JSTOR

Phage Lines and Cell Strains

The wild‐type T7 (T7⁺) bacteriophage (GenBank AY264774) has a 40‐kb dsDNA genome encoding around 60 proteins and infects several strains of Escherichia coli (Molineux 1999). Derivatives of three E. coli strains were used as hosts here: C, B, and K12. All three strains are infected by T7⁺. Yet these strains are known to differ in their lipopolysaccharide (Wright et al. 1980), surface molecules that serve as the receptors for T7 (Molineux 1999). They are thus plausible candidates for the evolution of discrimination by T7. Escherichia coli C was used as a permissive host (supporting phage replication) in all adaptations. An E. coli B strain IJ1958 trxA::Kn was the nonpermissive cell in the T7_B⁻ adaptation to avoid B, while IJ1517 trxA::Kn, a K12 strain, played that role in the T7_K⁻ adaptation to avoid K12. Both nonpermissive strains allow infection but abort all T7 progeny because they lack the host factor thioredoxin. Adsorption assays used E. coli C, BL21 (a permissive B strain), and IJ113 (a permissive K12 strain). For optimal foraging assays, a strain designated E. coli C_tet‐R (E. coli C carrying a tetracycline‐resistant pBR322 plasmid) was used as the good host, while IJ1142 (a tetracycline‐sensitive K12 otherwise nearly identical to IJ1133) acted as the poor host.

Selection for Host Discrimination Mutants

Wild‐type T7⁺ was transferred in a mix of permissive C and either nonpermissive K12 or nonpermissive B in Luria‐Bertani (LB) broth (10 g NaCl, 10 g Bacto tryptone, 5 g Bacto yeast extract/L) at 37°C. In general, a large excess of nonpermissive cells was used, with enough permissive cells to maintain the phage population. A somewhat heuristic approach was used to select avoidance of K12, in which a culture of phage was often challenged with pure nonpermissive cells before adding permissive ones; a more standardized protocol of serial transfer into flasks with 95% nonpermissive cells was used to select avoidance of B. The stringency of the negative selection likely affects the rate of adaptation to avoid a host but is not expected to alter the outcome.

Construction of T7_choosy Phage

The phage that evolved to avoid K12 (T7_K⁻) carried a substitution in its tail fiber gene (17) but also carried at least one other new substitution in gene 17.5. To separate the effects of the 17 substitution from the effects of other changes, site‐directed recombination was used to create a wild‐type phage carrying just the 17 substitution. This protocol involved creating plasmid pRS36255, a pUC18 vector carrying the gene 17 fragment of T7_K⁻ spanning nucleotides 36054–36360. Phage T7⁺ was plated on Topo10 cells (Invitrogen) bearing pRS36255 and was incubated at 37°C. Several plaques were suspended in LB broth, and the suspension of phage was enriched for recombinants bearing the substitution by a 2‐h incubation with IJ1517 trxA::Kn (which preferentially destroyed phages that adsorbed to K12). The culture was then treated with chloroform to kill cells and was plated on E. coli C. Plaques were isolated and sequenced to check for the presence of the desired substitution in gene 17 and the lack of the substitution in 17.5. This phage, in which the only genetic change should be one that decreases adsorption to K12 (although it may have epistatic effects), is called T7_choosy.

Adsorption Rate Assays

Cells were grown from frozen aliquots for 1 h to a density of approximately 10⁸ cells/mL. The culture was infected with 10⁶ plaque‐forming units of the requisite phage from a lysate less than 5 days old. At 5 min, the culture was plated from centrifuged and uncentrifuged aliquots to determine the density of unadsorbed phages (N_u) and total phages (N_t). Adsorption rate α (mL/min) was calculated from , where C is the density of cells per milliliter, also determined by plating (Bull et al. 2004).

Optimal Foraging Assays

A 50:50 mix of T7_choosy (infects C, avoids K12) and T7⁺ (an indiscriminate phage that infects C and K12) was added to a flask containing, in 10 mL of LB broth, a mix of good‐quality hosts (tetracycline‐resistant C_tet‐R) and poor hosts (tetracycline‐sensitive K12, IJ1142). Flasks were placed in a water bath with continual shaking (200 rpm) to maintain cell suspension and aeration. Cells were added from frozen aliquots at volumes expected to yield the cell densities reported at 1 h, but cell densities were measured by plating in each set of trials to confirm densities. Tetracycline at varying concentrations was added 10 min before phage addition to decrease the productivity of K12 infection. After phage growth for 30 min and before phage densities exceeded cell densities, a portion of this flask was added to another flask with the requisite cell densities. This was repeated so that each phage mixture was adapted for a combined 2 h.

