Ongoing Phenotypic and Genomic Changes in Experimental Coevolution of RNA Bacteriophage Qβ and Escherichia coli

According to the Red Queen hypothesis or arms race dynamics, coevolution drives continuous adaptation and counter-adaptation. Experimental models under simplified environments consisting of bacteria and bacteriophages have been used to analyze the ongoing process of coevolution, but the analysis of both parasites and their hosts in ongoing adaptation and counter-adaptation remained to be performed at the levels of population dynamics and molecular evolution to understand how the phenotypes and genotypes of coevolving parasite–host pairs change through the arms race. Copropagation experiments with Escherichia coli and the lytic RNA bacteriophage Qβ in a spatially unstructured environment revealed coexistence for 54 days (equivalent to 163–165 replication generations of Qβ) and fitness analysis indicated that they were in an arms race. E. coli first adapted by developing partial resistance to infection and later increasing specific growth rate. The phage counter-adapted by improving release efficiency with a change in host specificity and decrease in virulence. Whole-genome analysis indicated that the phage accumulated 7.5 mutations, mainly in the A2 gene, 3.4-fold faster than in Qβ propagated alone. E. coli showed fixation of two mutations (in traQ and csdA) faster than in sole E. coli experimental evolution. These observations suggest that the virus and its host can coexist in an evolutionary arms race, despite a difference in genome mutability (i.e., mutations per genome per replication) of approximately one to three orders of magnitude.


Introduction
Host-parasite coevolution has been a topic of intense research interest in various fields from basic science of molecular evolution to agricultural and medical applications [1][2][3][4][5].According to the Red Queen hypothesis or arms race dynamics, coevolution leads to complex but continuous change, adaptation, and counteradaptation of the phenotypes of interacting organisms [2,6,7].Futuyma and Slatkin suggested that investigation of coevolution could raise and help provide answers to many questions regarding the history of evolution, e.g., whether parasites tend toward specialization or toward benign or even mutualistic relationships with their hosts [8].
There have been many previous observational and theoretical studies on natural host-parasite dynamics.With regard to the relationships between bacteria and phages, Rodrı ´guez-Valera et al. proposed the constant-diversity dynamics model in which the diversity of prokaryotic populations is maintained by phage predation [9].Moreover, an observational study supported the model by analyzing the dynamics of bacteria and phages in four aquatic environments using a metagenomics method and showed that microbial strains and viral genotypes changed rapidly [10].In addition, experimental models in simplified environments have been employed to analyze the ongoing process of coevolution.Various pairwise combinations of bacteria and phages and one with Caenorhabditis elegans and bacteria have been subjected to longterm laboratory cultivation [11][12][13][14][15].These studies indicated that coevolution proceeded on a laboratory time scale [11][12][13][14], accelerated molecular evolution of parasites [16,17], and broadened the host range of parasites [14].However, the changes in genetic information and phenotype of parasites and their hosts through coevolution remain to be elucidated, and the changes in host specificity and virulence of the parasites through the arms race have not been determined in sufficient detail because ongoing adaptation and counter-adaptation in simplified experimental model systems have not been analyzed at the levels of population dynamics and molecular evolution.
To examine the ongoing changes driven by host-parasite interactions, we have constructed a coevolution model consisting of Escherichia coli and the lytic RNA bacteriophage Qb (Qb) in a spatially unstructured environment.Qb is a simple RNA bacteriophage that infects and lyses E. coli cells, taking about 1 h for its burst without escaping into a lysogenic state.It has a singlestranded RNA genome of 4,217 bases encoding four genes for A2, A1 (read-through), coat protein, and RNA replicase b subunit [18].Due to a high misinsertion rate and lack of a proofreading mechanism, ribovirus RNA replicase (including that of Qb) has a high mutation rate [18][19][20][21][22], which allows us to monitor the evolutionary changes on a laboratory time scale.
Here, we report that in coevolution through 54 daily copropagations of the parasite and its host, E. coli first evolved partial resistance to infection and later showed acceleration of its specific growth rate, while the phage counter-adapted by improving release efficiency with a change in host specificity and a decrease in virulence.Fitness analysis indicated that these phenotypic changes occurred within an arms race, i.e., accompanied with a monotonic fitness increase of either the parasite or its host.Whole-genome analysis indicated that the phage accumulated 7.5 mutations mainly in the A2 gene 3.4-fold faster than in Qb propagation evolution where the phage was transferred daily to freshly prepared E. coli cultures, while E. coli showed fixation of two mutations (in traQ and csdA) faster than in sole E. coli experimental evolution.The results indicated ongoing adaptation and counter-adaptation through a host-parasite arms race.

