Recent investigations of gene amplification in E. coli have focused on amplification of plasmid-borne genes. A phenotypically leaky F'-borne mutation, ϕ(lacIX13-lacZ), gives rise to Lac⁺ revertants bearing amplified arrays of 40–80 copies of the lac region [4]. Lac⁺ revertants of F'lac bearing the +1 frameshift allele ϕ(lacI33-lacZ), extensively employed in studies of adaptive mutation, consist mainly of one-base deletions in runs of iterated bases [5,6], but clones bearing amplified arrays appear at a lower rate as well [7,8]. Properties of lac amplification have generally supported the duplication-expansion model. (i) An engineered duplication of the frameshift mutant lac locus amplifies at a greatly elevated frequency [9], as predicted by the idea that duplication is the rate-limiting step (and as had been seen in the case of chromosomal ampC [2]). (ii) Amplification is dependent upon recBCD and ruvABC, as well as recA, indicating an important role for homologous recombination [10].

Expansion of a pre-existing repeat has also been studied primarily on plasmids. In one study, a pBR322 derivative was constructed with two directly repeated tetA genes, each bearing an inactivating mutation, but arranged in such a way that a single unequal crossover would generate an array of three copies, one of which was a functioning gene. In this system, expansion was reduced only five-fold in a recA mutant; expansion was elevated in strains bearing mutations in dnaQ, dnaE, dnaB, or dnaN [11].

Expansion of a pre-existing duplication was compared with amplification of a single copy of F'-borne ϕ(lacI33-lacZ) in another study [10]. Expansion was found to be increased in a polA mutant, and unaffected by overexpression of xonA, while amplification from a single copy was inhibited by both of these conditions. It was concluded that the amplification defects caused by the polA mutant and by xonA overproduction were in duplication, not expansion.

This study was undertaken to characterize expansion of a repeated sequence in the bacterial chromosome. A duplication of chromosomal ϕ(lacI33-lacZ) was constructed (Fig. ​(Fig.1).1). As expected, it expands at a high rate under selection for function. The effects of mutations in various recombination, replication, DNA repair, and stress response genes on expansion of the duplication were tested. The findings support the idea that expansion occurs via homologous recombination, and suggest as well that many of the recombination events leading to expansion take place at replication forks.

The Lac⁺ colonies appearing in these experiments can arise either by expansion or by mutation. A strain bearing even a single chromosomal copy of the un-frameshifted lacI-lacZ fusion grows well on lactose minimal medium. However, most or all of the excess Lac⁺ colonies produced by the duplication-bearing strain contain expanded arrays of the structure diagrammed in Fig. ​Fig.1.1. The amplified sequences in these clones are readily visualized as specific bands in restriction enzyme digests of total cellular DNA [4]. In tests of 28 Lac⁺ revertants from 22 cultures bearing the duplication, including four colonies which appeared on day two, all had expanded; none of 10 single-copy revertants contained amplified arrays (Table ​(Table1).1). Quantitation of total DNA, and DNA in the amplified bands, as described in the methods section, from 10 Lac⁺ clones, indicated a mean lac copy number of 72, with a standard deviation of 17. Expansion of this magnitude would be expected to result in β-galactosidase production comparable to a single-copy un-frameshifted gene, as the frameshift mutation reduces β-galactosidase production 100-fold [12]. The modest variability of lac copy number in the Lac⁺ clones presumably reflects a sort of optimum or equilibrium, in which the benefit of more β-galactosidase is balanced by the cost of extra DNA in the chromosome (70 copies of the repeat make the chromosome roughly 10% larger). Cultures of the expanded-array variants contain unknown numbers of point-revertant lac genes. Under continued selection for lac function, it is possible that the population would eventually be taken over by revertants which, bearing a single good copy of lac, have the benefit of sufficient β-galactosidase without the cost of more DNA. However, it is likely that such a changeover would take many generations because the Lac⁺ revertant has, if anything, only a small selective advantage – it does not have a noticeably faster growth rate on lactose minimal medium, for example. Hastings et al. [8] tested this idea, and found that amplified ϕ(lacI33-lacZ) clones kept under selection for lac function do not form revertants readily.

The clones appearing as Lac⁺ colonies acquire their ability to grow on lactose while under selection. Colonies re-streaked on lactose minimal plates form colonies visible to the unaided eye in 24 hours or less, regardless of whether they were picked on day 2 or day 7. Reversion and amplification of F'-borne ϕ(lacI33-lacZ) are adaptive, in that they occur only in the presence of lactose, not when the bacteria are simply starved [8,13]. The leakiness of the mutant allele is critical for adaptive mutation: residual lactose metabolism is enough to power the replication/recombination/repair processes involved, though not enough for cell division [12]. The experiment graphed in Fig. ​Fig.33 shows that expansion of a chromosomal ϕ(lacI33-lacZ) is similarly adaptive. Cultures were plated on minimal medium containing no available carbon source. Lactose was added after two or four days by injection under the agar slab. There was no sudden burst of colonies appearing two or three days later, as would be expected if expanded lac arrays had accumulated during starvation. Rather, the kinetics of appearance of Lac⁺ colonies resembled that seen in the cells initially plated on lactose, with a delay of either two or four days, and at declining rates, suggesting that the starving cells gradually lost their potential to expand.

