al. 1997 ; BJEDOV et al. 2003 ).

We described earlier a novel strategy for obtaining lacZ reversions in nondividing cells of E. coli, to the exclusion of those that occur in dividing cells (REDDY and GOWRISHANKAR 1997A ). The lacZ parental strain also carried a galE(Ts) mutation and was grown at 42° on lactose-supplemented plates so that the Lac⁺ revertants arising during growth (but not the Lac^- parental cells) were killed by intracellular galactose released from lactose hydrolysis. The population of Lac^- cells in stationary phase was then shifted to 30° so that any new Lac⁺ revertants arising in it would survive and be identified as papillae. With this approach, we showed that a variety of mutations (including an insertion) in chromosomal lacZ can revert to Lac⁺ in the central zones of nondividing cells in colonies growing on a solid medium.

In the present study, we have adopted a modified version of the method above to characterize, in nondividing cells, the mutator sensitivity of spontaneous reversions of a chromosomal lacZ(Am) allele. Our results indicate that alkylation- and some oxidative-damage repair functions, but not the functions of methyl-directed mismatch repair or SOS repair, are important for genome defense in nondividing cells. We have also examined a forward mutational target, namely the lacI/lacO repressor/operator genes, in nondividing cells and have found that the corresponding mutational spectrum is quite distinct from that in dividing cells.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Bacterial strains, plasmids, and media:
E. coli K-12 strains and their genotypes are described in Table 1 and in the legends to Fig 1 and Fig 2. Routine growth media were Luria-Bertani (LB) and minimal A (MM) with 0.2% glucose (Glc), as described (MILLER 1992 ). Plates that were incubated for extended lengths of time were enclosed in a humidified box, and the medium in them was supplemented with nystatin at 20 µg/ml to inhibit the growth of fungal contaminants. Concentrations of other antibiotics used were as described earlier (REDDY and GOWRISHANKAR 1997A , REDDY and GOWRISHANKAR 1997B ).

View larger version (73K): In this window In a new window Download PPT slide   Figure 1. Lac+ papillation in galE(Ts) strain with F'128-borne lacI-Z33 (WT) and in its recA and lexA3 derivatives. Strains employed are as described in Table 2. Representative pairs of plates are shown. Plates in the top row were incubated continuously at 28°, and those in the bottom row were subjected to temperature shift from 42° to 28°.

View larger version (42K): In this window In a new window Download PPT slide   Figure 2. Lac+ papillation in galE(Ts) strain with chromosomal lacZ(Am) (WT) and in its indicated mutant derivatives. Strains employed are as described in Table 2. Representative pairs of plates are shown. Plates in the top row were incubated continuously at 28°, and those in the bottom row were subjected to temperature shift from 42° to 28°.

Genetic and DNA methods:
The viable count of bacteria in a lawn was estimated from the number of colonies obtained (at appropriate dilutions) from the suspension of an agar plug removed from the lawn with the aid of a 1-ml micropipette tip. Transfer of mutations between strains was by P1 transduction (GOWRISHANKAR 1985 ). Dominance or recessivity of chromosomal lacI/lacO mutations was tested by introducing an F'128 derivative (carrying lacI⁺ lacO⁺ and lacZ::Tn10dKan as a selective marker) from donor strain GJ2479 into the mutants by conjugation (MILLER 1992 ) and scoring for, respectively, blue or white colony phenotypes on Glc-MM supplemented with 25 µg/ml Xgal. Mutagenesis with nitrosoguanidine was performed as described (MILLER 1992 ). Standard protocols were followed for experiments involving recombinant DNA (SAMBROOK et al. 1989 ).

galE(Ts)-based papillation method for lacZ reversions in nondividing cells:
In the method previously described (REDDY and GOWRISHANKAR 1997A ), colonies of the galE(Ts) lacZ parental strain had been grown on the Lac⁺ papillation medium at 42° and then shifted to 30°, and papillae arising in the central zones of the colonies following the temperature shift were taken to represent reversions that had occurred in nondividing cells. The parental strain had also carried (i) mutations in the galactose-transporter loci mgl and galP, to preclude bystander killing of Lac^- cells following release of galactose by the lysis at 42° of Lac⁺ revertants in their vicinity; and (ii) a multicopy Amp^r plasmid carrying the galE(Ts) gene, to provide sufficient GalE function at the permissive temperature for papillary growth of the Lac⁺ revertants.

Three modifications of this method were adopted in the present study: (i) the parental strain had only the chromosomal galE(Ts) allele and did not carry the multicopy-Amp^r plasmid with galE(Ts); (ii) 10⁷ cells of the parental lacZ galE(Ts) galP mgl strain or its mutator derivatives were plated in a soft agar overlay on Lac⁺-papillation medium [MM supplemented with 0.5% Casamino acids (CAA), 0.1% lactose, and nystatin] in a 35-mm petri plate and incubated at the restrictive temperature (42°) for 72 hr, during which time the parental strain had grown as a lawn to stationary phase and all growth-associated Lac⁺ revertants in the population were eliminated; after which (iii) the plate was shifted to the permissive temperature of 28° (rather than 30°) at which temperature the haploid galE(Ts) allele conferred sufficient GalE activity for growth of the Lac⁺ papillae. After 7–10 days at 28°, 100 µg of Xgal was added in a soft agar overlay for visualization of Lac⁺ papillae. As controls, Lac⁺ papillae were also visualized on plates that had been continuously incubated at 28°.