Experimental evolution system
Evolution experiments were carried out with copropagation of E. coli and Qb and with propagation of Qb only (Figure 1A).In the copropagation experiment, the ancestral E. coli strain HL2 (Anc(C)) and Qb derived from cloned Qb cDNA [23] (Anc(P)) were mixed, cultivated, and diluted so that the next daily culture was initiated at approximately 1610 7 E. coli cells/ml.We calculated the replication generations of Qb genome as the cumulative generations of each passage, (N final /N initial ) = 2 g , where N final and N initial represent final and initial free phage density of each passage in plaque forming units (PFU/ml), respectively, and g represents replication generation.We also calculated E. coli cell generations as the cumulative generations of each passage, (N f / N i ) = 2 n , where N f and N i represent the final and initial colony forming units (CFU/ml) of each passage, respectively, and n represents cell generation.In the very early phase of the copropagation experiment, the cell generation was underestimated due to cell lysis by infection.The copropagation experimental population was divided into two on day 18, equivalent to 59 replication generations and 62 cell generations.Culture was continued to a total of 54 days (lines 1 and 2), equivalent to 163 replication generations and 163 cell generations for line 1, and 165 replication generations and 164 cell generations for line 2 (Figure 1B).Two Qb propagation experiments, lines 3 and 4, were conducted in parallel for 18 days, equivalent to 169 and 168 replication generations where the phage population was separated daily by centrifugation from the host and transferred into fresh logarithmic cultures of the host Anc(C) (Figure 1B).
The population dynamics of the copropagation experiment demonstrated the coexistence of E. coli and Qb (Figure 2), although Qb is lytic and has no lysogenic state.The daily Qb population density fluctuated over the course of the copropagation experiment, while the E. coli population density was stable probably due to the constant initial density of the host at each daily coculture.The degree of phage amplification in the copropagation experiment (2-20-fold per single coculture) was substantially lower than that in the Qb propagation experiment (approximately 1,000-fold), even though the initial multiplicity of infection (MOI) in each passage was approximately 0.5 (approximately 10 7 phages/ml over 2610 7 E. coli cells/ml) for the Qb propagation experiment and was not higher than that for the copropagation experiment (Figure 2B and 2C).These observations suggested that the biotic environment for phage amplification, i.e., the cellular state of the host E. coli, changed during the copropagation experiment.

Fitness analysis of the parasites and hosts evolved in the copropagation experiment
Cross-cocultures were conducted to determine the changes in fitness of E. coli and Qb in the copropagation experiment.Four hosts (Anc(C), M54(C), M163(C), and M165_2(C)), and the four corresponding phages (Anc(P), M54(P), M163(P), and M165_2(P)) at the 1 st , 54 th , 163 rd , and 165 th replication generations in the copropagation experiments of lines 1 and 2 were cocultured to measure fitness in each pairwise combination.Here, the fitness of E. coli is defined as the ratio of the initial to the stationary optical density at 600 nm (OD 600 ), while the fitness of the phage is the ratio of the initial to the stationary free phage density (PFU/ml) (Table 1, Table 2 and Figure S1).

