A Novel Plasmid-Encoded Serotype Conversion Mechanism through Addition of Phosphoethanolamine to the O-Antigen of Shigella flexneri

Shigella flexneri is the major pathogen causing bacillary dysentery in developing countries. S. flexneri is divided into at least 16 serotypes based on the combination of antigenic determinants present in the O-antigen. All the serotypes (except for serotype 6) share a basic O-unit containing one N-acetyl-d-glucosamine and three l-rhamnose residues, whereas differences between the serotypes are conferred by phage-encoded glucosylation and/or O-acetylation. Serotype Xv is a newly emerged and the most prevalent serotype in China, which can agglutinate with both MASF IV-1 and 7,8 monoclonal antibodies. The factor responsible for the presence of MASF IV-1 (E1037) epitope has not yet been identified. In this study, we analyzed the LPS structure of serotype Xv strains and found that the MASF IV-1 positive phenotype depends on an O-antigen modification with a phosphoethanolamine (PEtN) group attached at position 3 of one of the rhamnose residues. A plasmid carried gene, lpt-O (LPS phosphoethanolamine transferase for O–antigen), mediates the addition of PEtN for serotype Xv and other MASF IV-1 positive strains. These findings reveal a novel serotype conversion mechanism in S. flexneri and show the necessity of further extension of the serotype classification scheme recognizing the MASF IV-1 positive strains as distinctive subtypes.


Introduction
Shigella flexneri is the major pathogen causing bacillary dysentery in developing countries. It is estimated that Shigella is responsible for approximately 164.7 million shigellosis cases annually worldwide, resulting in 1,100,000 deaths, with the majority involving children under five years old [1].
S. flexneri is divided into various serotypes based on the combination of antigenic determinants present in the O-antigen of the cell envelope lipopolysaccharide (LPS). To date, at least 16 serotypes have been recognized [2,3,4,5]. Except for serotype 6, all have the same basic repeating tetrasaccharide unit, comprised of R2)-a-L-Rhap III -(1R2)-a-L-Rhap II -(1R3)-a-L-Rhap I -(1R3)-b-D-GlcpNAc-(1R [4]. The basic O-antigen structure is referred to as serotype Y characterized by a single group 3,4 antigenic determinant. The addition of glucosyl and/or O-acetyl residues to different sugars in the tetrasaccharide unit results in the type-(i.e. I, II, III, IV, V, IC) and group-(i.e. 3,4; 6; 7,8) specific antigenic determinants [3]. The O-acetyl on Rha I defines group 6 determinant in serotype 3a, 3b, 1b, and 4b strains [6,7]. Determinants specific for type I, IC, II, IV, V and group 7,8 antigens are associated with glucosylation on various sugar residues in the tetrasaccharide unit. The genes responsible for the O-antigen glucosylation [gtrA, gtrB, and gtr (type)] have been identified [3,8,9,10,11]. They are arranged in a single operon known as the gtr cluster [12]. gtrA and gtrB are highly conserved and interchangeable, while gtr (type) is unique and encodes the glucosyltransferase responsible for the attaching of a glucosyl group to a specific sugar in the tetrasaccharide repeat unit of the O-antigen [3,12]. The O-acetylation of Rha I depends on the presence of the oac gene for O-acetyl transferase [6,7]. All the Oantigen modification genes known to date are encoded by seven temperate bacteriophages or prophages (SfI, SfIC, SfII, Sf6, SfIV, SfV and SfX), which are integrated into the conserved sites of the host S. flexneri genome [3,6,8,9,10,11,13,14].
Since serotype Xv can agglutinate with both group antigenspecific monoclonal antibodies MASF IV-1 and 7,8, it was initially named as serotype 4X and then as 4c [15,16]. Compared to typical serotype X, serotype Xv strains carry an additional new Oantigenic epitope MASF IV-1 (also called E1037, an antigenic determinant specific for MASF IV-1 antibody) [2]. Such antigenic determinant has also been reported to be presented on some S. flexneri serotypes 4a, Y and 6 strains [15,17]. The genes responsible for the group 7,8 antigen have been identified on a SfX prophage integrated into the host genome [2]. However, the factor(s) responsible for the presence of MASF IV-1 or E1037 antigenic determinant in serotype Xv and other MASF IV-1 positive serotypes had not yet been identified.
In this study, we analyzed the O-antigen structure of serotype Xv strains and found a phosphoethanolamine (PEtN) residue attached at position 3 of Rha II , which is absent from the typical serotype X O-antigen. This modification was shown to confer the MASF IV-1 positive phenotype in serotype Xv strains. The gene named as lpt-O, encoding an LPS phosphoethanolamine transferase for O-antigen was identified to be responsible for carrying out the PEtN modification in serotype Xv and other MASF IV-1 positive strains and was found to be carried on a plasmid.

