Emergence of a Competence-Reducing Filamentous Phage from the Genome of Acinetobacter baylyi ADP1

ABSTRACT

Bacterial genomes commonly contain prophage sequences as a result of past infections with lysogenic phages. Many of these integrated viral sequences are believed to be cryptic, but prophage genes are sometimes coopted by the host, and some prophages may be reactivated to form infectious particles when cells are stressed or mutate. We found that a previously uncharacterized filamentous phage emerged from the genome of Acinetobacter baylyi ADP1 during a laboratory evolution experiment. This phage has a genetic organization similar to that of the Vibrio cholerae CTXϕ phage. The emergence of the ADP1 phage was associated with the evolution of reduced transformability in our experimental populations, so we named it the competence-reducing acinetobacter phage (CRAϕ). Knocking out ADP1 genes required for competence leads to resistance to CRAϕ infection. Although filamentous bacteriophages are known to target type IV pili, this is the first report of a phage that apparently uses a competence pilus as a receptor. A. baylyi may be especially susceptible to this route of infection because every cell is competent during normal growth, whereas competence is induced only under certain environmental conditions or in a subpopulation of cells in other bacterial species. It is possible that CRAϕ-like phages restrict horizontal gene transfer in nature by inhibiting the growth of naturally transformable strains. We also found that prophages with homology to CRAϕ exist in several strains of Acinetobacter baumannii. These CRAϕ-like A. baumannii prophages encode toxins similar to CTXϕ that might contribute to the virulence of this opportunistic multidrug-resistant pathogen.

IMPORTANCE We observed the emergence of a novel filamentous phage (CRAϕ) from the genome of Acinetobacter baylyi ADP1 during a long-term laboratory evolution experiment. CRAϕ is the first bacteriophage reported to require the molecular machinery involved in the uptake of environmental DNA for infection. Reactivation and evolution of CRAϕ reduced the potential for horizontal transfer of genes via natural transformation in our experiment. Risk of infection by similar phages may limit the expression and maintenance of bacterial competence in nature. The closest studied relative of CRAϕ is the Vibrio cholerae CTXϕ phage. Variants of CRAϕ are found in the genomes of Acinetobacter baumannii strains, and it is possible that phage-encoded toxins contribute to the virulence of this opportunistic multidrug-resistant pathogen.

INTRODUCTION

Prophages, genomically integrated copies of viral DNA, are common in bacteria. They can comprise significant portions of bacterial genomes (1) and critically impact host cells (2). Some prophages are derived from recent infections with temperate phage and are still capable of replicating and reinfecting cell populations if induced stochastically or by environmental triggers. However, many prophages are believed to be cryptic, having accumulated mutations that disrupt their ability to produce infective phage particles. Both intact and cryptic prophages can confer benefits to their hosts, such as by providing proteins that aid in stress responses (3), by selectively killing lysogen-free competitors (4), or by promoting horizontal gene transfer (5). Prophages can also carry virulence factors that are necessary for bacteria to adopt pathogenic lifestyles, as is the case for the CTXϕ prophage of Vibrio cholerae (6).

Naturally competent bacteria are capable of transforming DNA without any artificial chemical or electrical treatments (7). DNA uptake is most often accomplished through the use of a type IV pilus structure that binds and translocates extracellular DNA into the cell (8). The ease of transforming designed DNA sequences into naturally competent species makes them attractive targets for genome-scale engineering. For instance, the naturally competent Gram-negative bacterium Acinetobacter baylyi ADP1 has been proposed as a platform for synthetic biology due to its metabolic versatility and high natural competence (9, 10). Unlike other bacterial species that become competent only under specific environmental conditions or in which only a fraction of cells in a population become competent—such as Vibrio cholerae (11) or Bacillus subtilis (12)—all or most ADP1 cells are constitutively competent during exponential growth (13).

We and others have found that mutants with reduced transformability arise and take over populations of ADP1 during long-term evolution experiments (14, 15). This result was thought to occur due to a strong selective pressure to lose competence, either a direct cost for producing the DNA uptake apparatus or a deleterious effect from active DNA uptake. We report here that, while studying the evolutionary dynamics in these populations in more detail, we discovered that a filamentous prophage in the ADP1 genome had become reactivated during these experiments. This phage inhibits the growth and transformability of ADP1, but it has no effect on strains deleted for genes required for competence. Therefore, we have named it the competence-reducing acinetobacter phage (CRAϕ). The timing with which CRAϕ emerged in the evolution experiment suggests that loss of ADP1 competence was selected because it protects against phage infection.

MATERIALS AND METHODS

Strain and growth conditions.

Strains in this study were derived from an A. baylyi ADP1 evolution experiment (15). Growth conditions were incubation at 30°C with orbital shaking at 140 rpm over a 1-inch diameter in LB (10 g/liter tryptone, 5 g/liter yeast extract, and 10 g/liter NaCl). Population and clone samples to be studied were revived by inoculating 2 μl of a −80°C glycerol stock (20% [vol/vol] glycerol) into 5 ml of LB in a sterile test tube and allowing overnight growth. Then, they were preconditioned by transferring 10 μl of this culture into 10 ml of LB in a sterile 50-ml Erlenmeyer flask and growth for 24 ± 1 h. ADP1 clones from evolved populations were isolated by plating portions of preconditioned populations diluted in sterile saline solution on LB agar (1.6% [wt/vol]) plates, growing the plates overnight at 30°C, and then picking single colonies. ADP1 knockout strains were obtained courtesy of the collection maintained at Genoscope (Évry, France) (16), except that the pilB knockout was constructed through chromosomal replacement of this gene with a Kan^r-tdk cassette using methods described elsewhere (15).