Author Summary
To examine the ongoing changes driven by host-parasite interactions, we have constructed a coevolution model consisting of Escherichia coli and the lytic RNA bacteriophage Qb (Qb) in a spatially unstructured environment.In coevolution through 54 daily copropagations of the parasite and its host, E. coli first evolved partial resistance to infection and later accelerated its specific growth rate, while the phage counter-adapted by improving release efficiency with a change in host specificity and a decrease in virulence.Whole-genome analysis of E. coli and Qb revealed accelerated molecular evolution in comparison with Qb propagation in this study and E. coli sole passage reported previously.The results of the present study indicated that, despite the large difference in mutability of their genomes (approximately one to three orders of magnitude difference), a host with larger genome size (4.6 Mbp) and a lower spontaneous mutation rate (5.4610 210 per bp per replication) and a parasite with a smaller genome size (4,217 bases) and a higher mutation rate (1.5610 23 to 1.5610 25 per base per replication) were capable of changing their phenotypes to coexist in an arms race.that the host population evolved, increasing its fitness, to become almost oblivious to the phages.
Despite the development of partial resistance by the host, the phage also increased its fitness through changes in host specificity (Table 2).The evolved phage M54(P) and M163(P) showed higher fitness on the evolved host M54(C) with partial resistance than    The fitness of Qb was calculated as (Fitness) = Log 10 (PFU/ml _max /PFU/ml _0h ), where PFU/ml _max is PFU/ml of free phage in which we adopted the higher value between the values of 7 h or 22-29.5 h after infection, and PFU/ml _0h is PFU/ml of free phage just after infection.The growth curves are shown in Figure S1.NA, not available.doi:10.1371/journal.pgen.1002188.t002 Anc(P) on the same host (one-way ANOVA F 2,3 = 19.7,P,0.05; post hoc Tukey test, P,0.05; Table 2).In line 1 and line 2, the most evolved Qb, M163(P) or M165_2(P) showed the highest fitness on corresponding E. coli, M163(C) or M165_2(C), respectively.Briefly, there was a significant difference in fitness among the host-parasite combinations (host: M163(C) or M165_2(C), parasite: Anc(P), M54(P), M163(P), or M165_2(P), one-way ANOVA F 2,3 = 60.8,P,0.01; post hoc Tukey test, P,0.05 for M163(C); F 2,3 = 37, P,0.01; post hoc Tukey test, P,0.05 for M165_2(C); Table 2).The phage evolved through natural selection to show greater amplification on the corresponding host, although the amplification ratio itself decreased from approximately 10 4 to 10 1 .The phage, while responding adaptively to the evolutionary changes of its host, showed a decrease in amplification ratio on the ancestral host strain, leading to a decrease in virulence.The amplification ratios of the phage on the host Anc(C) gradually decreased over the course of the copropagation experiment (host: Anc(C), parasite: Anc(P), M54(P), M163(P), and M165_2(P), oneway ANOVA F 3,4 = 17.4,P,0.01; post hoc Tukey test, P,0.05; Table 2).A decline in phage amplification was also observed as a reduction in plaque size (Figure 3).Consequently, the evolved phage showed less cell killing effect against the ancestral strain Anc(C), resulting in better growth of the ancestral bacterial strain (host: Anc(C), parasite: Anc(P), M54(P), M163(P), and M165_2(P), oneway ANOVA, F 3,4 = 340, P,0.01; post hoc Tukey test, P,0.01; Table 1 and Figure S1A, left), i.e., the phage showed a decrease in virulence.In addition, the phage evolved in the Qb propagation experiment (S94_3(P)) showed the greatest amplification ratio (host: Anc(C), parasite: Anc(P), M54(P), M163(P), M165_2(P), and S94_3(P), one-way ANOVA F 4,5 = 37.8, P,0.01; post hoc Tukey test, P,0.05; Table 2 and Figure S1B, left) and similar virulence against Anc(C) with Anc(P) (host: Anc(C), parasite: Anc(P) and S94_3(P), Welch's t test, t = 12.7, P = 0.45; Table 1 and Figure S1A, left).These results suggest that the decrease in virulence was not due to simple degeneration through the long-term passage experiment, but was probably at the expense of increasing the fitness of the phage in the arms race with its host.

