Exploitation of a new flagellatropic phage of Erwinia for positive selection of bacterial mutants attenuated in plant virulence: towards phage therapy

Authors


  • Present address Department of Life Sciences, Division of Biology, Faculty of Natural Sciences, Imperial College London, South Kensington Campus, London, UK

George P.C. Salmond, Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge, CB2 1QW, UK. E-mail: gpcs@mole.bio.cam.ac.uk

Abstract

Aims:  To isolate and characterize novel bacteriophages for the phytopathogen, Erwinia carotovora ssp. atroseptica (Eca), and to isolate phage-resistant mutants attenuated in virulence.

Methods and Results:  A novel flagellatropic phage was isolated on the potato-rotting bacterial species, Eca, and characterized using electron microscopy and restriction analysis. The phage, named ΦAT1, has an icosahedral head and a long, contractile tail; it belongs to the Myoviridae family. Partial sequencing revealed the presence of genes with homology to those of coliphages T4, T7 and Mu. Phage-resistant transposon mutants of Eca were isolated and studied in vitro for a number of virulence-related phenotypes; only motility was found to be affected. In vivo tuber rotting assays showed that these mutants were attenuated in virulence, presumably because the infection is unable to spread from the initial site of inoculation.

Conclusions:  The Eca flagellum can act as a receptor for ΦAT1 infection, and resistant mutants are enriched for motility and virulence defects.

Significance and Impact of the Study:  ΦAT1 is the first reported flagellatropic phage found to infect Eca and has enabled further study of the virulence of this economically important phytopathogen.

Introduction

Bacteriophages are obligate intracellular parasites of bacteria. Ten times more abundant than bacteria, there are an estimated 1031 tailed phages globally. Consequently, they represent a massive and untapped genetic diversity; newly sequenced phage genes frequently have few homologues in existing databases (Breitbart et al. 2002).

The host range of a phage largely depends on the presence of its receptor at the bacterial cell surface; this is one of several important determinants in phage typing methods of bacterial classification (Pitt and Gaston 1995). Diverse surface molecules can serve as phage receptors. For example, a number of T-even-like phages require OmpA for infection (Skurray et al. 1974), phage T5 requires FhuA (Braun et al. 1973), phage EPS7 requires BtuB (Hong et al. 2008), and phage λ requires LamB (Randall-Hazelbauer and Schwartz 1973). In addition to membrane proteins, a large proportion of phages require lipopolysaccharide (LPS) as their receptor (Petty et al. 2006b, 2007), a fact that reflects the dominance of this molecule on the Gram-negative bacterial surface as well as the absolute requirement for bacteria to synthesize at least a core LPS structure. Another macromolecular structure that is known to bind phages is the flagellum. The first flagellatropic phage was identified in 1936 (Sertic and Boulgakov 1936), which infects a number of enteric strains. Since then, phages of both Gram-negative and Gram-positive bacteria have been shown to require the flagellum, including the coliphage χ (Schade et al. 1967) and PBS1 (Joys 1965), which infects Bacillus subtilis. Additionally, ΦCbK depends on a functional flagellum for maximal infection of Caulobacter crescentus (Bender et al. 1989). Nonetheless, flagellatropic phages remain unusual, and none, to our knowledge, have been reported for the phytopathogen, Erwinia carotovora ssp. atroseptica (Eca).

Eca is a Gram-negative, motile, nonsporulating, facultative anaerobe (Pérombelon 2002), belonging to the same taxonomic family as Escherichia coli, the Enterobacteriaceae. It is an opportunistic pathogen of potato and the main aetiological agent of blackleg and soft rot (Toth et al. 1999) – important causes of crop loss in temperate climates (Pérombelon 2002). There are no chemical control measures available, so understanding its virulence is of crucial importance for devising pest control strategies. The ability to cause disease is largely dependent on the secretion of plant cell wall degrading exoenzymes, which include pectinases, cellulases and proteases (Pérombelon 2002). The production of exoenzymes is cell-density dependent and is regulated by a quorum sensing system (Whitehead et al. 2002). Additional factors that affect virulence are auxotrophy, growth rate, motility and LPS structure (Toth et al. 1999).