Phage isolation.

Phage genomic DNA was isolated from standard overnight cultures of clone P5-C and ancestor ADP1 by following an M13 DNA isolation protocol (17). Briefly, 1 ml of culture was pelleted at 15,000 × g for 5 min, after which the supernatant was transferred to a new tube where 200 μl of 20% polyethylene glycol (PEG)–2.5 M NaCl was added. After gentle mixing followed by incubation at room temperature for 15 min, phage particles were precipitated by centrifuging the mixture at 15,000 × g for 5 min at 4°C. The supernatant was then removed, leaving a small and visible pellet in the case of the P5-C clone. No pellet was visible for ancestral ADP1. DNA was purified from this pellet by phenol-chloroform extraction followed by ethanol precipitation.

Transformation assays.

Measurements of transformation frequency were conducted as previously described (15). Briefly, we combined 1 ml fresh LB, 100 ng AB-KAN genomic DNA, and 70 μl preconditioned culture in a sterile test tube, incubated the resulting culture for 18 to 24 h under standard growth conditions, and then plated dilutions in sterile saline solution onto selective (kanamycin [Kan]) and nonselective (no antibiotic) agar plates. The ratio of CFU on selective plates to CFU on nonselective plates after overnight incubation at 30°C was used to determine the frequency of transformation of the kanamycin resistance marker into the ADP1 genome.

The ability of a culture's supernatant to inhibit transformation was assayed by combining 500 μl of spent medium from an overnight culture that was sterilized by passing through a 0.22-µm-pore-size filter with 500 μl of fresh LB, 100 ng AB-KAN genomic DNA, and 70 μl preconditioned ancestor ADP1. As before, transformations were incubated for 18 to 24 h before plating. Filter-sterilized supernatant from the ancestral ADP1 strain was used as a control. Spent media were collected from cultures grown for 18 to 24 h under standard conditions which were then pelleted via centrifugation to collect supernatant. Where appropriate, cell-free filtered supernatant was boiled by immersing a 1.7-ml Eppendorf tube in a beaker of boiling water for 15 min and then cooled to room temperature before use.

Growth curves.

To test whether an overnight culture's supernatant inhibited ADP1 growth, 100 μl of a preconditioned ADP1 culture was used to inoculate 9 ml of LB mixed with 1 ml of filter-sterilized spent medium in a 50-ml flask. Each culture was grown under standard conditions for 6 h, removing 1 ml every hour to measure optical density (absorbance) at 600 nm (OD₆₀₀). All conditions were tested in triplicate. Boiled supernatant was prepared as described earlier.

Genome sequencing.

Genome sequencing of A. baylyi ADP1 clones was carried out as previously described (15). Briefly, genomic DNA was isolated from strains of interest and sheared to an average size of ∼550 bp using a Covaris S2 ultrasonicator. Libraries were prepared using a NEBNext prep kit and sequenced on an Illumina HiSeq 2000 instrument at the University of Texas at Austin Genome Sequencing and Analysis Facility (GSAF). Mutations were identified by comparing DNA sequencing reads to the ADP1 genome (GenBank accession no. NC_005966.1) (18) using the breseq computational pipeline (19, 20). The same methodology was used to sequence DNA isolated from phage particles collected from clone P5-C. FASTQ files of DNA sequencing reads have been deposited in the NCBI Sequence Read Archive (see below).

Phage protein analysis.

For electron microscopy and mass spectrometry (MS) analyses, phages were PEG precipitated from a 1-liter culture, as described above, and then further purified to remove cell debris. PEG-precipitated pellets were resuspended in 10 mM Tris·Cl (pH 7.6) with 10 mM MgCl_2. The phages were then purified by the use of a CsCl density gradient as described by Casjens et al. (21). An aliquot of the purified sample was precipitated with trichloroacetic acid (TCA), run on a 15% SDS-PAGE gel, and Coomassie stained. The Research Technology Support Facility (RTSF) at Michigan State University performed digestion of protein samples extracted from gel bands using trypsin and identified proteins based on liquid chromatography-tandem mass spectrometry (LC-MS/MS) matches to unique peptide fragments.

Electron microscopy of negatively stained CRAϕ.

A 3.5-μl aliquot of purified CRAϕ was applied to a continuous carbon support film (Ted Pella, Redding, CA) that had been plasma cleaned in a Fischione model 1020 plasma cleaner. The sample was briefly washed with distilled water and then stained with 1% aqueous uranyl formate. Micrographs were recorded on a DE-20 camera (Direct Electron, LP, San Diego, CA) in a Jeol 2200FS microscope at a nominal magnification of ×30,000 (1.72 Å per pixel) at an acceleration voltage of 200 keV.

Detection of circularized phage episome and mutations in CRAϕ.

