INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

We and others have been studying the dynamics of phenotypic and genomic changes in 12 populations of Escherichia coli B while they have evolved in, and adapted to, a minimal glucose medium for more than 20,000 generations (8, 23, 24, 30, 33, 35, 38). As part of this research, the performances of the derived lines and the common ancestor were quantified on a large set of diverse substrates to determine if losses of catabolic function accumulated during evolution in an environment in which glucose was the sole carbon source (8). We observed a general decline in catabolic niche breadth, which was largely attributable to subtle reductions in function on several carbon sources. However, absolute losses of function were rare, with one conspicuous exception: all 12 populations lost the ancestral ability to catabolize D-ribose. In this study, we examined the mechanisms responsible for these parallel losses, including both population genetic processes and the underlying molecular genetic events.

Parallel evolution of a trait across multiple lineages is often used as an indicator that the change is adaptive and has been shaped by natural selection (9, 18, 19, 29, 34). Therefore, we hypothesized that the loss of the ability to use ribose may have improved fitness in the glucose-limited environment. The alternative explanationthat mutations accumulated in genes affecting ribose function, but without those mutations enhancing fitness on glucoseseemed less likely on two grounds. First, other catabolic functions also experienced relaxed selection but did not exhibit such losses. Second, given typical mutation rates and the duration of the evolution experiment (12, 38), we would expect the accumulation of neutral mutations to cause sporadic but not parallel losses of function in the replicate populations.

In this paper, we report experiments that examined (i) the phenotypic dynamics of loss of ribose function in each population, (ii) the underlying rate of mutation from Rbs⁺ to Rbs, (iii) the affected loci and mutational mechanism responsible for the losses of ribose function, and (iv) the effects of the losses on competitive fitness in glucose medium. We used these data to examine quantitatively the contributions of mutation and selection to the evolutionary dynamics that drive this systematic loss of function.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Long-term evolution experiment and culture media. The design of the long-term evolution experiment is described in detail elsewhere (24). Briefly, 12 populations evolved from a single clone of E. coli B (strain Bc251 T6^r Str^r rm₁₁₁ ara (see [22]) also http://myxo.css.msu.edu/ecoli/strainsource.html). This ancestor is unable to grow on arabinose (Ara) and was used to found six lines; the other six lines were founded with a spontaneous Ara⁺ mutant of the ancestor. The Ara marker is neutral in the experimental environment (24), which consists of daily 1:100 dilution into fresh Davis minimal broth supplemented with limiting glucose at 25 µg/ml. The 100-fold dilution and regrowth allows ~6.6 (=log₂ 100) generations per day. The 12 populations have been evolving under the same conditions for more than 20,000 generations (3,000 days); the experiments described here focus on the first 2,000 generations, during which time all 12 populations lost catabolic function on D-ribose. Every 500 generations, samples from the populations were stored in glycerol at 80°C, where they are available for study at any time.

Phenotypic screening. The loss of D-ribose function was initially discovered in experiments with Biolog microtiter plates used to detect changes in catabolic functions during the evolution experiment (8). These plates contain different carbon sources in each of 95 wells plus a tetrazolium indicator dye. Three clones from each of the 12 population samples stored at generation 2000 were evaluated using these plates; all but 1 of the 36 clones (the exception being a clone from population Ara+2) showed almost no growth after 48 h in the well containing D-ribose (8).

Direct measurement of the mutation rate from Rbs⁺ to Rbs. In preliminary experiments, we observed that the ancestral Rbs⁺ strain generated many Rbs mutants. We then performed a Luria-Delbrück fluctuation test (21, 26, 27, 36) to estimate the mutation rate among 56 replicate cultures. Independent cultures were each founded from a small number (~50) of cells of the Rbs⁺ ancestor in flasks containing 10 ml of the same glucose-supplemented Davis minimal medium. After 24 h, each culture was diluted, several hundred cells were spread on TR agar, and the Rbs⁺ and Rbs colonies were counted. These data yielded the total number of cells and the frequency of Rbs mutants in each culture, from which the rate of mutation from Rbs⁺ to Rbs was estimated (21, 27, 36) (further details are provided in Results).