Phage resistance, as a positive selection marker for mutant bank screening, is a known tool in the identification of bacterial surface virulence factors. This rationale has been used successfully to isolate reduced virulence mutants of Erwiniae. Schoonejans et al. (1987) observed that some ΦEC2-resistant mutants of Erwinia chrysanthemi, a species related to Eca, had a dramatically altered LPS composition. These strains also showed impaired virulence in the plant host, the African violet (Saintpaulia ionantha). Similarly, LPS has been found to affect bacterial virulence in the spontaneous ΦA5-resistant mutant of Eca strain SCRI1043, A5/22 (Toth et al. 1999). However, unlike mutants isolated by Schoonejans et al. (1987), the mutation in A5/22 is pleiotropic, affecting motility, exoenzyme levels and outer-membrane protein profile in addition to LPS structure (Toth et al. 1999).

Here, we report the first flagellatropic phage known to infect Eca and show that phage-resistant mutants of this commercially important pathogen are attenuated in virulence.

Materials and methods

Bacterial strains, plasmids and growth conditions

Bacterial strains, bacteriophages and plasmids used in this study are listed in Table 1. Escherichia coli and Erwinia strains were grown at 37 and 25°C, respectively. Bacterial growth phase on Luria Broth (LB) contained the following, per litre: 5 g NaCl, 5 g yeast extract and 10 g tryptone. Solid medium was prepared by supplementing LB with 1·6% (w/v) agar. For agar overlays, 0·35% (w/v) agar was used. Minimal medium consisted of phosphate buffer (0·04 mol l−1 K2HPO4, 0·015 mol l−1 KH2PO4, pH 7·0), 0·2% (w/v) (NH4)2SO4, 0·2% (w/v) sucrose, and 0·1 mol l−1 MgSO4. When required, antibiotics were added at the following final concentrations: kanamycin (Kn) 50 μg ml−1, ampicillin 50 ml−1.

Table 1.   Bacterial strains, plasmids and bacteriophages used in this study
 Relevant featuresKey referenceEfficiency of plaquing*
  1. *Efficiency of plaquing is defined as the number of plaques obtained on the test strain divided by the number of plaques obtained on EMS6.1; data refer infectivity of ΦAT1.

Bacterial strains
 Erwinia carotovora ssp. atroseptica
  SCRI1043 Wild-type Eca isolated in UK by SCRIHinton et al. (1989) 
  EMS6.1 SCRI1043, LacHinton et al. (1989)1
  A5/22 SCRI1043, Rvi, ΦA5RToth et al. (1999)0·02
  S32RSCRI1043, rfbIA. Ind, unpublished data0·14
  Eca 3·37SCRI1043 spontaneous SmR; waaJ::CmRT. Evans, unpublished data0·11
  SCC34Undefined mutant, CmRS. Coulthurst, unpublished data2·4
  M231 EMS6.1, fliA::Tn5, KmR, Mot, RviShih (1998)<10−9
  F35 EMS6.1, fliC::Tn5, KmR, MotShih (1998)<10−9
  O22 EMS6.1, flgL::Tn5, KmR, MotShih (1998)<10−9
  XX17 EMS6.1, flgC::Tn5, KmR, MotShih (1998)<10−9
  D414 EMS6.1, hexY::Tn5, KmRShih et al. (1999) 
  ATX1 EMS6.1, waaJ::Tn5, KmR, RviThis study 
  ATX3 EMS6.1, fliF::Tn5, KmR, Mot, ΦAT1R, RviThis study4 × 10−6
  ATX4 EMS6.1, ECA1735::Tn5, KmR, ΦAT1R, RviThis study1 × 10−3
 Escherichia coli
  S17-1 λ-pirrecA, thi, pro, hsdRR+, recA::RP4-2-Tc::Mu-::Tn7, λpir, AmpR, SpR, SmRDe Lorenzo et al. (1990) 
Plasmids
 pBluescript-II KS+colE1 replicon, AmpRStratagene 
 pUTmini-Tn5lacZ1 Delivery plasmid for mini-Tn5lacZ1, AmpR, KmRDe Lorenzo et al. (1990) 
Bacteriophages
 ΦM1 General transducing bacteriophage, LPS-dependentToth et al. (1997) 
 ΦA5 LPS-dependent bacteriophageToth (1991) 
 ΦAT1Flagellatropic phageThis study 
 ΦAT2–ΦAT11 Uncharacterized bacteriophagesThis study 
Primers
 T3ATTAACCCTCACTAAAGGGA  
 T7TAATACGACTCACTATAGGG  
 LL-68TGCATCAATCGCTGCTTTTACCTG  
 LR-68CAGCAACACCTTCTTCACGAGG  
 M13FTGTAAAACGACGGCCAGT  
 M13RCAGGAAACAGCTATGAC  