Sequencing

The DNA for sequences was obtained from polymerase chain reaction (PCR) products, either of isolates or of populations. Genomes of the two phages adapted to avoid a cell type were sequenced over genes 17 (tail fiber; Dunn and Studier 1983) and 17.5 (a holin thought to affect lysis; Vukov et al. 2000; Wang et al. 2000). Phage T7_K⁻ was also sequenced over genes 11 and 12, which encode other tail proteins. All reported nucleotide numbers presented are from T7⁺ (GenBank V01146; Dunn and Studier 1983). The population at the end of each optimal foraging assay, as well as the original 50:50 mix of T7⁺ and T7_choosy, was sequenced across base 36254 in the tail fiber gene, at which the phages differ. The relative peak height of the two bases was used to estimate frequencies by dividing the height of one peak by the total combined height of both peaks. The known 50:50 mix was found to yield peak heights of very close to 50% each (50.3% T7_choosy), validating the accuracy of this method. Sequences were obtained from PCR products with ABI Big Dye mix (ver. 3.1) and an ABI3100, analyzed with AB Sequencing Analysis Software, version 5.1.1, and DNA Star, version 4.05.

Evolution of Discrimination

Wild‐type T7⁺ was independently adapted in two mixes of Escherichia coli strains: C with either B or K12. In both adaptations, C was the permissive host, while the other (B or K12, depending on the adaptation) aborted T7 infections due to deletion of a host gene needed for viral replication. Both adapted phages evolved to largely avoid the nonpermissive host but maintained a high adsorption rate to C; the evolved phages are designated T7_B⁻ and T7_K⁻, respectively (fig. 1; the adsorption of T7_K⁻ is represented with T7_choosy, a recombinant that carries the relevant adsorption mutation). The selective environment in both treatments was confined to just two of the three possible hosts (C, B, and K12). However, avoidance of the nonpermissive host evolved without greatly compromising adsorption to either the permissive host or the unselected host (a slight reduction on the permissive host was observed for TK_K⁻).

(21 KB)

Figure 1: Adsorption rates are shown for different phage isolates (T7⁺, T7_choosy, and T7_B⁻) on three hosts (C, B, and K12, indicated by letters above each bar). Wild‐type T7⁺ adsorbed well to all cell types. Up (down) arrows indicate selection to grow on (avoid) a particular host strain. Adapted phages’ adsorption was low on hosts they were selected to avoid, while there were varying (but less extreme) effects on adsorption to other cell strains. Phage T7_choosy carries the mutation in the tail fiber gene that caused adsorption changes during the adaptation to avoid K12 and is statistically indistinguishable from T7_K⁻. All error bars represent 1 SE. An upper limit for adsorption rate, based on physical properties, is considered to be 10⁻⁸ (Adams 1959), so values greater than 10⁻⁹ are considered to indicate moderately high adsorption rates.

Open New Window

The adsorption rates observed here are small, and if one is not accustomed to quantitative aspects of phage dynamics, the implications just claimed may not be obvious. We thus offer some basics. The maximum attainable adsorption rate for a generic phage, in which nearly all phage‐bacterial collisions result in adsorption, is approximately 10⁻⁸ mL/min (Adams 1959). The low magnitude of this value is due to the small sizes of the phage and bacterium and the consequent improbability of a random collision per minute between the two individuals in a milliliter of media. Adsorption rates of most lab‐adapted phages on their lab hosts are within the range of and are hence close to this limit. A drop of even 10‐fold, while appearing to be small relative to 10⁻⁸, has a profound effect on infection rates. For example, under the conditions used in this design (10⁸ cells/mL), a change in the adsorption rate from to to would mean that after 10 min, the fraction of phage adsorbed would drop from 95% to 25% to 3%. Thus, the evolved declines in adsorption rates have a substantial effect in reducing the phage’s use of that host.

