The immunodominant lipopolysaccharide is a key antigenic factor for Gram-negative pathogens such as salmonellae where it plays key roles in host adaptation, virulence, immune evasion, and persistence. Variation in the lipopolysaccharide is also the major differentiating factor that is used to classify Salmonella into over 2600 serovars as part of the Kaufmann-White scheme. While lipopolysaccharide diversity is generally associated with sequence variation in the lipopolysaccharide biosynthesis operon, extraneous genetic factors such as those encoded by the glucosyltransferase (gtr) operons provide further structural heterogeneity by adding additional sugars onto the O-antigen component of the lipopolysaccharide. Here we identify and examine the O-antigen modifying glucosyltransferase genes from the genomes of Salmonella enterica and Salmonella bongori serovars. We show that Salmonella generally carries between 1 and 4 gtr operons that we have classified into 10 families on the basis of gtrC sequence with apparent O-antigen modification detected for five of these families. The gtr operons localize to bacteriophage-associated genomic regions and exhibit a dynamic evolutionary history driven by recombination and gene shuffling events leading to new gene combinations. Furthermore, evidence of Dam- and OxyR-dependent phase variation of gtr gene expression was identified within eight gtr families. Thus, as O-antigen modification generates significant intra- and inter-strain phenotypic diversity, gtr-mediated modification is fundamental in assessing Salmonella strain variability. This will inform appropriate vaccine and diagnostic approaches, in addition to contributing to our understanding of host-pathogen interactions.
Bacterial pathogens frequently evolve mechanisms to vary the composition of their surface structures. The consequence is enhanced long-term survival by facilitating persistence and evasion of the host immune system. Salmonella sp., cause severe infections in a range of mammalian hosts and guard themselves with a protective coat, termed the O-antigen. Through genome sequence analyses we found that Salmonella have acquired an unprecedented repertoire of genetic sequences for modifying their O-antigen coat. There is strong evidence that these genetic factors have a dynamic evolutionary history and are spread through the bacterial population by bacteriophage. In addition to this genetic repertoire, we determined that Salmonella can and often do employ stochastic mechanisms for expression of these genetic factors. This means that O-antigen coat diversity can be generated within a Salmonella population that otherwise has a common genome. Our data significantly enhance our appreciation of the genetic and regulatory characteristics underpinning Salmonella O-antigen diversity. The role attributed to bacteriophage in generating this diversity highlights that Salmonella are acquiring an extensive repertoire of O-antigen modifying traits that may enhance the pathogen's ability to persist and cause disease in mammalian hosts. Such genetic traits may make useful markers for defining new epidemiological and diagnostic tools.
Citation: Davies MR, Broadbent SE, Harris SR, Thomson NR, van der Woude MW (2013) Horizontally Acquired Glycosyltransferase Operons Drive Salmonellae Lipopolysaccharide Diversity. PLoS Genet 9(6): e1003568. https://doi.org/10.1371/journal.pgen.1003568
Editor: Diarmaid Hughes, Uppsala University, Sweden
Received: January 12, 2013; Accepted: May 1, 2013; Published: June 20, 2013
Copyright: © 2013 Davies et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by Wellcome Trust Grants WT080086MA (MWvdW) and WT076964 (NRT) (http://www.wellcome.ac.uk/funding/index.htm). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
There are two recognized Salmonella species, S. bongori and S. enterica, which can be divided into ∼2600 recognized Salmonella serovars according to the Kaufmann-White-Le Minor (KW) classification scheme , . Over half of these serovars are represented by S. enterica subspecies enterica (S. enterica subspecies I), which constitute 99% of human clinical Salmonella infections. The KW Scheme is governed by differences in antigenicity of the O-antigen of the lipopolysaccharide (LPS) and the flagella (H factor). The KW O-serotype is dependent on the reactivity of immune sera against immunodominant epitopes of the O-antigen, which in turn is dictated by the O-antigen structure. The KW Salmonella classification scheme is based on a panel of antisera recognizing 58 different O-antigen epitopes that allows classification of Salmonella into 46 different O-serogroups with some serogroups having multiple O-antigen epitopes , . More than 58 epitopes have been identified, but are not employed for classification purposes within the current scheme.
O-antigen diversity derives largely from differences in the carbohydrate composition and structure of the basal units of the O-antigen. This has been attributed to horizontal gene transfer events within the O-antigen biosynthesis locus, rfb (reviewed by , ). Further O-antigen structural and antigenic modifications have been identified that are driven by genes residing outside of the rfb loci such as the O-acetyltransferase gene, oafA that modifies the serotype through acetylation , wzy-like genes that alter linkages between O-antigen units  and glycosyltransferase (gtr) operons that add sugars onto the basic O-antigen structure thus changing the O-serotype –.
The best studied example of O-antigen glucosylation in Salmonella is due to bacteriophage P22 lysogenization of S. enterica subsp. enterica serovar Typhimurium (S. Typhimurium) strains. Phage P22 encodes a gtr operon, formerly designated as “con” or “a1”, that results in the addition of glucose to the galactose moiety of the O-antigen basal unit by a 1–6 linkage, leading to seroconversion and the additional recognition by O:1 typing sera –. A second gtr cluster has been identified in the chromosome of S. Typhimurium . Formerly called “oafR” , this gtr operon glucosylates the same galactose residue as the P22 gtr operon, but with a different (1–4) linkage . This modification yields a change from O:12 to sub-type O:122 serotype. However, this ‘sub-type’ change is not identified as part of standard Salmonella serotyping carried out in centers using the KW classification scheme. Furthermore, according to the KW classification scheme, both O:1 and O:122 expressing S. Typhimurium strains are collectively designated as the same serogroup, serogroup B. Thus, Salmonella strains that are identified as having the same clinical O-serogroup or O-serotype may in fact differ in their detailed O-antigen structures.
The gtr gene clusters described above consist of three genes, referred to in Salmonella as gtrABC. Shigella have similar clusters, and combining genetic and biochemical analysis for both species, a model has been proposed for the role of the three gene products (reviewed by ). GtrA and GtrB are proposed to be the bactoprenol-linked glucosyltranslocase or ‘flippase’ and the bactoprenol glucosyltransferase respectively, while GtrC is defined as the serotype-specific glycosyltransferase. There is little sequence homology between the known GtrC sequences , , probably reflecting the variety of O-antigen substrates recognized and the differing enzymatic activities.
The role of the O-antigen for the biology of pathogens is diverse and not fully understood. The O-antigen may be required for virulence as is the case for Escherichia coli, whereas other pathogens can succeed in the absence of an O-antigen. In general, the composition, structure and length of the O-antigen as determined by the rfb cluster, will influence antigenicity and can affect direct interactions with host cells, facilitate molecular mimicry, and alter the efficacy of the innate immune response , . In salmonellae, the composition may contribute to persistence and sequential infection of serovars . O-antigen diversity in Salmonella has also been implicated in host adaptation by altering susceptibility to digestion by intestinal amoebae , .
Likewise, the importance of O-antigen modification as mediated by non-rfb genetic mechanisms such as gtr and oafA remains to be fully elucidated, but studies indicate an impact on various aspects of Salmonella biology. O-antigen modification can alter phage susceptibility when the O-antigen is a primary or secondary receptor, with very specific modifications affecting specific phage. This was shown recently in S. Typhimurium, with expression of the P22 gtrABC decreasing P22 infectivity . In addition, the phase variable expression of the P22 gtr operon  confers transient resistance to phage SPC35 . The immune response to the O-antigen is also a key factor in determining spread of Salmonella and therefore, immune evasion or ‘sero-conversion’ as a result of changes in antigenicity due to modification of the O-antigen, may contribute to Salmonella persistence and dissemination , , . Furthermore, O-antigen modification may have a role in gut persistence in a mouse model of S. Typhimurium infection . The general effects associated with rfb-dependent composition may also be affected by O-antigen modification, but this has received little attention to date.
