Clonal Adaptive Radiation in a Constant Environment

See allHide authors and affiliations

Science  28 Jul 2006:
Vol. 313, Issue 5786, pp. 514-517
DOI: 10.1126/science.1129865


The evolution of new combinations of bacterial properties contributes to biodiversity and the emergence of new diseases. We investigated the capacity for bacterial divergence with a chemostat culture of Escherichia coli. A clonal population radiated into more than five phenotypic clusters within 26 days, with multiple variations in global regulation, metabolic strategies, surface properties, and nutrient permeability pathways. Most isolates belonged to a single ecotype, and neither periodic selection events nor ecological competition for a single niche prevented an adaptive radiation with a single resource. The multidirectional exploration of fitness space is an underestimated ingredient to bacterial success even in unstructured environments.

Abundant bacterial variety occurs in most environments (1) and is believed to have arisen from adaptive radiation to the myriad of structured environmental and biotic niches and specialization on alternative resources (2, 3). Thus, an environment with multiple niches results in more obvious diversification of experimental bacterial populations than a homogeneous environment (4, 5). Yet the other inference from this ecological view of adaptation, that a uniform environment with a single resource leads to low sympatric diversity, has not been tested exhaustively. Understanding the rapidity and limits of bacterial diversification in defined environments would be of benefit in modeling everything from infection progression to the stability of large-scale industrial fermentations.

Although many mutations occur in large bacterial populations, a low phenotypic diversity is expected because of purges involving fitter mutants, called periodic selection events (6, 7). Mutational periodic selection events correlate with purifying sweeps by strongly beneficial mutations (8). Still, it is questionable whether mutational selections maintain low diversity in clonal populations (9), and several long-term experimental evolution cultures resulted in stable coexistence of clones (10, 11). The generation of new ecotypes and niches is one possible source of radiation in a restricted environment (12), with the evolution of cross-feeding polymorphisms as an example (10). The temporal and nutrient-concentration fluctuations in other intensively analyzed experimental evolution experiments (11) also cannot eliminate the possibility of specialization for unidentified niches. We explored the metabolic, phenotypic, genotypic, and ecotypic divergence in an evolving chemostat population with a constant unstructured environment and a single resource.

A screen for diversification was developed by using an Escherichia coli strain with a reporter gene that can detect several types of regulatory change, resulting in improved transport properties and large fitness benefits in glucose-limited chemostats [strain BW2952 (1315)]. As described in (16), the uniformity of populations could be tested by assaying the malG-lacZ fusion activity, glycogen staining, and methyl α-glucoside (α–MG) sensitivity, which detects increased glucose uptake (17). Mutations affecting rpoS, mlc, malT, and ptsG influence one or several of the assayed traits to differing extents (13). Bacteria were grown at a dilution rate (D) of 0.1 hour–1 (a doubling time of 6.9 hours) with a population size of 1.6 × 1010 and sampled over 28 days. rpoS mutations, which confer a large fitness advantage (18), initially swept the population and led to a rapid elimination of parental rpoS+ bacteria (Fig. 1). Several identified (13) and unidentified sweeps followed the rpoS sweep in the dominant rpoS subpopulation, but the uniformity of the sampled isolates stayed high in the first 2 weeks of culture. Subsequently, and especially after 17 days, the population diversified to reveal multiple combinations of properties (Fig. 1). Simultaneously, rpoS+ bacteria again became a substantial proportion of the population. Similar diversity and recovery of rpoS+ bacteria was found in three other replicate populations (respectively, 41%, 38%, 70%, and 35% rpoS+) and prompted a more detailed analysis of late culture samples.

Fig. 1.

Time course of changes in an evolving bacterial population. E. coli strain BW2952 was grown at D = 0.1 hour–1 in a glucose-limited chemostat as described in (13). Daily samples were analyzed for changes in rpoS (circles) (16, 34). The diversity index had its basis in assays of α-MG sensitivity and rpoS status (both yes-no traits) and in the malG-lacZ activity shown in table S1 as the third trait. The fusion activity was divided into low (<200 units, wild type), intermediate (300–600 units), and high (700 units) ranges. The number of combinations of shared characters in 40 isolates tested is shown at each time point (squares).