Phage-related methods

Lysates were prepared as described by Sambrook et al. (1989). Phages were titrated by pouring a top lawn containing 4 ml 0·35% agar, ∼108 bacteria and 10 μl of serially diluted phage. Phage lysates were prepared by adding 4 ml of phage buffer [0·01 mol l−1 Tris–HCl pH 7·4, 0·01 mol l−1 MgSO4, 0·01% (w/v) gelatine] to plates showing confluent lysis and scraping off the top agar. This was vortexed for 2 min in the presence of 400 μl NaHCO3-saturated chloroform and centrifuged at 2200 g for 20 min. The supernatant was stored over a few drops of chloroform. Phage titres were determined by counting plaques on plates where individual plaques could be seen. The number of plaque forming units per ml phage lysate was determined. Generalized transduction was carried out as described by Toth et al. (1997).

Phage isolation

Sewage treatment plant effluent was either filter sterilized (0·2-μm filter, Millipore, Watford, UK) or treated with NaHCO3-saturated chloroform to neutralize bacterial contamination. Enrichment experiments were then set up as follows: 20 ml 2× LB, 20 ml treated effluent and 2 ml overnight A5/22 bacterial culture were mixed and incubated for 5 days at 25°C with shaking. Samples of 1 ml were taken and centrifuged (10 000 g, 10 min). The supernatants were tested for the presence of infectious particles by inoculating agar overlays seeded with Eca SCRI1043. When present, individual plaques were picked and resuspended in 0·2 ml of phage buffer [0·01 mol l−1 Tris–HCl pH 7·4, 0·01 mol l−1 MgSO4, 0·01% (w/v) gelatine].

Electron microscopy

High titre phage lysates were loaded onto charge discharged, carbon-coated copper grids, by allowing the sample to adsorb for 30 s. The grids were then washed three times in sterile deionized water and stained with 1% phosphotungstic acid (pH 7·5) for 1 min. The samples were examined using a Philips CM100 Transmission Electron Microscope.

DNA-related methods

DNeasy Tissue kit, QIAquick gel extraction kit and Qiaex II gel extraction kit (Qiagen Ltd., Crawley, UK) were all used in this study, following the protocols supplied by the manufacturer. Eppendorf Phase Lock Gel DNA purification system was used for ΦAT1 DNA extraction according to the manufacturer’s specifications.

Restriction digests were carried out using the enzymes stated in the text, supplied by New England Biolabs (Beverly, MA, USA), according to the manufacturer’s specifications. For phage DNA digests, 5 μg DNA was digested for 16 h.

Transposon mutagenesis

Transposon mutagenesis of Eca SCRI1043 was carried out essentially as described by De Lorenzo et al. (1990). Briefly, a biparental mating was set up by mixing 1·5 × 109 bacteria of the recipient Eca SCRI1043 and the donor E. coli S17-1 λpir, carrying the pUTmini-Tn5lacZ1 plasmid. The cells were resuspended in 100 μl of LB and spotted directly onto an LB plate. ΦAT1-resistant transconjugants were selected by resuspending the mating patch and mixing with 4 ml 0·35% minimal medium agar containing 50 μg ml−1 Kn and 200 μl ΦAT1 bacteriophage at ∼1010 plaque-forming units per ml.

Single specific primer site PCR

Genomic DNA preparations from the mutant strains were made using the DNeasy kit and according to the supplied protocol. Genomic DNA was then digested with XbaI and HindIII (New England Biolabs). The 5′ phosphate groups were removed using Shrimp Alkaline Phosphotase (USB, Cleveland, OH, USA). The fragments obtained were ligated into XbaI–HindIII-digested pBlueScript-II KS+, using T4 DNA ligase (Invitrogen, Carlsbad, CA, USA).