In contrast to the large body of work detailing the biochemical diversity and serotypic association of Salmonella O-antigens, relatively little insight has been accumulated regarding the genetic systems that promote Salmonella O-antigen heterogeneity. Here we explore the gtr repertoire in salmonellae based on genome sequence analysis and provide initial characterization to further our understanding of the role of these gene clusters in shaping the evolution, transmission and virulence of this important pathogen.
High prevalence ofgtr-like operons within Salmonella serovars that cluster into 10 distinct GtrC ‘families’
BLASTn was used with phage P22 gtrA, gtrB and gtrC genes as query sequences to identify gtr-like operons in 57 Salmonella genomes constituting the two known Salmonella species, S. enterica (n = 28) and S. bongori (n = 29) and four O-antigen modifying Salmonella bacteriophage genomes (P22, ST104, ST64T and epsilon34). Based on this analysis, a total of 59 gtr-like operons were identified: 52 in the 22 S. enterica subspecies enterica (subspecies I) genomes, 3 in 3 S. bongori serovars and 4 in the 4 Salmonella bacteriophage genomes (Tables S1 and S2). No gtr operons were identified in the singular genome sequences representing S. enterica subspecies salamae (subspecies II); arizonae (subspecies IIIa); diarizonae (subspecies IIIb); houtenae (subspecies IV) and indica (subspecies VI). While the three S. bongori genomes harbored a single gtr operon, 20 of the 22 S. enterica subspecies I genomes carried between two and four different gtr operons (Table S1). Thus, despite the important role that O-antigen structure plays in assigning the serogroup according to the Kaufmann-White Salmonella classification scheme, this genomic repertoire of gtr operons might suggest the existence of a level of antigenic complexity that is not defined in the standard typing scheme.
Since the gtrC gene product is purported to be responsible for the serotype-specific modification of the O-antigen, this gene product was used to define the functional relationships of the 59 gtrABC operons. Maximum likelihood approaches and Bayesian clustering analyses were used to plot the genetic relationships of all 59 GtrC products included in this study , . Figure 1 shows that the 59 GtrC's form 10 distinct clusters, herein defined as ‘GtrC families’ (Figure 1, ‘GtrC’). Phylogenetic analyses of the cognate GtrA and GtrB gene products identified 5 and 7 clusters respectively (Figure 1, ‘GtrA’ and ‘GtrB’). Only gene clusters represented by GtrC families IV and VIII were also represented in the same, single gtrA, gtrB cluster (Figure 1). In general, sequence conservation between any two GtrC families was very low. For example, the GtrC of phage P22, gtrP22 (GtrC family I) and S. Typhimurium gtrLT2_I (GtrC family III) share less than 18% amino acid identity, yet both add a glucose residue to the same galactose moiety of the S. Typhimurium O-antigen but via different linkages –. In contrast, orthologous GtrC proteins within the same GtrC family generally exhibit greater than 96% amino acid identity. Unlike the 10 GtrC families, the amino acid identity between any two GtrA and GtrB families are much higher, 84% and 90% respectively. Conserved internal gene deletions were found in the gtrB gene of all family II gtr operons and premature stop codons are predicted to occur within three GtrC gene products: One from GtrC family IV (S. Choleraesuis gtrCSCB67_IV) and two from GtrC family III (S. Paratyphi gtrC9150_II and S. Paratyphi gtrC12601_II).
A total of 59 gtr-like operons were identified from BLAST analysis of 22 complete S. enterica subspecies I complete genome sequences, 28 draft S. bongori genome sequences and 4 phage genome sequences. The grey bar indicates clusters of sequences that segregated on the basis of genomic sequence as determined by the Bayesian algorithm software, BAPS , . 10 GtrC clusters (‘GtrC families’) are identified by clustering and colored accordingly. This GtrC family color scheme is applied to the GtrB and GtrA trees to examine evolutionary relationships between the 3 gene products. GtrB and GtrA segregate into 7 and 5 clusters accordingly and the overall clusters are more closely related compared to GtrC gene products (scale bar). The lack of congruent clustering between GtrC, GtrB and GtrA gene products reflect different evolution histories of the gene products. Maximum likelihood rectangular phylograms were derived from ClustalW2 alignments using PhyML. Horizontal distances are proportional to sequence differences per site relative to the scalebar shown for each tree. Bootstrap values, out of 1000 trials, are shown on the tree branches. Each terminal node is labeled with a Gtr acronym assigned in Table S1. Asterisk denotes sequences predicted to contain deletions or frameshift mutations. The six S. Typhimurium genome sequences examined in this study (Table S1) share 100% sequence identity in their family III and family IV GtrC sequences and are collectively represented in this figure as STM_I and STM_II respectively.
Dynamic and ongoing evolutionary history of thegtr operon
The general lack of congruency between gtrA, gtrB and gtrC clustering (Figure 1) raises the possibility that recombination has shaped the evolutionary history of the gtr operon. Two approaches were employed to investigate this further. Comparative alignments of the three KW serogroup C1 genomes S. Paratyphi C, S. Choleraesuis and S. Infantis, showed that S. Choleraesuis possessed an extra gtr operon, gtrSCB67_I (Figure 2A). While the GtrB and GtrC proteins of S. Choleraesuis gtrSCB67_I exhibited over 98 percent amino acid identity to the GtrB and GtrC proteins of S. Paratyphi C gtrRKS4594_I, the GtrA protein of these operons were substantially lower at 75 percent (Figure 2A). These data suggest that allelic exchange may have occurred between the gtrA or gtrBC of a gtrSCB67_I-like and gtrRKS4594_I-like progenitor. A second approach using homoplasic SNPs as a marker of recombination identified extensive clusters of homoplasies scattered throughout the gtrA and gtrB sequences (Figure 2B). In the case of S. Agona gtrSeAg_I and S. Hadar gtr18_I the recombination event was not restricted by gene boundaries. These data illustrate a dynamic and complex picture of ongoing gtr evolution across Salmonella genomes and raise the possibility that a degree of functional redundancy may exist in the Gtr system with new combinations of gtrA, gtrB and gtrC sequences arising through recombination.
A) Schematic representation of a BLAST comparison between the gtr operons from three Kaufmann-White serogroup C1 genomes; S. Paratyphi C, S. Infantis and S. Choleraesuis. Red shaded regions indicate gene products with >98% protein similarity; blue regions <90% similarity; yellow regions <80% similarity and Grey regions <50% similarity as determined by ClustalW2 amino acid alignments . A possible gtr recombination is evident between the gtrA or gtrBC of S. Choleraesuis gtrSCB67_I and the gtrA or gtrBC of S. Paratyphi C gtrRKS4594_I gtr operons. S. Choleraesuis gtrSCB67_I is encompassed within a full-length P22-like prophage termed Scho1 . Genome integration sites as depicted in Figure 3 are represented above the Figure. IS refers to insertion-like sequences. B) Mapping of homoplasic SNPs within gtrA and gtrB gene sequences. On the left of the figure is the mid-point rooted phylogeny of the gtrA and gtrB gene sequences with gene names colored according to GtrC Family designation (Figure 1). Colored vertical lines depict homoplasic bases (red:A, blue:T, green:C, orange:G) that differ from the ancestral sequence and are shown relative to the concatenated gtrAB gene sequence shown on top. The pattern of lines represents a homoplasic ‘barcode’ of similarity between strains and is used to indicate recombined regions. High homoplasic SNP density across both gtrA and gtrB genes indicates a dynamic and complex evolutionary history driven by extensive recombination. Some examples of recombined regions are indicated by blue boxes. In the case of S. Agona gtrSL483_I and S. Hadar gtr18_I/S. Newport gtrSL254_I, recombination is not restricted by gtrAB gene boundaries.
