Epigenetic Control of O-Antigen Chain Length: A Tradeoff between Virulence and Bacteriophage Resistance

Summary

A tradeoff can increase the adaptive capacity of an organism at the expense of lowering the fitness conferred by specific traits. This study describes a tradeoff that confers bacteriophage resistance in Salmonella enterica at the expense of reducing its pathogenic capacity. Phase variation of the opvAB operon creates two subpopulations of bacterial cells, each with a distinct lipopolysaccharide structure. One subpopulation is large and virulent but sensitive to phages that use the lipopolysaccharide O-antigen as receptor, while the other is small and avirulent but phage resistant. In the presence of a phage that targets the O-antigen, only the avirulent subpopulation survives. However, phase variation permits resuscitation of the virulent opvABOFF subpopulation as soon as phage challenge ceases. This transient tradeoff may illustrate the adaptive value of epigenetic mechanisms that generate bacterial subpopulations in a reversible manner.

Introduction

The study of differentiation in bacterial species that undergo developmental programs has played a historic role in biology [1,2,3]. In addition, phenotypic differences between colonies [4] and within colonies [5,6] were described many years ago in bacterial species that do not undergo development. Despite their technical limitations, these early studies contributed to bring about the idea that phenotypic heterogeneity might be a common phenomenon in the bacterial world [7]. This view has been confirmed by single cell analysis technologies [8,9,10,11,12]. Furthermore, theoretical analysis has provided evidence that phenotypic heterogeneity can have adaptive value, especially in hostile or changing environments [13,14,15]. In certain cases, the adaptive value of subpopulation formation is illustrated by experimental evidence [16,17,18].

Formation of bacterial lineages is governed by diverse mechanisms, including programmed genetic rearrangement [19] and contraction or expansion of DNA repeats at genome regions [20,21]. In other cases, however, lineage formation is controlled by epigenetic mechanisms: certain cell-to-cell differences serve as physiological signals, and signal propagation by a feedback loop generates an inheritable phenotype [12,22]. Cell-to-cell differences can be a consequence of environmental inputs or result from the noise intrinsical to many cellular processes [10,12,15]. In turn, the feedback loops that propagate the initial state beyond can be relatively simple or involve complex mechanisms like the formation of inheritable DNA adenine methylation patterns in the genome [12,23,24]. Some feedback loops are stable enough to cause bistability, the bifurcation of a bacterial population into two distinct phenotypic states [22].

If a feedback loop is metastable, reversion of the epigenetic state will occur after a certain number of cell divisions. Reversible bistability is usually known as phase variation, and typically involves reversible switching of gene expression from OFF to ON or from low to high expression [25,26,27]. Examples of phase variation have been described mostly in bacterial pathogens, and subpopulation formation is frequently viewed as a strategy that may facilitate evasion of the immune system during infection of animals [25,26]. This view is supported by the observation that phase-variable loci often encode envelope components or proteins involved in modification of the bacterial envelope [25,26].

Some phase-variable envelope modifications controlled by DNA adenine methylation play roles in bacteriophage resistance. For instance, phase variation in the gtrABC1 cluster protects S. enterica against the T5-like phage SPC35, probably by an indirect mechanism [28]. In Haemophilus influenzae, DNA adenine methylation controls phase-variable resistance to bacteriophage HP1c1 but the underlying mechanism remains hypothetical [29]. Phase variation can also contribute to phage resistance without alteration of the bacterial surface. For instance, certain genes encoding restriction-modification systems show phase variation [30,31].