Primers used for detecting active phage DNA and for sequencing specific CRAϕ regions are provided in Table S1 in the supplemental material. Briefly, phage episomes were detected by performing PCRs across a junction that is unique to the circularized phage genome. PCRs across the junctions of the genomically located prophage were conducted in parallel to serve as a positive control. PCRs were conducted with Taq polymerase in ThermoPol buffer (New England BioLabs) under standard conditions with purified genomic DNA as the template. Mutations in CRAϕ isolates were detected by Sanger sequencing of PCR products amplified from genomic DNA purified from evolved ADP1 cells. Sanger traces were aligned against the ADP1 genome sequence using Geneious (version 6.1.6) software (22).

Identification of CRAϕ-like prophage in other Acinetobacter genomes.

A microbial TBLASTN analysis of CRAϕ restricted to Acinetobacter genus members with full genomes was performed; homologs that had an E value lower than 1E−5 were marked as present. No significant homologies were found for psh, cep, orfU, ace, ACIAD1851, or ACIAD1850. Phage genetic structure calls indicate whether the proximal arrangements of genes are consistent with the presence of a CRAϕ-like phage genome. Strains with partial phage structures (i.e., a section of CRAϕ but not an identifiable complete phage structure) were excluded from the results.

Accession number(s).

FASTQ files of DNA sequencing reads have been deposited in the NCBI Sequence Read Archive (accession no. SRP074541).

RESULTS

Evolved ADP1 cells secrete a factor that inhibits transformation and growth.

While investigating the results of a long-term laboratory evolution experiment performed with A. baylyi ADP1 (15), we noticed a large discrepancy between the transformability of a population (designated P5) which had been archived after 1,000 generations of evolution and that of clonal isolates from this population. The whole-population transformation frequency was 330-fold lower than the average transformation frequency of 15 randomly selected clones tested individually (Fig. 1A). We observed a similar but weaker trend for a second population (designated P3). Here, the 1,000-generation whole-population sample was 3.6-fold less transformable than expected from taking the average of 15 clone measurements. These discrepancies might mean that the selected clones were not representative of the true mixtures of genotypes in these populations or that rare subpopulations in each mixed sample were reducing the transformability of other cells in the same culture.

FIG 1.

Laboratory-evolved A. baylyi ADP1 strain P5-C produces a factor that inhibits natural transformation and growth of wild-type ADP1 cells. (A) Profiling the transformability of two evolved ADP1 populations (designated P5 and P3) at 1,000 generations unearthed an anomaly showing that transformation of the mixed population was significantly less efficient than expected from averaging measurements of 15 randomly selected clones tested independently. Error bars show estimated 95% confidence intervals. For the ancestral strain (Anc) and mixed-population samples (Pop), averages and confidence limits were calculated from log-transformed replicate measurements (n = 3). For the mixed-population values (Clone Avg), confidence limits on the averages of the clone transformation frequencies were estimated from 1,000 bootstrap resamplings of these 15 measurements. Each of the 15 clone transformation frequencies used in these calculations was itself the average value of log-transformed replicate measurements (n = 3). (B) Adding the cell-free supernatant from one of these 15 clones (P5-C) inhibited transformation of wild-type ADP1, whereas cell-free supernatant from this ancestral ADP1 strain (Anc) had no effect on transformation. Boiling the P5-C supernatant ameliorated this inhibition. Error bars are 95% confidence limits estimated from log-transformed transformation frequency measurements (n = 3). (C) Cell-free supernatant (Sup) from clone P5-C slows the growth rate of wild-type ADP1 cells. Again, cell-free supernatant from ancestral ADP1 had no effect on its own growth, and the inhibitory effect of P5-C supernatant was lost upon boiling.

Pursuing the second possibility, we hypothesized that a transformation-reducing minority subpopulation might secrete a diffusible factor that affected DNA uptake by other cells in the same culture. Therefore, we tested whether sterile-filtered supernatant from each of the 15 clones isolated from P5 was able to inhibit transformation of the ancestral, wild-type ADP1 strain. We found that addition of supernatant from 1 of the 15 clones (designated P5-C) reduced transformation by ∼10-fold (Fig. 1B). The addition of P5-C supernatant also markedly reduced the growth rate of the ancestral strain of ADP1 (Fig. 1C). Boiling the P5-C supernatant completely eliminated its ability to inhibit transformation and growth (Fig. 1B and C), suggesting that the unknown secreted factor either was a heat-labile small molecule or had an essential proteinaceous component.

Evolved clone P5-C produces phage particles.

To identify the inhibitory factor present in P5-C supernatant, we turned to genetic evidence: what mutations had this strain accumulated during the evolution experiment? Sequencing the genomes of P5-C and other noninhibitory clones from the same P5 population revealed several mutations in a putative prophage region that were unique to P5-C. First, we observed a 12.1-fold increase in the read-depth coverage of this region relative to the rest of the genome (Fig. 2A). Individual reads were found that spanned a new DNA sequence junction, not found in the reference genome, connecting the ends of this region to one another in a way that is consistent with either circularization of prophage DNA or the existence of tandem duplications of this prophage in the P5-C genome (Fig. 2B). Three point mutations and a 75-bp deletion were also present in most copies of the prophage region genes in the DNA sample from clone P5-C (Fig. 2B).

FIG 2.