With the modified method, the need to demarcate central and peripheral zones of nondividing and dividing cells, respectively, for each colony was avoided, since the entire lawn on the temperature-shifted plate could now be treated as a population of nondividing cells; furthermore we had observed that the earlier method was associated with variable spontaneous loss of the multicopy galE(Ts) plasmid from cells in the colonies (and consequently with loss of the ability of Lac⁺ cells to grow as papillae), presumably because of the loss of selection for the plasmid following enzymatic inactivation of the ampicillin in the medium.

galE(Ts)-based papillation method for lacI/lacO forward mutations in nondividing cells:
The strategy involved exposure of a population of cells of a galE(Ts) galP mgl strain with an inducible lac operon to phenyl-ß-D-galactoside (PG) at 42°, so that all mutants expressing ß-galactosidase constitutively (that is, with lacI/lacO mutations) were killed by the Gal released from PG hydrolysis. The population was then shifted to 28° to allow for growth as papillae of new lacI/lacO mutants that had arisen among the nondividing cells. The parental strain had the mutations lacI^q (increased lacI expression) and PL8 (reduced lac promoter strength), both of which served to reduce the level of leaky background expression of lacZ in the absence of inducer (MILLER 1992 ). The lacI coding region was either wild type (WT) (lacI⁺) or one (lacI204) in which the triple-tetranucleotide hot spot site for frameshift mutation had been destroyed by synonymous codon substitution mutations (SCHAAPER and DUNN 1991 ).

An inoculum of 10⁶ cells of the parental strain was grown as a lawn at 28° for 36 hr to stationary phase on MM supplemented with 0.2% CAA, 0.05% Glc, nystatin, and 25 µg/ml Xgal in a 35-mm petri plate (to a final cell density of 4 x 10⁹ cells), following which PG was added to 0.05% in a soft agar overlay and the plate was shifted to 42° for 72 hr. Reconstitution experiments were undertaken to demonstrate that all PG⁺ mutants that arose during growth of the lawn would be eliminated during the period of incubation at 42° (data not shown). The plate was then returned to 28° and scored daily over the succeeding 10–15 days for blue PG⁺ papillae. The galE(Ts) derivatives of both the lacI⁺ and lacI204 parental strains yielded around three PG⁺ papillae per 10⁹ cells on the temperature-shifted plates.

lacI/lacO forward mutations in dividing cells:
Independent cultures were initiated, each with an inoculum of 100 cells, and grown at 28° to an A₆₀₀ of 0.05 in 3 ml of LB or Glc-MM. The cells were washed in MM, plated on 0.1% PG-MM, and incubated at 28° for 48–72 hr. One randomly chosen PG⁺ mutant from each culture was saved.

Screening for mutations at lacI CTGG repeat hot spot:
All PG⁺ mutants that were obtained from the lacI⁺ parent were first screened for presence in lacI of expansion or contraction at the triple tetranucleotide repeat sequence (FARABAUGH et al. 1978 ). The lacI region spanning the repeat was PCR amplified from each mutant with primers 5'-AGCGGGCCCATTAAGTTCT-3' and 5'-TCGGCTGAATTTGATTGCGA-3', and the PCR products were electrophoresed on a 20% polyacrylamide gel. It was thus possible to distinguish the PCR products of the expansion mutants (87 bp) and contraction mutants (79 bp) from those of mutants that had no size change in this region (83 bp).

Sequence analysis of lacI/lacO mutants:
The complete lacI/lacO locus from each mutant (that did not have its mutation at the CTGG repeat hot spot) was PCR amplified with primers 5'-CCCGACACCATCGAATG-3' (upstream of lacI, top strand) and 5'-GCCTCTTCGCTATTACGCCA-3' (in proximal lacZ, bottom strand), and its sequence was determined with the aid of these two primers and the series of forward and reverse sequencing primers described earlier (ERFLE et al. 1996 ). Numbering schemes for nucleotides in lacI and lacO are as in FARABAUGH 1978 and GILBERT et al. 1973 , respectively.

Statistical comparisons between lacI/lacO spectra:
The statistical significance of differences between lacI/lacO mutational spectra was determined with the aid of the hypergeometric test algorithm previously described (ADAMS and SKOPEK 1987 ; PIEGORSCH and BAILER 1994 ). The software for performing the test is described in CARIELLO et al. 1994 and is available at http://www.ibiblio.org/dnam/des_hypg.htm. The data were subjected to 1700 Monte Carlo simulations, and the proportion of the simulated tables that exhibited a distribution more improbable than that observed was taken as the estimate of the P value of the observed table under the null hypothesis.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

galE(Ts)-based strategy for lacZ reversions in nondividing cells:
A modified version of the previously described galE(Ts)-based strategy (REDDY and GOWRISHANKAR 1997A ) was employed to examine sensitivity to various mutators for reversion of a chromosomal lacZ(Am) allele in nondividing cells. The parental lacZ galE(Ts) strains were either grown as a lawn to stationary phase on Lac⁺ papillation medium at the permissive temperature (28°) and maintained at 28° or grown at the restrictive temperature for lacZ⁺ revertants (42°) and then shifted to 28°. Viable count measurements (data not shown) indicated that the parental cells (both WT and mutators) attained a density of 3 x 10⁹ cells/plate at either temperature in the stationary phase and did not change much thereafter.

For each strain, the Lac⁺ papillation frequencies (per 10⁹ viable cells) on plates continuously incubated at 28° (A) and on the temperature-shifted plates (B) were determined as averages of at least three experiments, each with duplicate cultures; the variation between the individual values was <20%. Since each papilla represents an independent mutational event, the papillation frequency itself provides a direct estimate of the mutation frequency in the strain. The first value A is a measure of the spontaneous Lac⁺ reversion frequency in the strain while the second value B is a measure of the reversions occurring in nondividing cells. However, because of the possibility that the absolute value of B may be artifactually elevated (particularly in the mutator strains) even if a small fraction of the mutants arising during growth had escaped killing at the restrictive temperature (see DISCUSSION), we have chosen to use the B/A ratio as a measure of the relative proportion of reversions that had arisen in nondividing cells.

B/A ratios for F' lacI-Z33 and chromosomal lacZ::Tn10dKan reversions:
That the B/A ratio is a valid indicator of reversions in nondividing galE(Ts) cells was established by two sets of experiments. First, reversion of the F'-borne lacI-Z33 allele was reduced by recA or lexA3 mutations (that abolish SOS functions) both on the temperature-shifted plates and on plates continuously incubated at 28° (see Fig 1; papillation frequencies listed in Table 2), and also by a multicopy mutL⁺ plasmid (data not shown), just as has been reported earlier for adaptive reversions in FC40 (FOSTER 1999 ; ROSENBERG 2001 ). Second, the frequency of precise excision of a Tn10dKan insertion in lacZ was increased 10-fold in uup or ssb mutants on plates continuously incubated at 28°, but was largely unaffected by the mutations on temperature-shifted plates (Table 2); these data are consistent with previous findings that the ssb and uup mutations predominantly increase precise excision of Tn10 derivatives in dividing cells (REDDY and GOWRISHANKAR 1997A , REDDY and GOWRISHANKAR 1997B ).

