Significance

Abstract

Results and Discussion

In the experimental populations, phages had no effect on bacterial densities (Fig. S1A) despite evidence of ongoing phage lysis in all replicate populations of the phage treatment (Fig. S1B). At the end of the experiment, free virions of all phages were detected in four of six populations, whereas in the other two populations, only 3 and 4 virions were detected (Fig. S1C). We observed high rates of lysogeny [i.e., integration of prophage(s) into the bacterial chromosome] in five of six populations, but the phages differed in their propensity to form lysogens: lysogens of the transposable phage 4 approached fixation in five of six populations, whereas lysogenization of bacteria by the other phages was less common, and, where observed, was typically as a polylysogen in combination with 4 (Fig. S2). Thus, lysogeny, and indeed polylysogeny, was rapidly established in our experimental populations; moreover, lysogeny appears to have been essential for the long-term maintenance of phages in populations.

Open in a separate window

Fig. S1.

Bacterial-free phage dynamics through time when grown in ASM. (A) Colony-forming units per milliliter in each population, separated by treatment. (B) Total free phage (plaque-forming units per milliliter) in phage-treated populations. (C) Relative densities of free virions for each of the LES phages in end point populations from the phage treatment, where relative density is calculated as log₁₀ (copies per microliter + 1), as determined by qPCR. Boxplots represent the median and interquartile range.

Open in a separate window

Fig. S2.

Lysogen dynamics over time in populations evolving with phages. Each population is represented by a single area plot denoting the frequency of bacterial isolates lysogenic for each specified combination of LES prophages through time.

To determine the fitness response to selection, we competed each evolved population against the ancestral PAO1 in ASM, and because lysogens may have higher fitness simply due to phage-mediated killing of susceptible competitors (29, 31, 32), we also performed competitions against a phage-resistant PAO1ΔpilA, an isogenic knockout mutant strain lacking the gene encoding the type IV pilus protein PilA [all of the temperate phages used here infect via the type IV pilus (33)]. We observed that evolved populations from both treatments were fitter relative both to PAO1 and PAO1ΔpilA [Fig. 1; one-sample t test (alt = 0), all significant at an α level of 0.0125]. Populations evolved with phages had higher fitness than populations evolved without phages relative to PAO1, but this fitness advantage of evolving with phages was lost when competing against PAO1ΔpilA (Fig. 1; treatment × competitor interaction: F_1,20 = 8.54, P < 0.01; simple effect of treatment against competitor PAO1: F_1,10 = 7.12, P < 0.025; and simple effect of treatment against competitor PAO1ΔpilA: F_1,10 = 1.53, P = 0.24). Together these data confirm that populations in both treatments adapted to the sputum-like environment and that lysogenised hosts had enhanced competitiveness against phage-susceptible competitors.

Open in a separate window

Fig. 1.

Fitness response to selection in populations evolving with and without phages. Data points represent the mean ± SE fitness calculated as selection rate for populations evolved with (filled symbols) or without (open symbols) phages in competition against either ancestral PAO1 or an isogenic phage-resistant competitor, PAO1ΔpilA.

To determine the genetic basis of the observed evolutionary adaptation, we performed whole genome sequencing on population samples containing 40 random clones pooled per population from the end of the experiment. All populations contained SNPs and small insertions or deletions (indels), and all replicate populations that had evolved with phages contained integrated prophages. At the genome-wide scale, populations evolved with or without phages did not differ in abundance or frequency of SNPs and indels (excluding insertions caused by prophage integrations) and both groups had high variance of polymorphic sites: between 16 and 173 among the phage-free populations and 17–176 among the phage-containing populations (Table S1).