The construct was used as a template in PCR experiments, as described by Shyamala and Ames (1989), using T7 and one of the following primers: LL-68, LR-68 or T3. Reactions were carried out using the Long Template System (Roche, Basel, Switzerland). PCR products were separated by 1% agarose gel electrophoresis. The most prominent bands were excised, and the DNA purified by using the QIAquick gel extraction kit. These samples were sequenced at the DNA Sequencing facility, Department of Biochemistry, University of Cambridge.

Adsorption assays

In a shaking water bath (25°C, 250 rev min−1), 25 ml of bacterial culture was grown to early stationary phase (OD = 1·4–2·0), and ΦAT1 was added to give a multiplicity of infection of 1. One millilitre samples was taken after 5, 10, 25, 55 and 105 min, and 20 μl of NaHCO3-saturated chloroform was added. The samples were vortexed and centrifuged (10 000 g, 10 min). The supernatant was titrated as described by Sambrook et al. (1989) to determine the number of unadsorbed phage particles in the culture supernatant.

Motility assays

Swimming motility was assayed using tryptone swarm agar [0·3% (w/v) Bacto agar, 1% (w/v) Bacto tryptone and 0·5% (w/v) NaCl]. Four microlitres of an overnight bacterial culture diluted to an OD600 of 0·2 was spotted onto the agar. The diameter of the halo was measured following overnight incubation at 25°C.

Potato tuber virulence assays

Potato virulence assays were performed as described by Walker et al. (1994). Briefly, tubers (Maris Piper cultivar) were washed with distilled water and immersed for 30 min in 5% chloros (disinfectant), rinsed in distilled water and air-dried. Three uniform bores were made in each tuber using a sterile Gilson pipette tip (0–200 μl). These were inoculated with 102 bacteria and sealed with silica grease. The tubers were wrapped in alternate layers of damp tissue paper and cling-film and incubated at 25°C. The amount of macerated tissue was weighed at 24-h intervals.

Phage genome sequencing

Two micrograms of phage DNA was digested overnight using, separately, the following restriction enzymes: EcoRV, HincII and SmaI. Reactions were cleaned up using the PCR clean up protocol of the QIAquick gel extraction kit. Fragments were subsequently ligated into pBlueScript-II KS+, which had been linearized using the corresponding restriction enzyme. Recombinant plasmids were recovered using QIAprep Spin Miniprep kit (Qiagen Ltd.). ΦAT1 genomic inserts were sequenced using primers M13F and M13R at the DNA Sequencing facility, Department of Biochemistry, University of Cambridge.

Results

Isolation and initial characterization of the novel phage, ΦAT1

Novel phages of Eca were isolated by incubating effluent from a sewage treatment plant with a bait culture. We have found that the majority of phages isolated in our laboratory use LPS as their cell surface receptor (Petty et al. 2006b;Petty et al. 2007; T.J. Evans, unpublished data; T.R. Blower, unpublished data), probably because this is the dominant surface molecule, with 106 molecules covering up to 50% of the bacterial outer membrane in Salmonella typhimurium (Smit et al. 1975). Therefore, a pleiotropic LPS mutant of Eca, A5/22 (Toth et al. 1999), was used as bait in these enrichment experiments to selectively enrich for non-LPS-binding phages. Eleven phages were isolated, and all were capable of infecting A5/22 as well as the wild-type strain, Eca SCRI1043. Of the 11 phages isolated, ΦAT1 was arbitrarily chosen for further study.

Transmission electron microscopy studies revealed that ΦAT1 was a tailed phage, composed of an icosahedral head with a diameter of 90 nm; a 6-nm-long neck; and a 95-nm-long tail, of which the sheath, when contracted, measured 55 nm and the exposed tail core measured 42 nm (Fig. 1). The outer sheath of the tail was contracted in most phage particles. The overall length of the virion is approximately 191 nm.

Figure 1.

 Transmission electron micrograph of ΦAT1 negatively stained with phosphotungstic acid. Scale bar: 500 nm.

ΦAT1 genomic DNA was purified and found to be susceptible to digestion by a variety of restriction endonucleases. By determining the size of EcoRV fragments, the size of the genome was estimated to be approximately 90 kb (data not shown). Given the morphological features of the phage and the fact that it has a double-stranded DNA genome, the phage was classified as a member of the Myoviridae family.