Gtr families localize to four genomic locations withinSalmonella serovars
To determine whether gtr operons localize to defined genomic loci, 55 gtr operons representing 9 of the 10 Salmonella GtrC families were mapped onto the S. Typhimurium strain D23580 genomic backbone to show relative locations. All gtr operons identified in this study localized to four genomic regions (Figure 3, locus 1 to 4). These four regions are localized alongside tRNA genes and constitute the P22 (locus 1) and P2 (locus 4) integration sites, and the stably maintained phage derived genomic regions SPI-16 (locus 2) and SPI-17 (locus 3) . While locus 3 and 4 contain a single GtrC family (GtrC family II and IV respectively), locus 1 and 2 harbor representatives of three and seven different GtrC families, respectively (Figure 3).
Circular representation of the serogroup B S. Typhimurium D23580 genome (accession number FN424405) indicating that 9 GtrC families localize to four genomic regions (locus 1–4) relative to the D23580 genome. The italicized locus 3 (SPI-17) site is not represented in KW serogroup B genomes and thus, its relative location in S. Typhimurium D23580 is shown. Triangular gtr operons point the common location of GtrC family members. Pentagonal gtr operons are associated with full-length prophage. Dotted pentagonal operons with the phi symbol ‘Φ’ indicate gtr operons associated with full-length prophage that do not localize with their ‘parent’ GtrC family members. Gtr acronyms for these three gtr operons are shown (Table S1). Color code of the pentagon/triangles relate to the GtrC family clustering defined in Figure 1. The circular S. Typhimurium D23580 genome depicts forward open reading frames in blue, reverse open reading frames in red and tRNA in black.
Four gtr operons, S. Schwarzengrund gtrCVM19633 (family I), S. Dublin gtrCT02021853_I (family I), S. Typhimurium gtrD23580_BTP1 (family II), and S. Choleraesuis gtrSCB67_I (family V), did not localize to the same genomic location as their other family members (Figure 3). All four of these operons appear to be located on full-length prophage (Figure 3). This is consistent with lysogenic phage carrying gtr operons having alternative site-specific integration sites in the Salmonella genome. Indeed, the P22-like GtrC family, family I, is inserted at the P22 attachment site (locus 1). In contrast, a lambda-like prophage from S. Schwarzengrund CVM19633 harboring the family I GtrC, GtrCCVM19633 and a P22-like prophage from S. Dublin carrying two gtr operons, GtrCCT02021853_I (family I) and GtrCCT02021853_II (GtrC) are integrated at locus 2 (Figures 3, 4), which is also the integration site for the GtrC family I and III operons carried on a novel P22-like prophage .
Represented is an alignment and sequence comparison of the cysS to pheP genomic region encompassing locus 2 (SPI-16, Figure 3) from nine gtr positive Salmonella serovars. The phage-associated locus 2 is defined by a pair of direct repeats (pink) and a 5′ Arg-tRNA . Black coding sequences refer to those located within locus 2 boundaries, while white coding sequences are located outside locus 2. The gtr operon located within locus 2 is colored on the basis of GtrC family clustering (Figure 1). A different colored GtrC family likely reflects a functionally different gtr operon. S. Schwarzengrund contains a gtr positive lambda-like prophage and S. Dublin harbors a large remnant P22-like phage  within locus 2. Regions of genetic similarity between genome sequences are shaded in grey and were determined by BLASTn analysis using Easyfig .
The correlation between GtrC repertoire and serotype
The presence of a specific GtrC family member at locus 2 (Figures 3, 4) appears to correlate with the O-antigen substrate of the host genome. For example, in Kaufmann-White serogroup A, B and D1 genomes, which share a common tri-saccharide O-antigen backbone , , locus 2 contained a GtrC family III operon. In contrast, the same locus 2 site in KW serogroup C1 serovars and C2 were occupied by two adjacent gtr operons encoding GtrC family V and VI members, and a family VII GtrC, respectively (Figure 4). The absence of a family III GtrC, in serogroups C1 and C2 is consistent with the absence of the galactose moiety in the O-antigen that is the receptor for the family III GtrC .
Overall however, the repertoire of gtr operons did not correlate with the Kaufmann-White O-serogroupings. For example, serovars represented by serogroups B, C1 and D1 possessed different numbers of gtr operons (Table S1). Furthermore, different strains of a single serovar can encode different numbers of gtr operons. For example, S. Typhimurium strains LT2, DT2, SL1344 and 14028s have two gtr operons compared to S. Typhimurium strains DT104 and D23580, which possess three gtr operons (Table S1).
In combination with the recombinogenic signature of genes within the gtr operons, these data suggest that phage-mediated acquisition and transfer of gtr operons is ongoing. Acquisition is likely to be restricted by O-antigen substrate, plausibly because the O-antigen is a known (co)-receptor for phage attachment . This and the availability of a suitable genomic integration site will influence the gtr diversity in these bacterial genomes.
O-antigen modification and the GtrC families
gtr functionality has previously been identified for single representatives of GtrC families I , , III  and VIII . Ascribing function to additional GtrC families is an important step in understanding the biological contribution of gtr to Salmonella. Thus, as a first step we determined whether the occurrence of LPS modification could be shown for gtr families present in serovars S. Typhimurium (KW serogroup B; GtrC families III and IV), S. Infantis (KW serogroup C1; GtrC families IV, V and VI) and S. Typhi (KW serogroup D1; GtrC families II and III).
Salmonella strains were constructed that were devoid of all gtr-like operons, which are referred to further as ‘basal’ strains . As we have previously shown, phase variation can occur for some gtr operons . Therefore, strains were also constructed containing a single gtr operon expressed from a well-characterized, constitutive promoter to ensure consistent expression , . LPS banding profiles from wild type and these mutant strains were compared and a shift in the LPS ladder rungs was used as indicator of O-antigen modification and GtrC functionality (Figure 5).
A) S. Typhimurium strain ST4/74 lipopolysaccharide (LPS) comparing the banding profiles of wildtype (WT) LPS to LPS from an ST4/74 strain devoid of all known LPS modification genes termed ‘basal’ (delta gtrSTM_I, delta gtrSTM_II and delta oafA); and mutant strains expressing a single gtr operon gtrSTM_I (GtrC family III) and gtrSTM_II (GtrC family IV) driven by its native promotor, pWT, or the constitutive promoter pTac. B) S. Infantis strain S1326/28 LPS profiles of GtrC family V and family VI operons in comparison to LPS profiles from WT and basal (delta gtrS1326/28_I, gtrS1326/28_II and gtrS1326/28_III) S. Infantis strains. LPS was extracted from overnight bacterial cultures and visualized on silver stained TSDS-PAGE gels. A change in the size of LPS banding patterns is used to infer functionality of gtr operons.
Modification mediated by family III and family IV GtrC was assessed from the representative operons present in the S. Typhimurium strain ST4/74. Consistent with our description of phase variation of the gtrSTM_I operon and the described bias to the OFF phase , no detectable change in the size of the LPS rungs is observed comparing basal and wild type LPS (Figure 5A, compare lanes 1 and 2). However, a shift is observed when the gtrSTM_I operon is expressed from the constitutive promoter (Figure 5A, compare lanes 3 and 2). These data are consistent with the recent report that this locus mediates the addition of glucose to the galactose moiety of the S. Typhimurium O-antigen via an alpha 1–4 linkage (locus STM0557 - STM0559 ). In contrast, even when ensuring constitutive chromosomal expression of the gtrSTM_II operon, representing GtrC family IV, no visible LPS size differences were found in comparison to the ‘basal’ STM strain (Figure 5A, compare lanes 4 and 2).