Parallel evolution at a particular locus, where independent mutations are observed more often than expected by chance, is strong evidence for positive selection. For example, in the absence of selection, the probability of observing a mutation in two populations at the conclusion of the experiment is only P = 0.003, and P = 0.0002 if observed in three populations (for an average 1,004-bp protein-coding sequence in the PAO1 genome). Thus, to identify loci likely to have been under selection during experimental evolution, we concentrated our analyses on the subset of genes that had been targeted by mutations in at least two replicate populations per treatment (Fig. 2 and Tables S2 and ​andS3).S3). A greater degree of parallel evolution was observed in the presence of phages (measured as the probability of randomly drawing a pair of mutated genes from different populations, with phages 0.056 ± 0.016 SE and without phages 0.024 ± 0.007 SE, P < 0.05 by bootstrap test). Some parallel targets of selection were shared among treatments, including genes involved in the type IV pilus motility, flagellar motility, biofilm formation, metabolism, and regulation, suggesting that these mutations were beneficial in the sputum-like environment per se. Interestingly, some loci were more likely to evolve in the presence of phages. In particular, mutations affecting the quorum sensing (QS) transcriptional regulators lasR and mvfR were each observed in five of six replicate phage-treated populations compared with only one of six replicate phage-free populations. In addition, three of six populations evolving with phages vs. one of six evolving without phages contained mutations in fha1, which encodes the forkhead-associated (FHA) domain protein that posttranslationally activates type VI secretion (34). A further indication of stronger selection due to phages is that parallel selected loci displayed higher allele frequencies in the phage treatment (mean allele frequency = 34.33 ± 3.2 SEM/40) compared with the control treatment (mean allele frequency = 22.67 ± 3 SEM/40), suggesting that selective sweeps in the phage treatment were closer to fixation (discussed in more detail in SI Materials and Methods and SI Results and Discussion).

Open in a separate window

Fig. 2.

Genetic loci under positive selection as indicated by parallel genomic evolution in populations evolving with and without phages. Each concentric circle corresponds to a replicate population in either the control (without phages) or treatment group (with phages), as indicated. Positions around each concentric circle, starting at the 12 o'clock position and in a clockwise direction, correspond to positions around the published P. aeruginosa PAO1 single circular chromosome. The smaller circles around each concentric circle indicate the positions of variants in each replicate that were observed in an ORF under positive selection, i.e., mutated in parallel in at least one other replicate. Only variants in fha1 were at precisely the same position and were likely to be homologous. Variants are also listed in Table S6.

A key difference between the populations evolving with, vs. without, phages is that a substantial fraction of ORFs under positive selection (8 of 26; Fig. 2) contained mutations caused by prophage integration in the phage-containing treatment. Prophage 2 and 3 were found exclusively at the intergenic sites homologous to their positions in the LESB58 chromosome from where they were isolated (4,629,220 and 4,103,724 bp in the PAO1 chromosome) and were therefore excluded from analyses. In contrast, 4 integrations were observed at 19 different positions across the six replicates, suggesting that 4 integration provided an additional source of genetic diversity for selection to act on. Although 4 is closely related to D3112 and therefore likely to integrate randomly throughout the chromosome (10, 27, 33), we tested for hotspots of integration with an exhaustive search of sequence motifs using the Multiple Em for Motif Elicitation (MEME) and Motif Alignment and Search Tool (MAST) algorithms (35, 36) in the 1,000-bp region around each of the integration sites. Only very weakly conserved motifs were identified (Table S4), suggesting that 4 has the low integration specificity characteristic of transposable phages like D3112. Despite the very low integration specificity of 4, all of the 19 integration sites were located within or adjacent to only seven different operons, which is consistent with positive selection for integrations at those regions (Fig. 2 and Tables S5 and ​andS6).S6). The functions most commonly predicted to be disrupted by positively selected 4 integration events were type IV pilus motility and QS. Thus, transposable temperate phages like 4 may commonly alter the mode of bacterial evolution by increasing the supply of mutations available to natural selection.

Because all of the temperate phages used here infect via the type IV pilus (33) and PAO1ΔpilA mutants showed higher fitness compared with PAO1 against lysogenized bacteria evolved in the phage treatment (Fig. 1), we hypothesized that disruption of type IV pilus motility associated genes may have been selected to prevent superinfection and lysis of 4 lysogens. Notably, although 4 lysogens have strong superinfection immunity against 4, they remain susceptible to infection and subsequent lysis by 2 and 3 (33), suggesting that loss of type IV pili could be beneficial in the presence of a diverse phage community. In support of this, we observed higher rates of phage resistance in populations that evolved with vs. without phages (Fig. 3A; Mann–Whitney test; W = 24.0, n₁ = n₂ = 6, P = 0.02). Correspondingly, type IV pilus-dependent twitching motility was lost more rapidly in phage-containing populations than in phage-free populations, suggesting that the loss of type IV pilus function was more strongly selected in the presence of phages (Fig. 3B). To determine whether loss of type IV pilus twitching motility phenotype was associated with 4 integration, we tracked allele frequency dynamics in two replicate populations. Specific PCR primer sets (Table S7) were used to detect integrated 4 prophage in type IV fimbral biogenesis protein encoding genes fimU and pilV. In both cases, there was a positive association between the allele frequency dynamics and the rise in frequency of twitching motility-deficient mutants (Fig. 3C). We next contrasted the allele frequencies of SNPs, indels, and 4 prophage-integration mutations occurring at type IV pilus associated loci. Across six phage-treated populations, 12 mutations of the type IV pilus-associated genes were detected in parallel, the majority (n = 9) were caused by 4 prophage integration, and half of them (n = 6/12) occurred at high frequencies (≥10/40 clones per population). However, in the absence of phages, although more mutations were detected in these genes in parallel (n = 16), only 13% of these mutations occurred at high frequency (Table S3). Consistent with the phenotypic data (Fig. 3 B and C), this suggests that there was stronger positive selection for mutations disrupting type IV pilus-associated genes in the presence of phages, driving faster selective sweeps, compared with type IV pilus disrupting mutations occurring in the absence of phages. The exception to this pattern is replicate 3 of the phage treatment, where the impairment of twitching motility in 35 out of 40 isolates can be explained by a single frame-shift deletion variant in pilY1 (PA4554; encodes a type IV pili biogenesis protein; Fig. 2). It is notable that a low frequency of lysogeny was observed in this population, unlike all other phage treatment replicates where lysogens approached fixation (Fig. S2).