B/A ratios for chromosomal lacZ(Am) reversions in mutator strains:
The chromosomal lacZ(Am) mutation under study here is in codon 18 of the gene and is the same as that first used by CAIRNS et al. 1988 to identify the phenomenon that is now referred to as adaptive mutation. The earlier study, however, employed an episomally located allele. Lac⁺ revertants in these strains can arise by all six possible types of base substitutions at lacZ or the suppressor tRNA loci.

The papillation frequency values A and B (as well as the B/A ratio) for the WT chromosomal lacZ(Am) galE(Ts) strain were similar to those for its recA derivative (Table 2). Furthermore, as expected (DRAKE 1991 ; ECHOLS and GOODMAN 1991 ; MILLER 1992 ; FRIEDBERG et al. 1995 ; HUTCHINSON 1996 ; RUPP 1996 ; SLUPSKA et al. 1996 ), various mutator derivatives of the WT strain exhibited 2- to 100-fold increases in Lac⁺ papillation frequency on plates continuously incubated at the permissive temperature (that is, in the values of A); however, the B/A ratios for these strains exhibited no correlation with their mutator magnitude. The papillation plates for some representative strains are shown in Fig 2, and all the papillation values are listed in Table 2. On the basis of their B/A ratios, the mutator strains could be grouped into three categories: very high (0.9; ada ogt), very low (0.01–0.075; mutH/L/S, dam, uvrD, dnaQ [mutD5], and mutM/T), and roughly equivalent to WT (0.27–0.45; mutY, nth nei, and mutA). These three categories of mutations may therefore be taken to exert their predominant mutator effects, respectively, in nondividing cells, in dividing cells, and in both dividing and nondividing cells. The B/A ratio value for the mutYM double mutant was intermediate between that for either single-mutant strain (Table 2). The ung and miaA mutants were modest mutators, with little or no increase in Lac⁺ papillation frequency on the temperature-shifted plates (data not shown). The values of A and B/A were not significantly altered also by mutations in lexA, dinB, recB, recD, recF, recG, recQ, ruvABC, polB, xthA, mfd, rep, dps, nfo, soxR, or sodB or by overexpression of MutL and MutS from a plasmid in the mut⁺ galE(Ts) strain (data not shown).

Three sets of control experiments (data not shown) were performed to exclude artifactual explanations, such as differences in viability of the populations under selection or in efficiency of elimination of growth-associated Lac⁺ mutants at the restrictive temperature, for the different B/A ratios in the various mutants. First, no significant differences were observed in viability of the populations under selection. Second, Lac⁺ revertants of the ada ogt strain (which exhibited the highest B/A ratio) were shown in reconstitution experiments to both form papillae at 28° and be killed at 42°, just as efficiently as Lac⁺ revertants from the WT or the mutS strain (which had one of the lowest values for B/A). Finally, when mixed cultures of differentially marked mutH and mutY strains (that are comparable in their mutator magnitude but have different B/A ratios) were plated on the Lac⁺ papillation medium, each strain continued to exhibit its distinctive B/A ratio, indicating that the low frequency of Lac⁺ in the mutH strain following temperature shift is not a consequence of massive bystander cell lethality at the restrictive temperature.

Additional support for the correlation between the different mutator mutations and their distinct B/A ratios came from an experiment in which we subjected the lacZ(Am) galE(Ts) strain GJ2231 to nitrosoguanidine mutagenesis, identified new mutator derivatives by screening for increased Lac⁺ papillation in colonies continuously incubated at 28°, and then classified the mutators on the basis of their B/A ratios. Three mutants with a 5- to 16-fold increase in papillation and a B/A ratio between 0.35 and 0.46 were shown to be mutY, and six others with an 8- to 12-fold increase in papillation and a B/A ratio between 0.05 and 0.15 were mapped to mutH (four), mutS (one), and uvrD (one; data not shown).

lacI/lacO mutational spectra in nondividing cells:
Just as lactose can select for or against Lac⁺ strains that are, respectively, galE⁺ or galE, so too can the lactose analog PG select for or against lacI/lacO strains (which express ß-galactosidase constitutively) that are galE⁺ or galE. A lawn of cells of the parental galE(Ts) strain (with a functional lac repressor/operator system) was first exposed to PG at 42°, so that all lac-constitutive mutants in the population were eliminated and then shifted to 28°. Papillae obtained after the temperature shift were assumed to represent mutations in lacI or lacO that had arisen within the nondividing cells.

To determine the spectrum of lacI/lacO mutations in nondividing cells, we purified 95 PG⁺ papillae from temperature-shifted plates of the galE(Ts) strain GJ2422 with the lacI⁺ coding sequence; this number was composed of all papillae that arose from each of nine independently plated cultures. As a control, 130 independent PG⁺ mutants were obtained from early exponential-phase cultures, for determination of the spectrum for dividing cells. The lacI/lacO lesion in each mutant from either collection was then identified as described in MATERIALS AND METHODS.

A comparison between the two mutational spectra (see Table 3; details are presented in supplemental Table 4 at http://www.genetics.org/supplemental/) revealed several differences, even though the frequency of expansion (44% of total) or contraction (6% of total) of the CTGG repeat at the hot spot site in lacI⁺ was similar. Statistical analysis of the data by the hypergeometric test also established that the spectra were different from one another at P < 10^-6. [The log-phase spectrum obtained in this study was more or less identical with the lacI/lacO spectra described earlier for WT strains (SCHAAPER et al. 1986 , SCHAAPER et al. 1987 ; HALLIDAY and GLICKMAN 1991 ).]