In this study, we describe a phase variation system that confers resistance to bacteriophages that use the lipopolysaccharide (LPS) O-antigen as receptor. The genome of Salmonella enterica contains a horizontally-acquired locus, known as opvAB or STM2209-STM2208 [32]. The opvA and opvB genes form a bicistronic operon [32] and encode inner membrane proteins [32]. OpvA is a small peptide of 34 amino acids, and OpvB is a larger protein of 221 amino acids with homology to the Wzz superfamily of regulators of LPS O-antigen chain length [32]. We show that expression of the S. enterica opvAB operon confers resistance to bacteriophages P22 (Podoviridae), 9NA (Siphoviridae), and Det7 (Myoviridae) by modification of the phage receptor, the LPS O-antigen. Because expression of opvAB is phase-variable, bacteriophage resistance occurs in the subpopulation of opvAB^ON cells only. This subpopulation, which is extremely small, preadapts Salmonella to survive phage challenge albeit at the cost of reduced virulence. However, because the opvAB^ON state is reversible, the virulence payoff is temporary, and a virulent bacterial population resuscitates as soon as phage challenge ceases.

Results

Localization of the OpvA and OpvB proteins in the S. enterica envelope

OpvA and OpvB were previously shown to be inner membrane proteins involved in LPS synthesis [32]. Because the LPS is known to have a helical distribution in the cell envelope [33], the OpvA and OpvB subcellular localization was investigated. For this purpose, a chromosomal opvB::mCherry fusion was constructed downstream of the opvB gene (so that the strain remains OpvAB⁺). In a wild type background, expression of opvB::mCherry was low in most cells (Fig 1A). However, rare cells with high levels of expression of opvB::mCherry were detected (Fig 1A), an observation consistent with the occurrence of phase variation skewed towards the OFF state [32]. Expression of opvB::mCherry was also monitored in an opvAB-constitutive (opvAB^ON) strain engineered by elimination of GATC sites upstream of the opvAB promoter [32]. In an opvAB^ON background, all cells displayed high levels of fluorescence, similar to those of the rare fluorescent cells visualized in a wild type background (Fig 1B). In fluorescent cells, OpvB was seen forming helical intertwining ribbons in the inner membrane (Fig 1A and 1B).

The subcellular distribution of OpvA was examined using a plasmid-borne opvA::mCherry fusion. This experimental choice was based on the consideration that construction of an mCherry fusion in the upstream gene opvA would likely prevent opvB expression because of a polarity effect. In the strain carrying plasmid-borne opvA::mCherry, intense fluorescence was observed in all cells (Fig 1C), presumably because opvA::mCherry overexpression from the multicopy plasmid abolished phase variation. This construction was useful, however, to permit clear-cut observation of helical intertwining ribbons formed by OpvA (Fig 1D). The evidence that OpvA and OpvB may have a similar or identical distribution in the bacterial envelope is consistent with the physical interaction previously described between OpvA and OpvB [32].

Roles of OpvA and OpvB in control of O-antigen chain length

Constitutive expression of opvAB leads to the production of a particular form of O-antigen in the S. enterica LPS, with a modal length of 3–8 repeat units [32]. A diagram of LPS structure is presented in S1 Fig, together with an electrophoretic separation of O-antigen chains and a diagram of the differences in LPS structure between opvAB^OFF and opvAB^ON pubpopulations. To investigate the role of individual OpvA and OpvB proteins in control of O-antigen chain length, non-polar mutations in opvA and opvB were constructed in the wild type and in an opvAB^ON background. In the wild type, lack of either OpvA or OpvB did not alter the electrophoretic profile of LPS (Fig 2), an observation consistent with two known facts: the subpopulation of cells that express opvAB in wild type Salmonella is very small [32], and an OpvAB⁻mutant displays an LPS profile identical to that of the wild type [32]. In contrast, OpvA⁻opvB^ON and opvA^ON OpvB⁻mutants showed differences with the parental opvAB^ON strain and also with the wild type:

• Absence of OpvB (opvA^ON) yielded a seemingly disorganized LPS with no clear modal length (Fig 2), reminiscent of the LPS produced in the absence of the modal length regulators Wzz_ST and Wzz_fepE [34,35,36].

• Absence of OpvA (opvB^ON) yielded an LPS with the modal lengths typically conferred by Wzz_ST and Wzz_fepE [34] but also showed a preferred opvAB^ON-like modal length in the lower weight band region (Fig 2).