Fitness assays. The protocol used for estimating the relative fitness of two strains during direct competition is described in detail elsewhere (15, 24) and summarized here. We estimated the fitness of seven independently isolated Rbs mutants, each relative to that of the Rbs⁺ ancestor (their progenitor) and with fivefold replication. Equal culture volumes of the mutant and the ancestor were mixed and diluted in flasks containing fresh medium. These cultures were propagated by daily serial transfer over the course of 6 days (as opposed to a 1-day competition) to better detect small fitness differences. Initial and final (day 6) densities of both types were enumerated on indicator agar. Fitness is expressed as the ratio of the realized population growth rates for the mutant and the ancestor while they competed for the common pool of nutrients. The relative fitness of a mutant will be >1.0 if it grows more quickly than does the ancestor and <1.0 if the mutant grows more slowly.

DNA handling. Genomic DNA was prepared from 3-ml cultures using standard methods (31). DNA fragments used as probes were cold labeled, and hybridizations were performed with the digoxigenin-labeling and detection kit sold by Roche. All hybridizations and washings were done at 68°C under high-stringency conditions.

PCR experiments. The sequences adjacent to IS150 were cloned from HincII-digested genomic DNA of the ancestor by inverse PCR. Genomic DNA of the ancestor was digested with HincII, and the resulting fragments were separated on a 0.8% agarose gel with PstI- and HindIII-digested lambda DNA used as size markers. The gel fraction containing the expected insertion element (IS) was cut, and DNA was purified using a modification of the Geneclean kit of Bio-101 (6). The fragments were self-ligated with T4 DNA ligase (Roche) at 5 to 10 µg/ml, and the ligated mixtures were used as templates in PCR experiments. Primers used for inverse PCR to amplify sequences adjacent to IS150 were G5, 5' GAT CCT GTA ACC ATC ATC AG 3', and G6, 5' CTG AAG GAT GCT GTT ACG G 3'. Both sequences lie near the IS150 extremities and are directed outward. All PCRs were performed with Expand Taq DNA polymerase (Roche) according to the manufacturer's recommendations. The PCR product containing the IS150-adjacent sequences was cloned into the pCRII-Topo vector (Invitrogen), which contains no HincII and two EcoRI restriction sites located on either side of the inserted DNA in the multiple cloning sites. After transformation of E. coli TOP10 competent cells (Invitrogen), the plasmid content of white colonies was digested with EcoRI and HincII, giving rise to three fragments. One fragment contained only the vector sequence, while each of the others contained one of the two IS150-adjacent sequences. The inverse PCR was performed with HincII-digested DNA, such that a single HincII restriction site was present in the PCR product; in this particular case, there were no EcoRI restriction sites in the product. The two IS150-adjacent sequences were used as probes in hybridization experiments and they were also sequenced.

Restoration of Rbs⁺ phenotype by transduction. P1 transduction (28) was used to restore the functional rbs operon to two of the seven Rbs mutants of the ancestor. Each Rbs⁺ transductant was then competed against its corresponding Rbs mutant, with sixfold replication.

Construction of a non-IS150-mediated deletion in the rbs operon. A deletion of 5,563 bp of the rbs operon was constructed in the ancestor strain without involving the upstream IS150 element. This deletion includes about 90% of rbsA, all of rbsC, rbsB, rbsK, and rbsR, and about 20% of yieO (which lies downstream of the rbs operon).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Differences in ribose utilization in E. coli K-12 and B strains. E. coli K-12 is able to catabolize D-ribose, and it has a single functional rbs operon located at 83 min on the chromosome (25). The particular strain of E. coli B with which we began the long-term evolution experiment is also able to grow on ribose. However, other widely used strains of E. coli B are unable to catabolize ribose and have two defective rbs operons located at 2 and 83 min (1, 25). Possible explanations for the differences between E. coli B strains in ribose function are considered in Discussion.

