OxyR-dependent formation of DNA methylation patterns in OpvAB^OFF and OpvAB^ON cell lineages of Salmonella enterica Open Access

Abstract

Phase variation of the Salmonella enterica opvAB operon generates a bacterial lineage with standard lipopolysaccharide structure (OpvAB^OFF) and a lineage with shorter O-antigen chains (OpvAB^ON). Regulation of OpvAB lineage formation is transcriptional, and is controlled by the LysR-type factor OxyR and by DNA adenine methylation. The opvAB regulatory region contains four sites for OxyR binding (OBS_A-D), and four methylatable GATC motifs (GATC_1–4). OpvAB^OFF and OpvAB^ON cell lineages display opposite DNA methylation patterns in the opvAB regulatory region: (i) in the OpvAB^OFF state, GATC₁ and GATC₃ are non-methylated, whereas GATC₂ and GATC₄ are methylated; (ii) in the OpvAB^ON state, GATC₂ and GATC₄ are non-methylated, whereas GATC₁ and GATC₃ are methylated. We provide evidence that such DNA methylation patterns are generated by OxyR binding. The higher stability of the OpvAB^OFF lineage may be caused by binding of OxyR to sites that are identical to the consensus (OBS_A and OBS_c), while the sites bound by OxyR in OpvAB^ON cells (OBS_B and OBS_D) are not. In support of this view, amelioration of either OBS_B or OBS_D locks the system in the ON state. We also show that the GATC-binding protein SeqA and the nucleoid protein HU are ancillary factors in opvAB control.

INTRODUCTION

For decades, bacteriological research was based on the study of large populations of bacterial cells in batch cultures. This experimental approach assumed that the value of any parameter measured in the population would reflect a unimodal distribution around the average value in individual cells. This may be true for many cellular parameters. However, in the last two decades single cell analysis has shown that clonal populations of bacteria, even when growing in homogeneous environments, can exhibit phenotypic heterogeneity between individual cells (1–3). In certain cases, phenotypic heterogeneity reflects the occurrence of bistability, the formation of two subpopulations with distinct patterns of gene expression (4,5). Phenotypic diversity can be also generated by reversible ON-OFF switching of gene expression at high frequencies, a phenomenon known as phase variation (6,7). In bacterial pathogens, phase variation often occurs at loci that encode envelope structures and may be viewed as a strategy to generate programmed polymorphism (6,7). Indeed, lineage formation can help to evade the host immune system and to protect bacterial subpopulations against bacteriophage infection, among other potential adaptive advantages (7).

The molecular mechanisms of phase variation are diverse. Some are genetic, such as site-specific recombination (8) and slipped-strand mispairing in tracts of repetitive DNA sequences (9). In other cases, however, the formation of bacterial lineages has epigenetic origin, without alteration of the DNA sequence (10–12). Some of the best known examples of epigenetic phase variation involve the formation of heritable DNA adenine (Dam) methylation patterns (10–12). The list includes the pap operon of uropathogenic Escherichia coli, which encodes fimbrial adhesins for adherence to the urinary tract epithelium (10,13), the agn43 aggregation gene of E. coli (14) and the glycosyltransferase operon gtr of Salmonella enterica (15). In all these cases, a transcriptional regulator binds a regulatory region that contains GATC sites, which become non-methylated because binding of the regulator hinders Dam methylase activity. The OFF and ON states of the phase variation locus thus differ in the methylation state of critical GATC sites (12). Because DNA base methylation often prevents or restrains binding of proteins to DNA, non-methylation can increase binding of the regulatory protein, thus generating a positive feedback loop that propagates the epigenetic state (12). However, in all phase variation systems both the OFF and ON states are metastable, which permits phenotypic switching after a number of generations (6,7). The switching frequencies are idiosyncratic for each phase variation locus, and may vary depending on culture conditions (12).

The opvAB locus of S. enterica serovar Typhimurium, previously annotated as STM2209-STM2208, is a Salmonella-specific locus that encodes cytoplasmic membrane proteins involved in control of O-antigen chain length (16). The opvA and opvB genes form a bicistronic transcriptional unit, which is transcribed from a canonical, σ⁷⁰-dependent promoter under the control of the LysR-type factor OxyR (16). Expression of opvAB is phase variable, and S. enterica batch cultures contain subpopulations of OpvAB^OFF and OpvAB^ON cells. Each subpopulation harbors a distinct type of O-antigen, and OpvAB-mediated modification renders OpvAB^ON cells resistant to bacteriophages that use the O-antigen as receptor (17). However, the OpvAB^ON subpopulation shows sensitivity to serum, reduced capacity to proliferate in macrophages and attenuation in the mouse model (16,17).

In this work we describe the epigenetic mechanism responsible for the formation of OpvAB^OFF and OpvAB^ON cell lineages in S. enterica. Each lineage shows a distinct pattern of GATC methylation at the opvAB regulatory region. We present evidence that such patterns are generated by differential OxyR binding at the opvAB regulatory region. We also show that the GATC-binding protein SeqA and the nucleoid protein HU are ancillary factors for opvAB lineage formation. Finally, we present a model of opvAB phase variation, partly based upon experimental evidence, partly inspired by literature data and containing some speculative elements as well.

MATERIALS AND METHODS

Bacterial strains, bacteriophage, media and culture conditions

The strains of S. enterica used in this study (Supplementary Table S1) belong to serovar Typhimurium, and originate from strain ATCC 14028. For simplicity, S. enterica serovar Typhimurium is routinely abbreviated as S. enterica. E. coli CC118 λ pir [phoA20 thi-1 rspE rpoB argE(Am) recA1 (λ pir)] and E. coli S17–1 λ pir [recA pro hsdR RP4–2-Tc::Mu-Km::Tn7 (λ pir)] were used for directed construction of point mutations. E. coli M15 [pREP4] (Qiagen, Valencia, CA, USA) was used for 6×His-OxyR^C199S production. Plasmid pTP166 (18) was kindly provided by Martin G. Marinus, University of Massachusetts, Worcester, MA, USA.