The isolates from day 26 were screened for eight further phenotypic and genotypic characteristics (table S1). All of these characteristics changed under prolonged glucose limitation. Growth yields on glucose were tested because of changes in other studies (15) as were outer membrane profiles (Fig. 2A) associated with altered porins and antibiotic susceptibilities (19). Aminotriazole (AT) sensitivity probed altered concentrations of ppGpp, an important alarmone of E. coli (20), because of possible effects of this global regulator on rpoS-related expression (21) and because of ppGpp-related spoT changes in other evolving populations (22). Variation in the aggregation of cultures was also observed (Fig. 2B) and compared. The characterizations included sequencing of the rpoS and mgl mutations (13) to differentiate alleles in the one culture. Table S1 summarizes 11 separate properties of the parental strain and 41 isolates.

Fig. 2.

Phenotypic changes in evolved isolates after 26 days of selection. (A) The outer membrane proteins of 41 evolved clones of E. coli under glucose limitation were analyzed by using the SDS-urea electrophoresis method described in (35). The positions of porin protein bands OmpF and OmpC were identified by comparing outer membrane protein (OMP) profiles of mutants lacking either OmpF or OmpC. One example from each of the five groups of OMP patterns observed in the 41 isolates (table S1) is shown. The OmpF control was strain MH513 and the OmpC control was MH225. The ancestral strain produces pattern I, and chemostat-evolved isolates BW3767, pattern II; BW4001, pattern III; BW4011, pattern IV; BW4027, pattern V; and BW4003, pattern I. (B) The sedimentation of cultures was initially observed as shown in the cuvettes. (C) The rapidly sedimenting BW4001 culture was observed by phase-contrast microscopy (Olympus BH, Olympus Optical Company, Tokyo, Japan) directly from a chemostat culture (left), as well as after 6 hours of standing (center). (Right) The parental BW2952 strain after 6 hours of standing. (D) The different rates of sedimentation of representative isolates from table S1. Error bars indicate standard deviations based on three replicants.

The most prominent outcome of this analysis was the level of diversity among coevolved strains. Incorporating the data in Table S1 into a nearest neighbor–joining dendrogram revealed multiple branched clusters in the population (Fig. 3). One broad group included all rpoS mutants, with different rpoS alleles in the subclusters. The rpoS+ isolates differed in α–MG and AT sensitivity, so that population partitioned into three distinct approaches to global gene regulation associated with RpoS and ppGpp. The magnitude of the diversification in Fig. 3 suggests that a clonal bacterial population essentially became a collection of individual lineages in about 90 bulk generations under nutrient stress. The closest approximation to this level of diversification is the Lenski long-term E. coli lineages, where extensive insertion sequence rearrangements were observed over 10,000 generations (23) without recognition of the extensive phenotypic divergence described here.

Fig. 3.

Phylogenetic relationships among E. coli isolates evolved in a single population. Sympatric divergence of evolved clones was analyzed by the neighbor-joining method rooted to the ancestral strain, BW2952 (black box and white type), as described in (36). The dendrogram has its basis in the 11 characteristics of the 41 isolates shown in table S1. Isolates printed in bold or italic type within the large rpoS cluster indicate isolates with distinct rpoS alleles (see table S1 for details). Numbers on the branches indicate bootstrap values based on 1000 replications.

The growth yields of the 41 isolates on glucose varied markedly (table S1). The distinct yields suggested at least four parallel approaches to glucose conversion into biomass and associated metabolic adaptations. The different yield classes produced different amounts of acetate, CO2, and biomass from the equivalent amount of glucose; only one isolate produced acetate in a glucose-limited chemostat; and no other fermentation product was evident in pure cultures of other isolates (16). A cross-feeding polymorphism (10) was not common in this population within 90 generations.

Increased outer membrane permeability is an adaptive strategy for bacteria growing with low extracellular glucose concentrations (24). Analysis of outer membranes revealed five combinations of protein changes in individuals of the one population (table S1 and Fig. 2A). The changes were associated with increased membrane permeability, because there was an increased sensitivity to one or more antibiotics and detergent (table S1) (19). Individually inactivating the genes encoding outer membrane proteins (OmpF, OmpC, LamB, PhoE, and OmpG) in representative isolates indicated that the five groupings based on protein banding patterns underestimated the number of permeability-related changes in the population (table S2). Two class II isolates responded differently to ompF knockouts, indicating distinct mutational histories. Inactivation of each of the above proteins individually (table S2) did not reduce the elevated susceptibility to rifampicin. As with the regulatory alterations, there was a multiplicity of parallel permeability adaptations in a single population and even within a single cluster (compare BW3767 with BW4004).