Open in a separate window

Fig. 3.

The evolution of resistance to phages and pilus-dependent twitching motility traits. (A) Boxplot of phage resistance in end point populations. The thick horizontal line denotes the median frequency of isolates in a population resistant to one or more LES phages for each treatment. Asterisks denote outliers and narrow horizontal lines denote the upper and lower quartiles. (B) Frequency of bacterial isolates in each population through time displaying impaired twitching motility in the control (gray circles, solid line) and phage treatment (black triangles, dotted line). (C) Allele frequency dynamics of LES4 integrated into fimU and pilV for populations P7 and P11, respectively, and loss of twitching motility in these populations. Closed black circles and open white diamonds represent populations P7 and P11, respectively, solid gray lines denote loss of twitching motility data, and dashed black lines denote allele frequency data.

Table S7.

Primers used in the study

Protocol	Primer name	Sequence	Target	Length of amplicon (bp)	Cycling conditions	Reference
Detection of prophage (multiplex PCR assay)	LESp2F	CTCCACTTCTCGGTTGCTTC	LES 2	206	95 °C, 4 min then 30 cycles: 95 °C, 30 s; 58 °C, 30 s; 72 °C, 30 s; final extension step: 72 °C, 7 min.	This study
	LESp2R	ACTAGCCCCGTATCCGAGTT
	LESp3F	TCAGGAAAACCTTGCCATTC	LES 3	384
	LESp3R	GTCTTCTGGTGGTCGGTGAT
	LESp4F	AGTTACGCCTGCTGGTGAGT	LES 4	506
	LESp4R	CCTCAGTCGTGCCTTCTTTC
Determination of LES4 integration site in PAO1 chromosome	7leftF	TTCGAGTTGGATCCGGCC	fimU LJ	247	95 °C, 4 min then 30 cycles: 95 °C, 30 s; 55 °C, 30 s; 72 °C, 30 s; final extension step: 72 °C, 7 min.	This study
	7leftR	GTCGTGCTGTGCTGATCTTT
	7fimUrF	CGCTCATTCCGTGCCAATTA	fimU RJ	358
	7fimUrR	TCAATGCGATGCTGCAGTAC
	7leftF	TTCGAGTTGGATCCGGCC	fimU intact insertion site	328
	7fimUrR	TCAATGCGATGCTGCAGTAC
	11leftF	CTTCTTCAAGGCCAAGGGGT	pilV LJ	458
	11leftR	CGATGGCGATACGGTGATGA
	11rightF	GACGAGGTGCTTAGACGGAG	pilV RJ	610
	11rightR	ATCATGGACAGGCCCGATTG
	11leftF	CTTCTTCAAGGCCAAGGGGT	pilV intact insertion site	349
	11rightR	ATCATGGACAGGCCCGATTG
qPCR to detect relative copies of free LES phages	2totF	AGTAGCCGACCCAGACCTTT	LES2	141	95 °C, 10 min then 40 cycles: 95 °C, 10 s; 60 °C, 15 s; 72 °C, 30s.	(33)
	2totR	ATGGAAGCAACCGAGAAGTG
	3tot1F	CGCAGGTACCACCAGACTTT	LES3	122
	3tot1R	CATGTCCAGCAGGTTCAAAA
	4tot1F	GCTCATGAGTGGCTGACAAC	LES4	105
	4tot1R	TCTTGGGCAGAGAACCATTC
Confirmation of complete degradation of bacterial DNA	PS1	ATGAACAACGTTCTGAAATTCTCTGCT	P. aeruginosa oprI lipoprotein gene	249	94 °C, 5 min then 30 cycles: 94 °C, 1 min; 60 °C, 1 min; 72 °C, 2 min; final extension step: 72 °C, 5 min.	(84)
	PS2	CTTGCGGCTGGCTTTTTCCAG