Similar to the findings of family IV, we were not able to demonstrate visible shifts as a result of expression of family II GtrC in S. Typhi strain BRD948 (data not shown). We think it unlikely that this lack of visible modification is due to transcriptional regulation as the promoter is a constitutive promoter, but we cannot rule out post-transcriptional regulation. A more likely explanation is that for these family II and family IV operons the modification does not yield a detectable shift in this assay, or that the enzymatic target or substrate is absent. The former could be due to either the nature of the moiety added or if gtr mediates a change in sugar linkage.
Finally, the functionality of three gtr operons in S. Infantis strain S1326/28 was examined. While sharing the GtrC family IV with KW serogroup B and C2 serovars, this serogroup C1 serovar also harbors two KW serogroup C1 specific gtr operons, family V (gtrS1326/28_I) and family VI (gtrS1326/28_I1) (Figure 1). The LPS ladder pattern from a WT strain showed clear differences compared to that of the basal mutant (Figure 5B, compare lane 4 and 5), indicating at least one gtr operon mediated modification. This was apparently not due to the S. Infantis GtrC family IV (gtrS1326/28_III) as no shift was observed even when it was constitutively expressed (Figure 5B, compare lanes 8 and 9). The lack of visible modification by S. Infantis GtrC family IV and the S. Typhimurium family IV GtrC (Figure 5A, lanes 1, 2) is therefore unlikely to be due to KW serogroup context. In contrast, expression of the other two families mediated S. Infantis LPS shifts. Family V Gtr (gtrS1326/28_I) when expressed from both the native promoter (pWT) and constitutive promoter (pTac) resulted in visible modification (Figure 5B, compare lanes 1 to 3), whereas for GtrC family VI (gtrS1326/28_I1), modification was only detectable when constitutively expressed (Figure 5B, compare lanes 6 to 8). The latter indicates that some gtr operons may not be expressed under all conditions, either due to phase variation or environmental control. Thus, for the first time we show detectable functionality for GtrC family V and family VI. Overall, of the six GtrC families examined, detectable functionality has now been shown for four GtrC families (I, III, V, VI).
Modified LPS and changes inSalmonella O-serotype
The structure of the O-antigen is inherently linked with the O-serotype used in the current Kaufmann-White Salmonella classification scheme. Thus, gtr-dependent modifications may alter the outcome of O-serotyping according to clinical diagnostic practices. Using the strains described above, we determined whether S. Infantis O-antigens modified by GtrC family V (gtrS1326/28_I) and family VI (gtrS1326/28_II) (Figure 5B, lanes 3 and 7) alter the S. Infantis strain serotype in comparison to the WT and gtr negative ‘basal’ strains (Figure 5B, lanes 4 and 5). All four strains were O-serotyped by a Salmonella reference laboratory and were determined to be O:6,7 positive, consistent with the O:6,7 designation in the KW scheme for serogroup C1 serovars. Additional agglutination analyses with O:14 and O:61 sera, two factors associated with O:6,7 sero-converting phage, did not yield a positive agglutination test in any of the four strains. Thus, the family V and VI gtr –dependent modification of the O-antigen within KW serogroup C1 observed here (Figure 5B) is not identified by current serotyping protocols. These findings highlight the limitations of sero-diagnostic methodologies in differentiating between sero-converting Salmonella isolates that have modified their immunodominant O-antigen.
Dam-dependent phase variable control ofgtr expression is widespread but not universal
We recently determined that expression of three gtr operons representing GtrC families I, II, and III is controlled by phase variation resulting in heterogeneous expression of the operon in a genetically clonal population . This phase variation mechanism is epigenetic and involves Dam, a DNA methyltransferase, and OxyR a transcriptional regulator. Signature sequence elements at the gtr operon that are required for this regulation were identified as four Dam target sequences (GATC) and two overlapping OxyR binding motifs that each contain two GATC sequences .
The conservation of this phase variable promoter architecture was examined by analysis of the sequence 128 bases upstream of the gtrA transcription start site, as previously defined . Based on the 59 operons analyzed here, 43 operons distributed among 8 of the 10 GtrC families (GtrC families I, II, III, V, VI, VII, IX and X) had recognizable signature sequences suggestive of phase variation (Figure S1). Of these 43, 37 had all 4 GATC sites and OxyR-like motifs. The signature sequences of specific GtrC family III, V and VI gtr operons however contained various sequence variations, including some of the gtr operons analyzed for O-antigen modification described above (Figure 6). For example, S. Infantis gtrS1326/28_I lacked GATC-1, S. Paratyphi C gtrRKS4594_I lacked both GATC-1 and GATC-2 and S. Choleraesuis gtrSCB67_I contained only the similar -35 (encompassing GATC-4) and -10 promoter sequences. Of the 59 operons, 16 lacked sequence resembling the phase variation signature sequence, representing all GtrC family IV members and the phage epsilon34 encoded gtr (family VIII).
. Top of the figure is a schematic representation of the phase variable gtr regulatory region showing the location of three OxyR dimer binding sequences (thick black line), GATC sequences (open rectangle), −35 and −10 promoter sequences, transcription start site (+1) and the gtrA open reading frame. The presence of these elements is a signature of gtr phase variation. Below the schematic is an alignment of seven gtr regulatory region sequences showing consensus OxyR dimer binding sequence, GATC sequences in bold and the −35 of the promoter shaded in grey. Sequence “SIN S1326/28_III” is shown as comparison to illustrate lack of phase variation sequence elements. This sequence was aligned based on conserved sequence surrounding the +1 site (not shown). Acronyms represented: STM, S. Typhimurium; SCH, S. Choleraesuis; SIN, S. Infantis. Figure S1 shows the alignment and conserved nucleotides of 128 bp of regulatory sequence from 43 gtr operons.
To confirm the sequence-based prediction of phase variation, transcriptional regulation was examined using a single copy, chromosomal reporter fusion (gtr-lacZ). Phase variation is determined by the occurrence of both lacZ expressing (Lac+/“On”) and non-expressing (Lac−/“Off”) colonies . As predicted, based on the presence of the signature sequence, S. Infantis gtrS1326/28_II (family VI) and S. Choleraesuis gtrSCB67_III (family V) phase varied (Table 1, Figure 6). Consistent with the lack of the signatures sequences, expression of S. Choleraesuis gtrSCB67_I, gtrSCB67_II and gtrSCB67_IV did not phase vary (Table 1, Figure 6).
To address the impact of the rare mutations that were identified at some gtr operons, a similar analysis was carried out for S. Infantis gtrS1326/28_I (family V) with a mutated GATC-1 sequence. This sequence still allowed for phase variation with similar switch frequencies as gtr operons with an intact GATC-1 sequence (Table 1; ). To improve the predictive power of sequence analysis for this regulation, a mutation was introduced into GATC-1 (GATC-1 to CATC-1) for two known phase varying gtr operons, S. Typhimurium gtrLT2_I and gtrP22 . Interestingly, the mutations did not abrogate phase variation in context of the gtrLT2_I sequence (sMV510), but abrogated phase variation in context of the gtrP22 sequence (sMV511, Table 1). This analysis shows that the effect of point mutations in the signature sequence on gtr expression will depend on the sequence context.
Taken together, our experimental data combined with sequence analyses shows that the presence of the Dam/OxyR regulatory sequence elements can be used as a signature for Salmonella phase variation. Our findings highlight that not only are the makeup and complement of gtr operons fluid and variable within the Salmonella species, but there is an additional layer of complexity introduced by phase variation in determining the O-antigen phenotype in salmonellae.