P5-C contains mutations in a putative filamentous prophage that is integrated into the A. baylyi ADP1 genome. (A) Illumina sequencing of DNA isolated from P5-C cells showed increased read-depth coverage of the prophage region relative to the remainder of the genome. Individual reads also spanned a new sequence junction that is consistent with the replication of circular single-stranded DNA (ssDNA) phage genomes by this strain as a result of reactivating the phage. (B) Genes in the ADP1 prophage region are arranged in a manner that is consistent with it encoding a filamentous phage, which we have named the competence-reducing acinetobacter phage (CRAϕ). Putative functions of CRAϕ genes (shown above) were assigned based on homology to V. cholerae CTXϕ (shown below) (6, 24) and the general structure of filamentous phages (23). The CRAϕ prophage genes include putative homologs of genes (cep, ace, and zot) encoding three bacteriophage proteins that act as virulence factors in Vibrio cholerae, but CRAϕ does not encode a homolog of the cholera toxin that is present in CTXϕ (ctxAB). Flanking copies of the repetitive sequence (RS) region containing the rst genes (ACIAD1861 to ACIAD1857 and ACIAD1850 to ACIAD1846) have identical sequences in the wild-type ADP1 genome (indicated by shading in the same color). Mutations in this region in clone P5-C were localized to the RS1 copy. Circularization produces molecules consistent with the connection illustrated with a dashed line.

The organization of this ADP1 prophage region resembles that of a filamentous phage (23), with the greatest similarity in gene content and layout to the Vibrio cholerae CTXϕ phage (6, 24). We therefore hypothesized that the increased DNA copy number of this region and the inhibitory effect of clone 5-C supernatant were due to the production of infective phage particles. We found that a polyethylene glycol (PEG) precipitation procedure designed to purify phage particles yielded a visible pellet from P5-C cell-free supernatant but not from the supernatant of wild-type ADP1. High-throughput sequencing of DNA isolated from the precipitated particles showed that they were further enriched for reads mapping to the prophage relative to the original genomic DNA sample from P5-C cells, with an 84.9-fold-increased read-depth coverage of the phage region versus the rest of the genome. Filamentous phages have circular single-stranded DNA genomes, so all of our results are consistent with clone 5-C having evolved to produce intact viral particles.

To establish that viral particles with a protein capsid were being produced by P-5C, we imaged the phage after further purification via a cesium chloride gradient (Fig. 3A). We observed elongated structures that resembled the morphology of other filamentous phages. To confirm that these structures were derived from the ADP1 prophage region that was amplified in clone P5-C, we separated proteins in the purified phages by SDS-PAGE and identified the proteins in the major bands using mass spectrometry (Fig. 3B). The two most abundant protein bands at <10 kDa both corresponded to the product of the ACIAD1855 prophage gene, which is positioned in the prophage sequence where one would expect to find the major coat protein (pVIII). A much fainter protein band at ∼12 kDa was assigned to the product of the ACIAD1857 prophage gene, which is expected to act as a DNA binding protein (pV).

FIG 3.

Visualization and biochemical analysis of CRAϕ particles. (A) A transmission electron micrograph of purified CRAϕ shows elongated structures that resemble those of other filamentous phages. (B) SDS-PAGE gel of CRAϕ proteins. Three bands were excised and identified as either the putative pVIII major coat protein encoded by ACIAD1855 (dark bands labeled 1 and 2) or the putative pV DNA binding protein encoded by ACIAD1857 (faint band labeled 3) from their trypsin fragments by LC-MS/MS (see Table S2 in the supplemental material).

We refer here to this ADP1 phage as CRAϕ (competence-reducing acinetobacter phage). We were unable to find conditions under which CRAϕ formed visible plaques on lawns of ADP1 cells. This result is not unexpected for a filamentous phage, as these phages bud from the surfaces of live cells rather than lysing their hosts (23). For filamentous phage infections, there is typically still a noticeable fitness cost for cells that are actively infected due to the resource drain of viral replication. The reduction in the growth rate of wild-type ADP1 after exposure to P5-C supernatant is consistent with this sort of fitness cost (Fig. 1C). Thus, we used growth inhibition of ADP1 as a proxy for CRAϕ infection in further experiments.

Organization and evolution of the CRAϕ genome.

The integrated CRAϕ genome is similar in genetic structure to the prophage form of CTXϕ that is present in the genomes of type II V. cholerae strains at the El Tor insertion site (24). On this basis, we were able to assign putative functions to CRAϕ genes (Fig. 2B). Like these CTXϕ prophages, the CRAϕ integrant is flanked by two copies of a repetitive sequence (RS) region containing the rst genes that mediate phage replication (25). In CRAϕ, the upstream RS1 copy and the downstream RS2 copy have identical nucleotide sequences. Both CRAϕ RS regions lack the antirepressor rstC gene found in certain CTXϕ RS regions that act as satellite phages (26). RstC is not necessary for prophage induction, but it can greatly increase CTXϕ production.

The CTXϕ prophage encodes several proteins with roles in the life cycle of a filamentous phage that also act as Vibrio cholerae virulence factors. We found evidence that homologs of some of these genes are also present in the CRAϕ prophage in A. baylyi ADP1. CRAϕ encodes a protein with amino acid sequence similarity to the CTXϕ zonula occludens toxin (zot) (27, 28). On the basis of its genetic organization, CRAϕ also appears to harbor genes similar to those encoding two other CTXϕ virulence factors: the accessory cholera enterotoxin (ace) (29) and the core encoded pilin (cep) (30). CRAϕ does not appear to carry genes with homology to the cholera toxin genes (ctxA and ctxB), which are located downstream of zot in CTXϕ. Zot, Ace, and Cep are thought to fulfill the roles of filamentous phage pI, pVI, and pVIII proteins (Fig. 2B) (6). During viral replication, the pI (Zot) and pVI (Ace) proteins are involved in virion transport and assembly at the cellular membrane, and pVIII (Cep) is the pilin-like major coat protein of the viral capsid.