ΦAT1 is flagellatropic

To identify the receptor for ΦAT1, phage-resistant mutants were isolated by transposon mutagenesis; ΦM1-mediated trasduction of these mutations into the wild-type strain demonstrated that the ΦAT1-resistant phenotype was co-inherited with antibiotic resistance, thus creating strains ATX3 and ATX4 (see Table 1). The mutations were mapped using a single specific primer site PCR protocol, followed by sequencing of the resultant amplicon. Because Eca strain SCRI1043 has been genomically sequenced (Bell et al. 2004), it was possible to identify the gene disrupted by the transposon insertion, as well as the genomic context of that gene.

During the screening, we identified two mutants with insertions in the genetic region containing genes primarily involved in motility and chemotaxis. In one case, the gene encoding for the major component of the M ring –fliF– was disrupted. ECA1735, a gene of unknown function located 10 kb downstream of fliF, was found to be mutated in another mutant. These observations pointed to the flagellum being the receptor for ΦAT1; we therefore tested the flagellar mutants that had been isolated in separate studies (strains M231, F35, O22, XX17; Shih 1998) for their ability to sustain ΦAT1 growth. All were found to be resistant to phage infection, despite not having been previously exposed to the phage (Table 1).

To test further the hypothesis that ΦAT1 is flagellatropic, 21 spontaneous ΦAT1-resistant mutants were isolated and assessed for motility on tryptone swarm agar. None of these mutants were motile (data not shown), presumably because of the defects in flagella assembly. As a result of these 21 spontaneous mutants remaining sensitive to infection by the well-characterized transducing phage ΦM1, which requires LPS for infection, it can be assumed that gross LPS defects had not occurred. The correlation between phage resistance and lack of motility provides further evidence that ΦAT1 requires the flagellum for adsorption to the bacteria.

It has been shown that the flagellatropic coliphage χ cannot adsorb to nonflagellated mutants (Schade et al. 1967). The corollary of this is that phage adsorption to a hyperflagellated strain is expected to be increased. Therefore, the ability of ΦAT1 to adsorb to wild-type Eca, the hyperflagellated mutant, D414 (Shih et al. 1999), and the nonflagellated mutant, M231 (Shih 1998) was investigated. As shown in Fig. 2, 99·9% of all ΦAT1 particles had adsorbed to the hyperflagellated mutant within 5 min of mixing, compared to 81·5% of phage in the wild type. This represents a greater degree of adsorption that also occured more rapidly. No adsorption at all was observed to the nonflagellated mutant. The number of free phages in suspension was found to increase between 25 and 55 min postinfection, probably because of a lytic burst, suggesting that the viral life cycle is completed within this time frame.

Figure 2.

 Adsorption of ΦAT1 to ( • ) SCRI1043; ( ▮ ) the nonflagellated mutant, M231 and ( bsl00066 ) the hyperflagellated mutant D414.

Partial sequencing of ΦAT1

To partially sequence this phage, fragments from restriction digests were ligated into a cloning vector and sequenced. In total, about 16 kb were sequenced, representing nearly one-fifth of the entire genome. A number of nucleotide blast searches of some of the sequences revealed homology to the E. coli 0157-specific phage, rv5. This phage has been sequenced, but an analysis of this genome has not yet been published. However, database searches for many of the nucleotide sequences did not find homologues, a common feature of phage genes (Breitbart et al. 2002). Consequently, the sequence data that we obtained were translated in all six reading frames, and the amino acid sequences of putative ORFs were analysed with fugue– a bioinformatic tool that recognizes distant homologues by sequence–structure comparisons (Shi et al. 2001). The results are summarized in Table 2 and include a number of hits to structural proteins and enzymes involved in DNA replication in the coliphages T4, T7 and Mu; an endonuclease with homology to one found in Φsp01, a temperate phage that infects B. subtilis; a peptidoglycan hydrolase; and a class III ribonucleotide reductase. In addition to these hits, eight further unique sequences failed to return any significant hits.

Table 2.   Partial sequence data of ΦAT1
CloneAccession no.HitSpeciesZ-value*
  1. Eight additional clones provided sequences for which no homologues were identified by fugue analysis; these sequence data have been submitted to the DDBJ/EMBL/GenBank databases under accession numbers FN396594, FN396595, FN396596, FN396597, FN396598, FN396599, FN396600 and FN396601.