Ethics Statement
This study was reviewed and approved by the ethics committee of National Institute for Communicable Disease Control and Prevention, the Chinese CDC. S. flexneri strains were acquired with the written informed consent of the diarrhea patients and with the approval of the ethics committee of National Institute for Communicable Disease Control and Prevention, according to the medical research regulations of Ministry of Health (permit number 2007-17-3).

Bacterial Strains, Plasmid and Culturing Condition
Serotype Xv strains 2002017 [2] and 2003055 were used for LPS structure analysis. Serotype X strain 51580 (amp s ) and 4a strain NCTC 9725 (amp s ) were used as hosts for lpt-O gene expression. pMD20-T Vector (TaKaRa) was used for DNA sequencing and expression vector. E. coli JM109 was used for plasmid propagation. S. flexneri strains used for plasmid profiling and lpt-O gene detection analysis were isolated from diarrheal patients in China or purchased from National Collection of Type Cultures (NCTC), UK. Strains were grown in a 37uC incubator or orbital shaker in Luria-Bertani broth (LB) supplemented with ampicillin (100 mg ml 21 ) when appropriate.

Isolation of Lipopolysaccharides and Polysaccharides
LPS were isolated from dried bacterial cells by the phenol-water method [18]. The crude extract without separation of the layers was dialyzed against distilled water, nucleic acids and proteins were precipitated by adding aqueous 50% trichloroacetic acid at 4uC to reduce pH to 2, the supernatant was dialyzed against distilled water and freeze-dried to give purified LPS in yields of 5.6-7.6% of dried cells mass. Delipidation of the LPS was performed with aqueous 2% acetic acid (6 ml) at 100uC until precipitation of lipid A. The precipitate was removed by centrifugation (13,0006g, 20 min), and the supernatant was fractionated by gel-permeation chromatography on a column (5662.6 cm) of Sephadex G-50 Superfine (Amersham Biosciences, Sweden) in 0.05 M pyridinium acetate buffer, pH 4.5, monitored with a differential refractometer (Knauer, Germany). Highmolecular mass O-polysaccharides were obtained in yields of 27-44% of LPS mass.

NMR Spectroscopy
Samples were deuterium-exchanged by freeze-drying twice from 99.9% D 2 O and then examined as solutions in 99.95% D 2 O at 40uC. NMR spectra were recorded on an Avance II 600 spectrometer (Bruker, Germany) using internal sodium 3-(trimethylsilyl) propanoate-2,2,3,3-d 4 (d H 0) and acetone (d C 31.45) as references. Two-dimensional NMR spectra were obtained using standard Bruker software and parameters set as described [19]. TopSpin 2.1 program was used to acquire and process the NMR data.

Slide Agglutination Analysis
The serotypes of S. flexneri strains were determined by slide agglutination test using two serotyping kits specific for all S. flexneri type-and group-factor antigens: (i) a commercially available monovalent antisera kit (Denka Seiken, Japan) and (ii) monoclonal antibody reagents (Reagensia AB, Sweden), as described in the manufacturers' protocols.
PCR products using primers lpt-O-1were purified and cloned into pMD20-T T-A cloning vector (TaKaRa, Japan) to generate the lpt-O expression vector pSQZ. The pSQZ was further transformed into ampicillin sensitive strains 51580 (serotype X) and NCTC 9725 (serotype 4a).