The mutations found in CRAϕ in P5-C are two nonsynonymous single-base-pair changes in the orfU gene (ACIAD1854), a synonymous base change in the hypothetical ACIAD1848 reading frame, and a 75-bp deletion overlapping rstR (ACIAD1849) and ACIAD1848 in the upstream RS1 copy of the rst region (Fig. 2B). orfU encodes the pIII protein that facilitates filamentous phage binding to the type IV pilus and other host cell receptors (23). Mutations in phage tail-fiber proteins, which likewise mediate host cell recognition and infection, are often found in laboratory evolution experiments when there is coevolution between a lytic virus and its bacterial host (31–33). It is possible that the pIII (OrfU) mutations expand the host range of CRAϕ phage produced by the P5-C clone, such that it has an improved ability to infect the ADP1 variants that had evolved by 1,000 generations in the evolution experiment. As the evolved P5-C CRAϕ phage still inhibits growth and transformation of wild-type (ancestral) ADP1, these pIII mutations do not seem to have restricted its host range.

The 75-bp deletion in P5-C CRAϕ includes the first 16 bp of the RS1 copy of the rstR open reading frame (ORF) (ACIAD1849) and all of the intergenic space between rstR and ACIAD1848. RstR is the phage repressor protein (25), but this disruption of the gene is not expected to affect CRAϕ production because the genomic copy of rstR (ACIAD1860) in the RS2 rst copy is still intact. Instead, the key impact of this mutation is probably in removing RS1 RstR operator sites that control expression of the divergent rstA promoter, as suggested by the regulatory architecture in the V. cholerae CTXϕ prophage (34). Loss of phage repressor binding sites in this manner is known to create “ultravirulent” variants of other temperate phages that are able to superinfect cells that would normally be protected by genomic copies of the repressor protein (35). Indeed, the synonymous point mutation in the hypothetical ACIAD1848 reading frame is also positioned where it could alter the RS1 rstA promoter, regardless of whether ACIAD1848 truly encodes a protein. Overall, the mutations in the P5-C ADP1 clone are consistent with its apparent role as an overproducer of an ultravirulent CRAϕ phage with an expanded host range.

Noncompetent ADP1 mutants are resistant to CRAϕ.

Given the significant inhibitory effect of CRAϕ on the growth rate of the ancestral A. baylyi ADP1 strain, we next investigated whether any bacterial mutations that occurred during our evolution experiment protected cells from infection by reactivated phage. We tested whether P5-C supernatant inhibited the growth of five strains (Fig. 4), each constructed from wild-type ADP1 by deleting one gene that was commonly found to be inactivated by mutations in the evolution experiment (15). Deletion of pilB, comC, or barA made ADP1 completely resistant to growth inhibition by CRAϕ in our assay. Loss of pgi conferred partial resistance. Loss of ACIAD3148 (a gene of unknown function) had no effect on CRAϕ inhibition of growth.

FIG 4.

Knockout of ADP1 genes affecting competence confers resistance to CRAϕ. Growth curves show the effect of sterile-filtered supernatant from the ancestral ADP1 strain (Anc) versus that from the CRAϕ-producing clone (P5-C) on the growth of various single-gene-knockout mutants of wild-type ADP1. Loss-of-function mutations in pilB, comC, barA, pgi, and ACIAD3148 were observed during the original evolution experiment in which reactivated CRAϕ emerged from the genome. The comP knockout was tested because this gene is known to be required for competence. Each strain-supernatant combination was tested in triplicate. Error bars show 95% confidence intervals for OD measurements, but some are obscured by the symbols.

Extracellular DNA uptake by ADP1 requires a competence apparatus that is related to a type IV pilus (8), the same kind of structure that is targeted by other filamentous phage (23). The complete resistance of the pilB and comC deletion strains to CRAϕ suggested that the competence apparatus itself might be the receptor for CRAϕ infection. PilB is thought to function in biogenesis of the competence pilus (18, 36), and ComC appears to be a cell surface protein involved in DNA binding and/or uptake by ADP1 (37). We have previously shown that insertion sequence (IS) element insertions into pilB and comC reduce transformation frequencies by 100- and 1,000-fold, respectively (15). We further tested whether CRAϕ was able to inhibit the growth of a strain deleted for the gene comP, which encodes one of the core pilins in the competence apparatus (38) but was not found to be mutated in our evolution experiment. Deletion of comP renders ADP1 completely noncompetent (39). As expected, the comP deletion mutant was also completely resistant to growth inhibition by CRAϕ (Fig. 4), providing further evidence that the competence apparatus serves as the phage receptor.

In addition to the competence apparatus, A. baylyi ADP1 also expresses a type IV thick pilus that is involved in twitching motility and a type I thin pilus involved in adhesion (40, 41). Strains defective in genes required for competence, including comC and comP, still produce both other types of pili and remain fully motile (37–39). Thus, the type IV pilus involved in competence is believed to function independently from the thick and thin pili in ADP1 (36), and these results seem to rule out any role for these other pili in CRAϕ infection.