Mechanism of fitness improvement
To examine how E. coli and Qb improved their fitness during coevolution, free phages, infected E. coli cells, and total E. coli cells from the copropagation of line 1 were monitored hourly by determining the numbers of PFUs in the supernatant and pellet after centrifugation and CFU, respectively (see Materials and Methods).
E. coli was found to first evolve partial resistance to Qb, which was followed by a later increase in the specific growth rate.After 3 hours of incubation with Anc(P), almost all of the ancestral host Anc(C) cells were infected, while infection ratios of the evolved hosts M54(C) and M163(C) were only 0.03% and 0.08%, respectively (Figure 4A, 4B, and 4D, left).The observed partial resistance was likely due to a very low adsorption rate of E. coli cells to the phage (Figure 5).As most of the evolved host cells remained uninfected, they were able to proliferate, while the ancestral cells could not.In addition, the uninfected cells of the most evolved hosts, M163(C) and M165_2(C) for lines 1 and 2, respectively, showed higher specific growth rates than those of M54(C) (see legend of Figure 6 for specific growth rates, ANCOVA, F 2,24 = 18.0,P,0.001; post hoc Tukey test, P,0.001).Although the OD 600 /CFU seemed to have changed over the copropagation experiments (Figure 6), the same conclusion was obtained using specific growth rates based on CFU values (data not shown).Briefly, M54(C) eliminated Anc(C) from the population by developing partial resistance to phage infection, and M163(C) and M165_2(C) finally took over the population due to acceleration of specific growth rate.
The phages evolved to show increased release efficiency, i.e., the number of phages released from a single infected cell per unit time.

Molecular evolution of host and parasite
We performed whole genome sequence analyzes of all of the Qb populations indicated in Figure 1B to determine how molecular evolution of the phage proceeded in response to the adaptation of hosts.First, mutations were shown to be accumulated in a biased manner in the A2 gene, which encodes a multifunctional protein related to infection and cell lysis (Figure 1B and Table 3).The A2 gene, accounting for 30% of the whole genome, accumulated 65.5% of all mutations, and this bias was shown to be statistically significant (P,0.05,two-tailed binomial test).A similar substantial accumulation of mutations in genes related to host infection was demonstrated previously in an evolution experiment using the DNA bacteriophage W2 [16].The mutation fixation rate in phage was higher in the copropagation experiment (1.0610 25 66.0610 27 per base per generation) than that in the Qb propagation experiment (3.2610 26 65.1610 27 per base per generation) (two-tailed Welch's t test t = 4.3, P,0.01), suggesting that the phage showed accelerated molecular evolution through coevolution with its host (Figure 7).
Whole-genome analysis of E. coli revealed the process of molecular evolution in the host cells.We analyzed the whole genome sequence of M163(C) using an Illumina Genome Analyzer IIx (GAIIx; Illumina, San Diego, CA) and confirmed the mutations in M163(C) together with Anc(C) and M54(C) by the dideoxynucleotide chain termination sequencing method [25].A single mutation in traQ (S21P) encoded on the F plasmid was detected in M54(C) and an additional mutation was detected in csdA (D340N) in M163(C) (Table S1, Table S2).As discussed below, the protein products from these genes may contribute to resistance to phage infection and the increase in fitness of E. coli.