Deletions [both at a specific hot spot site previously described (1108-lacO +4; HALLIDAY and GLICKMAN 1991 ) and at others] as well as IS1 insertions were underrepresented in the spectrum for nondividing cells. Two mutations (GC-to-AT transitions at positions 56 in lacI and +5 in lacO, respectively) that were prominent hot spots in nondividing cells (with each accounting for 10% of the total mutant numbers) were not represented at all among the log-phase mutants.

If one considers just the subset of lacO mutations, a characteristic feature of all WT spectra so far reported is a ratio of <0.05 for +5 (GC-to-AT) to +6 (AT-to-GC) transitions, and indeed all the lacO mutations that occurred in dividing cells were at +6. On the other hand, this ratio was 2.8 in the spectrum obtained from nondividing cells; furthermore, even apart from the lacO +5 site, several other positions in lacO recorded hits in the spectrum for nondividing cells but not in that for dividing cells (Fig 3).

View larger version (28K): In this window In a new window Download PPT slide   Figure 3. Frequency distribution of substitution mutations in the lacO locus of dividing (red) and nondividing (blue) cells of galE(Ts) lacI+ (GJ2422, solid bars) or galE(Ts) lacI204 (GJ2432, striped bars) strains. Total numbers of lacI/lacO mutants obtained in the different categories are indicated in parentheses. At any individual site in lacO, all the substitutions were identical: +5 and +11, GC to AT; +6 and +16, AT to GC; +8, AT to CG; and +10, GC to TA.

In control experiments, the lacO +5 or the lacI 56 mutants, which were identified as occurring more frequently on the temperature-shifted plates, were indistinguishable from other mutants such as lacO +6 or lacI +195 with regard to three features that were tested (data not shown): (i) constitutivity of lacZ expression, as measured by ß-galactosidase assays; (ii) sensitivity to PG at 42°; and (iii) ability to form papillae on PG-containing medium at 28°. The latter two findings were established from reconstitution experiments in which pairs of the mutants (differentially marked) that were mixed in defined ratios were plated at 100 cells/plate along with a lawn of the lac strain GJ2218.

It was recognized that the validity of the individual differences observed between the two spectra may be limited by the fact that multiple hypotheses were being tested. Accordingly, a second comparison of the lacI/lacO mutational spectra was undertaken, this time in the galE(Ts) lacI204 strain GJ2432 in which the CTGG tetranucleotide repeat expansion/contraction hot spot in lacI has been destroyed without altering the primary structure of the encoded protein (SCHAAPER and DUNN 1991 ). (This strain was chosen to reduce the numbers of mutants for analysis, since the previous comparison had shown no difference between the two spectra for mutations at the repeat site.) Around 30 PG⁺ mutants each from the exponential-phase cultures and from the temperature-shifted plates were studied, and several features from the first comparison were reproduced in the second (see Table 3 and Fig 3; detailed information is in supplemental Table 4): (i) a reduction in the proportion of deletions (at both 1108-lacO +4 and other sites) and of IS1 insertions, (ii) a marked increase in the proportion of GC-to-AT transitions at +5 of lacO, and (iii) increased mutability at additional lacO positions (except +6) in the spectrum for nondividing cells relative to that for dividing cells. The ratio, in the mutants from nondividing cells, of +5 to +6 alterations in lacO was 2.0, whereas once again the only lacO mutants isolated from the population of dividing cells were at +6. By the hypergeometric test, the lacI/lacO spectra for dividing and nondividing cells of the lacI204 strain were different at P = 0.005 (95% confidence limits, 0.001–0.008).

Greatly increased frequency of dominant mutations in late-arising PG⁺ papillae of the galE⁺ strain:
As mentioned above, the majority of investigators studying mutations in nondividing cells have collected mutants at late time points from populations subjected to a nonlethal selection. To compare this approach with the galE(Ts)-based one employed in the present work, we collected PG⁺ papillae as a function of time from the galE⁺ lacI204 parental strain grown on CAA-Glc [that is, the same medium as that employed above for the galE(Ts)-based approach] and also by following the identical temperature-shift protocols. In addition, PG⁺ papillae were collected from plates in which 0.2% acetate was substituted for CAA-Glc as the utilizable carbon source. Since in both instances the parental strain was galE⁺, there was no selection against lacI/lacO mutants even at 42°, and the papillation frequency was 16/10⁹ cells.

Plates were examined daily for up to 14 days; papillae were marked on the days that they became visible, and all of them were purified at the end of the experiment. Since our interest was in determining the conditions, if any, in which GC-to-AT transitions at lacI +56 and lacO +5 occurred [that were the apparently characteristic features of the spectrum for nondividing cells obtained by the galE(Ts)-based approach], and since both mutations are dominant to lacI⁺ lacO⁺, we first classified the new mutants into dominant and recessive categories and then sequenced only the former.

The data for PG⁺ mutants that were grouped into three classes are presented in Fig 4, namely as early-, intermediate-, and late-arising papillae. On lawns grown on either carbon source (CAA-Glc or acetate), the proportion of dominant mutations was low among the early-arising papillae [similar to that reported earlier for spontaneous lacI/lacO mutations in the lacI204 strain (SCHAAPER and DUNN 1991 )], but rose progressively to approach nearly 100% among the late-arising papillae. The preponderance of dominant lacI/lacO mutations among late-arising papillae was apparent even when the lawns (both on CAA-Glc and on acetate) were not subjected to temperature shift but were continuously incubated at 28°, and also when 0.2% Glc (in MM) or LB medium was used as the carbon source for growth of the lawn at 28° or 42° (data not shown).

View larger version (44K): In this window In a new window Download PPT slide   Figure 4. Frequency distribution of recessive and dominant lacI/lacO mutations (bottom and top bars, respectively, in each pair) in PG+ papillae from the lacI204 galE+ strain GJ2474, as a function of their time of appearance following PG addition to plates with different utilizable carbon sources (as indicated). Dominant mutations were further classified into four categories, as per the key shown. The three ascending time intervals marked in each part correspond to the designations early, intermediate, and late, respectively, that have been used in the text. N, total number of mutants tested. For palmitate and succinate sections, mutants were pooled from plates that were either subjected to temperature shift or continuously incubated at 28°; for the other two sections, mutants were from temperature-shifted plates.