Open in a separate window

Our genomic data suggest that temperate phages promoted the loss of QS with positive selection of 4 prophage integrations, SNPs, and indel mutations at the mvfR and lasR loci. Mutations to lasR lead to disruption of the acyl-homoserine-lactone (AHL) signaling system (37), whereas mutations to mvfR lead to disruption of the Pseudomonas quinolone signal (PQS) system (38), suggesting large-scale alterations to QS cell–cell signaling in populations evolving with phages. To test whether QS deficient bacteria have higher fitness in the presence of phages, we competed PAO1 against PAO1ΔlasR in ASM with and without the temperate phages. There was no effect of phages on the fitness of PAO1ΔlasR (Fig. S3; two-sample t test, t₁₀ = −0.7989, P = 0.44), which was substantially fitter relative to PAO1 in both the presence [one-sample t test (alt = 0), t₅ = 5.0331, P < 0.01] and absence [one-sample t test (alt = 0), t₅ = 6.7065, P = 0.001] of phages. These data suggest that lasR mutations are beneficial in ASM per se. This interpretation is consistent with the observation that single populations in the phage-free treatment also acquired mutations in QS genes, but suggests that the rate of evolution at these loci was higher in the presence of phages. Second, we compared the rates of spontaneous phage lysis of 4 lysogens constructed in both the PAO1 and PAO1ΔlasR backgrounds: there was no significant difference in production of free phages in stationary phase cultures (median free phage density: PAO1, 3.4 × 10⁸ pfu/mL, PAO1ΔlasR, 3.3 × 10⁸ pfu/mL; Mann–Whitney test; W = 92.0, P = 0.345). Thus, although direct interaction between temperate phages and bacterial QS has been reported in other systems, via QS induced lysis by phages (39) or QS mediated alteration of phage receptor expression by bacteria (40), this does not appear to be an important factor in our study. Phages may have simply increased the supply of large effect mutations available to natural selection, notably via 4 prophage integrations into mvfR (Fig. 2). Alternatively, there may have been epistatic interactions between the fitness effects of QS mutations and other positively selected mutations, which strengthened selection for loss of QS in the presence of phages. Mutations in QS regulators are commonly observed to accumulate over time in CF chronic infection (41). Both AHL and PQS signaling are required for full virulence in P. aeruginosa (42), suggesting that temperate phage selection may accelerate the loss of virulence in chronic infections.

Open in a separate window

Fig. S3.

Fitness of QS-deficient mutants (PAO1ΔlasR) relative to WT PAO1 in the presence and absence of temperate phages in ASM. We find no effect of phage presence on the relative fitness of PAO1ΔlasR (two-sample t test, t₁₀ = −0.7989, P = 0.44), which was fitter than WT in both the presence [one-sample t test (alt = 0), t₅ = 5.0331, P < 0.01] and absence [one-sample t test (alt = 0), t₅ = 6.7065, P = 0.001] of phages. r = 0, fitness of PAO1ΔlasR = PAO1; r > 0, fitness of PAO1ΔlasR > PAO1. Data are means (n = 6) ± SEM.