The sedimentation of many isolates in a static culture (Fig. 2B) was also indicative of surface changes. Microscopic investigation indicated a tendency to aggregate in sedimenting bacteria (Fig. 2C). Nevertheless, in stirred chemostats, the isolates stayed in suspension as individual or dividing cells (Fig. 2C), so it is unlikely they formed a novel niche under the selection conditions. The clumping phenotype was not linked to any regulatory or growth-yield grouping (table S1) and so was not associated with a particular growth strategy. Surprisingly, 7 of 41 isolates remained in suspension better than the ancestor (Fig. 2D), indicative of at least four parallel surface or density changes. It remains to be established whether the sedimentation changes constitute a benefit under glucose limitation.

Did ecological specialization determine overall diversity? There were constant environmental conditions in the chemostat culture and no obvious fractionation of the population spatially or through adhesion. None of the 41 bacterial isolates showed increased adhesion to glass (16), and no obvious wall growth was observed in the selection chemostats. Temporal fluctuations were at best 20 to 25 s, between the drops added at the slow dilution rate used in the culture. There was no lysis (transient reductions in density) in the chemostats. We did not detect the production of secondary metabolites besides acetate in one isolate (16). In the absence of alternative niches, evidence for possible divergence through ecotypic changes was sought from competition experiments between members of the population (4, 7).

Isolates from various clusters in Fig. 3 were tested in direct competition experiments with ancestral bacteria. The marker used to resolve competing populations did not change the frequency of ancestor (Fig. 4A) (BW2952). Of 10 isolates tested, 8 displaced the parental strain from a glucose-limited chemostat, indicating fitness in the original selection niche. The displacement by BW3767 was not due to cross-inhibition of ancestor, because there was no difference in the growth of either strain in spent medium from a BW3767 culture. Elimination is compatible with the starvation of ancestor for glucose and washing out in the presence of fitter isolates.

Fig. 4.

Competitive interactions between different isolates under glucose limitation. (A) The population composition was followed when two strains, grown independently for 24 hours under glucose limitation (D = 0.1 hour–1), were mixed in equal proportions and the culture maintained under glucose limitation at the same dilution rate. The strains were differentiated by using T5 resistance as a neutral marker in the ancestral strain. T5-sensitive BW2952 and its T5-resistant derivative (BW3494) were equally fit during these experiments (solid squares). The composition of the cultures when BW3494 was mixed with strains from the various clusters in Fig. 3 is shown. Each experiment was repeated two to three times with similar patterns obtained. (B) Each of the isolates was competed pairwise against each of the other three strains. The inocula and mixing were performed as in (A). Duplicate competitions were initiated with chemostat-adapted bacteria mixed at 1% and 99% ratios of each strain. The arrows point away from the 1% or low-abundance start point for each isolate. The reported s values, or selection coefficients (14), are shown for each direction for each pair, with positive values showing increasing proportion or higher fitness in the competition experiments. Negative values indicate decreasing proportions in the competitions. The population estimates, as described in (16), have their basis in the means of the strain counts, including those obtained with reciprocal markers; the variation found was <7.5% standard deviation.

Interestingly, BW4005 was less fit than the ancestor and likely to belong to a new ecotype. Presumably BW4005 used resources in the long-term culture absent from the reconstituted chemostat. BW4005 and BW4003 were in the same phenotypic cluster in Fig. 3 but had different fitness properties, indicating yet unidentified divergence(s) between the strains. Because BW4003 did not fully eliminate the ancestor, it was possible that this strain also used an alternative resource. To further test whether BW4003 and isolates from other phenotypic groupings showed fitness properties consistent with ecotype divergence, we prepared competitions between evolved strains from the different clusters in Fig. 3 (Fig. 4B). The selection coefficients in Fig. 4B are shown for pairwise competitions between four isolates when the isolates start at 1% and 99% abundances, respectively. No two strains had identical fitness properties, consistent with ongoing population shifts evident from Fig. 1. BW4003 was the least fit against all strains, consistent with the weaker competition against ancestor (Fig. 4A). Only BW4001 exhibited frequency-dependent fitness properties in competition with each of the other strains. Applying the definition of an ecotype (7), that an adaptive mutant from within an ecotype outcompetes to extinction other strains of the same ecotype, we find that three of the four divergent strains in Fig. 4B are competing for a single niche. We have not identified alternative niches, but bacteria such as BW4005 and BW4001 may be evolving toward alternative ecotypes. Also, the relatively small sample in Fig. 4B may not have uncovered other variations within the large clusters in Fig. 3.

