Effects of Environment on Compensatory Mutations to Ameliorate Costs of Antibiotic Resistance

See allHide authors and affiliations

Science  25 Feb 2000:
Vol. 287, Issue 5457, pp. 1479-1482
DOI: 10.1126/science.287.5457.1479


Most types of antibiotic resistance impose a biological cost on bacterial fitness. These costs can be compensated, usually without loss of resistance, by second-site mutations during the evolution of the resistant bacteria in an experimental host or in a laboratory medium. Different fitness-compensating mutations were selected depending on whether the bacteria evolved through serial passage in mice or in a laboratory medium. This difference in mutation spectra was caused by either a growth condition–specific formation or selection of the compensated mutants. These results suggest that bacterial evolution to reduce the costs of antibiotic resistance can take different trajectories within and outside a host.

Among the major factors determining the frequency of resistance in a bacterial population are (i) the volume of antibiotic use, (ii) the costs of resistance to bacterial fitness, and (iii) the ability of bacteria to genetically compensate for such costs (1, 2). Generally, both plasmid- and chromosomally conferred resistances cause fitness losses, even though exceptions are known. When resistance has a cost, compensatory mutations can ameliorate these costs, commonly without loss of resistance (3, 4).

To determine whether the costs of resistance are compensated by different mutations under different growth conditions, we examined two types of antibiotic resistance in Salmonella typhimuriumstrain LT2: streptomycin resistance (SmR) caused by mutations in therpsL gene, which encodes ribosomal protein S12 (5), and fusidic acid resistance (FusR) caused by mutations in the fusA gene, which encodes elongation factor G (EF-G) (6). All resistant mutants studied grow slowly in laboratory media because of a decreased rate of protein synthesis (5, 6). As shown here for the FusR mutants and previously for the SmR mutants (4), they were also slow-growing in mice. Independent lines of the resistant bacteria were evolved by serial passages in a laboratory medium (LB, Luria Bertani broth) or in mice in the absence of antibiotic. By this procedure, spontaneous mutants were selected by virtue of their faster growth rates (7). The occurrence of compensated mutants was determined by plating bacteria from each cycle on agar plates; when fast-growing compensated mutants became the dominant population, the experiments were terminated. One random fast-growing clone from each lineage was examined. For the SmR mutants, the number of generations of growth and the population sizes were similar in mice and in LB. Thus, the opportunity for compensated mutants to arise and evolve was similar under the two growth conditions. For the FusR mutants, the number of generations of growth was fewer in mice than in LB (7). The compensatory mutations were located by sequencing of the gene with the resistance mutation and of extragenic targets that have previously been implicated in compensation (8).

For the SmR mutants, all LB-selected compensated mutants contained extragenic suppressor mutations in either the rpsDor rpsE gene, whereas all mouse-selected mutants had one specific intragenic compensatory mutation. For the FusR mutants, compensation was preferentially by intragenic suppression for bacteria grown in LB and mainly by true reversion for bacteria evolved in mice (Table 1). Compensated mutants commonly retained resistance. Thus, 17 of the 18 types of compensated FusR mutants and all of the compensated SmR mutants kept their resistance. Two tests were performed to ensure that the compensatory mutations were specific adaptations. When control lines of the sensitive parent were serially passaged under both growth conditions, no compensatory mutations were selected, which shows that these mutations are not generally beneficial. Furthermore, whenrpsD compensatory mutations were placed in the sensitive background, they did not increase fitness (4).

Table 1

Spectra of compensated mutants isolated after evolution in LB and in mice.

View this table:

We considered two models to explain the different spectra of compensatory mutations: The difference was either in the rate of formation of the compensatory mutations or in the selection of the compensated mutants. For the FusR mutants, selection accounted for the difference in compensatory mutations. Full compensation of the growth defect in LB can be conferred either by reversion (which is rare) or by intragenic compensatory mutations (which are common because they can occur by many different substitutions). Thus, intragenic mutants were preferentially found in LB because they were more common. In contrast, when the relative fitness of the compensated bacteria was determined by competition assays against the wild-type strain in mice, the revertants were fully compensated, whereas the intragenically compensated mutants showed only partial compensation. The 15 different compensated mutants evolved in LB showed relative fitness values from 0.33 to 0.94, whereas the true revertants had a relative fitness of 1.0 (Table 2). Thus, even though the revertants (because of their smaller genetic target size) were rarer than the second-site suppressors, they were predominantly selected in mice because they had fully restored fitness.