Expansion of host range in tailed phages often stems from changes in tail fiber genes (Drexler et al. 1989, 1991; Hashemolhosseini et al. 1994), so it was anticipated that the T7 discrimination would also stem from changes in the tail fiber gene (gene 17). Sequencing of 17 indeed revealed one coding substitution in each adaptation, at nine and 33 residues from the carboxyl (distal) end of the 533 amino acid protein (table 1); this end of the protein is involved in interaction with the host, so changes in this region plausibly affect adsorption (Steven et al. 1988). Site‐directed mutagenesis/recombination was used to introduce the 17 change from T7_K⁻ to T7⁺, and this phage (T7_choosy) was found to have adsorption statistically indistinguishable from that of T7_K⁻ on both C and K12. Thus, the discrimination between C and K12 can be attributed to the single substitution in the tail fiber gene. It is likely that the (different) gene 17 change in T7_B⁻ caused the discrimination between C and B, but the effect of this substitution alone was not tested.

Table 1: Genetic evolution of T7_K⁻ (adapted to avoid K12) and T7_B⁻ (adapted to avoid B)

Open New Window

An amino acid change was also detected in the adjacent gene 17.5 of each evolved phage. This gene has a role in lysis timing. Lysis time was measured in one of the evolved phages, T7_K⁻, and was significantly earlier than in T7_choosy, which was wild‐type, except for the gene 17 substitution (methods as in Heineman et al. 2005; data not shown). Because only part of that phage was sequenced, it cannot be ruled out that unidentified changes were responsible for the difference in lysis time, but the change in lysis time is compatible with an effect of the 17.5 change on lysis.

Predictions of Optimal Foraging Theory

The conditions used to select discrimination were extreme and thus merely tested whether T7 could easily evolve to discriminate between a pair of host strains. Given that host discrimination mutants exist, optimal foraging theory makes three testable predictions (Emlen 1966; MacArthur and Pianka 1966; Pyke et al. 1977; Bull 2006). For simplicity, we consider optimal foraging theory as it applies to a phage growing on a population of two hosts of different qualities (one host is good, the other poor for phage reproduction). (1) Dependence on relative host quality: holding host densities constant, discrimination will be increasingly favored as the difference in quality of the two hosts is amplified. (2) Dependence on good‐host density: discrimination will be selected against at low densities of the good host but will be favored at high densities of the good host (holding density of the poor host constant). (3) Lack of dependence on poor‐host density: the density of poor hosts affects the strength of selection on avoidance but not the direction.

These predictions rest on several assumptions: (i) the two host types are encountered randomly, in proportion to their abundances; (ii) hosts outnumber phages; and (iii) avoiding one host type has no effect on the rate of encountering another. The first two assumptions are easily met by the experimental conditions: a liquid culture in which the two hosts are continually mixed by shaking and from which phages are transferred before their densities outnumber cell densities. The third assumption is met by virtue of the kinetics of phage‐bacterial encounters, as reviewed above (Adams 1959).

Tests of Optimal Foraging Theory

The predictions were tested by competing phages with different host ranges on mixed populations of two hosts. The two hosts were engineered to present different benefits to the phage by supplementing the media with low levels of the antibiotic tetracycline and endowing one of the hosts with a tetracycline‐resistant plasmid; impairing bacterial physiology and protein synthesis with this drug directly impairs phage physiology because the phage depends on many bacterial functions. Specifically, 1:1 mixes of T7⁺ and T7_choosy were introduced to different densities of E. coli C (C_tet‐R, tetracycline resistant) mixed with K12 bacteria (tetracycline sensitive), and changes in phage frequencies were monitored to assess relative fitness. These tests could also have employed T7_B⁻ as the discriminating phage (on a mix of E. coli C and B hosts), but the apparently similar molecular bases of host discrimination by T7_K⁻ and T7_B⁻ led us to focus on just one phage. Furthermore, when the engineered strain T7_choosy was utilized, the difference in host range was known to be the single amino acid difference in the tail fiber gene, and confounding effects of other substitutions were thereby avoided.

Phage T7_choosy infects C and avoids K12, whereas T7⁺ infects both at high rates (fig. 1). Prediction 1, that larger differences in quality between good and poor hosts will increasingly favor discrimination, was tested by competing the two phages under conditions imposing different disadvantages to infecting the poor host (K12, tetracycline sensitive) by using different concentrations of tetracycline. Increasing levels of tetracycline may have also affected fitness of C_tet‐R cells somewhat, but for the model, it is the increasing disparity between the two host types that is important. The same density of cells was used in all experiments (approximately 5% C_tet‐R and 95% K12, with a total density of 10⁸ cells/mL).