Overall, the 41 isolates contained at least three regulatory, four metabolic, four aggregation-related, and five different membrane permeability solutions. The observed combinations must reflect the multiplicity of physiological choices available to bacteria adapting to a glucose limitation. Evidently, an adaptive radiation can take place in a near-constant environment, and bacterial populations do not fully subscribe to the competitive exclusion principle (2). A rare, highly beneficial mutational advantage may still arise in long-term chemostat populations and purge diversity, but analysis of four parallel populations did not exhibit such sweeps beyond the first 2 weeks of culture.

How do we explain the number of mutations in the 26-day population? Judging from the sequence changes in table S1 and that the MG and AT sensitivities and different outer membrane changes are due to distinct mutations, most isolates contain at least four or five beneficial as well as an unknown number of neutral mutations. The multiplicity of mutations collected within 26 days does not agree with the Drake estimate of long-term bacterial mutation rates [0.0033 mutations per division (25)], which would yield only 0.3 mutations per genome in 90 generations. The mutations occurred in the absence of detected mutator cells in this culture (16), but two factors contributed to high mutation supply. First, the mutation frequency is about 30-fold elevated in a glucose-limited chemostat above that in nutrient-excess bacteria (13). Secondly, the successful bacteria will have undertaken more than 90 divisions, because the spread of each beneficial mutation must lead to transiently faster growth rates within the chemostat population (Fig. 4A).

There is extensive individuality of the isolates by day 26 (Fig. 3). Yet the initial sweeps did not introduce phenotypic diversity into the population (Fig. 1), although several alleles of mutated genes [as in rpoS, mgl: (table S1)] were present. Consistent with longer-term studies of bacterial adaptation, the major gains in fitness, such as the rpoS sweep, occurred early in the life of a population (26). Clonal interference, or competition between clones with different beneficial mutations in the population (27), may have slowed the emergence of multiple types in the first weeks. Beyond the early gains, incremental changes to properties such as membrane permeability improved fitness, but probably in small steps. The recovery of rpoS+ bacteria did not occur as a rapid sweep, and the mutations accumulating in the two lineages of rpoS+ bacteria were not beneficial enough to displace coevolving genotypes. Purifying periodic selections are inherently unlikely when weakly beneficial mutations arise in large populations subject to persistent selection (7, 9). To explain the increasing diversity, we propose that mutations of small fitness benefit swamp clonal interference or purifying periodic selection events. With isolates such as BW4001, negative frequency-dependent selection also operates to maintain diversity in the chemostat population, as discussed elsewhere (28).

The major radiation apparent in a single clone may seem inconsistent with the observation that bacteria in nature can be classified into groupings discernable as species. It needs to be remembered that our study excluded the acquisition of foreign genes. Lateral gene transfer (29) probably does superimpose a purging effect and maintenance of some order in bacterial relationships, because such transfers are rarer than the mutational sources of genetic change studied above. Periodic selection events in nature may indeed mostly depend on lateral gene transfer. It is also true that the adaptive radiation in the chemostat involved evolutionary overspecialization, with its strong tradeoff costs. Many of the adaptations involved antagonistic pleiotropy. For example, the increased detergent sensitivity of many isolates reflects increased membrane permeability beneficial specifically under chemostat conditions (18). These outer membrane changes would result in killing by bile salts in the normal habitat of E. coli. The RpoS mutations and reduced ppGpp would also cause problems in the transition to more stressful environments (17). Nevertheless, the large pool of genetic variation available through clonal divergence may be the source material for rarer lateral transfer in times of stress. Much of the microevolution seen within bacterial species (30) may be ultimately sourced to clonal diversification.

The speed and scale of the observed radiation has implications for evaluating bacterial success in all situations. Multidirectional divergence is relevant to bacteria in populations during persistent infections (31), facing a new environment, or crossing to a new host. In this context, it is relevant that mutation supply is limiting in pathogen populations, which are not as large as the population we investigated, and that mutations in mutator genes are often associated with pathogenesis (32). Our results suggest ecological specialization for multiple niches is not essential for bacterial diversity (3, 33) and that mutational periodic selections are unlikely to ensure the purity of bacterial species in the absence of lateral gene transfer. Lastly, our results suggest that sharing of a niche by a large number of diversifying members of the same species is a feasible evolutionary strategy. A single fitness solution, or survival of the fittest, is not the only answer in a competitive environment.

Supporting Online Material

Materials and Methods

SOM Text

Table S1 to S3


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article