Phenotypic and genetic changes of Qb and E. coli
In the copropagation experiment, E. coli adopted a simple strategy with only two mutations, while the phage accumulated more mutations within its small genome as counter-adaptation against the evolutionary changes in the host.The host first developed resistance to phage infection via a non-synonymous mutation in traQ.This gene encodes TraQ, the conjugal transfer pilin, which is a component of the F pilus, and is a chaperone for inserting propilin into the inner membrane.Propilin was reported to be unstable in traQ 2 cells [26], and amino acid 21 of TraQ where the mutation was detected in this study interacts with propilin [27].F pilus assembly from membrane F-pilin requires many Tra proteins [28].As no mutations were detected in other Tra protein genes in the copropagation experiment, the mutation on TraQ may result in a decrease in the amount of inserted propilin, leading to the partial resistance observed in this study.E. coli then showed further mutation in csdA, which encodes CsdA, an enzyme related to Fe/S biogenesis and a new sulfur transfer pathway that is related to the fitness of these cells, especially in stationary phase [29].Therefore, this mutation could be beneficial as the host was passaged daily at the stationary phase in the evolutionary experiment.
On the other hand, the phage evolved to increase release efficiency by accumulating mutations mostly in the gene encoding the A2 protein.A2 is a multifunctional protein with roles in host cell lysis, adsorption to the F pilus of E. coli, RNA binding during capsid assembly, protection of the 39 terminus, penetration into the cytoplasm of the host, and blockage of cell wall biosynthesis by inhibiting the catalytic step from UDP-GlcNAc to UDP-GlcNAc-EP catalyzed by MurA [18,[30][31][32].Due to the cell lytic activity of A2, it is unsurprising that these mutations might have resulted in an increased burst frequency and release efficiency.In fact, the experimental lag period between infection and detection of the increase in free phage became shorter by approximately 1 hour in the cross-culture experiment (e.g., 3 h for Anc(P) and 2 h for M163(P) on M163(C), Figure 4D and 4F, right, respectively).It should be noted that the uninfected M163(C) reached the stationary phase, which was not susceptible to phage infection, approximately 1 hour earlier than the other hosts (see legend of Figure 6 for the time to reach the stationary phase, one-way ANOVA, F 2,3 = 693.8,P,0.001; post hoc Tukey test, P,0.001).Thus, it is possible that the increased burst frequency of M163(P) for the earlier phage release evolved as a counter-adaptation on the evolved host M163(C) due to the shorter period available for infection.Previous experiments using DNA bacteriophages indicated that shorter latent periods were favored in the presence of a high density of highly susceptible host cells [33,34].
It is of interest that Qb evolved to show reduced virulence toward the ancestral host.Many studies have indicated that phages with low or moderate virulence were favored in vertical transmission or in structured environments [35][36][37], while Qb has no lysogenic state and evolved reduced virulence in this experiment.The decrease in virulence observed in this study may have been a side effect of the increase in burst frequency.If fact, the evolved phage M163(P) with increased burst frequency on the evolved host M163(C) showed lower virulence and lower fitness on Anc(C) than Anc(P) on Anc(C) (Figure 4 and Figure S1 left), suggesting that Qb may have co-evolved to increase the burst frequency in reducing some benefits that can be gained if the host reverts to the Anc(C)-like phenotype.
The single non-synonymous mutation at position 221 of the A2 gene found in M54(P) seems to have resulted in reduced virulence Table 3.Nucleotide sequences in the ancestral and evolved phage genomes.*This mutation was AUG to GUG in the start codon.If this protein was functional, the GUG start codon would be used to encode Met.
To determine mutation fixation and cumulative mutation numbers in Figure 7, polymorphic sites (shown as 6) were counted as 0.5, and monomorphic sites (shown as +) were counted as 1.For position 221 in line 1, a count of 1 was given from Anc(P) to M54(P), 0.5 from M54(P) to M109(P), and 0.5 from M109(P) to M163(P).For position 221 in line 2, a count of 1 was also given from Anc(P) to M54(P) and 0.5 from M54(P) to M165_2(P).doi:10.1371/journal.pgen.1002188.t003 and a change in host specificity.As the non-synonymous mutation was only observed in the copropagation experiment and two others were also observed in Qb propagation and the deposited sequence (NCBI accession no.AY099114), the mutation at 221 and/or the combinations with the mutation and two other mutations may have resulted in the decrease in virulence and the change in host specificity observed in M54(P).