Horizontally acquired genes appear to play a key role in defining O-serovar identity, and retention of O-antigen modification genes such as gtr operons by salmonellae may convey significant biological advantages. In this study, we address the frequency, sequence diversity, functionality and regulatory characteristics of gtr operons in salmonellae.
Through BLAST approaches of 22 complete S. enterica subsp. enterica (subspecies I) genome sequences we identified 52 gtr operons. Each of the subspecies I genomes examined possessed between 1 and 4 gtr operons. It should be noted that the S. enterica genomes screened were of serovars representing Kaufmann-White serogroups A-D1, representing 7 of the 46 KW O-serogroups. Thus, the specific number of GtrC families identified here is not highly relevant, as we predict that more GtrC families will be identified as genomes from serovars of more diverse KW O-serogroups become available. The identification of 3 gtr operons from the draft genomes of S. bongori serovars from KW O-serogroups G and R also shows that gtr operons are not limited to O-serogroups A - D1 and are not restricted to S. enterica subspecies I serovars.
Phylogenetic and clustering analyses of the 55 gtr operons with an additional 4 gtr-like operons derived from publically available Salmonella phage genome sequences, identified 10 GtrC ‘families’ (designated GtrC families I to X). Of these, three GtrC families include strains of only a single Kaufmann-White O-serogroup, serogroup C1 and C2 (Figure 1, GtrC families V, VI and VII). This may indicate that these gtr operons target a moiety in the O-antigen that is specific to KW O-serogroup C1 and C2 serovars. In contrast, the GtrC families (I–IV) reside within genomes from multiple serogroups. This can be understood by considering the biochemical structure of the basal (un-modified) O-antigen. For example, serogroups A, B and D1 all share the same basic tri-saccharide basal O-unit [Gal-Rha-Man], but differ in their side 3,6 di-deoxyhexose sugars, namely paratose, abequose and tyvelose respectively (reviewed in ). We found that these three serogroups may harbor GtrC family I and III members, which can mediate addition of glucose to the galactose moiety via 1–6 (O:1 serotype) ,  and 1–4 linkages (O:122) , respectively. We predict that the Gtr family III will have the same modifying activity in all serogroup A, B and D1 strains, as they all possess the required alpha galactose receptor (ie. O:12 positive). This prediction is supported by the observation that family I and family III GtrC operons were not identified in serogroup C1 genomes, which have a basal O-unit that lacks the galactose moiety . In these genomes, we identified the serogroup C1 unique GtrC families V, VI and VII. In contrast, GtrC family II and family IV were not confined to serogroups with similar O-antigen biochemical structures. Interestingly, we did not find evidence that the O-antigen is a substrate for these two GtrC families based on our LPS gel-based assay and no activity has been assigned to any members of those two families. Collectively, these analyses highlight that the O-antigen structure is a defining factor for the repertoire of GtrC families that any one serovar can possess.
O-antigen modifying operons appear to be associated with bacteriophage genomes that use the O-antigen as their receptor. In fact, all 16 known O-antigen modification operons in Shigella flexneri are associated with mobile elements , . In Salmonella, gtrABC-like genes are present on at least 3 different P22-like temperate phage genomes . Our analyses support these observations and extend the presence of gtr operons across the Salmonella genus. Our genome analyses placed all Salmonella gtr operons within four previously defined ‘regions of difference’ that were purported to be of phage origin , . Consistent with these studies, we found that two of these regions, locus 1 (P22 locus) and locus 2 (SPI-16 locus) contain gtrABC operons within full-length prophage sequences (Figure 3). Locus 2 appears to be a hot-spot for gtr insertion with all gtr-positive serovars harboring gtr in this loci (Figure 3) and some strains containing multiple gtr operons (Figure 4). The identification of GtrC families and KW O-serogroup associations is consistent with the observation that phage are drivers of gtr dissemination, and as such, phage can only lysogenize when their O-antigen structural receptor is present.
Not all gtr-dependent modification is reflected in the current KW scheme. For example, we demonstrated that the apparent modification due to expression of GtrC families V and VI in the KW serogroup C1 serovar S. Infantis (Figure 5B, lanes 3 and 7) did not alter the O-serotype based on standard clinical typing practices or known serogroup C1 subtype sera (O:61). Although not completely unexpected as serotyping is based on a limited number of sera and the KW scheme eliminates many known phage based variations including some known gtr modifications , the findings reiterate that KW O-serogroups are more heterogeneous regarding the composition of expressed surface structures than previously thought. Therefore, the currently applied KW O-serotyping scheme and assignment of O-serogroups does not and cannot differentiate the extensive diversity in O-antigen repertoire. This is in contrast to the situation with Shigella flexneri, where phage associated gtr dependent modification has been retained in the serotyping scheme . Further functional characterization of the gtr families is ongoing, and should provide additional insight into antigenic variation that exists among Salmonella serovars.
The homology between any two GtrA and GtrB clusters was high, >84% and >90% amino acid similarity respectively. In contrast, inter-family GtrC homology was low, <46% similarity. This sequence variation between GtrC families is unlikely to be the result of high diversifying selection but suggests that either they perform distinct functions or they have convergently evolved to perform the same or related functions. Given the lack of sequence conservation, evolutionary linkages between GtrC families could not be inferred, as there is little genomic evidence that they share a common ancestor. Similar observations of sequence diversity have been observed for the GtrC-like O-antigen modification genes of Shigella , and in general, significant sequence divergence is not unusual among functionally related glycosyltransferases . The lack of substantial sequence divergence for GtrA and GtrB is consistent with the Shigella counterparts . Our clustering and recombination analyses identified a mosaic of gene shuffling events leading to new allelic variants of gtr gene sequences. Furthermore, our analysis of KW serogroup C1 genomes is suggestive of gene rearrangements occurring between chromosomal and phage-associated gtr operons resulting in ‘new’ combinations of gtrABC families. The generation of sequence diversity could be facilitated by lysogenization of a strain containing one or more gtr operons by phage with a complimentary gtr repertoire. This would also suggest that evolutionary history of the gtr operon is dynamic and ongoing which in turn blur any phylogenetic inferences.
The presence of premature stop codons in three GtrC gene products (S. Paratyphi A, GtrC family III and S. Choleraesuis, family IV) suggests that not all GtrC families contain functional products. The presence of these rare gtrC pseudogenes may be indicative of recent selection such as immune pressure. Similarly, the gtrB gene from GtrC family II shares a conserved deletion, yet the gtrC in all genomes remain intact. All family II GtrC genomes harbor a second gtr operon, which raises the possibility that this may enable functionality of the family II GtrC gene product. Alternatively, GtrB may not be a functional requirement for this operon. Support for the former possibility comes from analyses in Shigella showing that at least some GtrA and GtrB can be used interchangeably without affecting GtrC function , and our finding of extensive recombination within Salmonella gtrAB gene sequences that are not restricted by gene boundaries. Thus, there may a degree of redundancy in the Gtr system, but further studies are required to confirm this. The presence of multiple gtr operons and the general lack of gtrC mutations within S. enterica subspecies I serovars suggest a selective pressure to retain gtrC-dependent activity, which in turn promotes O-antigen diversity.