Bertani's lysogeny broth (LB) was used as standard liquid medium. Solid LB contained agar at 1.5% final concentration. Green plates (19) contained methyl blue (Sigma-Aldrich, St Louis, MO, USA) instead of aniline blue. The indicator for monitoring β-galactosidase activity in plate tests was 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal; Sigma-Aldrich, 40 μg/ml). Antibiotics were used at the concentrations described previously (20). To grow oxyR strains on LB agar, 75 μl of a 10 mg/ml catalase solution (Sigma-Aldrich, St Louis, MO, USA) were spread on the surface of the plates.

Transductional crosses using phage P22 HT 105/1 int201 (21) were used for construction of strains with altered chromosomal markers. The transduction protocol has been described elsewhere (22). To obtain phage-free isolates, transductants were purified by streaking on green plates. Phage sensitivity was tested by cross-streaking with the clear-plaque mutant P22 H5.

The oligonucleotides used in this study have either been described previously (16) or are listed in Supplementary Table S2. Gene disruption was achieved using plasmids pKD3, pKD4 and pKD13 (23) and oligonucleotides PS1, PS2 or PS4. Verification of the constructs was achieved using oligonucleotides E1 and E2. Antibiotic resistance cassettes introduced during strain construction were excised by recombination with plasmid pCP20 (23).

Directed construction of point mutations

Mutation of GATC sites within the opvAB regulatory region was achieved using previously described procedures and oligonucleotides (16). Additional primers are included in Supplementary Table S2. Antibiotic resistance cassettes from pKD3 and pKD4 were introduced in opvAB::lac and opvAB::gfp backgrounds using oligonucleotides delGATC-PS1 and delGATC-PS2, respectively (16). The resulting strains were used as intermediates in the construction of point mutations. Mutation of OxyR binding sites was achieved in the same way using primers labeled OxyRB and OxyRD (Supplementary Table S2).

β-galactosidase assays

Bacterial cultures were grown in LB until stationary phase (O.D.₆₀₀ ∼4). Levels of β-galactosidase activity were assayed using the CHCl₃-sodium dodecyl sulfate permeabilization procedure (24). All data are averages and standard deviations from more than three independent experiments.

Calculation of phase transition frequencies

Phase transition rates were estimated as described by Eisenstein (25). Briefly, a strain harboring an opvAB::lac fusion was plated on LB + X-gal. After 16 h growth at 37ºC, colonies displaying ON and OFF phenotypes were chosen, resuspended in phosphate buffered saline (PBS) and respread on new plates. Phase transition frequencies were calculated using the formula (M/N)/g where M is the number of cells that underwent a phase transition, N the total number of cells scored, and g the total number of generations that gave rise to the colony.

Flow cytometry

Bacterial cultures were grown in LB at 37°C until exponential phase (O.D.₆₀₀ ∼0.3). Cells were then diluted in PBS to a final concentration of ∼10⁷/ml. Data acquisition and analysis were performed using a Cytomics FC500-MPL cytometer (Beckman Coulter, Brea, CA, USA). Data were collected for 100 000 events per sample, and were analyzed with CXP and FlowJo8.7 software. Data are represented by a dot plot (forward scatter [cell size] versus fluorescence intensity [opvAB::gfp expression]).

Construction of plasmid pIZ1885 (pQE30::oxyR^C199S)

A DNA fragment containing oxyR^C199S (16) was amplified using oligonucleotides His-oxyR-BamHI-5 and His-oxyR-SalI-3, and cloned into pQE30 (Qiagen, Valencia, CA, USA) using the BamHI and SalI sites. The recombinant plasmid (pIZ1885) was verified by restriction analysis and DNA sequencing.

Purification of OxyR protein

For 6×His-OxyR^C199S purification, plasmid pIZ1885 was transformed into E. coli M15 [pREP4] (Qiagen, Valencia, CA, USA). M15/pIZ1885 was grown in LB broth containing ampicillin, and expression of 6×His-OxyR^C199S was induced with 1 mM isopropyl β-D-thiogalactopyranoside (IPTG). After 3 h of induction, cells were centrifuged and resuspended in 10 ml of lysis buffer (20 mM Tris, 300 mM NaCl, 10 mM imidazole) per g of pelleted cells, and were lysed by sonication. The suspension was centrifuged at 10 000 rpm for 30 min and the supernatant containing the soluble fraction of 6×His-OxyR^C199S was transferred to a HisTrap HP nickel affinity chromatography column (GE Healthcare, Wauwatosa, WI, USA). The column was washed with 4 ml of lysis buffer, 4 ml of washing buffer (20 mM Tris, 300 mM NaCl, 30 mM imidazole) and 4 ml of the same buffer with 50 mM imidazole. Protein elution was performed with 3 ml of elution buffer (20 mM Tris, 300 mM NaCl, 300 mM imidazole). Elution fractions enriched in 6×His-OxyR^C199S were selected and combined. Imidazole was removed by transferring to an Amicon^® ultra centrifugal filter (Merck Millipore, Darmstadt, Germany) and washing with storage buffer (20 mM Tris, 300 mM NaCl, 10% glycerol) or by dialyzing in cellulose membranes (Sigma-Aldrich, St Louis, MO, USA). 6×His-OxyR^C199S was either used immediately or frozen in liquid nitrogen and stored at −80ºC.