Sequence analysis of the collection of dominant lacI/lacO mutants from the temperature-shifted plates with CAA-Glc or acetate revealed the occurrence of +5 lacO mutations (and a ratio of 2.1 for the +5 to +6 lacO transitions) for only the late-arising papillae in populations grown on acetate as the carbon source, but not in the other categories such as early-arising papillae on acetate or both early- and late-arising papillae on CAA-Glc (Fig 4). From the other collections, of 11 dominant mutants obtained as PG⁺ papillae on or after day 5 on the acetate-grown lawn at 28°, 2 were again at lacO +5; on the other hand, none of 52 additional dominant mutants arising from the CAA-Glc-grown lawns was at this position.

The preponderance of dominant lacI/lacO mutations among the late-arising PG⁺ papillae was noted also for the galE⁺ lacI⁺ strain GJ2421 (that is, with the triple tetranucleotide hot spot site in the lacI coding region). Twenty-seven of a total of 37 papillae (73%) obtained on or after day 5 on lawns of this strain grown on CAA-Glc harbored dominant mutations (but they were not sequenced).

What feature of growth on acetate is correlated with mutation at the lacO +5 site?
The data above indicated that mutations at lacO +5 occur in late-arising papillae of acetate-grown but not in those of CAA-Glc-grown populations. Acetate is among the poorest of C sources for E. coli, whose utilization is dependent on full derepression of both the tricarboxylic acid cycle pathway and the glyoxylate bypass pathway (CLARK and CRONAN 1996 ). Acetate-grown cells are downregulated for the PhoP-PhoQ regulon (LESLEY and WALDBURGER 2003 ), and HALL 1998A has shown that the latter regulates adaptive mutagenesis. Acetate is also known to accumulate in cultures grown on Glc with inadequate aeration (for example, to high cell densities) and to be toxic under these conditions (KLEMAN and STROHL 1994 ).

To determine which feature of acetate metabolism might be associated with mutations at lacO +5, we obtained PG⁺ papillae of the galE⁺ lacI204 strain grown on palmitate (which feeds exclusively into the acetate metabolism pathway) or on succinate (which is also a poor carbon source but whose utilization does not invoke the glyoxylate bypass). In these two cases as well, we observed that the ratio of recessive to dominant mutations was reversed in the intermediate- and late-arising papillae when compared with that in the early-arising papillae. Furthermore, transitions at lacO +5 were obtained in both instances (just as with the acetate-grown cultures) among the late-arising papillae (Fig 4).

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

We employed galE(Ts) strains and an incubation protocol involving temperature shift, first to prepare populations of cells in which all growth-associated lacZ or lacI/lacO mutations were eliminated and then to obtain mutations in these chromosomal target loci that had arisen in nondividing cells. Since Lac⁺ reversions in the lacZ(Am) strain can occur only by a limited set of events, we exploited this system to study the effects in nondividing cells of the various known mutator mutations. On the other hand, since lacI/lacO forward mutations encompass a large variety of molecular events, we undertook a mutational spectrum analysis for this target in the nondividing cells.

In the case of adaptive Lac⁺ mutation in strain FC40, Roth and coworkers have suggested that the underlying mechanism is selection for amplification of the region including the leaky lacI-Z33 allele, resulting in slow-growing microcolonies within each of which a reversion to Lac⁺ is then likely to occur (ANDERSSON et al. 1998 ; HENDRICKSON et al. 2002 ; SLECHTA et al. 2002 ). In the present study as well, the alleles under selection for Lac⁺ or PG⁺ may be considered to be leaky, and therefore an important question that arises is whether a similar mechanism (of growth-associated mutations within microcolonies bearing amplifications of the lac locus) might explain the occurrence of lacZ or lacI/lacO mutations in the ostensibly nondividing populations. However, for the following reasons we do not believe this to be a likely mechanism operating in the galE(Ts) strains initially grown at the restrictive temperature (although, as discussed in a subsequent section, it may indeed be so for late-arising papillae in the galE⁺ strain).

First, the galE(Ts) mutant is Gal^- (in addition to being Gal^S) at 42° and hence cannot grow on, even if it is not killed by, the galactose that might be slowly released from PG hydrolysis at this temperature; even in the case of selection for Lac⁺, ANDERSSON et al. 1998 have shown that lac locus amplification cannot permit a leaky lacZ mutant to form microcolonies if the strain is also Gal^-. Second, the process of chromosomal amplification is RecA dependent, whereas the Lac⁺ and PG⁺ papillation frequencies on temperature-shifted plates of the galE(Ts) strains were RecA independent. Third, lacI/lacO mutations that arise within a tandemly amplified region to confer a PG⁺ phenotype would necessarily have to be dominant, whereas we found that a significant proportion of mutations recovered in PG⁺ papillae of the galE(Ts) strain after temperature shift were recessive (see below). Finally, the fact that the lacZ reversions and lacI/lacO mutations exhibited, respectively, distinct mutator specificities and distinct mutational spectra also suggests that they may not have arisen in the dividing cells within microcolonies.

Need for alkylation- and oxidative-damage repair, but not for methyl-mismatch repair or SOS/recombination functions, in nondividing cells:
For each lacZ(Am) galE(Ts) derivative with a mutator mutation, we determined the relative Lac⁺ papillation frequency both on plates continuously incubated at the permissive temperature (which is a measure of the overall mutator magnitude, designated A) and on plates shifted from the restrictive to the permissive temperature (which is a measure of mutation frequency in nondividing cells, designated B).

As explained above, we have chosen the B/A ratio (rather than B itself) as an index of the relative proportion of reversions to Lac⁺ that had arisen in nondividing cells. This was done to avoid the possibility of false-positive interpretations being made from papillae observed on the temperature-shifted plates that might actually represent growth-associated mutants that had escaped killing at the restrictive temperature. The upper limit of the estimate for the fraction of such putative surviving mutants is 0.01 (since the lowest observed B/A ratios were in this range).