In addition to diversity in the repertoire and sequences of GtrC families within Salmonella serovars, we previously found that clonal heterogeneity may arise through regulation of gtr expression by phase variation . Here we describe further evidence that phase variation is widespread among most GtrC families irrespective of Salmonella species, subspecies or serotype. Experimentally verified phase variation correlated with the presence of signature sequences for OxyR- and Dam-dependent regulation . The loss of one of the four Dam target sequences in the signature may abrogate phase variation, however, this is dependent on the sequence context. This likely is a result of the specific OxyR binding sequence yielding different OxyR binding affinity, which in turn will impact the epigenetic regulation , . The absence or presence of this characterized phase variation mechanism may still allow input from additional factors on gtr expression, even though we are yet to find growth conditions that significantly alter expression level, or induce or eliminate phase variation (unpublished). The relevance of the apparent lack of phase variation in specific gtr families and individual operons is not clear at present but as gtr-specific modifications and roles are elucidated, we may gain further insight. Interestingly, there is no evidence suggesting that phase variation controls expression of the Shigella flexneri gtr operon that has been shown to have a role in virulence . We furthermore propose that phase variable gtr expression contributes to the variability in biochemical glucosylation observed in some O-antigen NMR studies (reviewed in , ). This in turn underpins the weak and variable agglutination reactions that can be reported from diagnostic laboratories, and that are represented by brackets in the Kaufmann-White Salmonella classification scheme .
The large and variant repertoire of gtr operons in and among Salmonella serovars shows that there is an additional layer of LPS heterogeneity that is largely not being detected by current serotyping methodologies. The incorporation of phase variable elements into gtr expression promotes further complexity onto the structure of the immunodominant O-antigen. Full understanding of the impact of gtr expression will require comprehensive analyses that take into consideration the biochemistry of modification, serovar context and the related host and virulence variables. Based on the current insight, it is likely that these gtr-mediated strain differences are important for understanding Salmonella virulence, persistence and spread, and based on findings with Shigella  could impact on vaccine development. Taken together, utilizing the extensive occurrence and significant sequence variation between stochastically expressed GtrC families could aid in developing higher resolution molecular serotyping methods  for Salmonella diagnostics.
Materials and Methods
Identification ofgtr operons within Salmonella genome sequences
In order to examine the distribution of gtr operons, 24 publically available finished Salmonella genomes sequences were obtained from the European Bioinformatics Institute (listed in Table S1). To address gtr presence in non-subspecies I S. enterica serovars four draft genomes from The Wellcome Trust Sanger Institute (http://www.sanger.ac.uk/cgi-bin/blast/submitblast/salmonella) or The Genome Institute at Washington University (http://genome.wustl.edu/tools/blast) were screened using BLAST. Furthermore, draft genome sequences of 28 S. bongori genomes comprising of 8 different KW O-serogroups were also analyzed (Table S2, ). Four publically available Salmonella bacteriophage genome sequences that are known to encode for gtr operons, specifically Phi P22, Phi ST64T, Phi DT104 and epsilon 34, were also included –, , .
The presence of gtr genes in complete genome sequences was identified in two ways. The gtrC from bacteriophage P22 was used to search Salmonella genome sequences using BLASTn. Hits were considered to be gtr if the gtrC homolog sequence had neighboring gtrB and/or gtrA-like sequences. All identified gtrA, gtrB and gtrC genes were placed in a sequence file against which genome sequences were re-analyzed by BLASTn to confirm preliminary findings. Second, genomes representing the same KW serogroup were aligned using the Artemis Comparison Tool (ACT)  to examine site-specific integration of gtr operons. This also enabled characterization of regions flanking gtr operons for signatures of mobile genetic elements such as transposable elements or phage-like sequences. Examination of trans-membrane topology of the Gtr gene products were predicted using a hidden Markov model, TMHMM 2.0 .
For consistency, we adopt the O-antigen modifying glycosyltransferase (gtr) nomenclature used in our previous study , which is a modification of that proposed previously . Putative gtr-like operons identified through BLAST analyses were numerically assigned based on their genomic location in relation to the purported oriC. For example, S. Typhimurium gtrLT2_I represents the ‘first’ gtr operon (relative to the oriC) in the genome of S. Typhimurium strain LT2. Roman numeral designations were assigned based on ascending gene number and was employed to avoid conflict with strain numbering. Assigned acronyms are summarized in Table S1. This descriptive gtr gene nomenclature precisely defines the serovar and strain, and thus, is beneficial given the large variation in gtr repertoire within and among strains and serovars as identified in this study. It can be expanded upon without loss of clarity.
Multiple sequence alignment, phylogenetic, clustering and recombination analyses
To determine the relationship of the GtrA, GtrB and GtrC gene products, phylogenetic analyses were performed on aligned sequences. All sequence alignments were carried out on both nucleotide and translated nucleotide sequences for accuracy. Automated alignments using ClustalW2  were manually adjusted to account for differences in the size of gtr gene operons. The genome sequences represented in this study are biased towards serovars representing significant mammalian pathovars. For example, S. Typimurium is represented by 6 genome sequences. All conserved gtr operons identified within S. Typhimurium were identical when aligned. Thus, in order to reduce unnecessary over-representation of identical sequences, the gtr operons from serovar S. Typhimurium strains LT2, 14028s, SL1344, D23580, DT2 and DT104 are represented by one single acronym ‘STM’. As all S. Typhimurium genomes examined in this study contain a minimum of two conserved gtr-like operons, the final number of gtr operons used in phylogenetic analyses was reduced from 58 to 48. Phylogenetic trees were constructed for GtrA, GtrB and GtrC using both maximum likelihood and baysian interference methods. Maximum likelihood trees were estimated in PhyML using optimal WAG model for protein and GTR model for nucleotide sequences with rate heterogeneity estimated from the data. All other options were default. One thousand non-parametric bootstraps were applied for robustness.
To examine the clustering of the gtr genes, we performed a statistical genetic analysis using the software package Bayesian Analysis of Population Structure, BAPS , . Genetic clusters were designated using a nucleotide sequence alignment for the gtrA, gtrB and gtrC genes. BAPS clusters sequences based on shared polymorphisms using a non-reversible stochastic optimization algorithm that identifies the number of clusters that best fits the dataset. As GtrC is purported to convey serotype-specific modifications, the gtrC gene was used to define gtr clusters, designated in this study as GtrC families.
To examine the role of recombination within the evolutionary history of the gtr operon, homoplasic SNPs were identified for each branch of gtrA and gtrB phylogenetic trees and mapped onto concatenated gtrAB sequences using PAML . Clusters of matching homoplasies located along different branches of the phylogenetic tree were used as markers of recombination.
Bacterial strains and culture conditions
In order to examine O-antigen modifying activity of the gtr operons, Salmonella enterica subsp. enterica serovars Infantis (S. Infantis) strain S1326/28 and Typhimurium (S. Typhimurium) strain ST4/74 were used. All strains were a kind gift from Dr Rob Kingsley, The Wellcome Trust Sanger Institute, UK. Bacteria were routinely cultured aerobically at 37°C in Luria-Bertani (LB) broth or on LB with 1.75% agar. Strains harboring the temperature sensitive plasmids pKD46 and pSIM18 and were grown aerobically at 30°C. Where required, cultured strains carrying antibiotic resistant markers were supplemented with antibiotics (Sigma) at the following concentrations: ampicillin, 100 µg ml−1; kanamycin, 30 µg ml−1; hygromycin, 100 µg ml−1 and tetracycline, 12.5 µg ml−1.
Generation ofgtr mutants in Salmonella
S. Typhimurium and S. Infantis with defined mutations were constructed using the Lambda-Red recombination system , . gtr specific oligonucleotides were designed for site specific deletion of gtr operons through amplification of a kanamycin marker (kan) from plasmid pKD13  or the tetracycline (tetRA) cassette from transposon Tn10dTc . Primer sequences are shown in Table S3. Confirmation of integration and allelic exchange was performed by PCR as described previously ,  or by DNA sequencing.