Extracellular DNA might act as a competitive inhibitor of CRAϕ binding and protect cells from infection if DNA and CRAϕ utilize the same receptor on the type IV competence pilus. However, we found that the addition of even very high concentrations of Escherichia coli genomic DNA (up to 40 μg/ml), which are expected to saturate the DNA uptake sites on the surface of ADP1 (13), did not alleviate growth inhibition by P5-C supernatant (see Fig. S1 in the supplemental material). This result suggests that DNA binding is not able to block CRAϕ from accessing its receptor and gaining entry into the cell, perhaps because CRAϕ and DNA utilize distinct, nonoverlapping binding sites.

The complete resistance to CRAϕ infection exhibited by the barA deletion strain was unexpected (Fig. 4). BarA operates as a hybrid sensory histidine kinase with the UvrY response regulator. The function of this two-component system has not been characterized in A. baylyi ADP1, but in E. coli and other gammaproteobacteria it activates expression of Csr RNAs and thereby regulates carbon metabolism, motility, adhesion, biofilm formation, virulence genes, and other stress responses (42, 43) upon sensing acetate or formate in the environment (44). We found that deletion of barA reduced ADP1 transformability by ∼10-fold. Thus, the BarA/UvrY system appears to be necessary for fully inducing expression of competence in A. baylyi ADP1, and deletion of barA confers resistance to infection by CRAϕ for this reason.

Deletion of the pgi gene made ADP1 partially resistant to growth inhibition by CRAϕ (Fig. 4). The pgi gene encodes glucose-6-phosphate isomerase, which is involved in extracellular polysaccharide biosynthesis in this strain. Knockout of pgi is also known to cause ADP1 cells to aggregate (15). Changes to the cell surface associated with this mutation, possibly related to the production of polysaccharides that mask the phage receptor (45, 46) or that decrease the surface area accessible for phage adsorption, may lead to this partial resistance phenotype.

CRAϕ emergence and evidence for coevolution in two laboratory populations.

To learn more about the relationship between the phage and competence in the context of the long-term evolution experiment, we assayed CRAϕ emergence and population transformability at 100-generation intervals in two populations (P3 and P5). P3 was included to see if CRAϕ had also reactivated in other populations besides P5. To detect phage, we used a PCR assay specific for the recircularized CRAϕ episome (see Fig. S2 in the supplemental material). While developing this assay, we found that wild-type ADP1 produced a faint band after a sufficient number of PCR cycles, which may indicate that phage is stochastically activated in some cells in a wild-type ADP1 population. In line with this hypothesis, treatment with mitomycin C, a DNA-damaging agent that often induces prophage activation, increased the relative strength of the PCR band that is specific to the circular CRAϕ episome produced by active phage (see Fig. S3).

We detected CRAϕ emergence at as early as 300 generations in population P5 and 500 generations in population P3 (Fig. 5). The presence of the circularized phage episome correlates strongly with reduced population transformability before 600 generations in each population, which is consistent with the presence of infective CRAϕ phage inhibiting transformation. We previously observed that IS1236 insertions in the ADP1 genome that reduce or completely abolish competence arise and eventually become dominant within P3 (15). The timing with which cells with mutations in the competence apparatus reach high frequency in the P3 population (between 600 and 800 generations) is coincident with phage levels dropping below the detection limit of the assay before staging a comeback at 900 generations. In the P5 population, similar dynamics occurred later in the experiment. Noncompetent clones rose to detectable levels in the population by 1,000 generations, and phage also seemingly suffered a setback at 900 to 1,000 generations. Overall, these dynamics are suggestive of a coevolutionary arms race in which CRAϕ emerges and infects the population, resistant ADP1 cells arise and become dominant in the population, and then new varieties of CRAϕ evolve that are able to continue to exploit the evolved bacteria.

FIG 5.

CRAϕ activation correlates with reduced population transformability and the evolution of reduced competence. For two independent populations, P3 (A) and P5 (B), from a long-term evolution experiment (15), we assayed the overall transformability of cells in whole-population samples archived at 100-generation intervals (graphs). Error bars are 95% confidence intervals estimated from log-transformed measurements (n = 3). We further tested these samples for the presence of circularized CRAϕ DNA (phage detected) using a qualitative PCR assay (see Fig. S2 in the supplemental material), except in the cases of two samples for which it was not determined (n.d.). For P3, data on the population structure of competence (representing the percentage of fully competent clones) at 200-generation intervals are from a prior study (15). The value for P5 at 1,000 generations is from the current study. In each case, fully competent clones were defined as those having a transformation frequency of >0.0002, and ≥10 clones selected randomly from the whole-population sample were tested at each time point. −, no PCR band for circularized phage episome; +, weak band; ++, moderate band; +++, strong band.

To look for signatures of an evolutionary arms race and to better understand the genetic basis of prophage reactivation, we isolated CRAϕ-infected cells from archived samples of each population and examined their phage sequences. By screening for small bacterial colonies, we were able to isolate CRAϕ-infected ADP1 clones from population P3 at 500 and 1,000 generations and from population P5 at 300, 400, and 1,000 generations. We sequenced the regions in which we previously found mutations in the P5-C clone in these new CRAϕ isolates: the RS repeats and the orfU-ace-zot region (Table 1). The RS region was amplified with three different sets of PCR primers to distinguish among mutations in the RS1 and RS2 genomic copies and in the RS region contained within circularized phage episomes.