Strains combining the duplication with mutations in DNA transaction genes were tested for Lac⁺ colony formation (Fig. ​(Fig.4).4). A null mutation in recA reduced the rate nearly 1000-fold, nearly down to that of a strain bearing a single copy of ϕ(lacI33-lacZ) (labeled "sc" near the bottom of Fig. ​Fig.4).4). The recA mutation had no significant effect in the single copy background. The strong recA dependence of expansion in this experiment contrasts with the weak recA dependence seen in a previous study of expansion of a plasmid-borne duplication [11], but is not surprising, for two reasons: (i) The duplicated segment in this study was much larger, and recA dependence tends to increase with increasing homology lengths [14,15]. (ii) The assay employed in the previous study required only a single recombination event, whereas becoming strongly Lac⁺ by expansion of a chromosomal ϕ(lacI33-lacZ) duplication probably involves more than one recombination event.

Null mutations in recBCD and ruvC reduced Lac⁺ colony formation 100-fold and 10-fold, respectively. Other mutations eliminating single recombination functions, recF, recG, recN, recQ, recR, and ruvAB, had little overall effect. A ruvC recG double mutant was also tested. Like the recA mutation, and as in other homologous recombination events [16], it generated Lac⁺ colonies approximately 1000-fold less efficiently than wild type.

A disruption of the E. coli yfgL gene was reported to confer a strong recombination/repair deficiency phenotype [17]. As shown in Fig. ​Fig.4,4, however, a yfgL null mutation constructed for this study has little or no effect on expansion. It also confers no UV-sensitivity or transductional recombination phenotype, in either an MG1655 or an AB1157 strain background (not shown); others have found no recombination/repair phenotype associated with a yfgL null as well [18].

To test the hypothesis that the deficiencies of recA, recBCD, ruvC, and ruvC recG mutants in Lac⁺ colony formation are due to their inability to expand the duplication, two alternative explanations were considered and ruled out. (i) Lac⁺ revertants of these mutants could grow much more slowly than Lac⁺ revertants of wild type. Lac⁺ colonies were restreaked on minimal lactose plates on the days they arose. In the cases of wild type, recA, recBCD, and ruvC, all of the Lac⁺ clones formed colonies visible to the unaided eye by 24 hours after restreaking, independent of the day on which they arose. In the case of the ruvC recG double mutant, none of the 8 tested Lac⁺ revertants formed visible colonies by 24 hours, but all did so by 48 hours. However, counting the ruvC recG colonies after 8 days instead of 7 only increased the median from 0.029 to 0.056 colonies per million viable cells plated (data not shown). Thus, slow growth of revertants can account for only a small part of the deficiencies of the recombination mutants in forming Lac⁺ colonies. (ii) The mutants could be proficient at expansion, but deficient at survival on the selection plates. Survival of the mutants on lactose minimal medium was tested as described in the methods section. The results (Table ​(Table1)1) indicate that none of the mutants has a substantial survival defect relative to wild type.

The Lac⁺ revertants of the deficient mutants consist of varying populations of expanded and mutated clones (Table ​(Table1).1). In the case of recA, none of 10 tested clones had an expanded lac array. The frequency of expanded arrays among recBCD revertants was 9 out of 10; ruvC was 10 out of 10, recG was 18 out of 18, and ruvC recG was 6 out of 14.

The Lac⁺ reversion phenotype of a recG null mutant is more complex than the data in Fig. ​Fig.44 suggest. The mutant strain's rate of colony formation tends to increase sharply late in the experiment, with the new colonies tending to appear as satellites of older colonies (not shown). Factors influencing the timing and extent of this satellite-based population explosion include plating density, but are otherwise unknown. The recG mutant data shown in Fig. ​Fig.44 are from selected experiments, in which the plating density was low, and satellitism was not as strongly evident as in other experiments. Satellitism of this sort suggests that something produced by the older colonies on the plate stimulates recombination, perhaps via a genotoxic effect. It is consistent with the finding that overexpression of recG protects E. coli against weak organic acids [19].

The effects on expansion of varying RecA activity were tested by plating mutants affecting recA expression. The uninducible lexA3 mutation has been reported to reduce the frequency of a number of different homologous recombination events [20]; it also reduced Lac⁺ colony formation approximately five-fold. The SOS-constitutive lexA71::Tn5 mutation (in a sulA null background, to suppress its lethality) had no significant effect; neither did the recAo281 operator constitutive allele.

A number of replication genes were tested for roles in expansion. A null mutation in polA increased the rate of expansion 5-fold; null mutations in the other non-essential DNA polymerase-encoding genes polB, dinB, and umuDC, had little or no effect. Loss of the replication restart function priA caused a small decrease in expansion efficiency, while loss of rnhA caused a nearly 7-fold increase.

A strain lacking dam function exhibited a 35-fold elevated rate of Lac⁺ colony formation. Apparently, part of this elevated rate is due to the double-strand breaks which occur as the result of mis-directed mismatch repair in dam mutants [21,22]. As shown in Fig. ​Fig.4,4, a dam mutH double mutant exhibited an intermediate rate between those of wild type and the dam single mutant. The mutH null allele by itself had little or no effect.

Other DNA repair functions were tested for effects on expansion as well. A uvrD null mutant formed Lac⁺ colonies at a 10-fold elevated rate, while mutM, mutY, a mutM mutY double, and an ada null mutation had no significant effects.

The question of whether expansion of a chromosomal repeat occurs as part of a stress response, like amplification starting from a single episomal copy [23], was explored by testing null mutations in rpoS and relA. As shown in Fig. ​Fig.4,4, these mutations had little or no effect on Lac⁺ colony formation.

The Red recombination system of phage λ promotes RecA-independent recombination between chromosomes if at least one of the chromosomes is replicating, and RecA-dependent recombination between non-replicating chromosomes [27-30]. The RecA-independent expansion seen in red-substituted bacteria suggests that at least some of the Red-mediated recombination events involved in expansion take place at replication forks [30,31].