For construction of strains with multiple mutations in chromosomal genes, the tetRA cassette was replaced with an oligonucleotide (Table S3) using the Lambda-Red approach in combination with counter-selection for loss of tetracycline resistance ,  to generate in-frame deletions on the chromosome in the absence of a genetic scar. This same counter-selection procedure was employed to generate constitutive expression of gtr operons. Specifically, 154 bp of S. Infantis gtrS1326/28_I; 174 bp gtrS1326/28_II and 263 bp of gtrS1326/28_III upstream of the assigned gtrA start codon were first replaced with tetRA, after which tetRA was replaced with 116 bp containing the pTac promoter  derived from pMAL-c2x (New England Biolabs). pTac is a constitutive promoter in Salmonella since they do not encode the lac operon and repressor. Expression of gtr was confirmed by reverse transcriptase PCR or by LPS gel analyses.
Crude LPS extracts were prepared from 3 ml overnight bacterial LB cultures grown with aeration at 37°C. Cultures were pelleted by centrifugation at 3 700 g for 15 min at 4°C (Sorvall) and washed three times with 0.9% NaCl. Cells were resuspended in 125 µl ice-cold buffer solution (60 mM Tris-HCL, 1 mM EDTA, pH 6.8) containing 2% SDS and boiled for 5 mins. Lysed cells were diluted in 875 µl of buffer solution without SDS. Enzymatic digestion of nucleic acids was performed using RNase (Roche) and DNase (Promega) for four hours at 37°C, followed by proteinase K (100 µg) treatment overnight at 50°C. The crude LPS preparations were stored at 4°C for analysis.
Tricine-sodium dodecyl sulfate-polyacrylamide gel electrophoresis (TSDS-PAGE) analysis and silver staining
For analyses of LPS profiles from Salmonella mutants, crude LPS preparations were analysed on 20 cm×15 cm; 10.5% separating, 4% stacking TSDS-PAGE gels as modified from . An equal volume of 2× sample buffer (6% sodium dodecyl sulfate; 6% 2-mercaptoethanol; 10 mM dithiothreitol; 46% glycerol; 60 mM Tris-HCl, pH 8.0; 0.1% bromophenol blue) was added to the crude LPS preparations and boiled for 5 min. LPS was visualized by oxidative silver staining, modified from . Briefly, gels were fixed overnight in fixative solution (30% ethanol; 10% acetic acid) then oxidized with 0.7% periodic acid in fixative solution for 10 mins. Gels were washed 3×15 min in ddH2O before staining with 0.1% silver nitrate for 30 mins. Gels were washed briefly in excess ddH2O before developing in a 3% sodium carbonate, 0.02% formaldehyde solution until desired staining intensities were achieved.
Analysis of thegtr phase varion
The approach and methods used to analyze gtr phase variation were performed as described  using a CRIM-based , chromosomal, transcriptional fusion to lacZ in an S. Infantis S1326/28 or S. Typhimurium LT2 background. Primer sequences are shown in Table S3. Expression levels were determined using the β-galactosidase assay  and, where relevant, correction for % On in the culture was introduced by calculating and expressing the Miller units per 100% “On”.
Serotypic analysis ofgtr mutants
Salmonella mutants were serotyped by routine clinical diagnostic procedures at the Scottish Salmonella, Shigella and Clostridium difficile Reference Laboratory, Glasgow, where standard serum agglutination assays were performed using purified O-antigen serum routinely used for diagnostic procedures. Sub-type O:61 specific serum was a kind gift from Dr Peter Roggentin (Institute for Hygiene and Environment, Hamburg) and was used in-house in a standard agglutination assay.
Regulatory region sequence alignments from 43 gtr operons harboring sequence elements similar to that of the known phase variable gtrLT2_I regulon (GtrC family III) . 128 base pair regions upstream of the putative +1 transcriptional start site  were aligned using ClustalW2. Asterisks indicate conserved residues. Three putative OxyR binding sites are shown and the Dam methyltransferase recognition sequences (GATC) are represented in bold font and underlined as defined in . The +1 transcription start site and −35 and −10 promoter sequences  are indicated in grey shaded regions. The phase variable promoter architecture is conserved across most gtr regulatory regions. Strain acronyms are defined in Table S1.
Genome coordinates of 52 gtr operons identified in 22 Salmonella enterica subsp. enterica (subspecies I) serovar genome sequences.
S. bongori isolates+ screened for the presence of S. enterica-like gtr operons.
We would like to acknowledge P. Ashton, M. Lakins, T. Connor, and T. Taynton for technical support and R. Kingsley, G. Dougan, M. Susskind, and S. Casjens for discussions.
Conceived and designed the experiments: MRD SEB SRH NRT MWvdW. Performed the experiments: MRD SEB SRH. Analyzed the data: MRD SEB SRH NRT MWvdW. Wrote the paper: MRD SEB NRT MWvdW.
- 1. Grimont PAD, Weill F-X (2007) Antigenic Formulae of the Salmonella Serovars. Paris: WHO Collaborating Centre for Reference and Research on Salmonella, Institut Pasteur. Available: http://www.pasteur.fr/ip/portal/action/WebdriveActionEvent/oid/01s-000036-089.
- 2. Guibourdenche M, Roggentin P, Mikoleit M, Fields PI, Bockemuhl J, et al. (2010) Supplement 2003–2007 (No. 47) to the White-Kauffmann-Le Minor scheme. Res Microbiol 161: 26–29.
- 3. Butela K, Lawrence J (2010) Population Genetics of Salmonella: Selection for Antigenic Diversity. In: Robinson DA, Falush D, Feil EJ, editors. Bacterial Population Genetics in Infectious Disease. Hoboken, New Jersey: John Wiley & Sons, Inc. pp 287–319.
- 4. Reeves P (1993) Evolution of Salmonella O antigen variation by interspecific gene transfer on a large scale. Trends Genet 9: 17–22.
- 5. Slauch JM, Lee AA, Mahan MJ, Mekalanos JJ (1996) Molecular characterization of the oafA locus responsible for acetylation of Salmonella typhimurium O-antigen: oafA is a member of a family of integral membrane trans-acylases. J Bacteriol 178: 5904–5909.
- 6. Wang L, Andrianopoulos K, Liu D, Popoff MY, Reeves PR (2002) Extensive variation in the O-antigen gene cluster within one Salmonella enterica serogroup reveals an unexpected complex history. J Bacteriol 184: 1669–1677.
- 7. Bogomolnaya LM, Santiviago CA, Yang HJ, Baumler AJ, Andrews-Polymenis HL (2008) ‘Form variation’ of the O12 antigen is critical for persistence of Salmonella Typhimurium in the murine intestine. Mol Microbiol 70: 1105–1119.
- 8. Fukazawa Y, Hartman PE (1964) A P22 Bacteriophage Mutant Defective in Antigen Conversion. Virology 23: 279–283.
- 9. Pedulla ML, Ford ME, Karthikeyan T, Houtz JM, Hendrix RW, et al. (2003) Corrected sequence of the bacteriophage P22 genome. J Bacteriol 185: 1475–1477.
- 10. Vander Byl C, Kropinski AM (2000) Sequence of the genome of Salmonella bacteriophage P22. J Bacteriol 182: 6472–6481.
- 11. Villafane R, Zayas M, Gilcrease EB, Kropinski AM, Casjens SR (2008) Genomic analysis of bacteriophage epsilon 34 of Salmonella enterica serovar Anatum (15+). BMC Microbiol 8: 227.
- 12. Makela PH, Makela O (1966) Salmonella antigen 12-2: genetics of form variation. Ann Med Exp Biol Fenn 44: 310–317.
- 13. Allison GE, Verma NK (2000) Serotype-converting bacteriophages and O-antigen modification in Shigella flexneri. Trends Microbiol 8: 17–23.
- 14. Lukacova M, Barak I, Kazar J (2008) Role of structural variations of polysaccharide antigens in the pathogenicity of Gram-negative bacteria. Clinical microbiology and infection : the official publication of the European Society of Clinical Microbiology and Infectious Diseases 14: 200–206.