Ribose catabolism is rapidly lost from evolving populations. Figure 1 shows the dynamics of loss of the D-ribose catabolic phenotype in 12 experimental populations. After only 500 generations, Rbs mutants had taken over 7 of the 12 populations, and by generation 2000 the Rbs phenotype had become fixed, or nearly so, in all 12 of the populations. Previous studies indicated that several beneficial mutations swept through each of these populations in this interval, and these mutations conferred fitness gains of about 10% each (23, 24). The rapidity and reproducibility with which the losses of ribose catabolic function arose suggested to us that the Rbs mutations might therefore represent one of these strongly beneficial mutations.

View larger version (23K): [in this window] [in a new window]   FIG. 1.   Frequency of Rbs cells over time in the 12 evolving populations. At generation 2000 all the populations contained between 97 and 100% Rbs cells.

Rate of mutation from Rbs⁺ to Rbs is very high. An alternative explanation for the rapid parallel losses of ribose function is that this character is genetically unstable. This alternative was suggested by the unexpected finding that occasional spontaneous Rbs mutants were observed when ancestral cells were plated on TR indicator agar, which led us to perform a fluctuation test to measure the ancestor's rate of mutation from Rbs⁺ to Rbs. Among 56 independent cultures (each inoculated from ~50 cells), the average total population size was 4.2 × 10⁸ cells and the average frequency of Rbs mutants was 0.000512. We employed a local computer program (P. J. Gerrish, Los Alamos National Laboratory) (see also reference 35) to find the mutation rate consistent with these data. This program generates expected Luria-Delbrück distributions according to the method of Ma et al. (27) and then utilizes maximum likelihood to refine the mutation rate estimate (36) and calculates the 95% confidence interval for the mutation rate based upon the theoretical variance of the maximum likelihood estimate. This procedure generated a mutation rate of 8.56 × 10⁵ per cell per generation, with an upper confidence limit of 1.22 × 10⁴ and a lower bound of 5.99 × 10⁵. However, this program does not take into account the fitness benefit that we measured for the Rbs mutants over their Rbs⁺ progenitor (see below), which would lead to a slight overestimation of mutation rate. Using numerical simulations consistent with the median method of Lea and Coulson (21), and allowing the mutants to replicate with the measured selective advantage, we calculated a rate of mutation from Rbs⁺ to Rbs of 5.4 × 10⁵ per cell per generation, which is our best estimate of the true rate.

10

6

Rbs mutants are slightly more fit than their progenitor. Each of the evolved populations carries many mutations, including several that are beneficial (23, 24, 30, 33). It would therefore be inappropriate to use genotypes from these populations to measure the effect of the Rbs mutation on fitness, because they contain confounding mutations. Hence, we performed competition experiments between seven spontaneous Rbs mutants of the ancestor and their Rbs⁺ progenitor to quantify the effects of the Rbs mutations on fitness in the glucose minimal medium. All seven Rbs mutants were fitter than the Rbs⁺ ancestor (Fig. 2), with an average advantage of 1.4% ± 0.4% (95% confidence interval). Loss of D-ribose catabolic function therefore confers a small, but consistent, selective advantage in the environment that prevailed during the evolution experiment. Evidently, both selection for loss of function and a high underlying mutation rate contributed to the rapid and parallel evolution of the Rbs phenotype in the 12 experimental populations.

View larger version (15K): [in this window] [in a new window]   FIG. 2.   Fitnesses of seven spontaneous Rbs mutants relative to their progenitor, as measured by competition experiments. Error bars show 95% confidence intervals based on fivefold replication for each mutant.

An IS150 element adjacent to the rbs operon mediated deletions during the evolution experiment. Given the unusually high rate of mutation from Rbs⁺ to Rbs, we sought to identify the loci affected and the molecular basis of the mutations that were fixed during the experimental evolution. Restriction fragment length polymorphism analysis (30, 33) of HincII-digested genomic DNA from the ancestor and from clones sampled at generation 10000 using IS150 as a probe revealed parallel losses in all 12 populations of a 2.7-kb fragment that contained an IS150 element. The regions adjacent to this element were cloned from HincII-digested genomic DNA of the ancestor and then sequenced. The IS150 element was found to be located immediately upstream of the rbs operon (20) that is involved in ribose utilization (Fig. 3). The IS150 insertion also led to a 3-bp duplication of the target site.