These observations suggested that the function of OpvA might be to prevent the formation of normal O-antigen so that OpvB could then impose its preferred modal length. To test this hypothesis, LPS structure was analyzed in an OpvA⁻opvB^ON background in the absence of either Wzz_ST or Wzz_fepE. The results support the view that OpvB needs OpvA to prevent O-antigen formation by customary modal length regulators. In the absence of Wzz_ST, OpvB alone was able to produce an O-antigen similar to that found in the opvAB^ON strain (Fig 2). In contrast, lack of Wzz_fepE did not seem to facilitate OpvB function, suggesting that OpvB may mainly compete with Wzz_ST. This preference may be related to the fact that both Wzz_ST and OpvB convey relatively short preferred modal lengths: 3–8 for OpvB [32] and 16–35 for Wzz_ST [36,37,38] compared with >100 for Wzz_fepE [34].

Selection of opvAB^ON S. enterica cells upon bacteriophage challenge

The LPS O-antigen is a typical receptor for bacteriophages [39] and modification of the O-antigen can confer bacteriophage resistance [40]. On these grounds, we tested whether opvAB expression increased Salmonella resistance to the virulent phages 9NA [41,42] and Det7 [43,44]. We also tested the historic phage P22, using a virulent mutant to avoid lysogeny [45]. Three strains (wild type, opvAB^ON and ΔopvAB) were challenged with 9NA, Det7, and P22, which belong to different bacteriophage families and use the O-antigen as receptor. The experiments shown in Fig 3 were carried out by inoculating an exponential culture of S. enterica with an aliquot of a phage suspension at a multiplicity of infection (MOI) >10, and monitoring bacterial growth afterwards. The results can be summarized as follows:

• Growth of the opvAB^ON strain was not affected by the presence of 9NA, Det7, or P22, suggesting that the strain was resistant to these bacteriophages (S2 Fig).

• A culture of the wild type strain became clear 1–2 hours after P22 infection, suggesting that cell lysis had occurred. However, bacterial growth was observed around 4 hours after infection, and was interpreted as occurrence of P22 resistance. Infection with either 9NA or Det7 did not cause clearing but growth retardation, and growth resumed 4 hours after infection (Fig 3).

• Cultures of the ΔopvAB strain infected with 9NA, Det7, or P22 became clear or almost clear. Growth was detected later, albeit with significant delay compared with the wild type. Because the wild type, the opvAB^ON strain and the ΔopvAB strain show similar or identical growth rates in LB broth (S2 Fig), the explanation of this phenomenon is that growth of the ΔopvAB strain in the presence of 9NA, Det7, or P22 selects phage-resistant mutants (see below).

A tentative interpretation of these observations was that the wild type strain contained a subpopulation of opvAB^ON cells that survived phage challenge. Because opvAB phase variation is skewed towards the OFF state [32], the small size of the opvAB^ON subpopulation and the regular formation of phage-sensitive opvAB^OFF cells caused growth retardation (albeit to different degrees depending on the phage). In contrast, the opvAB^ON strain grew normally, an observation consistent with the occurrence of phage resistance in the entire bacterial population. This interpretation was supported by analysis of the LPS profiles of wild type and opvAB^ON strains grown in the presence of P22, 9NA, and Det7 until stationary phase (OD₆₀₀ ~4) (Fig 3). After phage challenge, the wild type contained an LPS different from the LPS found in LB (Fig 3D), and similar or identical to the LPS found in the opvAB^ON strain (Fig 2; see also [32]). In contrast, the LPS from the opvAB^ON strain did not change upon phage challenge (Fig 3D).

Confirmation that challenge of the wild type with P22, 9NA, and Det7 selected opvAB^ON S. enterica cells was obtained by flow cytometry analysis (Fig 4). Expression of opvAB was monitored using a green fluorescent protein (gfp) fusion constructed downstream opvB (so that the strain remains OpvAB⁺). In the absence of phage, most S. enterica cells expressed opvAB at low levels; however, a small subpopulation that expressed opvAB at high levels was also detected. Phage challenge yielded mostly S. enterica cells with high levels of opvAB expression. These observations provide additional evidence that phages P22, 9NA, and Det7 kill the opvAB^OFF subpopulation, and that opvAB^ON cells overtake the culture.