In summary, we showed that temperate phages enhanced parallel evolution in P. aeruginosa biofilms in a sputum-like environment. Our data suggest two ways in which this may have occurred: first, the transposable phage 4 mediated adaptive evolution by increasing the supply of positively selected mutations via insertional inactivation of genes caused by prophage integrations, particularly in type IV pilus and QS associated genes. Second, we present evidence that temperate phage strengthened selection, particularly for mutations in type IV pilus associated genes, accelerating the evolutionary loss of type IV dependent pilus motility presumably to avoid superinfection and subsequent lysis by phages which infect via the type IV pilus. A recent transposon sequencing study of P. aeruginosa PA14 shows that mutations in type IV pilus associated genes increase fitness in the murine lung (43). Moreover, loss of both motility (44, 45) and QS (46, 47) functions are known to frequently evolve in P. aeruginosa chronic infections of the CF lung. Temperate phages, including those used here, can be present at very high densities in the CF lung [exceeding bacterial densities by orders of magnitude (13)], which taken together with our findings suggests that temperate phages could play an important role in CF lung infections by driving the evolution of these clinically important traits in P. aeruginosa. In addition, our data suggest that living in a sputum-like environment per se selects for mutations in genes associated with motility, biofilm formation, metabolism and regulation. Similar mutations observed in CF lung isolated P. aeruginosa are therefore likely to be at least partially explained simply as adaptations to selection imposed by the sputum environment, but could have implications for susceptibility to antibiotics (48) or host immune responses (49) as correlated responses (50). Experimental evolution in clinically relevant infection models has the potential to enhance our understanding of the causal links between sources of selection and the evolutionary responses of pathogens in infections (51), advancing our understanding of within host pathogen evolution and our ability to direct this for improved patient health.

SI Materials and Methods

Materials and Methods

Twelve replicate microcosms (30-mL glass universals containing 5 mL ASM) were inoculated with 5 × 10⁷ cells P. aeruginosa strain PAO1 and grown as a biofilm [37 °C incubation with shaking at 60 rpm (Stuart SI500 16-mm orbital shaking incubator, Bibby Scientific Ltd.)]. LES phages 2, 3, and 4 were added to six microcosms after 24 h of growth to a total multiplicity of infection of 0.1 (phage treatment), and the remaining six were designated phage-free controls. Phages were added only once, at the beginning of the experiment. After a further 72-h growth, biofilms were homogenized with an equal volume of Sputasol and the homogenate transferred (1:100) into fresh ASM. Transfers were repeated every 4 d, to a total of 30 transfers (~240 bacterial generations). Every other transfer, bacterial and total free phage densities were enumerated, and every five transfers, the frequency of prophage carriage in the phage treatment was estimated with a multiplex PCR assay using primers targeted to each of the LES phages. At transfers 5, 15, and 30, 40 isolates per population were screened for the type IV pilus-mediated twitching motility phenotype using the agar stab method.

At the end of the experiment, DNA was extracted from 40 isolates per population and pooled, and the pooled DNA sequenced on an Illumina HiSeq 2000. Polymorphisms were called from reads aligned to the published PAO1 genome treating each sample as having ploidy = 40, reflecting the number of pooled isolates. 4 prophage insertion sites were estimated from the mapping locations of reads that mapped to the PAO1 chromosome and whose mate read mapped to the 4 prophage sequence. The MEME software suite (35, 36) was used for motif analysis. First, the 20 most conserved motifs at 4 prophage integration sites were selected using the MEME algorithm. Second, the entire host chromosome was searched for motif occurrences using the MAST algorithm (see SI Materials and Methods for details). Frequencies of each prophage integration site were estimated based on the number of read pairs split between prophage and reference chromosomes as a proportion of read pair depth in that region. ORF annotations from the published sequence were supplemented using the STRING v10 database (52). Counts of ORFs affected by mutations in more than one population per treatment were implemented using the BioPython library ver. 1.65 (53), and chromosome map plots were implemented using the svgwrite library ver. 1.1.3 in Python ver. 2.7.10. Included in the parallel selected loci were those exhibiting the “multidiverse” signature of unlinked polymorphisms in the same ORFs. The method is implemented in the CheckLinkage option of BAGA (https://github.com/daveuu/baga) (54) and discussed in SI Materials and Methods.

Relative free phage abundances in end point populations were estimated separately for each phage using a quantitative PCR (qPCR) assay of DNase-treated supernatants, using primers targeted to each phage and comparison with a set of standards of known concentration. Competitions were performed between end point populations against the ancestral PAO1 (labeled with a gentamicin resistance marker) or an isogenic LES phage-resistant competitor, PAO1ΔpilA (labeled with a tetracycline resistance marker), in conditions identical to one transfer of the selection experiment. The density of each competitor was determined by plating onto antibiotic selective and nonselective media. Fitness was calculated as the selection rate constant. Free phage densities were measured for LES4 lysogens in both PAO1 and PAO1ΔlasR. Ten independent lysogens were constructed in each host background (PAO1 and PAO1ΔlasR) and cultured in LB until the stationary phase, and free phage densities in the supernatant were quantified using a plaque assay. Full methods are included in SI Materials and Methods.

SI Results and Discussion

Acknowledgments

Footnotes

References