Accelerated molecular evolution rates of Qb and E. coli
In coevolution between Qb and its host, E. coli, the phage showed accelerated molecular evolution (Figure 7).In the Qb propagation experiment, the molecular evolution of the phage proceeded but seemed to slow down after the 94 th generation.On the other hand, the phage coevolving with E. coli retained a 3.4-fold faster molecular evolution rate throughout the copropagation experiment.The higher evolution rate may be attributable to the changes occurring in the host E. coli.If the host had stopped evolving, e.g., at the 54 th generation, the M163(P) or M165_2(P) phage would not have been fixed into the population as it had fitness similar to or less than that of M54(P), leading to deceleration of evolutionary rate.It should be noted that neutral mutations cannot be fixed in the copropagation experiment because 163 replication generations is too short for them to become fixed.The fixation of neutral mutations is known to require generations approximately as long as the effective population size (Ne) [38].The effective population size in the copropagation experiment was roughly estimated as the bottleneck size of the population (approximately 10 3 phages) assuming that 1% of the minimum initial 10 6 phages infect and burst to release approximately 10 7 phages.Thus, even synonymous mutations observed here were positively selected [38], consistent with the influence of the RNA secondary structure on Qb genome replication reported previously [39][40][41][42].Some synonymous mutations may have physiological impacts on phage growth because of genomic secondary structure; it has been reported that some synonymous mutations or mutations in intergenic regions show lethal effects in Qb [42].
It is noteworthy that the fixation rate of the E. coli genome in the copropagation regime (2.6610 29 per bp per generation) calculat-ed as 2 mutations in 4.73 Mbp per 163 generations was one order of magnitude higher than that under conditions of E. coli sole passage, maintaining log phase at 37uC (1.7610 210 per bp per generation) [43] or 20,000 generations (1.6610 210 per bp per generation) (Poisson distribution, P,0.01) [44].In summary, these observations indicated that molecular evolution rates of both the parasite and its host were accelerated through adaptation and counter-adaptation.

The arms race in the evolution experiment
Based on the observed fitness changes in the host E. coli and in the Qb phage, we propose a plausible coevolution path to depict the arms race between Qb and the host E. coli.As the order of phage fitness on the ancestral E. coli Anc(C) was Anc(P).M54(P).M163(P), the population in the coculture seemed to first take a route not in the direction of phage evolution (upward) but in the direction of host evolution (right), increasing host fitness by increasing its resistance to Qb (Figure 8).The arrows in Figure 8 reflect the experimentally determined finesses changes (Table 1 and Table 2).Arriving around the pair position of M54(C) vs. Anc(P), the population could take either the upward or rightward direction, but happened to take the direction of phage evolution due to the occasional appearance of a single nonsynonymous mutation at position 221 in the phage genome that was detected only under copropagation conditions (Table 3).The population of M54(P) and M54(C) could not fix a phage mutant like M163(P) with the same fitness as M54(P), but fixed the E. coli mutant M163(C) with fitness higher than that of M54(C).Due to the host change from M54(C) to M163(C) accompanied with an additional single non-synonymous mutation in csdA, the phage mutant M163(P) was fitter than M54(P) and was therefore fixed in the final population.Taken together, these findings indicated that the evolutionary path seemed to be an arms race involving adaptation of E. coli and counter-adaptation of the phage.
We showed that parasites, such as RNA viruses, and hosts, such as E. coli, have the potential to coexist even in an arms race.When a parasite encounters its host, the host may become extinct through the evolution of high parasite virulence, or the parasite may become extinct through the evolution of host resistance.However, both may also change their phenotypes by genomic mutation in a synchronized manner and thus coexist.The results of the present study indicated that a host with a larger genome size (4.6 Mbp) with a low spontaneous mutation rate (5.4610 210 per bp per replication) [45] and a parasite with a smaller genome size (4,217 bases) and a higher spontaneous mutation rate (1.5610 23 -10 25 per base per replication) [18][19][20][21][22], despite the large difference in mutability of their genomes (approximately one to three orders of magnitude difference), were capable of changing their phenotypes to coexist in an arms race.Further studies linking the phenotype mutability and genome complexity will help to elucidate the dynamic host-parasite relationship.