Reversibility of OpvAB-mediated bacteriophage resistance

If the above model was correct, we reasoned, cessation of phage challenge should permit resuscitation of a phage-sensitive subpopulation as a consequence of opvAB phase variation. This prediction was tested by isolating single colonies from cultures in LB + phage. After removal of phage by streaking on green plates, individual isolates were cultured in LB and re-challenged with P22, 9NA, and Det7 (≥ 20 isolates for each phage). All were phage-sensitive and their LPS profile was identical to that obtained before phage challenge. Representative examples are shown in Fig 5. Unlike the wild type, individual isolates of the ΔopvAB strain remained phage-resistant after single colony isolation and were considered mutants (see below).

Mutational bacteriophage resistance in the absence of OpvAB

Challenge of a ΔopvAB strain with phages P22, 9NA, and Det7 prevented growth for 5–6 h, and growth resumed afterwards (Figs 3 and 5). To investigate the cause(s) of phage resistance in the absence of OpvAB, individual colonies were isolated from stationary cultures of a ΔopvAB strain in LB + P22, LB + 9NA, and LB + Det7. Phage was removed by streaking on green plates. Independent isolates (each from a different culture) were then tested for phage resistance. Sixty seven out of 72 independent isolates turned out to be phage-resistant, thus confirming that they were mutants. Analysis of LPS in independent phage-resistant mutants revealed that a large fraction of such mutants displayed visible LPS anomalies (Fig 6). The few mutant isolates (5/67) that did not show LPS alterations may have LPS alterations that cannot be detected in gels or carry mutations that confer phage resistance by mechanisms unrelated to the LPS. Whatever the case, these experiments support the conclusion that resistance of S. enterica to phages P22, 9NA, and Det7 in the absence of OpvAB is mutational.

To determine whether isolates resistant to one phage were also resistant to other phages that target the O-antigen, cross-resistance was tested by growth in LB upon phage inoculation. Sixty seven mutants (24 P22-resistant, 24 9NA-resistant, and 19 Det7-resistant) were tested (S1 Table). The main conclusions from these experiments were as follows:

• Major alteration of LPS conferred resistance to all three phages.

• More subtle LPS alteration (as observed in 16/24 P22-resistant mutants) conferred incomplete resistance to 9NA but did not confer resistance to Det7.

Effect of opvAB expression on Salmonella enterica virulence

Because the LPS plays roles in the interaction between S. enterica and the animal host [34,46,47], we tested whether OpvAB-mediated alteration of O-antigen chain length affected Salmonella virulence. For this purpose, competitive indexes (CI's) [48] were calculated in the following experiments: (i) oral and intraperitoneal inoculation of BALB/c mice; (ii) infection of mouse macrophages in vitro; and (iii) exposure to guinea pig serum, an assay that provides reductionist assessment of the capacity of the pathogen to survive the bactericidal activity of complement [47]. As controls, CI's were also calculated in LB (Table 1). In all virulence assays, the CI of the opvAB^ON strain was found to be lower than those of the wild type and the ΔopvAB strain. Because the wild type, the opvAB^ON strain and the ΔopvAB strain show similar or identical growth rates in LB, the conclusion from these experiments was that expression of opvAB reduces Salmonella virulence (Table 1).

Discussion

A tradeoff is established whenever the adaptive capacity of an organism is increased at the expense of lowering the fitness conferred by specific phenotypic traits [49]. Tradeoffs have been mainly studied in sexually reproducing organisms but they occur also in microbes [50,51,52,53]. In pathogens, for instance, acquisition of mutational resistance to antimicrobial compounds often affects fitness [54,55], and may require loss of virulence as a payoff [56]. Bacteriophage resistance has been also shown to impair virulence in a variety of bacterial pathogens [57].