Gel mobility shift assay

A DNA fragment containing predicted OxyR binding sites in the opvAB regulatory region and labeled with 6-carboxyfluorescein (6-FAM) was prepared by polymerase chain reaction (PCR) amplification using primers FAMGATClargo-5 and FAMGATClargo-3 (Supplementary Table S2). The PCR product was purified with the Wizard^® SV Clean-Up System (Promega). The envR control fragment was prepared using primers envR-For-Dnase and envR-Rev-Dnase (26), and was kindly provided by Elena Espinosa. Thirty five nanogram were used for each reaction. The FAM-labeled probe was incubated at room temperature for 30 min with increasing concentrations of purified 6×His-OxyR^C199S in a final volume of 20 μl with 1× OxyR binding buffer [25 mM Tris–HCl pH 7.5, 50 mM KCl, 5 mM MgCl₂, 5% glycerol, 50 μg/ml bovine serum albumin (BSA), 1 mM DTT, 1 μg/ml poly(dI-dC)]. Protein–DNA complexes were subjected to electrophoresis at 4°C in a 5% non-denaturing polyacrylamide gel in Tris-glycine-ethylenediaminetetraacetic acid (EDTA) buffer (25 mM Tris–HCl pH 7.5, 380 mM glycine, 1.5 mM EDTA). The gel was then analyzed in a FLA-5100 Scanner (Fujifilm, Tokyo, Japan).

DNA methylation in vitro

PCR fragments were methylated in vitro using Dam methylase (New England Biolabs, Ipswich, MA, USA) according to the manufacturer's instructions and subsequently digested with MboI (New England Biolabs). The undigested product was purified using the Wizard^® SV Clean-Up system (Promega, Madison, WI, USA).

DNase I footprinting

DNA probes containing the opvAB promoter and the upstream regulatory region, labeled with 6-carboxyfluorescein (6-FAM) at the opposite ends, were prepared by PCR amplification using the primer pairs FAMGATClargo-5 + FAMGATClargo-3 and seqGATC-5 + FAMGATClargoconFAM-3. Dam-methylated versions of the probes were prepared as described above. DNase I footprinting was performed as described elsewhere (27) with minor modifications. DNase I footprinting reactions were performed in 15 μl reaction volumes containing 1× OxyR binding buffer and 2 μM 6×His-OxyR^C199S. The binding reaction was allowed to equilibrate at room temperature for 30 min. A total of 1 μl (0.05 units) of DNase I (Roche Farma, Barcelona, Spain) was then added, mixed gently and incubated at 37°C for 5 min. The reaction was stopped by addition of 2 μl EDTA 100 mM followed by vigorous vortexing and thermal denaturation at 95°C for 10 min. Digestion products were desalted using MicroSpin G-25 columns (GE Healthcare, Wauwatosa, WI, USA) and analyzed on an ABI 3730 DNA Analyzer along with GeneScan 500-LIZ size standards (Applied Biosystems, Foster City, CA, USA).

SMRT^® sequencing

Cultures of S. enterica were enriched for OpvAB^ON cells if needed (17). SMRTbell™ template libraries were prepared according to the instructions from Pacific Biosciences (Menlo Park, CA, USA), following the procedure and checklist for 1 kb template preparation and sequencing. Briefly, for preparation of 600 bp libraries, 4 μg of genomic DNA were sheared in microTubes using adaptive focused acoustics (Covaris, Woburn, MA, USA). Size range was monitored on an Agilent 2100 Bioanalyzer from Agilent Technologies, Santa Clara, CA, USA. DNAs were end-repaired and ligated to hairpin adapters applying components from the DNA Template Prep, Pacific Biosciences, Menlo Park, CA, USA. SMRTbell™ templates were exonuclease-treated for removal of incomplete reaction products. Conditions for annealing of sequencing primers and binding of polymerase to purified SMRTbell™ templates were assessed with the Pacific Biosciences’ Binding Calculator. Six movies were taken for both states on the PacBio RSII (Pacific Biosciences, Menlo Park, CA, USA) using P4-C2 chemistry at 2 h collection time. Secondly, stationary phase cultures were enriched for OpvAB^ON cells and libraries were prepared as given above. In this case five movies were taken using P4-C2 chemistry at 3 h collection time.

Resulting data were mapped to the complete genome sequence (GenBank accession number CP001363.1) of S. enterica subsp. enterica serovar Typhimurium strain ATCC 14028, using the BLASR algorithm (28) as implemented in Pacific Biosciences’ SMRT^® Portal 2.1.0 within the ‘RS_Modification_and_Motif_Analysis.1’ protocol applying default parameter settings. According to the setup of the experiment the secondary analysis jobs were named ‘OpvAB^OFF’ and ‘OpvAB^ON’. Besides the global methylation pattern, the methylation status of four GATC sites upstream of the opvAB operon was inferred using SMRT^® View, investigating the chromosomal positions 2 361 489 and 2 361 490 (GATC₁), 2 361 439 and 2 361 440 (GATC₂), 2 361 416 and 2 361 417 (GATC₃) and 2 361 366 and 2 361 367 (GATC₄). Results are shown in supplementary .csv files S1 and S2 (OpvAB-OFF basemod summary and OpvAB-ON basemod summary, respectively).