Plasmid Profiling and Southern Hybridization Analyses
A plasmid purification kit (Qiagen, Germany) was used to isolate the plasmids in accordance with the manufacturer's recommendations. Plasmid DNA was separated by electrophoresis with a Chef Mapper system (Bio-Rad) on a 1% SeaKem Gold agarose gel and visualized by ethidium bromide staining. Plasmids isolated from strain 2002017 were used as the positive control. Lambda DNA cleaved with HindIII (TaKaRa, Japan) was used as electrophoresis markers.
The plasmid DNA separated on Gold agarose gel were further transferred onto a nylon membrane (Amersham) using a Vacuum Blotter (Bio-Rad). Southern hybridization was performed using an ECL TM Direct Nucleic Acid Labelling and Detection System (Amersham) as recommended by the manufacturer. DNA product amplified from strain 2002017 using primer pair lpt-O-2 was prepared for a biotin-labeling DNA probe.

Identification of a Novel O-antigen Modification that gives Rise to the MASF IV-1 Reactivity in Serotype Xv Strains
To identify the MASF IV-1 determinant in serotype Xv strains, the isolated O-polysaccharide of serotype Xv strain 2002017 was analyzed by NMR techniques. The 1 H NMR and 13 C NMR spectra of the O-polysaccharide demonstrated a non-O-acetylated pentasaccharide O-unit containing one residue each of glucose (Glc) and N-acetylglucosamine (GlcNAc) and three residues of Rha ( Fig. 2B). However, comparison of the 1 H NMR and 13 C NMR spectra of the serotype X O-polysaccharide ( Fig. 2A) with those of serotype Xv showed additional signals for a PEtN group at d H 3.29 and 4.15, d C 41.4 and 63.2 in the latter (Fig. 2B). There are also other differences in the positions of some signals of Rha II (RII) as  Numbers refer to protons in sugar residues denoted as follows: G, Glc; GN, GlcNAc; RI, Rha I ; RII, Rha II ; RIII, Rha III . In the spectrum of the transformant 51580_Xv O-polysaccharide, the annotations for the signals of the minor bisphosphorylated O-unit are italicized. The most important difference between X and Xv is the presence in the latter of peaks for ethanolamine, which are annotated as EtN 1 and EtN 2. The different positions of the well as the intensities of the peaks for Rha I 1 (RI 1), Rha III 1 (RIII 1), and Rha II 1 (RII 1) between X and Xv (Fig. 2), particularly because the peaks for RI 1 and RIII 1 coincided in the spectrum of X, whereas the peaks for RII 1 and RIII 1 coincided in the spectrum of Xv.
The NMR spectra of the serotype Xv O-polysaccharide were assigned and analyzed using two-dimensional 1 H, 1 H and 1 H, 13   The O-antigen of another serotype Xv strain 2003055 was also analyzed and found to have the same structure as that of 2002017, suggesting that the PEtN modification is common among serotype Xv strains and responsible for the MASF IV-1 positive phenotype in serotype Xv.
A Plasmid Borne Gene (lpt-O) is Required for MASF IV-1 Reactivity in Serotype Xv Strains and Conversion of Serotype X to Serotype Xv Addition of PEtN to the inner core lipopolysaccharide (LPS) or lipooligosaccharide (LOS) has been reported in a number of bacterial species [22,23,24,25,26,27] and is carried out by various PEtN transferases. The known genes involved include lpt3 (NMB2012), lptA (NMB1638), lpt6 (NMA0408) (all Neisseria meningitides), cptA (STM4118), pmrC (STM429) (both Salmonella Typhimurium) and eptB (ZP_04872785) (Escherichia coli) [22,23,25]. We used these gene sequences to search the S. flexneri 2002017 genome including chromosome and plasmid sequences by tBLASTn and BLASTP for homologous proteins. Seven homologous proteins with 22-99% identity and 39-99% similarity were identified, including 6 from the chromosome and 1 from one of the plasmids, pSFxv_2. The 7 homologs in the 2002017 genome were further compared with other sequenced S. flexneri genomes, Sf301, 2457t and 8401, which are all MASF IV-1 negative phenotype signals for C3 of Rha II (non-phosphorylated in X and phosphorylated in Xv) are shown by a dotted line. doi:10.1371/journal.pone.0046095.g002 strains. All chromosomal homologs were found present in these 3 genomes, whereas the plasmid borne homolog (SFxv_5135) was unique, suggesting that SFxv_5135 may be responsible for the PEtN modification in serotype Xv strains.