The major interpretations arising from our data on B/A ratios for the various strains are that the genes involved in alkylation-damage repair (ada ogt) and methyl-mismatch repair are predominantly required for protection of genome integrity, respectively, in nondividing cells and in dividing cells. Mutations in all the five genes involved in mismatch repair (mutH, -L, -S, dam, and uvrD; RADMAN and WAGNER 1986 ; MODRICH 1991 ) behaved similarly in our assays. Of the genes involved in oxidative-damage repair, mutY and nth nei function in both dividing and nondividing cells whereas mutM is required only in dividing cells. mutT also appears to be a mutator predominantly in dividing cells, although in this case the values for B and B/A are sufficiently above the upper limit of false positivity to suggest that it does have a moderate mutator activity even in nondividing cells. Finally, unlike the case with adaptive reversions on the episome in the FC40 system (FOSTER 1999 ; ROSENBERG 2001 ), recombination and SOS functions are not involved in spontaneous mutagenesis or its avoidance at chromosomal lacZ in nondividing cells.

We have shown that the differences in B/A ratios for the different mutators are not because of differences in either viability of the parental population or growth of mutants to form papillae. Residual growth of the population, following temperature shift to 28°, may also be excluded as an explanation for the papillae that arose subsequently, because in that case one would have expected a direct correlation between the values for A (that is, mutator magnitude) and B/A for the different strains.

Our finding that mutations in ada ogt, mutY, and nth nei promote spontaneous mutagenesis on the chromosome in nondividing cells is in accord with the results obtained by other workers using a variety of mutational targets and assay systems. For example, ada ogt mutants exhibit increased frequencies of mutations following prolonged starvation (REBECK and SAMSON 1991 ; FOSTER and CAIRNS 1992 ; MACKAY et al. 1994 ; TAVERNA and SEDGWICK 1996 ; VIDAL et al. 1998 ) and so too do mutY mutants of both E. coli (BRIDGES et al. 1996 ) and Pseudomonas putida (SAUMAA et al. 2002 ). This implies that endogenous alkylation and oxidation damage to DNA (giving rise to mutable lesions) does occur in nondividing cells. There is evidence also that active oxygen species may contribute to spontaneous endogenous DNA lesions in nondividing cells (BENOV and FRIDOVICH 1996 ; BRIDGES and TIMMS 1998 ; BJEDOV et al. 2003 ).

On the other hand, there has been less consensus in the literature on the role of mismatch repair in nondividing cells (FOSTER 1999 ; ROSENBERG 2001 ). For example, different reports, albeit in diverse mutational assay systems, have suggested that mismatch repair functions are necessary (FOSTER and CAIRNS 1992 ; JAYARAMAN 1992 ; BRIDGES and TIMMS 1997 ; BJEDOV et al. 2003 ), that they are limiting (HARRIS et al. 1997 ), or even that they are dispensable (BREGEON et al. 1999 ) for avoidance of mutations in the stationary phase.

Distinctive features of the lacI/lacO mutational spectrum in nondividing cells:
In comparison with the log-phase lacI/lacO spectrum, that for nondividing cells exhibited (i) an unchanged frequency of expansions or contractions at the lacI tetranucleotide repeat, (ii) fewer deletions (at both 1108-lacO +4 and other sites) and IS1 insertions, (iii) GC-to-AT transition hot spots at lacO +5 and perhaps lacI 56, and (iv) increased mutability in lacO at sites other than +6. In this context, HALL 1998B , HALL 1999 has also reported small decreases in the proportion of IS1 insertions among the adaptive mutations at two chromosomal loci, bgl and ebgR; and BJEDOV et al. 2003 found that GC-to-AT transitions were increased in aging colonies at both the rpoB and lacZ loci.

The lacO +5 hot spot was identified in the spectra for nondividing cells (grown on CAA-Glc) from both the lacI⁺ and lacI204 parental strains, and the concomitant increase in the lacO +5 to +6 mutation ratio is the first reported for any of the spontaneous or induced mutational spectra in lacO (see, for example, FOWLER et al. 1986 ; FIX et al. 1987 ; SCHAAPER et al. 1987 ). From an analysis of the lacI/lacO mutants in the galE⁺ strain also grown on CAA-Glc (Fig 4), we have shown that these differences cannot be explained simply on the grounds that the mutants were obtained as papillae or from populations that had been subjected to temperature shift.

Unlike the hot spot at lacO +5, that at lacI 56 was observed in only one of the two spectra determined for nondividing cells, and even in this case all the mutations were obtained from just two of nine cultures (albeit as papillae, each of which represents an independent mutation; data not shown). Its significance for the present study is therefore unclear. Nevertheless, even if the data for this position were excluded from the comparison, the mutational spectra for dividing and nondividing cells of the lacI⁺ strain were judged to be significantly different (P = 0.002, 95% confidence limits 0–0.004).

Dominant lacI/lacO mutations in late-arising PG⁺ papillae and support for "instantaneous gratification" models of stationary-phase mutagenesis:
In the second approach to studying lacI/lacO mutations in nondividing cells, we examined late-arising PG⁺ papillae of the galE⁺ strain grown as lawns on different carbon sources and compared them with the early-arising papillae on these plates. Once again, transition mutations at lacO +5 were observed among only the late-arising papillae and furthermore only in cells initially grown on poor carbon sources. That growth on different carbon sources can modulate mutagenesis in aging colonies has been shown earlier (BJEDOV et al. 2003 ), but the underlying mechanism is not known. Whether there is a catabolite repression effect (AMBROSE and MACPHEE 1998 ), caused by Glc addition in the CAA-Glc plates, also remains to be determined. Nevertheless, the notion that lacO +5 is a specific target for spontaneous mutations in nondividing cells is reinforced by these data.

The other remarkable feature associated with the late-arising PG⁺ papillae was that virtually all of them harbored dominant lacI/lacO mutations (this time, irrespective of the carbon source employed for growth of the lawn), whereas mutations in the majority of early-arising papillae were recessive (similar to that reported in previous spectra; SCHAAPER and DUNN 1991 ). The dominant mutations were distributed over a number of sites, suggesting that it is the dominance per se that correlates with their occurrence. To our knowledge, this is the first study in which mutations occurring in nondividing cells subjected to a nonlethal selection were sought to be classified as dominant or recessive; all other studies done previously had employed assays only for dominant reversions.