Southern blot

Genomic DNA was isolated by phenol extraction and ethanol precipitation from stationary cultures in LB (O.D.₆₀₀ ∼4). A total of 16 μg of each DNA sample were digested with HaeIII and AccI (New England Biolabs, Ipswich, MA, USA), purified and divided into four fractions, three of which were subsequently digested with DpnI, MboI or Sau3AI (New England Biolabs). After digestion the samples were run in a 2% TAE-agarose gel at 100 V for 2 h. After electrophoresis, the DNA was denatured by treatment of the gel in acid conditions (0.25 M HCl, two washes 15 min each), followed by alkalinization (0.5 M NaOH, 1.5 M NaCl) and neutralization (0.5 M Tris, 1.5 M NaCl, pH 7.5; two washes, 30 min each). The gel was then washed in SSC 10× buffer (1.5 M NaCl, 150 mM trisodium citrate, pH 7) and the DNA was transferred by vacuum to an Amersham Hybond-N+ membrane (GE Healthcare, Wauwatosa, WI, USA) using a model 785 Vacuum Blotter (Bio-Rad, Hercules, CA, USA). The DNA in the membrane was then immobilized by UV crosslinking. A radioactive probe was prepared by PCR using dCTP [α-³²P] (Perkin Elmer, Waltham, MA, USA) and oligonucleotides 2208mut1DIRnuevo and 2208mut4INVnuevo (16). After the PCR reaction, non-incorporated nucleotides were removed by treatment in a Sephadex G-25 column (illustra MicroSpin G-25 columns, GE Healthcare, Wauwatosa, WI, USA) following the instructions of the manufacturer. Prior to hybridization the double-stranded DNA probe was denatured by heating at 95°C for 3 min, followed by incubation on ice. Hybridization with the probe was performed overnight at 42°C in hybridization buffer (0.5 M sodium phosphate pH 7.2, 10 mM EDTA, 7% sodium dodecylsulphate (SDS)). Excess probe was removed with washing buffer (40 mM sodium phosphate pH 7.2, 1% SDS) at 38°C (three washes, 30 min each). The membrane was developed using a FLA-5100 Scanner (Fujifilm, Tokyo, Japan).

RESULTS

Both the absence and the overexpression of Dam methylase increase opvAB expression and abolish phase variation

Genes under Dam methylation control fall into two categories. One includes genes in which methylation and non-methylation provide opposite signals (12). An example is the traJ gene of the Salmonella virulence plasmid, which is repressed by GATC methylation (29). In this class of genes, expression of the dam gene from a multicopy plasmid does not alter the wild-type phenotype (29). In other genes, however, a plasmid-borne dam gene does alter the gene expression pattern. This phenomenon is usually an indication that Dam dependent transcriptional control is more complex, and involves the formation of Dam methylation patterns (combinations of methylated and non-methylated GATC sites) (12). To ascertain whether opvAB belonged to the ‘simple’ or the ‘complex’ class of Dam methylation-dependent genes, the effect of introducing a dam gene carried on plasmid pTP166 was assayed. The results were as follows:

• In a wild-type background, an opvAB::lac translational fusion showed phase variation, and formed white (OpvAB^OFF) and blue (OpvAB^ON) colonies in the presence of X-gal. In a dam background, phase variation was abolished, and all colonies were Lac⁺ (OpvAB^ON). Plasmid pTP166 yielded an intermediate phenotype (Figure 1A), suggesting that formation of the OpvAB^OFF and OpvAB^ON subpopulations might involve the establishment of a DNA methylation pattern in the GATC sites of the opvAB control region, rather than methylation or non-methylation of the full set of GATC sites. A similar phenomenon occurs in the gtr operon (15) which is repressed in a dam background while introduction of a cloned dam gene results in an intermediate phenotype.

• Expression of opvAB::lac was also monitored by β-galactosidase assays (Figure 1B). Lack of Dam methylation increased expression of the opvAB operon as previously described (16). Introduction of the dam gene carried on the pTP166 plasmid yielded an intermediate opvAB expression level, as in the colonies described above.

• Expression of an opvAB::gfp transcriptional fusion was monitored by fluorescence analysis (Figure 1C). A major OpvAB^OFF subpopulation and a minor OpvAB^ON subpopulation were detected in the wild-type. In a dam background, a single population in the ON state was observed, in accordance with the results obtained with a opvAB::lac fusion. In the presence of a cloned dam gene (pTP166), a single population with intermediate levels of expression was detected and a shift toward the ON state remained visible (Figure 1C).

Altogether, the above observations suggested that DNA methylation patterns might be formed at the opvAB control region. This region, located upstream of the opvAB promoter, contains four GATC sites separated by 46, 19 and 46 nt and centered at the −172.5, −122.5, −99.5 and −49.5 positions upstream of the transcription start site (Figure 2A and Supplementary Figure S1). From now on, these GATC sites will be referred to as GATC₁ to GATC₄, the latter being closest to the −35 module of the opvAB promoter (Supplementary Figure S1).

Roles of individual opvAB GATC sites in the formation of OpvAB^OFF and OpvAB^ON cell lineages

To study the contribution of each GATC site to opvAB regulation, mutations were introduced by site-directed mutagenesis. The mutations were designed to change GATC sites so that they would no longer be a substrate for Dam methylation. Because OxyR is essential for opvAB expression (16), alteration of consensus sequences was avoided inside putative OxyR binding sites. CATC sites were thus introduced in place of GATC sites, and every combination of mutated and non-mutated GATC sites was produced.

The effect of GATC mutations on opvAB expression was first analyzed by comparing the β-galactosidase activity of an opvAB::lac translational fusion in dam⁺ and dam backgrounds (Figure 2B). Relevant observations were as follows:

• Mutation of GATC₁ and GATC₃ had a small effect on regulation by Dam methylation, although the absolute values of β-galactosidase activity were higher. Mutation of GATC₂ resulted in diminished regulation by Dam methylation. When GATC₄ was mutated, control by Dam methylation showed an inverted pattern (expression was higher in a dam⁺ background).

• As a general rule, combinations of two or more mutations seemed to have an additive effect. A remarkable case was the combination of mutated GATC₂ and GATC₄ which exacerbated the inversion of regulation by Dam methylation caused by mutation of GATC₄ alone. It is noteworthy that mutations in GATC₁, GATC₂ and GATC₃ together did not abolish Dam-dependent regulation, whereas a single mutation in GATC₄ inverted the pattern of Dam-dependent regulation.