To confirm this hypothesis, we first performed PCR assay with 9 more serotype Xv strains (2008129, 2008131, 05AH016, 05AH022, 05AH027, 05AH030, 06HN005, 06HN019 and 2008164 ) and 5 serotype X strains (06HN400, 05BJ002, 03HL001, 03HL020 and 2003055) to determine whether the lpt-O gene is carried exclusively by serotype Xv strains. The expected fragment of 1,098 bp can only be obtained from serotype Xv strains, but not from serotype X strains. Sequencing of the PCR products from these 9 serotype Xv strains showed that all were identical to that of 2002017.
Plasmid pSFXv_2 is a double-stranded circular plasmid of 6,850 bp in length. BLAST search revealed no homology with any plasmids or bacterial genomes in the NCBI database, except for 2 regions (4186-5525 and 6058-6847) showing high similarity (96% and 78% identity, respectively) to those of pJHCMW1, a multiresistance plasmid from Klebsiella pneumoniae [28]. There are 11 orfs coding for a mobilization protein, a replication initiation protein, a lipoprotein and a PEtN transferase discovered in this study as well as 7 proteins of unknown function (Fig. 1A).
To determine whether the lpt-O carrying pSFxv_2 plasmid is present in all serotype Xv strains, we analyzed the plasmid profiles of 59 serotype Xv and 40 serotype X strains by electrophoresis with 10 plasmid profiles (serotype Xv strains 2002017, 2008129, 2008131, 05AH022 and 05AH027, and serotype X strains 06HN400, 05BJ002, 03HL001, 03HL020 and 2003055 ) shown in Figure 1B. All strains presented a plasmid profile with 3-7 plasmids, from 2 kb to 10 kb. Note that plasmids larger than 20 kb were not resolvable on this gel and were not included. Compared to serotype X strains, all Xv strains had a 10-kb plasmid band (Fig. 1B). Southern hybridization using biotin-labeled lpt-O as a probe confirmed that the 10-kb plasmid band corresponds to the lpt-O-carrying pSFxv_2 of 6.85 kb as only Xv strains showed a positive signal (Fig. 1C). The 10-kb plasmid band of 2002017 was extracted from gel and used as template for overlapping PCR covering the circular plasmid. Results indicated that the plasmid sequence is identical to that of pSFxv_2 (unpublished data). The observation that the plasmid appeared larger in size on the gel than its actual size of 6.85 kb may be due to the migration difference of covalently closed circular plasmid DNA and the molecular marker of linear DNA. The data obtained suggest that all serotype Xv strains carry a pSFxv_2-like plasmid, which mediates the addition of PEtN to the O-antigen giving rise to the MASF IV-1 positive phenotype.
To elucidate the function of the lpt-O gene, we then cloned the entire lpt-O gene of 1,521 bp, together with 533 bp sequences up and downstream to cover its promoter and terminator sequences, from strain 2002017 into plasmid pMD20-T (pSQZ) and transformed it into serotype X strain 51580 (amp s ). Slide agglutination assay indicated that the host strain was converted to serotype Xv, with the transformant (51580_Xv) serologically  Table 2). Similar to that of serotype Xv strains, the transformant also reacted with monovalent antisera IV of Denka Seiken ( Table 2). The serological features of 51580_Xv were identical to that of the control strain 2002017 (Table 2).
A previous study has identified that serotype 4a strain G1668 contains a PEtN residue on Rha III at position 3 [29]. We confirmed serologically that G1668 is MASF IV-1 positive and thus is a serotype 4a variant hereafter called 4av. To examine whether the same gene can convert serotype 4a into serotype 4av, a serotype 4a strain NCTC 9725 (amp s ) was transformed with pSQZ. The transformant (NCTC 9725_4av) showed the acquired MASF IV-1 positive phenotype with the same serological feature as that of control serotype 4av strain G1668 (Table 2).