Table 2

Fitness in mice and in LB of FusR and compensated mutants.

View this table:

Why were certain compensatory mutations able to ameliorate fitness losses for bacteria grown in LB but not in mice? It has been shown that the FusR mutants have alterations in their levels of the nucleotides (p)ppGpp (9). Because these nucleotides are pleiotropic regulators of gene expression (10), alterations in their concentrations could affect the expression of virulence-related genes and thereby cause fitness differences in mice, without necessarily affecting growth in a laboratory medium.

For the SmR mutants, growth selection could not account for the different compensated mutants. From the competition experiments in mice, we did not detect any substantial difference in the competitive ability of the rpsD extragenic compensated mutants found after evolution in LB and the intragenic rpsL compensated mutants (AAC42 → AGA mutation) found after evolution in mice (Table 3). The rpsD-rpsL double mutants found showed relative fitness values from 0.91 to 1.0, and the intragenically compensated rpsL mutant showed a relative fitness of 1.0. Thus, the absence of rpsD compensatory mutations in the rpsL mutant evolved in mice could not be explained by poor fitness compensation. An alternative explanation was that the ability of the rpsD-rpsL double mutants to compete with the rpsL mutant in the evolution experiments was low when the mutant was present at a low frequency. Thus, in the competition experiments to examine the fitness of the evolved strains, the competing strains were mixed at a 1:1 ratio, whereas the compensated rpsD-rpsL mutants appearing in the evolution experiments initially represented a minor type among the majority noncompensated rpsL mutants. To test whether the competitive ability of the compensated mutants was decreased when they were present at a lower frequency than the competitor, one compensatedrpsD-rpsL mutant [Gln53 → Leu53 (Q53L), K42N (11) (Table 3)] was mixed with the rpsL mutant (K42N) at a low ratio (1:104) for competition in mice. Under these conditions also, the rpsD-rpsL mutant was able to compete out the rpsL mutant.

Table 3

Fitness in mice and in LB of SmR and compensated mutants.

View this table:

These results suggest that the different spectra of SmR compensatory mutations were caused by condition-dependent generation of the mutations. Considering the population size, the number of generations that the evolving bacteria spent in either mice or LB, and the fitness of the mutants, we calculated that the relative rate of the intragenic AAC → AGA mutation versus the extragenic rpsDmutations had to be several hundred times greater in mice than in LB in order to account for the difference in compensatory mutations (12). The specificity of the selected mutations in mice; i.e., only AAC42 → AGA mutations in preference to AAC42 → AAA reversion or extragenic rpsDcompensatory mutations, suggests the existence of a mutational mechanism that limits mutations to a particular nucleotide change within the rpsL gene. An increased repair of mismatched DNA heteroduplexes (template-induced mutations), as described in therIIB gene in bacteriophage T4, could explain the occurrence of the particular AAC42 → AGA mutation found in mice (13). Compatible with this idea was the finding that the mutation frequency in the codA, codB, orupp genes was higher for bacteria grown in mice than for those grown in laboratory medium. Thus, the median mutation frequency was about 13 to 28 times greater in bacteria recovered from mice than in bacteria grown in laboratory medium (14).

These findings about the specificity of the types of compensatory mutations found have several implications. First, they substantiate previous findings that reduced drug use might not result in a reduction of the frequency of resistant bacteria because of compensatory evolution and the maintenance of resistance in the compensated clones (2). Second, growth in mice and in laboratory medium imposes different constraints on the translation machinery (15). Third, and most important, the nature of the evolved compensatory changes is environment-dependent, which suggests that evolution to compensate for fitness losses caused by resistance mutations or other alterations, such as those associated with colonization and virulence, might occur by different mechanisms within and outside a host. In general terms, the rates and directions of molecular evolution may follow different trajectories because of the specific environment and its influence on mutation formation or selection. Finally, our data show that determinations of the amelioration of the costs of resistance should be performed in experimental hosts to allow a relevant assessment of the mechanism of compensation in resistant bacteria.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: Dan.Andersson{at}


View Abstract

Navigate This Article