Genomewide Patterns of Substitution in Adaptively Evolving Populations of the RNA Bacteriophage MS 2 Andrea

Experimental evolution of bacteriophage provides a powerful means of studying the genetics of adaptation, as every substitution contributing to adaptation can be identified and characterized. Here, I use experimental evolution of MS2, an RNA bacteriophage, to study its adaptive response to a novel environment. To this end, three lines of MS2 were adapted to rapid growth and lysis at cold temperature for a minimum of 50 phage generations and subjected to whole-genome sequencing. Using this system, I identified adaptive substitutions, monitored changes in frequency of adaptive mutations through the course of the experiment, and measured the effect on phage growth rate of each substitution. All three lines showed a substantial increase in fitness (a twoto threefold increase in growth rate) due to a modest number of substitutions (three to four). The data show some evidence that the substitutions occurring early in the experiment have larger beneficial effects than later ones, in accordance with the expected diminishing returns relationship between the fitness effects of a mutation and its order of substitution. Patterns of molecular evolution seen here—primarily a paucity of hitchhiking mutations—suggest an abundant supply of beneficial mutations in this system. Nevertheless, some beneficial mutations appear to have been lost, possibly due to accumulation of beneficial mutations on other genetic backgrounds, clonal interference, and negatively epistatic interactions with other beneficial mutations. EXPERIMENTAL evolution, or the study of adaptation in laboratory populations, provides a means of following adaptation in real time and in minute detail. Microbial systems, in particular, offer an opportunity to rigorously test theoretical models of adaptive evolution, as in these systems beneficial mutations can be readily observed and their effects measured in a controlled environment. Recent work in this area has addressed such questions as whether theory can accurately predict the distribution of fitness effects among beneficial alleles (Sanjuan et al. 2004; Rokyta et al. 2005; Barrett et al. 2006; Kassen and Bataillon 2006) and how interference alters this distribution among fixed beneficial alleles (Hegreness et al. 2006). Lenski and Travisano (1994) pioneered another kind of experimental evolution approach, which focuses on describing patterns of evolution in evolving lines. One generality that has emerged from these studies is that evolving populations tend to increase in fitness rapidly upon introduction to a new environment, but more slowly later (Elena and Lenski 2003). This slowdown in the rate of increase in mean fitness may be due to one of two causes or some mixture of the two. First, as a population approaches an optimal phenotype, the supply of beneficial mutations may become exhausted, and adaptation may be limited by an increasingly smaller mutation supply (Silander et al. 2007). Second, the rate of adaptation may slow if, as expected, mutations with large benefits tend to be fixed earlier than those with small benefits. Several population genetic and physiological factors may act together to ensure that large-effect mutations are fixed before mutations with small effects, particularly in large asexual populations. First, because these mutations with big benefits have shorter sweep times than other mutations, they will tend to be among the first mutations fixed (Kimura and Ohta 1969; Gerrish and Lenski 1998; Kim and Orr 2005). This is especially true in large populations, where an abundant mutation supply offers opportunity for competition between beneficial mutations (Gerrish and Lenski 1998; Kim and Orr 2005). Second, large-effect mutations have lower probabilities of stochastic loss (Haldane 1927; Gillespie 1991; Orr 2002), and correspondingly shorter waiting times until a successful mutation occurs, than smalleffect mutations. This may be particularly true in asexual populations, where successful mutations may need to have benefits large enough to overcome the effects of linked deleterious mutations ( Johnson and Barton Sequence data from this article have been deposited with the EMBL/ GenBank Data Libraries under accession nos. FJ799467–FJ99712. Address for correspondence: Institute of Evolutionary Biology, University of Edinburgh, King’s Bldgs., W. Mains Rd., Ashworth Labs, Room 123, Edinburgh, EH9 3JT United Kingdom. E-mail: [email protected] Genetics 181: 1535–1544 (April 2009) 2002). Third, diminishing returns epistasis may be common: the same beneficial mutation may have a larger effect if it occurs early in the course of adaptation rather than later (Hartl et al. 1985; Bull et al. 2000). Finally, later substitutions may be largely compensating for deleterious pleiotropic effects of earlier substitutions, and compensatory mutations may have smaller benefits on average than the mutations for which they compensate (Otto 2004). To date, only one experiment has directly examined whether or not earlier substitutions have larger benefits than later ones, independent of any reduction in the supply of beneficial mutations, with somewhat mixed results (Holder and Bull 2001). Here, I use experimental evolution to study adaptation in large populations of an RNA phage with a high mutation rate, MS2. MS2 is a single-stranded RNA bacteriophage of the family Leviviridae (reviewed in van Duin and Tsareva 2004). Like other phage of this family, MS2 has a small genome, consisting of 3.6 kb and four genes (Fiers et al. 1976). As a result, repeated whole-genome sequencing can be used to determine both the identity and the order of each adaptive substitution. In this experiment, I expose MS2 to a new environment—one that selects for rapid growth and lysis in Escherichia coli growing at a colder than usual temperature—and examine its response by determining the identity and fitness effects of each