Gain of the lpt-O Gene and MASF IV-1 Reactivity was Concomitant with Gain of the O-antigen Modification with PEtN
To confirm that the gain of the lpt-O gene and MASF IV-1 reactivity by serotype X strains correlates with the gain of PEtN, the O-polysaccharide of the transformant 51580_Xv was analyzed by NMR spectroscopy. The 1 H NMR and 13 C NMR (Fig. 2C) spectra of the 51580_Xv O-polysaccharide showed a structural heterogeneity. Analysis of the spectra using two-dimensional NMR techniques revealed the occurrence of two types of repeats in the ratio of ,1.5:1: the major O-unit typical of the Xv Opolysaccharide (Fig. 4B) and a minor bisphosphorylated nonglucosylated O-unit. Particularly, in addition to the major signals for the phosphorylated Rha II residue (d H-3 4.35 and d C-3 76.5), there were signals for the phosphorylated Rha III residue (d H-3 4.25 and d C-3 76.2). Accordingly, in the 1 H, 31 P HMBC spectrum (Fig. 3B) (Fig. 3A). Therefore, in serotype Xv strains phosphorylation interferes with glycosylation on Rha III giving rise to the O-units in which the lateral Glc residue is replaced with the second PEtN group. The bisphosphorylated O-units of this type are characteristic for some strains of serotype Y that may be called Yv (see below) ( Fig. 4C) (authors' unpublished data), which seems to be the same as reported serotype 4 s [33].
NMR spectroscopic analysis showed that the major O-unit of the NCTC 9725_4 av transformant (,50% of the total O-units) has the same structure as that of wild-type 4 av strains reported earlier [29] (Fig. 4D). Minor O-unit variants (each of ,25%) lack lateral Glc and have either one PEtN group on Rha II or two PEtN groups on both Rha II and Rha III (for the 1 H, 31 P HMBC spectrum see Fig. 3D). Again, a more careful analysis of the wild-type 4 av Table 2. Serological identification of transformants using the monoclonal antibodies of MASF scheme and monovalent antisera of Seiken by slide agglutination.

Strains
Serotype Reaction with monoclonal antibody of MASF Reaction with monovalent antisera of Seiken  polysaccharide revealed the presence of minor proportion (,10%) of the O-units with PEtN groups on Rha II (for the 1 H, 31 P HMBC spectrum of the O-polysaccharide from strain G1668 see Fig. 3C). Therefore, the lpt-O gene is required and sufficient for the addition of PEtN residue to the O-antigen in serotype 4 av strains too. These data also suggest that Lpt-O can mediate the attachment of PEtN to either Rha II or Rha III .

The Functional lpt-O Gene is Carried Only by MASF IV-1 Positive Phenotype Strains
To determine whether the lpt-O gene is more widely present in S. flexneri, another expanded PCR screening of 310 strains of 15 serotypes (Table 3) was performed using primer pair lpt-O-2. These strains were randomly selected from our collection, and their serotype characteristics have been determined earlier. A strict correlation was observed between the presence of functional lpt-O gene and MASF IV-1 positive reactivity. As expected, PCR products were found in all 59 serotype Xv strains tested. The expected PCR product (1,098 bp) was also obtained from 19 serotype Yv, 3 serotype 4 av (2002091, NCTC 8296 and G1668), and surprisingly, 4 serotype X (HN059, HN060, HN066 and HN378) and one serotype 4b (NCTC 9726) strains. Although the lpt-O gene can be amplified from these 5 strains, they all did not react with monoclonal antibody MASF IV-1. Plasmid profile analysis indicated that the pSFxv_2-like plasmid is present in these 5 strains. Sequencing of the lpt-O gene from these 5 strains revealed that all had a single base deletion at the position of 295, which resulted in a stop codon at amino acid 99, and abolishes the function of the gene without the presumed sulfatase domain (from aa165 to aa465) of Lpt-O. We cloned the defective gene and showed that it cannot convert serotype X strain 51580 to serotype Xv, confirming that the single base deletion inactivated the gene.
The lpt-O positive serotype Yv strains are serologically different from typical serotype Y with additional MASF IV-1 and monovalent antisera IV reactivity phenotype. Analysis of the Oantigen structure of a serotype Yv strain, HN006, demonstrated a PEtN modification at position 3 of both Rha II and Rha III (Fig. 4C) (details of structural studies on the Yv O-antigen will be reported elsewhere). The 3 lpt-O positive genotype 4 av strains differ from typical serotype 4a strains only in the MASF IV-1 positive phenotype ( Table 2)

Discussion
The O-antigen structure and its diversity of S. flexneri are well studied. Until recently, glucosyl and O-acetyl groups have been the only two known residues attached to S. flexneri O-antigen backbone Table 3. Distribution of the lpt-O gene in S. flexneri strains detected by PCR.