In this study, we describe a tradeoff that confers bacteriophage resistance at the expense of reducing virulence in the human pathogen Salmonella enterica. This tradeoff is however unusual because phage resistance is not mutational but epigenetic, and because the phage-resistant, avirulent phenotype is reversible.

The opvAB operon is present in most Salmonella serovars (S2 Table). Its products are inner membrane proteins that form intertwining ribbons (Fig 1) reminiscent of those formed by the LPS in the outer membrane [33]. Synthesis of OpvA and OpvB causes a decrease of long O-antigen chains and an increase of short O-antigen chains in the LPS (Fig 2; see also [32]). Genetic evidence presented in Fig 2 suggests that OpvA may prevent the formation of normal O-antigen, allowing OpvB to compete with the Wzz_ST modal length regulator. A similar phenomenon occurs in Pseudomonas aeruginosa, where the Iap transmembrane peptide encoded by bacteriophage D3 disrupts endogenous O-antigen biosynthesis allowing a phage-encoded O-antigen polymerase to produce a different O-antigen [58]. OpvB confers a predominant modal length of 3–8 units, while the wild type LPS shows modal lenghts of 16–35 units and of >100 units (Fig 2; see also [32]). As a consequence of the dramatic change in LPS structure caused by opvAB expression, S. enterica becomes resistant to bacteriophages 9NA, Det7, and P22 (Figs 3 and 4), an observation consistent with the fact that the O-antigen is the bacterial surface receptor used by these bacteriophages [39,59].

Expression of opvAB undergoes phase variation under the control of DNA adenine methylation and the transcriptional regulator OxyR [32]. Because opvAB phase variation is skewed towards the OFF state [32], S. enterica populations contain a major subpopulation of opvAB^OFF (phage-sensitive) cells and a minor subpopulation of opvAB^ON (phage-resistant) cells. In the presence of a bacteriophage that targets the O-antigen, the opvAB^OFF subpopulation disappears and the opvAB^ON subpopulation is selected (Figs 3 and 4). Hence, the existence of a small subpopulation of phage-resistant cells preadapts S. enterica to survive phage challenge. In OpvAB⁻S. enterica, acquisition of phage resistance is mutational only, and a frequent mechanism is alteration of LPS structure (Fig 6). Because the LPS plays major roles in bacterial physiology including resistance to environmental injuries and host-pathogen interaction [60], opvAB phase variation may have selective value by providing S. enterica with a non-mutational, reversible mechanism of phage resistance. This mechanism offers the additional advantage of protecting Salmonella from multiple phages, perhaps from all phages that bind the O-antigen (note that the phages used in this study belong to three different families: Podoviridae, Siphoviridae, and Myoviridae).

Acquisition of phage resistance in opvAB^ON cells requires a payoff: reduced virulence in both the mouse model and in vitro virulence assays (Table 1). In a phage-free environment, this payoff may be irrelevant because the avirulent subpopulation is minor as a consequence of skewed switching of opvAB toward the OFF state: 4 x 10⁻² for ON→OFF switching vs 6 x 10⁻⁵ for OFF→ON switching [32]. In other words, only 1/1,000 S. enterica cells can be expected to be avirulent in a phage-free environment. The virulence payoff is therefore enforced in the presence of phage only, and its adaptive value may be obvious as it permits survival. On the other hand, the fitness cost of OpvAB-mediated phage resistance can be expected to be temporary because phase variation permits resuscitation of the virulent opvAB^OFF subpopulation as soon as phage challenge ceases (Fig 5). Resuscitation may actually be rapid as a consequence of skewed switching towards the opvAB^OFF state.