Our findings provide support to a certain class of models that we refer to as instantaneous gratification models [the earliest of which was proposed by STAHL 1988 ; see FOSTER 1992 for review], to explain the mechanisms of mutagenesis in nondividing cells. In these models, a spontaneous mutagenic lesion occurring at random on a DNA strand in a nondividing cell (with sufficient cytoplasmic concentration of lac repressor protein) is most likely to be fixed as a mutation if it is immediately able to confer the ability to utilize PG, either (i) by virtue of its occurrence on the template strand for transcription, whereby it can direct the synthesis of a dominant negative repressor, or (ii) by rendering the operator incapable of binding repressor. The resulting mutations will necessarily be dominant.

Absence of bias against recessive lacI/lacO mutations obtained by the galE(Ts)-based approach:
The preponderance of dominant lacI/lacO mutations in nondividing cells was observed only when the mutants were identified as late-arising papillae from the galE⁺ strains, and not when they were obtained by the temperature-shift strategy from the galE(Ts) strains. In the latter case, for example, 13 of 28 mutations in the lacI204 background were recessive (Table 2), and the proportions did not vary with the time of appearance of papillae on the plates (data not shown). Furthermore, it is not likely that the recessive mutants were ones that actually arose during growth of the lawn but were not efficiently eliminated at 42°, since their molecular characterization suggests that the majority (if not all) of them are null for LacI function.

Although the reasons for absence of bias against recovery of recessive mutations in the galE(Ts)-based approach are not known, one possibility is that papillae on temperature-shifted plates of the galE(Ts) strain (3/10⁹ cells) represent just a small subset of all the late-arising papillae in the galE⁺ strain (16/10⁹ cells). In accord with earlier suggestions (LENSKI et al. 1989 ; HENDRICKSON et al. 2002 ), it is possible that the larger subset of the late-arising galE⁺ papillae represented clones with two sequentially acquired mutations, the first of which was growth associated, leading to formation of microcolonies, and the second in lacI/lacO. As explained above, such microcolonies can occur only in the galE⁺ strain, since their growth will not be supported in the galE(Ts) background at the restrictive temperature. If the first "mutation" involved amplification of the lac locus [which may occur by either RecA-dependent (HENDRICKSON et al. 2002 ) or RecA-independent (BZYMEK and LOVETT 2001 ) mechanisms], then the lacI/lacO mutation in the second step would necessarily have to be dominant.

Mechanism of mutation at the lacO +5 hot spot in nondividing cells:
The mechanism by which lacO +5 may be rendered mutable only in nondividing cells is not known. A spontaneous GC-to-AT transition such as that at lacO +5 can occur following C oxidation, C deamination, G alkylation, or incorporation during replication of the incorrect nucleotide, but these lesions are ordinarily repaired in wild-type cells by, respectively, the Nth Nei glycosylases (BLAISDELL et al. 1999 ; NAJRANA et al. 2000 ), Ung glycosylase (COULONDRE et al. 1978 ; DUNCAN and MILLER 1980 ), Ada Ogt methyltransferases (MACKAY et al. 1994 ), and the methyl-mismatch repair system (MODRICH 1991 ). Miller and coworkers have shown that dcm sites in lacI are hot spots for GC-to-AT mutation because deamination of the methylated C residues at these sites results in lesions that alone are blind to repair by Ung (COULONDRE et al. 1978 ; DUNCAN and MILLER 1980 ). Arguing by analogy, and on the basis of the B/A ratio data given above for lacZ reversions in the ada ogt (and nth nei) mutants, one could suggest that endogenous alkylation (or oxidative) damage is more or less randomly distributed on all G (or C) residues in nondividing cells and that in the wild-type strain the Ada and Ogt (or Nth and Nei) proteins might be able to repair this damage efficiently at all sites except lacO +5. Additionally, since the lacO +5 mutation is dominant, the "instantaneous gratification" models may also be relevant in explaining its occurrence.

Concluding remarks:
In summary, we have found in this study that the products of the ada ogt, nth nei, and mutY genes are essential for avoidance of spontaneous mutations in nondividing cells, but that the genes involved in methyl-directed mismatch repair are dispensable under these conditions. In WT strains, the lacO +5 site is apparently mutable only in nondividing cells. Finally, we have shown for the first time that there is a substantial bias against recessive mutations in late-arising mutants, providing support to models that propose that nucleotide lesions conferring an instantaneous gratification are likely to be fixed as mutations in nondividing cells.

	FOOTNOTES

¹ These authors contributed equally to this work.

	ACKNOWLEDGMENTS

We acknowledge V. Vamsee Krishna and T. Giri Babu for technical assistance, Mehar Sultana for primer synthesis, and N. Nagesh for DNA sequencing. We also thank Ivan Matic, Miroslav Radman, Francois Taddei, and members of the J.G. laboratory for advice and discussions. S.M.B. was a University Grants Commission Research Fellow. J.G. is an honorary faculty member of the Jawaharlal Nehru Centre for Advanced Scientific Research. This work was supported under project no. 2703-1 of the Indo-French Centre for the Promotion of Advanced Research/Centre Franco-Indien Pour la Promotion de la Recherche Avancée.

Manuscript received July 24, 2003; Accepted for publication October 21, 2003.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ADAMS, W. T. and T. R. SKOPEK, 1987  Statistical test for the comparison of samples from mutational spectra. J. Mol. Biol. 194:391-396.[CrossRef][Medline]

AMBROSE, M. and D. G. MACPHEE, 1998  Catabolite repressors are potent antimutagens in Escherichia coli plate incorporation assays: experiments with glucose, glucose-6-phosphate and methyl-alpha-D-glucopyranoside. Mutat. Res. 398:175-182.[Medline]

ANDERSSON, D. I., E. S. SLECHTA, and J. R. ROTH, 1998  Evidence that gene amplification underlies adaptive mutability of the bacterial lac operon. Science 282:1133-1135.[Abstract/Free Full Text]

BENOV, L. and I. FRIDOVICH, 1996  The rate of adaptive mutagenesis in Escherichia coli is enhanced by oxygen (superoxide). Mutat. Res. 357:231-236.[Medline]

BERLYN, M. K. B., 1998  Linkage map of Escherichia coli K-12, edition 10: the traditional map. Microbiol. Mol. Biol. Rev. 62:814-984.[Abstract/Free Full Text]

BJEDOV, I., O. TENAILLON, B. GERARD, V. SOUZA, and E. DENAMUR et al., 2003  Stress-induced mutagenesis in bacteria. Science 300:1404-1409.[Abstract/Free Full Text]

BLAISDELL, J. O., Z. HATAHET, and S. S. WALLACE, 1999  A novel role for Escherichia coli endonuclease VIII in prevention of spontaneous GT transversions. J. Bacteriol. 181:6396-6402.[Abstract/Free Full Text]

BREGEON, D., I. MATIC, M. RADMAN, and F. TADDEI, 1999  Inefficient mismatch repair: genetic defects and down regulation. J. Genet. 78:21-28.