  2. LPS, lipopolysaccharide.

  3. *A Z-value of ≥6 denotes certain homology (99% confidence); ≥4, likely (95% confidence); ≥3·5, marginal (90% confidence); >2, a guess (50% confidence).

 1FN396583Hnh homing endonuclease I –HmuIBacteriophage sp0117·8
 2FN396584Putative ATPaseCytophaga hutchinsonii ATCC 3340417·0
RuvBThermus thermophilus hb84·8
 3FN396585Central hub of baseplate; phage Mu gene product 44Coliphage Mu7·4
 4FN396586Ale-1 cell wall glycylglycine endopeptidaseStaphylococcus capitis4·4
 5FN396587DnaB monomerThermus aquaticus31·8
Helicase domain of T7 gene product 4Coliphage T725·8
RadB recombinaseThermococcus kodakaraensis8·9
RepA replicative helicasePlasmid rsf10108·8
 6FN396588Capsid proteinRous sarcoma virus3·6
Phage T4 short fibreColiphage T410·0
Phage T4 gene product 10, C-terminal domainColiphage T44·1
 7FN396589Ale-1 cell wall peptidoglycan hydrolaseStaph. capitis4·2
Putative methyltransferaseThermoplasma acidophilum4·4
 8FN396590Phage proteinBacillus cereus ATCC 145793·9
 9FN396591Endonuclease VIIColiphage T48·6
10FN396592Anaerobic ribonucleotide reductase NrdDColiphage T428·7
11FN396593Alpha subunit of replicative DNA polymerase IIIEscherichia coli4·6

Further characterization of ΦAT1-resistant mutants

In an attempt to determine whether ΦAT1 can be used to isolate Eca mutants with an attenuated virulence phenotype, phage-resistant transposon mutants were selected. Two phage-resistant mutants were examined in detail. These were named ATX3 and ATX4 (see Table 1 for genotypes). In particular, we tested the ability of these mutants to cause rot in potato tubers. As shown in Fig. 3a, ATX3, a fliF mutant, was significantly attenuated in virulence (P < 0·05, paired t-test), producing less than half the rot of the wild type. In contrast to ATX3, which was found to be nonmotile, ATX4 remained motile. Whilst ATX4 was isolated as being phage resistant, on more detailed investigation, it was found that this strain was capable of supporting ΦAT1 growth, but with a reduced efficiency of plaquing of 1 × 10−3 (Table 1). This mutant was also attenuated in the in vivo virulence assay (Fig. 3b), resulting in 34% less rot than the wild type (P < 0·05, paired t-test).

Figure 3.

 Tuber rot assay comparing virulence of SCRI1043 to ΦAT1-resistant mutants after 48 (a) and 72 h (b). Bars represent the mean of 10 replicates ± standard error of the mean. Reduction in rot was statistically significant in all cases (P < 0·05, paired t-test).

The LPS mutant ATX1 was also isolated in this study. Whilst this strain is not resistant to ΦAT1, it was included in the tuber virulence assays as a control, because other LPS mutants have previously been shown to result in decreased virulence (Toth et al. 1999). A statistically significant reduction in maceration of the tuber was indeed seen with this mutant (P < 0·05, paired t-test), and this decreased amount of rot was comparable to the values obtained for the phage-resistant mutants (Fig. 3).

Whilst in vivo virulence was found to be reduced in these mutants, an in vitro analysis found there to be no discernable difference in the production of the plant cell wall degrading exoenzymes, cellulase, gelatinase and pectate lyase, between these mutants and the wild type (data not shown). This demonstrates that the reduction in virulence is not mediated through altered exoenzyme production.

Discussion

In this study, a novel phage of the Myoviridae family was isolated. The majority of phages isolated in this laboratory require LPS as their cellular receptor (Petty et al. 2006a,b, 2007). By using an LPS mutant of Eca, we were able to isolate the first flagellatropic phage, ΦAT1, known to infect Eca.