Strains and media
The E. coli HL2 strain was used as the coculture host strain and A/l [46] was used as an indicator strain for the titer assay.The E. coli HL2 strain was constructed by conjugation with DH1DleuB::(gfpuv5-Km r ) [43] and HB2151 [47].We mixed logphase DH1DleuB::(gfpuv5-Km r ) and HB2151 for 2.5 hours and screened for kanamycin-resistant clones on LB agar medium supplemented with 25 mg/ml kanamycin.F9 retention of HL2 was checked by PCR with the primers TraU_f (59-ATGAAGCGA AGGCTGTGGCT-39) and TraU_r (59-GCAGCTTGAACGC  3. doi:10.1371/journal.pgen.1002188.g007CATGCGT-39) and the ability of HL2 to amplify Qb was confirmed.Before the evolution experiments, HL2 was grown in mM63gl (62 mM K 2 HPO 4 , 39 mM KH 2 PO 4 , 15 mM ammonium sulfate, 1.8 mM FeSO 4 ?7H 2 O, 15 mM thiamine hydrochloride, 2.5 mM MgSO 4 ?7H 2 O, 0.04% glucose, and 1 mM L-Leu) for several passages until the specific growth rate had become stable, and the strain with stable growth rate was used as the ancestor strain (Anc(C)).The OD 600 of stationary-phase Anc(C) cultured in mM63gl medium was approximately 0.4 (approximately 3610 8 CFU/ml) because of glucose limitation.Qb was kindly provided by Dr. Koji Tsukada (Osaka University, Japan), which was generated from Qb genomic cDNA [23].Qb particles were diluted with LB medium and plaque assay was performed according to the standard method [48].Polypropylene centrifuge tubes (15 ml, No. 430791; Corning Incorporated, Corning, NY) treated with 0.1% BSA for at least 15 minutes to prevent attachment of phages to the tube walls were used for all the experiments as culture tubes.

Experimental evolution system
Copropagation experiment: 4.8610 7 cells Anc(C) and 5.1610 7 PFU Anc(P) were mixed and copropagation was started in a culture volume of 3 ml at 37uC with shaking at 160 rpm.Mixed cultures were divided into 2 lines on the 18 th day, equivalent to 59 replication generations, and propagated independently for a further 36 days (line 1 and line 2 in Figure 1B).Serial transfer was conducted by daily transfer of the cultures with cells and phages.The portion of cultures calculated based on the final OD 600 were transferred into fresh medium with dilution to an initial OD 600 of 0.05.Daily culture samples were divided into thirds: one for preparing 280uC frozen stocks with 15% glycerol, one for CFU determination by dilution and spreading on low divalent cation mM63gl agar medium with 0.2 mM MgSO 4 ?7H 2 O, and the other for PFU analysis using the supernatant after centrifugation.Qb propagation experiment: Two lines (line 3 and line 4 in Figure 1B) were independently propagated from Anc(P) for 18 days, equivalent to 168-169 replication generations, at 37uC with shaking at 160 rpm.Serial passages consisted of infection of a host culture, followed by about 6 h of phage growth, and extraction of the phage from the culture.Each serial passage was performed as follows: uninfected Anc(C) cultures were grown at 37uC overnight and transferred into new medium with dilution to OD 600 of 0.03.When OD 600 became 0.06-0.07(approximately 1610 7 CFU/ml) after 2-2.5 h, cells were infected with phage to approximately 1.0-2.0610 7PFU/ml from the previous passage.The cultures were grown for about 6 h.E. coli cells were removed by centrifugation, and the supernatant was subjected to filtration with 0.2 mm syringe filters (Minisart RC15 filters; Sartorius Stedim Biotech, Goettingen, Germany), and phage solution was stored at 4uC for infection on the next serial passage.The replication generation number of the phage population (n) was calculated as n = ln 2 (N f /N i ), where N i and N f are the phage density (PFU/ml) at the initial and final time points of each passage, respectively.The initial value (N i ) was calculated by dividing the N f of the previous passage by the dilution rate.

Purification of E. coli and Qb phage from mixed cultures
The evolved E. coli populations (M54(C), M163M(C), and M165_2(C)) and Qb phage populations (M54(P), 163(P), and M165_2(P)) were purified from mixed cultures to analyze the phage genome sequence and to determine their fitness.To purify the evolved E. coli population, cultures stocked at 280uC including evolved E. coli and phage were streaked on mM63gl agar medium and then passaged several times in low divalent cation medium, 0.2 mM MgSO 4 ?7H 2 O mM63gl, to prevent further phage adsorption to E. coli.We checked the purity of evolved E. coli by confirming that no plaques were observed in the passaged and chloroform-treated cultures.To purify the evolved phage population, cultures stocked at 280uC including evolved E. coli and phage were cultured in mM63gl at 37uC with shaking at 160 rpm for 1 day and filtrated with 0.2 mm syringe filters (Minisart RC15 filters; Sartorius Stedim Biotech).These particles were used for RNA genome sequencing analysis as described below.For analysis of phage fitness, these filtrated particles and Anc(P) were dialyzed