TABLE 1.

Mutations found in CRAϕ-producing clones from P3 and P5^a

Population	Generation	Clone	Positionb	Mutation	Gene(s) affected
P3	500	A	(E) 1850875	Δ54 bp	ACIAD1850
	500	G	(E) 1850875	Δ54 bp	ACIAD1850
	1,000	M	N/A	None	None
P5	300	B	(E) 1850538	Δ75 bp	ACIAD1848, rstR
	300	L	(UE) 1850538	Δ75 bp	ACIAD1848, rstR
	300	M	N/A	None	None
	400	M	(E) 1850538	Δ75 bp	ACIAD1848, rstR
	1,000	C	1850343	G→A (P171L)	orfU
			1850395	C→T (D144N)	orfU
			(UE) 1850503	G→A (F14F)	ACIAD1848
			(UE) 1850538	Δ75 bp	ACIAD1848, rstR

^aMutations in clone C from P5 at 1,000 generations were identified by whole-genome sequencing. Sanger sequencing was used to identify mutations in the rst repeats and in the orfU, ace, and zot open reading frames of other clones.

^bMutations found in the rst region were localized to the upstream genomic copies (U) or to the episomal copies (E) or to both (UE). No mutations were found in the downstream genomic copy. N/A, not applicable.

Most CRAϕ phages in each population have a mutation in the replication region: either the 75-bp deletion overlapping the RS1 rstR and ACIAD1848 genes that was previously found in P5-C in population P5 or a 54-bp deletion overlapping ACIAD1850 in population P3. While the function of the ACIAD1850 gene is unknown, the proximity of the 54-bp deletion in this gene to rstR suggests that it may also affect phage repression, and this could be a parallel example of the evolution of an activated CRAϕ prophage in P3. We observed several cases where there was a putative activating mutation in the rst genes in the phage episome without a corresponding mutation in the genome. In these cases, we apparently found ADP1 cells that became CRAϕ overproducers after superinfection by evolved phage rather than after chromosomal mutations.

The phage isolated at 1,000 generations from a P3 clone did not have any mutations in the regions that we sequenced. Therefore, it must have had an independent evolutionary origin from earlier phages found in this population at 500 generations. In contrast, we found evidence of long-term coexistence and coevolution of CRAϕ in P5. Here, phages present at 1,000 generations were most likely descended from earlier phages detected at 300 and 400 generations, as these isolates all shared the 75-bp RS1 deletion. The pIII (orfU) mutations present in the P-5C clone isolated at 1,000 generations were not found in earlier isolates, further evidence that they appeared in response to the evolution of resistant or partially resistant ADP1 as part of an arms race. Again, we found a clone that had no mutation in the regions that we sequenced at 300 generations in P5. Infection of this ADP1 clone apparently resulted from spontaneous prophage activation, phage mutations outside of the sequenced regions, or mutations elsewhere in the ADP1 chromosome that caused this particular clone to become a CRAϕ producer.

CRAϕ-like prophage are present in pathogenic Acinetobacter baumannii strains.

To determine if CRAϕ was specific to ADP1 or more broadly conserved, we searched for similar filamentous prophage sequences in the genomes of other Acinetobacter species. We did not detect a similar prophage in the genomes of any sequenced Acinetobacter strains that have been demonstrated to be naturally transformable (47). However, we found eight strains of Acinetobacter baumannii with matches to key genes (rstA, zot, orfU, and ace) that enabled us to infer the presence of a CRAϕ-like prophage (Table 2). In contrast to A. baylyi, in which twitching motility and competence rely on distinct pili, twitching motility and competence appear to operate via the same type IV pilus in A. baumannii and other Acinetobacter species, as mutations that disrupt one activity also eliminate the other (48, 49). Thus, it is unclear whether these putative A. baumannii phages infect and operate in a way that is directly analogous to how CRAϕ functions in ADP1.

TABLE 2.

Putative CRAϕ-like prophages in Acinetobacter baumannii genomes

Strain	GenBank accession no.	Position (bp)a	Gene(s) matchedb	Organizationc
AB030	NZ_CP009257.1	3912890	zot	r > ϕ > r >
AB031	NZ_CP009256.1	2286583	zot	r > ϕ > r >
AB0057	NC_011586.1	2103151	zot, ACIAD1848, rstA, rstB	ϕ > r > r > ϕ > r > ϕ > r > ϕ > r > r > ϕ > ϕ > r >
AB5075-UW	NZ_CP008706.1	1989509	zot, ACIAD1848, rstA, rstB	r > ϕ > r < ϕ < r < ϕ < r < ϕ < r < r < r < r < r < r <
AbH12O-A2	NZ_CP009534.1	1913110	zot, rstA	r > ϕ > r > ϕ < r <
AYE	NC_010410.1	1899034	zot, rstR	r < ϕ < r < ϕ < r <
6200	NZ_CP010397.1	3481874	zot, rstA	r > ϕ >
CIP70.10	NZ_LN865143.1	491945	zot, rstA	ϕ < r <
D36	CP012952.1	1990179	zot, ACIAD1848, rstA, rstB	r < r < ϕ < r < r < r < ϕ < r <

^aPosition of either the first phage gene or the RS.

^bGenes with TBLASTN matches with E values of <10⁻⁵ from querying complete Acinetobacter genomes.