substitution. Results may differ from similar studies in DNA phage (Holder and Bull 2001) due to the more abundant supply of mutations in an RNA phage. Further, the data collected here allow the investigation of patterns of evolution associated with a combination of a large population size and high mutation rates. MATERIALS AND METHODS Phage and host strains: Phage in the genus Levivirus (family Leviviridae) consist of a single protein-coding RNA strand encapsulated by a virion coat. The coliphage of this group, including MS2, infect F1 E. coli cells through the pilus encoded by the F-plasmid. MS2 reproduces in host cells, using both its own gene products and host factors, and then lyses the host cell to release progeny phage. The small (3569 nucleotides) genome of MS2 includes four protein-coding genes: the maturase, coat, lysis, and replicase genes. The maturase gene is involved in infection and lysis; the coat protein dimerizes to form the repeating subunit that constitutes the virion’s protein coat; the lysis protein aids in lysis of the host cell; and the replicase combines with host factors to form the RNA-dependent RNA polymerase that copies the viral genome (van Duin and Tsareva 2004). In addition, the MS2 genome contains regulatory regions controlling the timing of translation of the coding regions; this regulation is mediated by the secondary structure of the singlestranded RNA chromosome (Klovins et al. 1997). For thisexperiment, a stock designated ‘‘MS2 ancestor’’ (kindly provided by J. Bollback) was obtained from a laboratory strain of MS2 (kindly provided by J. van Duin), by propagating the strain for 20 phage generations at 37 to preadapt it to laboratory conditions. A single plaque from this population was isolated and grown for 3 hr at 37 . The MS2 ancestor is thus clonally derived from a single phage and is expected to be largely genetically homogenous (among three sequenced genomes, 1 silent and 0 replacement differences were found; data not shown). Phage were grown on TOP 10 F9 E. coli cells (Invitrogen, Carlsbad, CA) in standard Luria–Bertania medium supplemented with 14 mg/ml tetracycline (Sigma, St. Louis) to maintain the presence of the F-plasmid. Serial passages: Three MS2 lines were independently propagated from the MS2 ancestor stock through a minimum of 50 serial passages at 30 (cold temperature). Serial passages consisted of infection of a host culture, followed by 130 min of phage growth, and extraction of the phage from the culture. The scheme used here was designed to allow tight control of population size (phage were bottlenecked every generation), to limit host–phage coevolution (naive hosts were provided for each passage), and to limit phage–phage coevolution [interaction between phage was limited by a low multiplicity of infection (MOI) of 1 phage/100 host cells]. Each serial passage was performed as follows: uninfected E. coli cultures were grown at 37 to an optical density of 0.25– 0.5 (OD measured at 600 nm, corresponding to a density of 2.5 3 10–5.0 3 10 cells/ml). Two 5-ml subsamples of this cell culture were physiologically adapted to growth in a 30 water bath for 20 min. One of these was infected with 1 3 10 phage (2.0 3 10 /ml) from the previous passage, and the other was grown alongside the infected culture and checked to confirm the absence of contaminate phage. The cultures were grown for 130 min at 30 , then E. coli cells were pelleted and removed by centrifuging for 20 min, and phage were extracted by removing the liquid lysate. Portions of this lysate were used (i) to measure the concentration of phage to determine the amount of lysate necessary for infecting the next passage, (ii) to infect the next serial passage, and (iii) to maintain a frozen ‘‘fossil record’’ of the evolving lines by archiving lysate in 10% DMSO at 80 . Phage concentration was estimated from the number of plaque-forming units (PFUs) in an appropriate dilution plated on host cells using a soft agar overlay method and grown overnight at 37 . Growth assays: I measured the growth rate of clonal phage populations containing each putatively adaptive mutation on the genetic background on which it occurred (i.e., the substitution of interest plus previous substitutions). To confirm that assayed phage had the genotype of interest with no additional mutations, I isolated and sequenced the genomes of phage from single plaques. One mutation, C3056U, could not be obtained in isolation and so was measured along with the next likely substitution, U1756A. To measure the fitness effect of each substitution, I performed 10 replicate growth assays under conditions that mimicked passaging conditions as closely as possible. To control for environmental variation, I used a paired design for the assays: each replicate includes a measurement of the growth rate of both MS2 ancestor and the genotype of interest taken at the same time and using the same host culture. Briefly, a single 60-ml E. coli culture was grown at 37 in a 250-ml flask to an OD of 0.25–0.5 and divided into two 25-ml cultures in 150-ml flasks, with an additional 5-ml culture used as a negative control. One of the 25-ml cultures was infected with MS2 ancestor, and the other was infected with phage of the genotype of interest. Cultures were then grown and phage extracted as in serial passages, with two exceptions. First, initial phage densities were lower ( 100 phage/ml), although in both cases phage were grown in a large excess of host cells. Second, phage were extracted using a 2-mm cellulose acetate spin filter (Costar, Cambridge, MA); in both cases, only phage outside host cells were extracted. Both changes to the protocol allowed for more accurate determination of phage numbers. After each assay, initial (N0) and final phage numbers (Nt) were determined from the initial and final phage concentration as detailed above, except that PFU counts were replicated (one to three independent dilutions plated four to six times 1536 A. J. Betancourt