A variety of lines of evidence support the hypothesis that ΦAT1 is flagellatropic. First, nonmotile flagella mutants were phage resistant, and phage-resistant mutants were nonmotile. Second, ΦAT1 was not able to adsorb to the nonflagellated mutant. In contrast, increased adsorption to a hyperflagellated strain was seen.

A number of Eca strains defective in different stages of flagella biosynthesis are available (Shih 1998; Shih et al. 1999). Comparing the infectivity of ΦAT1 on these strains (Table 1) allows speculations to be made regarding the mode of ΦAT1 infection. In enterobacteria, flagellar biosynthesis is coupled to its assembly. A fully formed hook–basal body structure, an intermediate in flagella assembly, is required for the expression of flagellin and several genes involved in chemotaxis (Chilcott and Hughes 2000). As shown in Table 1, a flagellin mutant (F35; fliC), despite having a fully formed hook–basal body complex, demonstrated the same degree of resistance as mutants without any flagellar gene expression (M231; fliA). From this observation, it can be inferred that flagellin could be the site of viral adsorption and that ΦAT1 does not bind to the basal body in a way that would allow infection. If true, it is possible that after binding, ΦAT1 would inject its DNA into the flagellum. This appears to be in contrast to what is known about the mechanism of infection of the flagellotropic coliphage χ, where attachment to the active flagellar filament is an intermediate step, improving the efficiency with which the phage binds to its ultimate receptor – the basal body (Schade et al. 1967). Nonethless, a χ-like mechanism of phage infection may still be at play here, but if so, direct attachment of the phage to the cell in the fliC mutant may simply be too inefficient to be observed under the conditions used. It is not clear why ATX3 (fliF) is phage sensitive – albeit at a very low efficiency of plaquing of 10−6 (Table 1) – because in other enteric bacteria, fliF mutants lack flagella (Kihara et al. 2001).

The amount of sequence data available for phage is much lower than that for bacterial genomes. A 2007 analysis showed that 403 phage genomes were available on the NCBI website, totaling 17·5 Mb (Lima-Mendez et al. 2007). This is equivalent to fewer than four E. coli genomes. This, in part, reflects the relative interests in general bacteriology compared with phage biology and because the genomes of some phages are difficult to sequence because of highly modified genomes.

A bioinformatic analysis of the sequence data obtained in this study suggests that many ΦAT1 proteins are homologues of the well-characterized phages T4 and T7, such as short tail fibres and a helicase that acts at the replication fork. The baseplate protein encoded by ΦAT1 is homologous to that of the temperate phage Mu. These three phages infect E. coli, which is in the enterobacterial family.

ΦAT1 appears to encode an endonuclease with homology to one encoded by the temperate phage, Φsp01, a phage known to lysogenize the Gram-positive species, B. subtilis (Landthaler et al. 2004). This endonuclease catalyses the movement of an intron from one gene to another, and the presence of this gene in ΦAT1 suggests that introns are present in this genome.

A putative ORF was identified, which has homology to Ale-1 – a peptidoglycan hydrolase encoded by the Gram-positive bacterium, Staphylococcus capitis. Ale-1 produced by Staph. capitis has a bacteriolytic activity on Staphylococcus aureus (Sugai et al. 1997), which is proposed to be an anti-competition strategy. In the phage context, this endopeptidase is likely to be responsible for cell lysis following viral replication. Phages are mediators of horizontal gene transfer and contribute to the mobile gene pool. Interestingly, ale-1 in Staph. capitis is plasmid encoded and therefore might be moved by phage-mediated generalized transduction.

Finally, a putative ΦAT1 protein was identified with homology to the large subunit of Aeromonas phage 65 anaerobic NTP reductase, NrdD. In E. coli, NrdD and NrdG comprise the class III nucleotide reductase, which supplies dNTP substrates for DNA replication under certain conditions. Phage T4 also encodes NrdD (formerly SunY), which notably contains an intron. Infection of E. coli by phage T4 under anaerobic conditions resulted in strong expression of this gene, presumably permitting rapid DNA synthesis prior to lysis (Young et al. 1994).

Apart from a general interest in phages themselves, an aim of this work was to isolate phage-resistant mutants that may be attenuated in virulence, to identify novel virulence determinants. Such a strategy has been used successfully before (see Introduction). In the current study, mutants ATX3 and ATX4 were tested for their ability to rot potato tubers. Virulence of both was found to be attenuated, compared to the wild type.