View larger version (28K): [in this window] [in a new window]   FIG. 3.   Losses of D-ribose catabolism in evolving populations caused by deletions in the rbs operon. (A) Map of the rbs operon in E. coli B ancestor, based on the genome sequence for E. coli K-12 (4). Arrows show the directions of gene transcription. An IS150 element is located upstream of the rbs operon. Thick lines indicate the left and right adjacent sequences used as probes. The primers used in PCR experiments (G5, G6, G76, and G77) and for constructing strain GBE127 (G266, G267, G268, and G269) are also shown, as are relevant HincII restriction sites. (B) The deletions in clones isolated from 11 of the 12 populations are shown as horizontal lines. Population Ara+2 is not depicted because an Rbs+ minority clone was sampled (see text). The right adjacent sequence of IS150 did not hybridize with any of the Rbs clones, whereas the left adjacent sequence hybridized in every case with a HincII fragment, whose size is shown in the column labeled Hyb. The sizes of the PCR products obtained from these clones using primers G76 and G77 are given in the column labeled PCR. The sizes, in the ancestral strain, of the hybridizing HincII fragment and PCR product are indicated at the bottom.

Similar deletions of the rbs operon occur in spontaneous Rbs mutants. We then performed the PCR assays used to identify deletion endpoints in evolved clones to characterize the seven spontaneous Rbs mutants whose fitnesses we had measured. In all seven mutants, the loss of D-ribose catabolism was caused by deletions of various sizes within and near the rbs operon. All had the same left endpoint of their deletion adjacent to the IS150 element; three had the right endpoint of the deletion in the rbsK gene, two in yieO, and one each in rbsA and rbsC. We reported in a previous section that these seven mutations all had beneficial effects on competitive performance, in each case increasing fitness by about 1 to 2% in glucose minimal medium. Thus, different underlying mutations conferred similar selective benefits, with the commonality being that the rbs operon was rendered nonfunctional by a deletion beginning adjacent to the IS150 element located just upstream of this operon. We also performed an experiment to measure the rate of reversion of the Rbs mutants back to the Rbs⁺ state. However, no Rbs⁺ revertants were observed, consistent with the rbs operon having been partially or completely deleted.

Restoration of the functional rbs operon reduces fitness. To confirm that the deletions in the rbs operon (and not some hypothetical mutation elsewhere) caused the observed fitness advantage of the Rbs mutants, we used P1 transduction to restore the functional rbs operon to two of the Rbs mutants. These two mutants differed in the extent of the deletion of the rbs operon: one lacked only part of the rbs operon, whereas the other lacked the entire operon plus part of yieO. In both cases, the restored functional operon significantly reduced fitness relative to the corresponding Rbs mutant (P < 0.01 in each case, based on six replicate competitions). Thus, restoration of the functional rbs operon counteracts the beneficial effect caused by its earlier deletion.

Non-IS150-mediated deletion of the rbs operon also increases fitness. As a final test of the relationship between ribose catabolic function and competitive fitness in glucose-limited medium, we constructed strain GBE127. In this strain, unlike others we examined, the promoter region and first gene of the operon, rbsD, are intact, but the rest of the genes in the rbs operon have been deleted. Like all seven spontaneous deletions (Fig. 2), this constructed deletion also enhanced competitive fitness, in this case by 2.1% (P < 0.0001 based on 10 replicate competitions). Therefore, neither the upstream IS150 element nor the promoter region needs to be involved in the deletion mutation in order to confer the beneficial fitness effect.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Twelve populations of E. coli B experienced rapid and parallel losses of D-ribose catabolic function during evolution in glucose minimal medium (Fig. 1). We showed that these losses were caused by similar deletions of the rbs operon; in all cases, one end of the deletion was immediately adjacent to an IS150 element located just upstream of the operon, whereas the other endpoint varied (Fig. 3). We further showed that mutations from Rbs⁺ to Rbs occurred at an unusually high rate, that these mutations were caused by deletions of the rbs operon similar to those during the evolution experiment, and that Rbs mutations consistently conferred a small but significant competitive advantage in glucose minimal medium (Fig. 2).