The overall conclusion from these experiments was that all four GATC sites are involved in Dam-dependent control of opvAB expression, and that the GATC₄ site may have an especially prominent role.

Even though disruption of OxyR binding sites had been avoided, GATC mutations affected opvAB expression irrespective of the presence or absence of DNA methylation, as observed in a dam background (Figure 2B). In the absence of Dam methylation, mutations in GATC₁ and GATC₃ increased opvAB expression whereas mutations in GATC₂ and GATC₄ resulted in lower opvAB expression. To separate such effects from those of Dam methylation itself, the β-galactosidase activity of opvAB::lac in a wild-type background was put in relation to the β-galactosidase activity in a dam background (Figure 2C). This representation leads to the interesting conclusion that mutations in GATC₂ and GATC₄ activate opvAB expression. Mutations in GATC₁ and GATC₃ show little effect on their own because opvAB expression is low in the wild-type, but they repress opvAB expression when combined with activating mutations in GATC₂ and/or GATC₄. Hence, the GATC sites in the opvAB regulatory region can be tentatively divided in two pairs: methylation of pair GATC₁ + GATC₃ seems to be associated with the OpvAB^ON state while methylation of pair GATC₂ + GATC₄ seems to be associated with the OpvAB^OFF state.

Analysis of fluorescence using an opvAB::gfp transcriptional fusion (Figure 3) allowed us to distinguish whether the differences in opvAB expression in GATC mutant backgrounds reflected differences in gene expression or differences in the sizes of the OpvAB^ON and OpvAB^OFF subpopulations. The main observations were as follows:

• In the wild-type, the OpvAB^ON subpopulation comprised ∼0.18% cells.

• Mutation of GATC₄ caused a drastic increase in the size of the OpvAB^ON subpopulation. Mutations in GATC₁, GATC₂ and GATC₃ had a smaller effect, which was more clearly seen when they were combined with each other and/or with a mutation in GATC₄.

• Two subpopulations were still distinguished when three GATC sites were mutated, provided that either GATC₃ or GATC₄ remained unaltered. The relative size of the OpvAB^OFF and OpvAB^ON subpopulations was however different in each case, with a predominant OpvAB^OFF subpopulation when GATC₄ remained unaltered and a predominant OpvAB^ON subpopulation when GATC₃ remained unaltered.

• Mutation of both GATC₃ and GATC₄ eliminated subpopulation formation regardless of the presence of mutations in GATC₁ and GATC₂, and yielded an OpvAB^ON population.

These observations are consistent with the gene expression analyses reported above, and permit to interpret the gene expression results in terms of subpopulation formation. Mutation of GATC₄ caused the most drastic increase in the proportion of OpvAB^ON cells, thereby confirming that methylation of the GATC₄ site may have a relevant role in the formation of the OpvAB^OFF subpopulation. Increase of OpvAB^ON subpopulation was likewise observed when a mutated GATC₄ was combined with other mutated GATC sites (Figure 3).

OxyR binds the opvAB regulatory region

Four putative OxyR binding half-sites are found in the regulatory region of opvAB centered in the −148, −116, −75 and −43 positions (Supplementary Figure S1), and sharing 10, 8, 10 and 7 nt respectively with the 10-nt consensus sequence (16). The OxyR binding half-sites upstream of the opvAB promoter will be from now on referred to as OBS_A to OBS_D, the latter being immediately upstream of the opvAB −35 promoter module (Figure 2A). Assuming a helical periodicity of 10.5 bp (30), the OBS are predicted to be spaced by one, two and one helical turns, which means that all the OxyR binding sites may be on the same face of the DNA helix. The distance between OBS_A and OBS_B, and between OBS_C and OBS_D as well, is canonical for binding of the reduced form of OxyR (31). GATC₂ and GATC₄ overlap with OBS_B and OBS_D, respectively (Figure 2A).

To test whether OxyR binds the opvAB regulatory region, an electrophoretic mobility shift assay (EMSA) was carried out using purified OxyR protein (Figure 4A). To avoid uncontrolled oxidation of OxyR and because it was previously shown that the oxidation state of OxyR is not relevant for opvAB regulation (16), we used a mutant version of the OxyR protein, OxyR^C199S, which cannot be oxidized but retains the properties of the reduced form of OxyR (31,32). Purified 6×His-OxyR^C199S protein (henceforth named OxyR for simplicity) was thus used. A DNA fragment containing the four regulatory GATC sites and the four OxyR binding half-sites was produced using a 6-FAM-labeled oligonucleotide and was incubated with increasing concentrations of OxyR. Binding was unambiguously detected. A DNA fragment from the regulatory region of an unrelated gene (envR) was used as a negative control, and binding was not detected (Figure 4A).

OxyR protects the opvAB regulatory region

To define the binding pattern of OxyR to the opvAB regulatory region, purified OxyR was used in a footprinting assay performed using 6-FAM-labeled DNA fragments and DNase I (Figure 4B). The same DNA fragment used in the EMSA assays, containing both the GATC sites and predicted OxyR binding sites, was labeled at the alternate ends and used in parallel experiments. Methylated and non-methylated DNA probes were used, as well as a probe in which GATC sites 1–4 had been converted to CATC sites by site-directed mutagenesis. The analysis confirmed the ability of OxyR to bind the opvAB regulatory region in vitro (Figure 4 and Supplementary Figure S2). Relevant observations were as follows:

• Protection from DNase I digestion was detected in a 133 bp DNA span, albeit with regional differences. GATC₁, GATC₂, GATC₃ are located in the protected region. Fragment-specific binding patterns were detected and the overall protection was less efficient when the DNA probe was either methylated or GATC-less.