Figure 1 .
Figure 1.Model evolution system.(A) In the copropagation regime, cultures including E. coli and Qb were passaged into fresh medium every day.In the Qb propagation regime, only Qb was isolated and used to infect fresh growing Anc(C).The values of PFU/ml or CFU/ml of cultures incubated for about 24 h (copropagation regime) or about 6 h (Qb propagation regime) were determined.The concentration of Qb and/or E. coli decreased by dilution in passage and increased by amplification or growth.(B) Phylogeny and nomenclature of the experimental lineages used in this study.Copropagation was conducted in two lines designated as lines 1 and line 2. Qb propagation was conducted independently in two lines: line 3 and line 4.The ancestral organisms were designated as Anc.The M and S represent the sample of copropagation and Qb propagation lines, respectively.The numbers in the boxes represent the replication generation numbers of Qb, and the numbers after the underbar represent the lineage line number.The mutations observed in the Qb genome in both propagation experiments are shown in the order of the sequence of the ancestral Qb genome, position on the genome, and the sequence of the evolved genome.The + and 6 in parentheses represent monomorphic and polymorphic sequences, respectively.Position 569 of Anc was heterogeneous with G and A, although the Anc Qb population was derived from cloned cDNA.doi:10.1371/journal.pgen.1002188.g001

Figure 2 .
Figure 2. Population dynamics in the copropagation or Qb propagation experiments.(A) Population dynamics of E. coli density (CFU/ml) in the copropagation experiments.Mixed cultures were divided into 2 lines on the 18 th day, equivalent to 59 replication generations, and propagated independently for a further 36 days (line 1 and line 2).On the 10 th and 17 th days, we restarted the copropagation experiment from stock cultures stored at 280uC.The open and closed triangles indicate the cell density of line 1 and line 2, respectively.(B) Population dynamics of free Qb density (PFU/ml) in copropagation experiments.Open and filled circles indicate the free phage density of line 1 and line 2, respectively.(C) Population dynamics of free Qb density (PFU/ml) in the Qb propagation experiment.The open and filled diamonds indicate the free phage density of line 3 and line 4, respectively.doi:10.1371/journal.pgen.1002188.g002

NA
Each experiment was conducted in duplicate and the two values are shown.The fitness of E. coli was calculated as (Fitness) = Log 10 (OD 600_7h /0.03),where OD 600_7h and OD 600 = 0.03 are the OD 600 values at 7 h after infection and initial OD 600 of the fitness assay experiment, respectively.The growth curves are shown in Figure S1.NA, not available.doi:10.1371/journal.pgen.1002188.t001

Figure 7 .
Figure 7. Time course of changes in cumulative mutations in the Qb genome.The cumulative mutations in line 1 ( N ), line 2 (#), line 3 (m), and line 4 (D) are shown.Cumulative mutation numbers were calculated as described in the legend of Table3.doi:10.1371/journal.pgen.1002188.g007

Figure 8 .
Figure 8. Estimation of evolutionary path.The direction of the arrowhead on each arrow indicates higher fitness.The gray line at M54(C) indicates that M54(P) and M163(P) showed almost equivalent fitness on M54(C).The red line indicates the plausible route for fixing M163(C) and M163(P) in this study.The genes in which mutations compared with the ancestor and detected only in the copropagation regime are shown in brackets, and the mutation positions in each gene for E. coli and in the Qb genome are shown in parentheses.doi:10.1371/journal.pgen.1002188.g008

Table 1 .
Fitness of ancestral and evolved E. coli.

Table 2 .
Fitness of ancestral and evolved Qb.
Each experiment was conducted in duplicate and the two values are shown.