BRIDGES, B. A., 1998  The role of DNA damage in stationary phase (‘adaptive’) mutation. Mutat. Res. 408:1-9.[Medline]

BRIDGES, B. A., 2001  Hypermutation in bacteria and other cellular systems. Philos. Trans. R. Soc. Lond. Ser. B 356:29-39.[CrossRef][Medline]

BRIDGES, B. A. and A. R. TIMMS, 1997  Mutation in Escherichia coli under starvation conditions: a new pathway leading to small deletions in strains defective in mismatch correction. EMBO J. 16:3349-3356.[CrossRef][Medline]

BRIDGES, B. A. and A. R. TIMMS, 1998  Effect of endogenous carotenoids and defective RpoS sigma factor on spontaneous mutation under starvation conditions in Escherichia coli: evidence for the possible involvement of singlet oxygen. Mutat. Res. 403:21-28.[Medline]

BRIDGES, B. A., M. SEKIGUCHI, and T. TAJIRI, 1996  Effect of mutY and mutM/fpg-1 mutations on starvation-associated mutation in Escherichia coli: implications for the role of 7,8-dihydro-8-oxoguanine. Mol. Gen. Genet. 251:352-357.[Medline]

BZYMEK, M. and S. T. LOVETT, 2001  Instability of repetitive DNA sequences: the role of replication in multiple mechanisms. Proc. Natl. Acad. Sci. USA 98:8319-8325.[Abstract/Free Full Text]

CAIRNS, J., J. OVERBAUGH, and S. MILLER, 1988  The origin of mutants. Nature 335:142-145.[CrossRef][Medline]

CARIELLO, N. R., W. W. PIEGORSCH, W. T. ADAMS, and T. R. SKOPEK, 1994  Computer program for the analysis of mutational spectra: application to p53 mutations. Carcinogenesis 15:2281-2285.[Abstract/Free Full Text]

CLARK, D. P., and J. E. CRONAN, JR., 1996 Two-carbon compounds and fatty acids as carbon sources, pp. 343–357 in Escherichia coli and Salmonella: Cellular and Molecular Biology, edited by F. C. NEIDHARDT, R. CURTISS, III, J. L. INGRAHAM, E. C. C. LIN, K. B. LOW et al. American Society of Microbiology, Washington, DC.

COULONDRE, C., J. H. MILLER, P. J. FARABAUGH, and W. GILBERT, 1978  Molecular basis of base substitution hotspots in Escherichia coli.. Nature 274:775-780.[CrossRef][Medline]

DRAKE, J. W., 1991  Spontaneous mutation. Annu. Rev. Genet. 25:125-146.[CrossRef][Medline]

DUNCAN, B. K. and J. H. MILLER, 1980  Mutagenic deamination of cytosine residues in DNA. Nature 287:560-561.[CrossRef][Medline]

ECHOLS, H. and M. F. GOODMAN, 1991  Fidelity mechanisms in DNA replication. Annu. Rev. Biochem. 60:477-511.[CrossRef][Medline]

ERFLE, H. L., D. F. WALSH, J. HOLCRAFT, N. HAGUE, and J. G. DE BOER et al., 1996  An efficient laboratory protocol for the sequencing of large numbers of lacI mutants recovered from Big Blue transgenic animals. Environ. Mol. Mutagen. 28:393-396.[CrossRef][Medline]

FARABAUGH, P. J., 1978  Sequence of the lacI gene. Nature 274:765-769.[CrossRef][Medline]

FARABAUGH, P. J., U. SCHMEISSNER, M. HOFER, and J. H. MILLER, 1978  Genetic studies of the lac repressor. VII. On the molecular nature of spontaneous hotspots in the lacI gene of Escherichia coli. J. Mol. Biol. 126:847-863.[CrossRef][Medline]

FIX, D. F., P. A. BURNS, and B. W. GLICKMAN, 1987  DNA sequence analysis of spontaneous mutation in a polA1 strain of Escherichia coli indicates sequence-specific effects. Mol. Gen. Genet. 207:267-272.[CrossRef][Medline]

FOSTER, P. L., 1992  Directed mutation: between unicorns and goats. J. Bacteriol. 174:1711-1716.[Free Full Text]

FOSTER, P. L., 1999  Mechanisms of stationary phase mutation: a decade of adaptive mutation. Annu. Rev. Genet. 33:57-88.[CrossRef][Medline]

FOSTER, P. L. and J. CAIRNS, 1992  Mechanisms of directed mutation. Genetics 131:783-789.[Abstract]

FOWLER, R. G., R. M. SCHAAPER, and B. W. GLICKMAN, 1986  Characterization of mutational specificity within the lacI gene for a mutD5 mutator strain of Escherichia coli defective in 3' 5' exonuclease (proofreading) activity. J. Bacteriol. 167:130-137.[Abstract/Free Full Text]

FRIEDBERG, E. C., G. C. WALKER and W. SIEDE, 1995 DNA Repair and Mutagenesis. American Society of Microbiology, Washington, DC.

GILBERT, W., N. MAIZELS, and A. MAXAM, 1973  Sequences of controlling regions of the lactose operon. Cold Spring Harbor Symp. Quant. Biol. 38:845-855.

GODOY, V. G. and M. S. FOX</