Phase variation systems that contribute to bacteriophage resistance have been described previously. For instance, certain restriction-modification systems show phase-variable expression [31]. However, protection by restriction-modification systems can be expected to be incomplete as only a fraction of infecting phage genomes are modified [61]. Phase variation can also confer phage resistance by preventing infection, and an interesting example is the gtrABC1 cluster which protects S. enterica against the T5-like phage SPC35 [28]. Although the receptor of SPC35 is the BtuB vitamin transporter, GtrABC-mediated glycosylation of the LPS O-antigen may reduce SPC35 adsorption by an indirect mechanism [28]. In Haemophilus influenzae, phase-variable resistance to bacteriophage HP1c1 may involve changes in LPS [29]. Because these studies did not investigate the impact of phase variation on bacterial fitness, it remains unknown whether the tradeoff associated with opvAB phase variation is unusual or commonplace. However, if one considers that envelope structures play multiple roles in bacterial physiology aside from serving as phage receptors, it is tempting to predict that phase-variable bacteriophage resistance may frequently involve fitness costs. Whatever the payoff, however, phase-variable resistance may have a crucial advantage over mutation by creating phenotypic heterogeneity in a reversible manner.

Methods

Bacterial strains

Strains of Salmonella enterica used in this study (Table 2) belong to serovar Typhimurium, and derive from the mouse-virulent strain ATCC 14028. For simplicity, S. enterica serovar Typhimurium is routinely abbreviated as S. enterica. For the construction of strain SV7643, a fragment containing the promoterless mCherry gene and the kanamycin resistance cassette was PCR-amplified from pDOC-R, an mCherry-containing derivative of plasmid pDOC [62] using primers HindIII-opvB-mCherry-5 and NdeI-opvB-mCherry-3. The construct was integrated into the chromosome of S. enterica using the Lambda Red recombination system [63]. For the construction of strains SV5675, SV6786, SV6791, and SV8020, targeted gene disruption was achieved using plasmid pKD13 [63] and oligonucleotides listed in S3 Table: wzzB5-PS4 + wzzB3-PS1 for wzz_ST disruption, fepE5-PS4 + fepE3-PS1 for wzz_fepE disruption, STM2209-PS4tris + STM2209-PS1 for opvA disruption, and STM2208-PS4 + STM2208-PS1 for opvB disruption. The kanamycin resistance cassettes were then excised by recombination with plasmid pCP20 [63]. For the construction of strain SV6727, a fragment containing the promoterless green fluorescent protein (gfp) gene and the chloramphenicol resistance cassette was PCR-amplified from pZEP07 [64] using primers STM2208stop-GFP-5 and STM2208stop-GFP-3. The fragment was integrated into the chromosome of S. enterica using the Lambda Red recombination system [63]. An opvB::gfp transcriptional fusion was formed downstream of the opvB stop codon, and the strain remained OpvAB⁺. For the construction of strains SV7645, SV8117, and SV8118, plasmid pKD46 was introduced in SV6401, and the PCR products used for construction of strains SV7643, SV8020 and SV5675 were integrated into the chromosome of SV6401 using the Lambda Red recombination system [63].

Culture media

Bertani's lysogeny broth (LB) was used as standard liquid medium. Solid LB contained agar at 1.5% final concentration. Green plates [65] contained methyl blue (Sigma-Aldrich, St. Louis, MO) instead of aniline blue. Antibiotics were used at the concentrations described previously [66].

Bacteriophages

Bacteriophages 9NA [41,42] and Det7 [43] were kindly provided by Sherwood Casjens, University of Utah, Salt Lake City. Bacteriophage P22 H5 is a virulent derivative of bacteriophage P22 that carries a mutation in the c2 gene [45], and was kindly provided by John R. Roth, University of California, Davis. For simplicity, P22 H5 is abbreviated as P22 throughout the text.

Construction of plasmid pIZ2011

A DNA fragment containing opvA and the native opvAB promoter was PCR-amplified using primers KpnI-opvA-plasmidoGFP-5 and KpnI-opvA-plasmidoGFP-3 (S3 Table). The amplification product was cloned into pDOC-R [62]. The resulting plasmid produces an OpvA-mCherry fusion protein.