• The OBS_A and OBS_C sites were fully protected, while OBS_B was partially protected.

• OBS_D, which contains the GATC₄ site, was not protected.

The relevance of these observations may be limited as methylated and non-methylated DNA probes were used, and evidence presented above had suggested that opvAB regulation involved both methylated and non-methylated GATC sites (Figure 1). With this caveat, footprinting experiments confirmed the ability of OxyR to bind the opvAB regulatory region and defined the DNA region protected by OxyR binding. An additional, interesting observation was that OxyR protection extended outside the OxyR binding sites, as previously described for other LysR-type factors (33–36) (see below).

OpvAB^OFF and OpvAB^ON subpopulations are characterized by inverse patterns of Dam methylation

Single-molecule real-time (SMRT^®) sequencing results showed that >97% of the total of 38 458 GATC sites present in the genome of S. enterica serovar Typhimurium are methylated, and that non-methylated sites are the exception. Within this set, several non-methylated GATC sites were detected upstream of the opvAB operon. In order to analyze them in more detail, position-specific base modification analyses were performed. Addition of the virulent P22 H5 phage to a culture of S. enterica results in selection of the OpvAB^ON subpopulation (17). Using this procedure, a culture was enriched in OpvAB^ON cells and the methylation state of the opvAB GATC sites was analyzed using SMRT^® sequencing (37). An ordinary culture, which contains >99% OpvAB^OFF cells (16,17), was also subjected to SMRT^® sequencing. A total of 246 373 (430 408) polymerase reads with a mean polymerase read length of 10 010 (8516) bp and mean sequence coverage of 178× (329×) were obtained for the OpvAB^ON (OpvAB^OFF) SMRT sequencing. The results from position-specific base modification analysis are shown in supplementary .csv files S1 and S2, and can be summarized as follows:

• In an ordinary OpvAB^OFF culture, GATC₁ and GATC₃ were non-methylated, whereas GATC₂ and GATC₄ were methylated (Table 1).

• In the OpvAB^ON culture, an inverse DNA methylation pattern was found: non-methylation of GATC₂ and GATC₄ and methylation of GATC₁ and GATC₃ (Table 1).

DNA modification status according to SMRT^® View for position specific base-modification analysis upstream of the opvAB operon^a

These observations confirm that establishment of the OFF and ON states of the opvAB locus involves the formation of DNA methylation patterns, as in other phase variation loci under Dam methylation control (10,13,15).

OxyR protects GATC sites from Dam methylation in vivo

OxyR has been previously described as a DNA methylation-blocking factor, able to induce the formation of non-methylated GATC sites (15,38). To test whether OxyR has a similar DNA methylation-blocking ability in the opvAB operon, the methylation state of the GATC sites in the opvAB regulatory region was tested in vivo. For this purpose, a Southern blot was performed using genomic DNA extracted from the wild-type strain and from an oxyR mutant. The methylation state of individual GATC sites was inferred from restriction analysis using enzymes that cut GATC sequences depending on their methylation state (MboI, DpnI and Sau3AI). GATC₁ and GATC₃ were found to be non-methylated while GATC₂ and GATC₄ were found to be methylated in the wild-type strain (Figure 5). In contrast, in an oxyR background, all four GATC sites were found to be methylated (Figure 5). These observations confirmed that OxyR has DNA methylation-blocking ability in vivo at the opvAB regulatory region.

Mutations in the OBS_B and OBS_D OxyR binding sites abolish phase variation

Of the four OxyR binding half-sites in the opvAB regulatory region, OBS_A and OBS_C are an absolute match (10 out of 10 nt) to the consensus sequences defined for OxyR binding (31). In contrast, OBS_B and OBS_D share only 8 and 7 out of 10 nt with the consensus sequence, respectively. The fact that opvAB phase variation is skewed toward the OFF state led us to hypothesize that the degree of OxyR binding site perfection played a role in such bias. To test our hypothesis, 1 nt change was introduced in OBS_B and two nucleotide changes in OBS_D so that their mutated versions would share 9 out of 10 nt with the consensus sequence. Construction of a perfect consensus sequence was avoided since it would inevitably destroy GATC₂ and GATC₄.

The consequences of OBS_B and OBS_D DNA sequence amelioration were analyzed using opvAB::gfp (Figure 6A) and opvAB::lac fusions (Figure 6B). Mutations in either OBS_B or OBS_D abolished opvAB phase variation, yielding a uniform OpvAB^ON population. In the case of OBS_B, a single nucleotide change led also to full expression of the operon. The mutation in OBS_D caused a smaller increase in expression and was epistatic to the mutation in OBS_B.

An interpretation of these observations is that OBS_B and OBS_D DNA sequence amelioration may ‘trap’ OxyR in the OpvAB^ON configuration. In support of this view, absence of Dam methylation had no effect on opvAB expression in these mutant backgrounds (Figure 6). Hence, the preference of OxyR for certain OxyR-binding sites may be a key factor in regulation of opvAB phase variation, and alternative binding of OxyR upstream of the opvAB promoter may generate the OpvAB^OFF and OpvAB^ON subpopulations.

SeqA contributes to the small size of the OpvAB^ON subpopulation

SeqA was considered a potential ancillary candidate for regulation of opvAB since it binds GATC sites (39) and is involved in regulation of other phase variation loci (40,41). Thus we analyzed the effect of a seqA mutation on opvAB expression and its influence on the formation of OpvAB subpopulations. A strain carrying a seqA null allele and an opvAB::lac fusion formed darker (Lac⁺) colonies on LB + X-gal than the wild-type, and displayed frequent sectoring. Nonetheless, two groups of differently colored colonies (light blue and dark blue) were still distinguishable (Figure 7A), which allowed calculation of phase transition frequencies. The OFF→ON transition rate was found to be 50-fold higher in a seqA background (3.0 × 10⁻³ compared with 6.1 × 10⁻⁵ in the wild-type), whereas the ON→OFF transition rates were similar (3.1 × 10⁻² compared to 3.7 × 10⁻² in the wild-type). Not surprisingly, the β-galactosidase activity of an opvAB::lac fusion was ∼10-fold higher in a seqA background (Figure 7B).