It may be instructive to examine the logic that led us to identify the evolutionary forces responsible for the loss of ribose function. The rapid and parallel evolution of the Rbs phenotype initially suggested to us that this change was adaptive. Repeatable change across lineages is widely taken as evidence of adaptation by natural selection (9, 18, 19, 29, 34). Furthermore, the fact that this loss of function occurred when adaptation to the new environment was most rapid (23, 24) reinforced our view that the loss was beneficial for the bacteria in glucose minimal medium. However, our observation that Rbs mutants were readily isolated from the ancestor led us to question our preconceptions and focus on the alternative possibility that the losses of ribose catabolic function might be caused by a hypermutable locus of some sort. Indeed, a fluctuation test confirmed this possibility. We then performed competition experiments between the spontaneous Rbs mutants and their Rbs⁺ progenitor, which showed that the mutants were also fitter than their progenitor in the glucose minimal medium. Evidently, both selection for the loss of ribose catabolic function and its hypermutable genetic basis contributed to the rapid and parallel phenotypic evolution that we observed.

The finding that all seven independent spontaneous Rbs mutants had a competitive advantage strongly suggested that the beneficial effect was caused by the rbs deletion, as opposed to some hypothetical secondary mutations that arose during the fluctuation test in which the mutants were generated. It is generally accepted that only a very small fraction of all mutations are beneficial in any particular environment, with the vast majority being either deleterious or neutral (13, 16). Thus, it is highly unlikely that secondary mutations would produce any improvement in fitness.

Nevertheless, to confirm absolutely that the rbs deletions per se were responsible for the fitness benefits observed in glucose minimal medium, we performed two additional types of strain constructions and corresponding competition experiments. First, we used P1 transduction to restore a functional rbs operon to two of the Rbs mutants. This restoration reduced their competitive fitness, which indicates that the deletion of the rbs operon itself was beneficial in the glucose environment. Second, we constructed de novo a deletion of most of the rbs operon that did not involve the upstream IS150 element or the rbs promoter region, and we used homologous gene replacement to introduce this into our progenitor. This deletion also produced a small but significant competitive advantage. Thus, three different genetic approachesspontaneous mutation, transduction, and gene replacementall demonstrate that the deletion of the rbs operon causes a beneficial effect in minimal glucose medium.

Molecular basis of the high rate of mutation from Rbs⁺ to Rbs. The same class of molecular events led to the loss of ribose function in the evolved lines and spontaneous Rbs mutants. A total of 18 Rbs genotypes (11 evolved lines and 7 spontaneous mutants) all showed deletions in which one endpoint was located precisely at the end of an IS150 element that was inserted upstream of the rbs operon. The extent of the deletion varied among the genotypes, but it always encompassed the promoter region and first gene (rbsD) and in some cases included all six genes in the rbs operon plus part of an adjacent gene of unknown function (yieO) (17). The fact that one endpoint of the deletion was always precisely located at the end of this IS150 element suggests that the mechanism of deletion involved first the transposition of an IS150 element (either the one upstream of rbs or any other) into the site corresponding to the other endpoint and in the same orientation as the one upstream of rbs. This transposition was presumably then followed by a recombination event between the new IS150 and the one upstream of rbs, thereby causing deletion of the intervening region. Whether the transposition and recombination events occurred simultaneously or successively is unknown; the fact that none of the spontaneous Rbs mutants showed a simple transposition that inactivated the rbs operon (without the associated deletion) suggests that the two events occurred in the same cell generation. Analysis of the nucleotide sequences at the right endpoints of the different deletions indicated no homology with the end of IS150 corresponding to the left deletion endpoint, which further suggests rearrangements associated with an initial transposition event. Examination of the nucleotide sequence of the presumptive target sites of the IS150 transpositions in the different genotypes failed to reveal any obvious preference for its insertion.

5

Quantitative analysis of the contributions of mutation and selection to the evolution of the Rbs phenotype. Both positive selection for loss of the rbs operon and its underlying mutability contributed to the evolutionary losses of ribose-catabolic function in the 12 experimental populations. Here, we examined mathematically their contributions as well as their interplay with one another and with selection at other loci.

5

5