In vitro analysis of the nonmotile strains ATX3 and ATX4 demonstrated that they are not deficient in the secretion of the cellulase, gelatinase or pectate lyase (data not shown). These enzyme virulence factors (Toth and Birch 2005) are responsible for the destruction of the plant cell well and the characteristic soft rot pathogenicity of this species. The reduction in tuber rot caused by these mutants therefore works via a different mechanism(s). The lack of motility may itself explain altered virulence because motility was shown to be necessary for full virulence by Mulholland et al. (1993), who isolated reduced virulence (Rvi) mutants that were nonmotile, although the possibility remains that other processes are also involved.

The function of ECA1735, disrupted in ATX4, is unknown (Bell et al. 2004), as is the function of its homologues XAC1689 and Rv1502 found in Xanthomonas axonopodis and Mycobacterium tuberculosis, respectively. ECA1735 activity is probably intracellular as it does not possess any known export signals and does not appear to have any transmembrane regions. Therefore, to gain an understanding of the function of ECA1735, its amino acid sequence was used in a fugue search (Shi et al. 2001). By this method, we identified potential structural similarities between ECA1735 and enzymes involved in carbohydrate metabolism, including β-fructosidase from Thermotoga maritima and levansucrase from Gluconacetobacter diazotrophicus. This finding suggests that ECA1735 may have a role in the synthesis or modification of carbohydrates, such as sugar moieties that act as a coreceptor for ΦAT1. Alternatively, this alteration in sugar metabolism may affect LPS structure, which can result in altered phage sensitivity. ECA1735 exists in a putative operon, with a further three genes downstream. Consequently, disruption of ECA1735 may ablate expression of ECA1736, ECA1737 and ECA1738. ECA1736 is a homologue of ECA1735 (40·9% identical and 55·7% similar over 319 amino acids), and a fugue analysis returned the same hits as ECA1735. ECA1737 and ECA1738 appear to encode an acyltransferase and a methyltransferase, respectively, and are involved in chemotaxis. In the case of phage χ infection of E. coli, not only must a flagellum be present, it must also be active. If mutation of ECA1735 prevents chemotactic responses and thus reduces motility, this may explain the reduced efficiency of plaquing seen in this mutant.

It was reported recently that motility of another important phytopathogen, Dickeya dadantii 3937 (formerly Erw. chrysanthemi 3937), is important for full virulence (Antúnez-Lamas et al. 2009). A number of motility and chemotaxis mutants were isolated; these were all attenuated in virulence, and the degree of attenuation corresponded to the decrease in motility. In vivo assays were carried out in four different plant species, demonstrating that motility is important at various stages in the infection process.

Because there are currently no effective methods for treating Eca infections, the use of virulent phages, such as ΦAT1, could be an attractive possibility. The possibility of phage therapy in plants and animals is of growing interest, and several phage-based products for tackling bacterial contamination have recently become available (discussed in Petty et al. 2006a). One criticism of the use of phages in the treatment of disease is that phage-resistant mutants may appear, permitting an uncontrolled spread of disease. However, in the case of ΦAT1, phage-resistant mutants are markedly impaired for virulence, because ΦAT1 is dependent on the flagellum for infection, and motility is important for full virulence. Thus, should phage-resistant Eca strains arise, disease progression should be stunted. Indeed, the same phenomenon was observed in mice, calves, piglets and lambs, when they were experimentally infected with E. coli (Smith and Huggins 1982, 1983). In the case of the mouse model, experimental infection used a strain of E. coli that had been isolated from the brain of a child suffering from meningitis. The K1 antigen, which has been shown to be important for invasiveness, is expressed by this strain. The phage used to treat the infection targeted this antigen, and phage-resistant mutants (which did arise in small numbers) were all found to be K1 and were almost completely avirulent. Thus, there are clear and interesting parallels in the treatment of plant and animal diseases using phage.

Acknowledgements

We thank Ian Foulds for help with phage isolation. This work was supported by the Biotechnology and Biological Sciences Research Council and performed under the Department for the Environment, Food and Rural Affairs plant health license. T.J.E. was supported by a Collaborative Award in Science and Engineering studentship from Leatherhead Food International.

Ancillary