Fluorescence assays showed that mutation of seqA caused an increase in the size of the OpvAB^ON subpopulation (Figure 7C). The effect was stronger in the presence of mutations in GATC₁ and/or GATC₂, and to a lesser extent in GATC₃ (Supplementary Figure S1). Interestingly, when GATC₄ was mutated, a mutation in seqA had an effect opposite to that observed in the wild-type: the OpvAB^ON subpopulation was reduced (Supplementary Figure S1). When both GATC₃ and GATC₄ were mutated, the seqA mutation did not have a significant effect (Supplementary Figure S1). These results seem to indicate that the main role of SeqA in the regulation of opvAB is the maintenance of a low OFF→ON transition rate (in other words, repression of OpvAB^ON subpopulation formation).

HU is essential for the formation of the OpvAB^ON subpopulation

HU is a nucleoid-associated protein known to regulate a large number of genes in E. coli and Salmonella (42–44). The HU protein can exist in three forms: the HU αβ heterodimer and the corresponding homodimers. The heterodimer is the predominant form in vivo (45). We deleted hupA and/or hupB, the genes encoding the two proteins forming the HU heterodimer and tested the effect of the mutations on the expression of an opvAB::gfp fusion (Figure 8A). The OpvAB^ON subpopulation was found to be reduced from ∼0.18% in the wild-type to 0.09% in single hupA and hupB mutants. Reduction of the OpvAB^ON subpopulation size was exacerbated in the double hupA hupB mutant: the OpvAB^ON subpopulation was virtually absent (Figure 8A).

When hupA and hupB mutations were introduced into an opvAB::lac background, a decrease in the β-galactosidase activity of the opvAB::lac fusion was likewise found (Figure 8B). In turn, when formation of Lac⁺ (OpvAB^ON) colonies was scored on LB + X-gal plates, blue (Lac⁺) colonies were still visible in the hupA and hupB single mutants but not in the double mutant hupA hupB background (Figure 8C).

When the effect of the hupA and hupB mutations was tested in OpvAB^ON-locked backgrounds, the population remained in the OpvAB^ON state (Figure 8D), although opvAB::lac expression was slightly lower (Figure 8E). Hence, HU seems to be necessary for maintenance of the OpvAB^ON state in the wild-type but not in mutants locked in OpvAB^ON state.

DISCUSSION

The S. enterica opvAB operon encodes membrane proteins that alter O-antigen chain length in the lipopolysaccharide (16). Expression of the opvAB operon is subject to phase variation, with switching frequencies of 6.1 × 10⁻⁵ (OFF→ON) and 3.7 × 10⁻² (ON→OFF) per cell and generation in LB medium (16). As a consequence of these disparate switching levels, S. enterica populations (e.g. batch cultures) contain a major OpvAB^OFF subpopulation (>99% cells) and a minor OpvAB^ON subpopulation (<1% cells).

The regulatory region upstream of the opvAB promoter, depicted in Figures 2 and 9, and Supplementary Figure S1, contains four half-sites for binding of OxyR (OBS_A-D), and 4 methylatable GATC motifs (GATC_1–4). OxyR is a LysR-type transcriptional regulator that also acts as a sensor of oxidative stress. Although OxyR was first described as an activator of genes responsive to oxidative damage (46), its function in opvAB regulation is unrelated to oxidative damage and independent of its own oxidation state (16). The same is true for other OxyR-dependent phase variation systems such as agn43 (47) and gtr (15). OxyR binds DNA as a tetramer (30).

SMRT^® sequencing data show that S. enterica OpvAB^OFF and OpvAB^ON subpopulations differ in their pattern of Dam methylation at the opvAB regulatory region (Table 1). The patterns found are actually opposite: in the OpvAB^OFF state, GATC₁ and GATC₃ are non-methylated, whereas GATC₂ and GATC₄ are methylated; in the OpvAB^ON state, GATC₂ and GATC₄ are non-methylated, whereas GATC₁ and GATC₃ are methylated. Combinations of methylated and non-methylated GATC sites have been previously described in other phase variation loci including pap and gtr (10,15). In these loci, GATC non-methylation is the consequence of DNA methylation hindrance upon protein binding. In an analogous fashion, we provide evidence that DNA methylation patterns at the opvAB regulatory region are generated by OxyR binding (Figure 5).

OxyR has been shown to bind alternative pairs of half-sites in gtr (15), and opvAB may constitute another example of the same phenomenon albeit with a different genomic architecture. In gtr, the sites bound by OxyR in the OFF and ON lineages have identical number of nucleotides in common with the consensus sequence (15), which may explain why the gtr locus has similar ON→OFF and OFF→ON transition rates. In contast, the OBS_A and OBS_C sites of opvAB are identical to the consensus sequence for OxyR binding while OBS_B and OBS_D share only 8/10 and 7/10 nt with the consensus, respectively. This difference may explain the higher stability of the OpvAB^OFF lineage, which results in a ∼600-fold difference in the ON→OFF and OFF→ON transition rates. The relevance of the nucleotide sequence of OxyR binding sites for opvAB regulation is illustrated by the observation that single nucleotide changes in OBS_B and OBS_D lock the system in the ON state (Figure 6). The lower increase in opvAB expression caused by a mutation in OBS_D (Figure 6) may be tentatively explained as a consquence of the OBS_D location immediately upstream of the −35 module: a mutation of OBS_D may impair the interaction between OxyR and the RNA polymerase. In agreement with this view, it has been proposed that RNA polymerase may contact OxyR and other LysR-type transcription factors within the DNA region occupied by the regulator (48). The fact that the mutation in OBS_D is epistatic over the mutation in OBS_B (Figure 6) may support this interpretation.