Results of 2‐h competitions were consistent with the optimal foraging model. The proportions of the two phages were calculated by comparing the height of sequence peaks at the nucleotide in the tail fiber gene that differed between the phage lines (the nucleotide known to be the basis of host discrimination). Avoidance of K12 was favored at the higher levels of tetracycline (1.5 μg/mL and above) when K12 was most impaired but was selected against at the lower levels of tetracycline (0.5 μg/mL; fig. 2). For all but one tetracycline concentration, only one phage base (and, hence, only one phage type) was detectable in the sequence profile from the 2‐h endpoint culture. At one intermediate tetracycline concentration (1 μg/mL), selection was weak enough that both phage types were maintained at intermediate frequencies throughout the 2‐h competition, although it appeared that discrimination had a slight advantage. Phage T7⁺ did grow on K12 impaired by tetracycline at 1.5 μg/mL (data not shown), so these results were not due to the trivial case that K12 simply prevented phage reproduction whenever discrimination was favored.

(25 KB)

Figure 2: Advantage of host avoidance depends on relative host quality. Phages were grown in a mixture of a good host (C_tet‐R) and a poor host (K12), with and 10⁸ cells/mL of each host, respectively. The quality of tetracycline‐sensitive K12 hosts was progressively reduced by increasing the concentration of tetracycline, as indicated on the horizontal axis. A 50:50 mixture of choosy phages that avoided K12 (T7_choosy) and indiscriminate phages that infected both strains (T7⁺) was used to initiate the competition; at 2 h, the final proportion of each phage type was measured. At least two assays were performed for each point. Confidence bars of 1 SE, smaller than the point, are provided.

Open New Window

Prediction 2, that increasing the density of good hosts favors discrimination, was tested as well. Using tetracycline at 1.5 μg/mL and a constant high density of K12 hosts (10⁸ cells/mL), we found that avoidance spread when the density of C_tet‐R was high but not when it was low (fig. 3). This result also supports the model. Again, selection was weak enough at one intermediate density of C_tet‐R that peaks from both phage types were evident in the sequence profiles from the end of the competition.

(22 KB)

Figure 3: Evolution of avoidance depends on the density of both good and poor hosts. When the relative productivity of good and poor hosts is held constant (1.5 μg/mL tetracycline) and the density of poor hosts is high (10⁸ K12/mL; solid line), increasing the density of good‐host cells (C_tet‐R) ultimately favors avoidance of poor hosts. When the poor host is at low density (10⁶ K12/mL; dashed line), the same pattern is expected as for the solid line, yet the discriminating phage is at a disadvantage, even when the good host is common. At least two assays were performed for each point, except the lowest‐density C_tet‐R 10⁸ K12 point. Confidence bars of 1 SE, sometimes smaller than the point, are provided.

Open New Window

From prediction 3, once good host density is high enough to favor discrimination, avoidance should evolve regardless of the density of the poor host. In contrast to the other successes of the optimal foraging model, this prediction was not supported: although discrimination was favored when C_tet‐R and K12 densities were and , respectively, discrimination was selected against when K12 density was dropped by two orders of magnitude to (tetracycline at 1.5 μg/mL in all cases; fig. 3). The basis for this model failure is suggested by the next result.

An implicit assumption of the optimal foraging model is that the only fitness cost of discrimination is realized from the missed opportunities of infecting the poor host (which may be advantageous or not, depending on the environment). Thus, the two phages should have equal fitness when exposed to just the good host. In contrast to this assumption, a competition of both phages on C_tet‐R (with 1.5 μg/mL tetracycline) revealed that T7⁺ grew faster than T7_choosy. Phage T7_choosy was only 38% of the population after 2 h (differing from a 50:50 ratio, with , two tailed). Thus, there is a modest cost to discrimination between C and K12, independent of differential host effects. This cost is likely to be caused by the lower adsorption rate of T7_choosy to C ( ; fig. 1) and is specific to the gene 17 substitution of T7_K⁻ (which can be inferred from the fact that T7_choosy was engineered to differ from T7⁺ by only the single substitution).

The choosy phage fared best (in competition with T7⁺) at high densities of poor hosts (10⁸), next best in the absence of poor hosts, and, paradoxically, slightly worse at intermediate densities of poor hosts (10⁶). However, the difference between T7_choosy performance with few poor hosts and with no poor hosts is not significant, and it seems likely that there is merely little difference between these two conditions.