Fluorescence microscopy

Bacterial cells from 1.5 ml of an exponential culture in LB at 37°C (OD₆₀₀ ~0.15) were collected by centrifugation, washed in phosphate saline buffer (PBS), and resuspended in 1 ml of the same buffer. Cells were fixed in 4% formaldehyde solution and incubated at room temperature for 30 minutes. Finally, cells were washed, resuspended in PBS buffer, and stored at 4°C. Images were obtained by using an Olympus IX-70 Delta Vision fluorescence microscope equipped with a 100X UPLS Apo objective. Pictures were taken using a CoolSNAP HQ/ICX285 camera and analyzed using ImageJ software (Wayne Rasband, Research Services Branch, National Institute of Mental Health, MD, USA). Z-stacks (optical sections separated by 0.2 μm) of mCherry fluorescence were taken with the same microscope. Maximal intensity projections are shown.

Electrophoretic visualization of LPS profiles

To investigate LPS profiles, bacterial cultures were grown in LB overnight. Bacterial cells were harvested and washed with 0.9% NaCl. The O.D.₆₀₀ of the washed bacterial suspension was measured to calculate cell concentration. A bacterial mass containing about 3 x 10⁸ cells was pelleted by centrifugation. Treatments applied to the bacterial pellet, electrophoresis of crude bacterial extracts, and silver staining procedures were performed as described by Buendia-Claveria et al. [67].

Flow cytometry

Bacterial cultures were grown at 37°C in LB or LB + phage (P22, 9NA, or Det7) until exponential (OD₆₀₀ ~0.3) or stationary phase (OD₆₀₀ ~4). Cells were then diluted in PBS. Data acquisition and analysis were performed using a Cytomics FC500-MPL cytometer (Beckman Coulter, Brea, CA). Data were collected for 100,000 events per sample, and analyzed with CXP and FlowJo 8.7 software.

Bacteriophage challenge

Overnight cultures were diluted 1:100 in 3 ml LB and grown in aeration by shaking at 37°C until they reached an optical density OD₆₀₀ ~0.3. One hundred μl of a bacteriophage lysate (P22 H5, 9NA, or Det7) were added (M.O.I. ≥10), and OD₆₀₀ was subsequently measured at 1 h intervals.

Virulence assays

Eight-week-old female BALB/c mice (Charles River Laboratories, Santa Perpetua de Mogoda, Spain) were inoculated with pairwise combinations of the wild type, an opvAB^ON strain, and a ΔopvAB strain at a 1:1 ratio. Bacterial cultures were previously grown overnight at 37°C in LB without shaking. Oral inoculation was performed by feeding the mice with 25 μl of PBS containing 0.1% lactose and 10⁸ bacterial colony-forming units (CFU). Intraperitoneal inoculation was performed with 10⁴ CFU in 200 μl of PBS. Bacteria were recovered from the spleen and the liver of infected mice at 2 days post-infection (intraperitoneal challenge) or 5 days post-infection (oral challenge). A competitive index (CI) was calculated as described elsewhere [48]. To permit strain discrimination, ATCC 14208 was tagged with trg::MudJ (Km^r), an allele that is neutral for virulence [68]. When necessary, cross-streaking on green plates with P22 H5 was used to discriminate phage-resistant isolates [65]. Infection of cultured J774 mouse macrophages, inoculation of guinea pig serum (Sigma-Aldrich), and calculation of competitive indexes in vitro followed previously described protocols [68]. The Student's t test was used to determine whether the CI's were significant.

Ethics statement

Animal research adhered to the principles mandatory in the European Union, as established in the Legislative Act 86/609 CEE (November 24, 1986) and followed the specific protocols established by the Royal Decree 1201/2005 of the Government of Spain (October 10, 2005). The protocols employed in the study were reviewed by the Comité Ético de Experimentación of the Consejo Superior de Investigaciones Científicas (CSIC), and were approved by the Consejería de Medio Ambiente, Comunidad de Madrid, Spain, on December 12, 2014 (permit number PROEX 257/14).

Supporting Information

Attachment 1

Attachment 2

Attachment 3

Attachment 4

Attachment 5