Preferential methylation of GATC₄ may be an additional factor contributing to the stability of the OpvAB^OFF lineage. The DNA sequences that flank GATC₁, GATC₂ and GATC₃ are predicted to be relatively poor Dam methylation substrates compared with the flanking sequences of GATC₄ (49). Rapid methylation of GATC₄ may thus contribute to perpetuation of the OpvAB^OFF state.

Our tentative model, based on a combination of experimental data, information from the literature and some speculation as well, proposes that the predominant OFF state involves binding of OxyR to the OBS_A and OBS_C sites, which protects GATC₁ and GATC₃ from methylation (Table 1 and Figure 5). In this configuration, GATC₂ and GATC₄ are unprotected and therefore are methylated by Dam. A caveat to this model is that, to our knowledge, OxyR has not been described to bind non-consecutive half-sites. However, such binding pattern is consistent with two lines of evidence: (i) only OBS_A and OBS_C are fully protected in the footprinting assay (Figure 4B); (ii) OxyR has been hitherto described to bind DNA as a tetramer (31,50). An alternative hypothesis is that OxyR dimers may bind independently the OBS_A and OBS_C sites.

DNA bending, which is commonly induced by OxyR (31), specifically by the reduced tetramer structure (51) and by other LysR-type regulators (52–55), may contribute to methylation hindrance in GATC₁ and GATC₃. A DNA bend is induced by OxyR in agn43 (47,56), another phase variation locus regulated by Dam methylation and OxyR. The occurrence of bending might thus help to understand why GATC₁ and GATC₃ are protected from methylation in the OpvAB^OFF configuration (Figure 5) despite their location outside OBS_A and OBS_c (Figure 9 and Supplementary Figure S1).

Another factor that might contribute to methylation hindrance in GATC₁ and GATC₃ in the OpvA^OFF lineage might be DNA wrapping, which has been proposed for other transcriptional regulators whose footprints extend outside the binding sites. Examples include the NtrC (33), RcnR (34) and NorR (35,36) transcription factors from E. coli. CarP, also called PepA, an E. coli transcription factor, specifically prevents methylation of a GATC site which is not included in the binding footprint (57). GATC₁ and GATC₃, which lie in the extended OxyR-bound region, may be protected from Dam methylation in an analogous fashion.

In the ON state, OxyR binds to the OBS_B and OBS_D sites. As a consequence, GATC₂ and GATC₄ are protected from methylation and remain non-methylated, whereas GATC₁ and GATC₃ are unprotected and are methylated (Table 1). In this configuration, RNA polymerase is successfully recruited to the opvAB promoter and transcription of opvAB takes place. OxyR has been shown to recruit RNA polymerase by direct contact with the C-terminal domain of the α subunit (58,50), and the inverse is also true: RNA polymerase can recruit OxyR (50), which might contribute to maintenance of the OpvAB^ON state.

Additional factors involved in the formation of OpvAB cell lineages are the GATC-binding protein SeqA and the nucleoid protein HU. SeqA contributes to the stability of the OpvAB^OFF lineage, acting as a repressor of the OFF→ON transition (Figure 7). SeqA action seems to be exerted mostly on GATC₃ and GATC₄ (Supplementary Figure S1). Because SeqA binds hemimethylated GATC sites (59), a tentative speculation is that it might favor DNA methylation over OxyR binding during DNA replication, as previously suggested for agn43 (40). In turn, HU contributes to formation of the OpvAB^ON lineage (Figure 8). Tentative interpretations may be that HU contributes to the establishment of the OpvAB^ON state either by inducing DNA bending or by stabilizing OxyR-mediated bending. The latter possibility may be more likely as HU often stabilizes bent DNA rather than bending DNA itself (60), and HU is not essential in OpvAB^ON-locked backgrounds (Figure 8). On the other hand, AT-rich DNA, such as that found in the opvAB regulatory region (which is 23% G + C only) is intrinsically prone to DNA bending (61,62).

In the model depicted in Figure 9, two OxyR tetramers are required to maintain the ON state but only one tetramer is necessary to maintain the OFF state (Figure 9). If true, this difference may be an additional factor to explain the high ON→OFF transition rate (together with the OxyR binding site differences and the preferential methylation of GATC₄). Upon passage of the DNA replication fork, the local concentration of OxyR will be halved, therefore facilitating the transition from ON (depending on two OxyR tetramers) to OFF (depending on one tetramer only).

We thank Simone Severitt and Nicole Mrotzek for excellent technical assistance, María A. Sánchez Romero and Elena Espinosa for advice, Martin Marinus for providing pTP166 and Marjan van der Woude and Khai Luong for discussions. We also thank Modesto Carballo, Laura Navarro and Cristina Reyes of the Servicio de Biología (CITIUS, Universidad de Sevilla) for help in experiments performed at the facility.

FUNDING

Ministerio de Economía y Competitividad of Spain and European Regional Fund [BIO2013-44220-R and CSD2008-00013 to J.C.]; Consejería de Innovación, Ciencia y Empresa, Junta de Andalucía, Spain [P10-CVI-5879 to J.C.]; German Federal Ministry of Science and Education through the German Center of Infection Research (DZIF) [8000-105-3 to J.O.]. Funding for open access charge: Ministerio de Economía y Competitividad of Spain and European Regional Fund [BIO2013-44220-R].

Conflict of interest statement. None declared.

REFERENCES