- 15. Lerouge I, Vanderleyden J (2002) O-antigen structural variation: mechanisms and possible roles in animal/plant-microbe interactions. FEMS Microbiol Rev 26: 17–47.
- 16. van der Woude MW, Baumler AJ (2004) Phase and antigenic variation in bacteria. Clin Microbiol Rev 17: 581–611.
- 17. Wildschutte H, Lawrence JG (2007) Differential Salmonella survival against communities of intestinal amoebae. Microbiology 153: 1781–1789.
- 18. Wildschutte H, Wolfe DM, Tamewitz A, Lawrence JG (2004) Protozoan predation, diversifying selection, and the evolution of antigenic diversity in Salmonella. Proc Natl Acad Sci U S A 101: 10644–10649.
- 19. Kim M, Ryu S (2012) Spontaneous and transient defence against bacteriophage by phase-variable glucosylation of O-antigen in Salmonella enterica serovar Typhimurium. Mol Microbiol 86: 411–425.
- 20. Broadbent SE, Davies MR, van der Woude MW (2010) Phase variation controls expression of Salmonella lipopolysaccharide modification genes by a DNA methylation-dependent mechanism. Mol Microbiol 77: 337–353.
- 21. Duerr CU, Zenk SF, Chassin C, Pott J, Gutle D, et al. (2009) O-antigen delays lipopolysaccharide recognition and impairs antibacterial host defense in murine intestinal epithelial cells. PLoS Pathog 5: e1000567.
- 22. Reeves P (1995) Role of O-antigen variation in the immune response. Trends Microbiol 3: 381–386.
- 23. Corander J, Marttinen P, Siren J, Tang J (2008) Enhanced Bayesian modelling in BAPS software for learning genetic structures of populations. BMC Bioinformatics 9: 539.
- 24. Corander J, Tang J (2007) Bayesian analysis of population structure based on linked molecular information. Math Biosci 205: 19–31.
- 25. Vernikos GS, Parkhill J (2006) Interpolated variable order motifs for identification of horizontally acquired DNA: revisiting the Salmonella pathogenicity islands. Bioinformatics 22: 2196–2203.
- 26. Knirel YA (2011) Structure of O-Antigens. In: Knirel YA, Valvano MA, editors. Bacterial Lipopolysaccharides. New York: Springer-Verlag/Wien. pp.41–116.
- 27. Lindberg AA (1973) Bacteriophage receptors. Annu Rev Microbiol 27: 205–241.
- 28. de Boer HA, Comstock LJ, Vasser M (1983) The tac promoter: a functional hybrid derived from the trp and lac promoters. Proc Natl Acad Sci U S A 80: 21–25.
- 29. Broadbent S, van der Woude M, Aziz N (2010) Accurate and simple sizing of primer extension products using a non-radioactive approach facilitates identification of transcription initiation sites. J Microbiol Methods 81: 256–258.
- 30. Sun Q, Knirel YA, Lan R, Wang J, Senchenkova SN, et al. (2012) A Novel Plasmid-Encoded Serotype Conversion Mechanism through Addition of Phosphoethanolamine to the O-Antigen of Shigella flexneri. PloS one 7: e46095.
- 31. Samuel G, Reeves P (2003) Biosynthesis of O-antigens: genes and pathways involved in nucleotide sugar precursor synthesis and O-antigen assembly. Carbohydr Res 338: 2503–2519.
- 32. Guan S, Bastin DA, Verma NK (1999) Functional analysis of the O-antigen glucosylation gene cluster of Shigella flexneri bacteriophage SfX. Microbiology 145: 1263–1273.
- 33. Wallecha A, Munster V, Correnti J, Chan T, van der Woude M (2002) Dam- and OxyR-dependent phase variation of agn43: essential elements and evidence for a new role of DNA methylation. J Bacteriol 184: 3338–3347.
- 34. West NP, Sansonetti P, Mounier J, Exley RM, Parsot C, et al. (2005) Optimization of virulence functions through glucosylation of Shigella LPS. Science 307: 1313–1317.
- 35. Theillet FX, Simenel C, Guerreiro C, Phalipon A, Mulard LA, et al. (2011) Effects of backbone substitutions on the conformational behavior of Shigella flexneri O-antigens: implications for vaccine strategy. Glycobiology 21: 109–121.
- 36. Wattiau P, Boland C, Bertrand S (2011) Methodologies for Salmonella enterica subsp. enterica subtyping: gold standards and alternatives. Appl Environ Microbiol 77: 7877–7885.
- 37. Fookes M, Schroeder GN, Langridge GC, Blondel CJ, Mammina C, et al. (2011) Salmonella bongori provides insights into the evolution of the salmonellae. PLoS Pathog 7: e1002191.
- 38. Mmolawa PT, Schmieger H, Tucker CP, Heuzenroeder MW (2003) Genomic structure of the Salmonella enterica serovar Typhimurium DT 64 bacteriophage ST64T: evidence for modular genetic architecture. J Bacteriol 185: 3473–3475.
- 39. Tanaka K, Nishimori K, Makino S, Nishimori T, Kanno T, et al. (2004) Molecular characterization of a prophage of Salmonella enterica serotype Typhimurium DT104. J Clin Microbiol 42: 1807–1812.
- 40. Carver TJ, Rutherford KM, Berriman M, Rajandream MA, Barrell BG, et al. (2005) ACT: the Artemis Comparison Tool. Bioinformatics 21: 3422–3423.
- 41. Krogh A, Larsson B, von Heijne G, Sonnhammer EL (2001) Predicting transmembrane protein topology with a hidden Markov model: application to complete genomes. J Mol Biol 305: 567–580.
- 42. Larkin MA, Blackshields G, Brown NP, Chenna R, McGettigan PA, et al. (2007) Clustal W and Clustal X version 2.0. Bioinformatics 23: 2947–2948.
- 43. Yang Z (1997) PAML: a program package for phylogenetic analysis by maximum likelihood. Comput Appl Biosci 13: 555–556.
- 44. Datsenko KA, Wanner BL (2000) One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products. Proc Natl Acad Sci U S A 97: 6640–6645.
- 45. Yu D, Ellis HM, Lee EC, Jenkins NA, Copeland NG, et al. (2000) An efficient recombination system for chromosome engineering in Escherichia coli. Proc Natl Acad Sci U S A 97: 5978–5983.
- 46. Karlinsey JE (2007) lambda-Red genetic engineering in Salmonella enterica serovar Typhimurium. Methods Enzymol 421: 199–209.
- 47. Bochner BR, Huang HC, Schieven GL, Ames BN (1980) Positive selection for loss of tetracycline resistance. J Bacteriol 143: 926–933.
- 48. Lesse AJ, Campagnari AA, Bittner WE, Apicella MA (1990) Increased resolution of lipopolysaccharides and lipooligosaccharides utilizing tricine-sodium dodecyl sulfate-polyacrylamide gel electrophoresis. J Immunol Methods 126: 109–117.
- 49. Kittelberger R, Hilbink F (1993) Sensitive silver-staining detection of bacterial lipopolysaccharides in polyacrylamide gels. J Biochem Biophys Methods 26: 81–86.
- 50. Haldimann A, Wanner BL (2001) Conditional-replication, integration, excision, and retrieval plasmid-host systems for gene structure-function studies of bacteria. J Bacteriol 183: 6384–6393.
- 51. Miller J (1972) Experiments in Molecular Genetics. Cole Spring Harbor, NY: Cold Spring Harbor Laboratory.
- 52. Sullivan MJ, Petty NK, Beatson SA (2011) Easyfig: a genome comparison visualizer. Bioinformatics 27: 1009–1010.