Serotypes
Number of strains tested Number of serological reactive strains with 1 lpt-O gene PCR results 2  [23], E. coli [31], Pasteurella multocida [35] and Campylobacter jejuni [32]. However, the PEtN modification in these cases occurs in the LPS or LOS core and does not involve in serotype conversion except in P. multocida [35]. Our study revealed other serotypes of S. flexneri modified by PEtN through PCR screening, and potentially there are many variants of the currently known serotypes due to PEtN modification. As reported previously [2], serotype Xv emerged as a new serotype and rapidly became the predominant serotype in China to cause a large number of shigellosis cases, the other variants also have the potential to flourish and change the dynamics of S. flexneri epidemics. The elucidation of this new O-antigen modification mechanism provides avenue for further study of S. flexneri serotype diversity and epidemiology. The presence and location of non-carbohydrate substituents in LPS, such as PEtN, are critical for the immune recognition of LPS by antibodies in some bacterial species [27,35]. The PEtN-3 constituent on the b-chain heptose (HepII) of the inner core LPS of N. meningitides is the target epitope of protective antibody mAb B5+ [27]. In P. multocida, the serological classification of strains into serotypes 2 and 5 is dependent on the presence or absence of PEtN on Hep II of LPS [35]. In this study, the MASF IV-1 positive reactivity was found to be dependent on the PEtN modification of the O-antigen, which confers the specific MASF IV-1 or E1037 antigen on the host. Since human immune response to S. flexneri infection is serotype specific with protection against further infection by the same serotype only [36], the presence of a new surface epitope in S. flexneri would expect to confer the bacterium an adaptive advantage to evade human immune killing. This may account for the emergence of serotype Xv in China, which appeared in one province initially and expand to most provinces of China during a short period of time, surpassing serotype 2a as the predominant serotype [2]. Apart from the multidrug resistance of the new serotype [2], the immune evasion offered by the new PEtN modification may also play an important role in the spread of the pathogen.
In this study, the lpt-O gene required for the PEtN addition was also identified. Its causal role is supported by the following evidences: 1) the protein encoded by the lpt-O gene is homologous to other PEtN transferases; 2) the cloned lpt-O gene can transform serotype X to serotype Xv and serotype 4a to serotype 4 av; 3) the PEtN modification correlates strictly with the presence of the functional lpt-O gene; 4) naturally occurring mutation (one base deletion) abolishes the function of the lpt-O gene without the presumed sulfatase domain and lost the reactivity with MASF IV-1 antibody. Future studies are necessary to characterize the biochemical process of the PEtN modification.
The known O-antigen modifications through glucosylation or O-acetylation in S. flexneri are mediated by serotype-converting bacteriophages or prophages with 7 identified up to date [3,6,8,9,10,11,13,14]. In contrast, our study demonstrated a plasmid-encoded serotype conversion mechanism, which has not been reported in S. flexneri earlier. This finding changed the landscape of S. flexneri serotype diversity and serotype conversion. As an easily transferable factor, the lpt-O carrying plasmid can spread among different stains and mediate the serotype conversion. Serotype Xv strains are clearly originated from serotype X strains by gaining a lpt-O carrying plasmid since previous PFGE analysis showed that serotype Xv strains were grouped together with serotype X, and many serotype Xv strains shared the same PFGE profiles as serotype X strains [2].
In conclusion, the chemistry and genetics of a new O-antigen modification in S. flexneri serotype Xv have been elucidated. Attachment of a PEtN group to Rha residues gives rise to the MASF IV-1 positive phenotype in this and several other serotypes. The lpt-O gene, required for PEtN addition is carried by a pSFxv_2 like plasmid in all MASF IV-1 positive phenotype strains. This novel mechanism of modification has profound implications in S. flexneri epidemiology as it may offer a significant advantage to the bacterium which is demonstrated by the prevalence of serotype Xv in China. Additionally, these findings, together with published data [2,15,17,33], show a necessity of further extension of classification scheme for S. flexneri by adding MASF IV-1 positive strains as new subtypes of serotypes 4, X and Y.