^cLetters designate putative prophage (ϕ) and rst repetitive sequence (r) copies based on amino acid sequence matches to specific genes and conservation of gene organization. The orientation of each unit on the top (>) or bottom (<) genomic strand is also indicated.

Most A. baumannii genomes contain most or all of the genes required for competence, but this trait appears to have been lost in parallel in many Acinetobacter lineages very recently (50). In particular, there is evidence that deletions of comP often arise in A. baumannii strains. Recurrent loss of this key competence gene is reminiscent of how reactivation of CRAϕ selected against cells that maintained competence in our evolution experiment; it suggests that loss of competence may evolve in nature to avoid infection by filamentous phages.

Interestingly, all of the CRAϕ-like A. baumannii prophages that we identified are located near genes related to pilus assembly. Several cases of prophage sequences interrupting competence genes in naturally transformable bacteria were highlighted recently (51). In that study, these events were taken as evidence that phage indirectly benefit from targeting these sites to knock out competence: it prevents further transformation of homologous DNA derived from a noninfected genome that could purge the prophage sequence from an infected genome. Our work offers an alternative explanation. If a phage—such as CRAϕ—requires the competence pilus as a receptor, then variants that knock out competence when they integrate into the host genome could gain an advantage because this prevents superinfection by their competitors.

CRAϕ-like prophages and their RS integration remnants often exist in many tandem copies in A. baumannii genomes. Strain AB5075-UW, for instance, has six complete prophage copies, and strain AB0057 has four. The various copies of these integrated CRAϕ-like phages in AB5075-UW and AB0057 differ somewhat in their gene content and orientation, suggesting that they are not genome assembly artifacts (which could have arisen from the inadvertent inclusion of circular phage genomes in the sequenced DNA samples). All of the A. baumannii CRAϕ-like prophages carry genes with sequences that are orthologous to those of the known V. cholerae CTXϕ pathogenicity factor genes zot and ace. As is the case in V. cholerae, these prophage-encoded genes may impact host colonization. This possibility warrants the further study of these prophage proteins in pathogenic A. baumannii strains, to determine if they play a role in virulence.

DISCUSSION

Few Acinetobacter phages have been studied experimentally (52–55), although prophage sequences are common in their genomes (50). Here, we report the discovery of CRAϕ, a filamentous Acinetobacter phage that inhibits transformation and growth of A. baylyi ADP1. Filamentous phages are known to utilize type IV pili as their cell surface receptors (23), and CRAϕ appears to target the ADP1 pilus involved in natural competence. Loss-of-function mutations in ADP1 genes required for DNA uptake via the competence apparatus confer complete resistance to CRAϕ. Mutations affecting extracellular polysaccharide production that increase aggregation of ADP1 cells also result in partial phage resistance. In sum, much of the phenotypic instability previously observed in long-term ADP1 cultures may be driven by the evolution of resistance to CRAϕ infection after the reemergence of this virus from its genome.

We were unable to unambiguously determine what mechanism links the observed reduction in competence with CRAϕ infection. As adding high concentrations of purified DNA was not found to inhibit phage infection, it is unlikely that phage and DNA directly compete for binding to the competence apparatus. One possibility is that CRAϕ infection reduces the expression of competence pili, resulting in fewer competence assemblies per cell and reducing the overall amount of DNA uptake by infected cells. This situation would resemble how E. coli phage M13, which infects via the type IV pilus involved in F-plasmid conjugation, represses expression of this receptor to prevent superinfection (56). If this hypothesis is correct, then the genomically integrated copy of CRAϕ in wild-type ADP1 appears to be incapable of this inhibitory function.

Selection for resistance to costly (though not fatal) infection by CRAϕ creates an evolutionary pressure for ADP1 to lose competence. We observed this outcome in our evolution experiment. Natural populations of bacteria could be under similar selection pressure to acquire mutations that deactivate or delete components of the competence apparatus in order to evade infection by other filamentous phages like CRAϕ. Alternatively, the existence of CRAϕ-like phages could also favor cells that tightly control the expression of their competence genes until circumstances arise in which the potential benefits of DNA uptake (e.g., for recombination and gene acquisition) (57, 58) outweigh the risk of possible exposure to phage infections (such as under stressful conditions or when adapting to a new environment). Thus, the discovery of this type of phage that apparently targets the DNA uptake machinery of bacteria could help to explain both the rarity of constitutive expression of competence genes in naturally transformable species and the apparent widespread loss of competence in many bacterial lineages.

Due to its unusually high natural transformability, ADP1 has been proposed as a next-generation chassis for genome and metabolic engineering (9). The emergence of CRAϕ in multiple experimental populations suggests that it represents a key vulnerability to using ADP1 as a model organism. In order to prevent CRAϕ from essentially nullifying the defining trait of ADP1, namely, its competence, it will need to be deleted from the genome. Thus, this report demonstrates how studying adaptive evolution can be an important tool for guiding strain improvement: learning how strains evolve in undesirable ways can lead to specific strategies to counteract them (59). Another aspect of laboratory evolution that it illustrates is the discovery of new aspects of microbiology that may go unnoticed in experiments with shorter durations. The reemergence of this filamentous phage suggests that similar phages may limit horizontal gene transfer by the uptake of environmental DNA in nature and that CRAϕ-like prophage may contribute virulence factors to A. baumannii in a way that is reminiscent of V. cholerae CTXϕ.