DNA polymorphism analysis of Brucella lipopolysaccharide genes reveals marked differences in O-polysaccharide biosynthetic genes between smooth and rough Brucella species and novel species-specific markers

Background The lipopolysaccharide is a major antigen and virulence factor of Brucella, an important bacterial pathogen. In smooth brucellae, lipopolysaccharide is made of lipid A-core oligosaccharide and N-formylperosamine O-polysaccharide. B. ovis and B. canis (rough species) lack the O-polysaccharide. Results The polymorphism of O-polysaccharide genes wbkE, manAO-Ag, manBO-Ag, manCO-Ag, wbkF and wbkD) and wbo (wboA and wboB), and core genes manBcore and wa** was analyzed. Although most genes were highly conserved, species- and biovar-specific restriction patterns were found. There were no significant differences in putative N-formylperosamyl transferase genes, suggesting that Brucella A and M serotypes are not related to specific genes. In B. pinnipedialis and B. ceti (both smooth), manBO-Ag carried an IS711, confirming its dispensability for perosamine synthesis. Significant differences between smooth and rough species were found in wbkF and wbkD, two adjacent genes putatively related to bactoprenol priming for O-polysaccharide polymerization. B. ovis wbkF carried a frame-shift and B. canis had a long deletion partially encompassing both genes. In smooth brucellae, this region contains two direct repeats suggesting the deletion mechanism. Conclusion The results define species and biovar markers, confirm the dispensability of manBO-Ag for O-polysaccharide synthesis and contribute to explain the lipopolysaccharide structure of rough and smooth Brucella species.


Background
The members of the genus Brucella are gram-negative bacteria that cause brucellosis, a zoonotic disease of great importance worldwide. Currently, several Brucella species are recognized [1]. B. abortus, B. melitensis, B. suis, B. neotomae, B. ovis, and B. canis have been known for a long time and are traditionally distinguished according to their preferential host, biochemical tests and cell surface characteristics [2]. In addition, Brucella strains isolated from cetaceans and pinnipeds during the last fifteen years have been grouped into B. ceti and B. pinnipedialis, [3]. Very recently, some Brucella strains have been isolated from the common vole and a new species, B. microti, proposed [4]. B. abortus, B. melitensis and B. suis have been classically subdivided into biovars according to H 2 S production, CO 2 -dependence, dye sensitivity and distribution of the A and M epitopes (see below) [2]. However, because these tests are difficult to standardize, molecular markers have been investigated [5][6][7][8][9].
Wild type B. melitensis, B. abortus, B. suis, B. neotomae, B. ceti, B. pinnipedialis and B. microti express a smooth (S)type lipopolysaccharide (LPS) formed by an O-polysaccharide connected to a core oligosaccharide which, in turn, is linked to lipid A, the section embedded into the outer membrane. However, both B. ovis and B. canis lack the O-polysaccharide and, accordingly, their LPS is termed rough (R) (R-LPS). Brucella LPS is of great interest not only because of these species differences but also because it is the foremost diagnostic antigen and a major virulence factor [10]. Despite this, the structure and genetics of Brucella LPS is only partially understood. The Opolysaccharide is a homopolymer of N-formyl-perosamine in α (1-2) or in α (1-2) plus α (1-3) linkages [11], and these variations relate to the main serovars in  [12]). Region wbo encodes two putative glycosyltransferases (wboA and wboB) and region wbk contains the genes putatively involved in perosamine synthesis (gmd [GDP-mannose 4, 6 dehydratase] and per [perosamine synthetase]), its formylation (wbkC) and polymerization (glycosyltransferases) (wbkA and wbkE), as well as those for bactoprenol priming (wbkD and wbkF) and O-PS translocation (wzm and wzt). In addition, wbk contains genes (manA O-Ag , manB O-Ag , manC O-Ag ) which may code for the enzymes that furnish mannose, the perosamine precursor. Intriguingly, wbkB and manB O-Ag do not generate R phenotypes upon disruption [12,13], and B. ovis and B. canis carry wbk genes despite the absence of the O-polysaccharide [14]. Much less is known on the Brucella core oligosaccharide. Reportedly, it contains 2keto, 3-deoxyoctulosonic acid, mannose, glucose, glucosamine and quinovosamine [12,15] but the structure is unknown. Thus far, only three genes have been proved to be involved in core synthesis: pgm (phosphoglucomutase, a general biosynthetic function), manB core (mannose syn-thesis) and wa** (putative glycosyltransferase) [12]. Obviously, genetic analysis encompassing a variety of strains could shed light on the differences behind the phenotypes of S and R species, confirm or rule out a role for known genes, and identify differences that could serve as serovar or biovar markers. With these aims, wbkE, manA O-Ag , manB O-Ag , manC O-Ag , wbkF, wkdD, wboA, wboB, wa** and manB core were analyzed for polymorphism in the classical Brucella spp., B. ceti, and B. pinnipedialis. Figure 1 shows the organization of LPS genes in B. melitensis 16 M [12]. PCR amplification of wbkE, manB O-Ag , manA O-Ag , manC O-Ag , wkdD, wbkF, wboA and wboB, wa** and manB core was conducted on representative strains of each of the Brucella species included in this study and their biovars with attention to the LPS characteristics (i.e. S versus R; and A dominant, M dominant, or A = M for the S-LPS). Except for wboA and wboB in B. ovis, all genes were successfully amplified in the strains of all Brucella species and biovars tested. These results confirm the absence of the wbo region in B. ovis [16,17]. They also suggest that conservation of wbk extends beyond those genes (wbkA to wbkC) examined in a previous work [14] and that wa** and manB core were are also conserved in the genus. Further analyses were then conducted to examine these possibilities.

Polymorphism in core LPS genes
Despite using six restriction enzymes, all brucellae displayed the same RFLP pattern for the manB core amplicon.
In silico, the four genomes available showed low polymorphism. A single nucleotide deletion at position 812 was detected in B. ovis, which should modify the C-terminal sequence of the protein ( Figure 5). Similarly, a low degree of polymorphism was observed in wa**. With the exception of B. suis biovar 2, one PstI pattern was specific of B. suis. Biovar 2 also lacked an AvaII site, which could be considered as a biovar marker. With Hinf1, a pattern was specific of B. ovis ( Figure 2, Table 1).

Discussion
Despite the high DNA homology of brucellae, gene polymorphism and species-and biovar-specific markers have been consistently found. Concerning outer membrane molecules, both have been found in genes of proteins [16,18,19] but not in the LPS genes examined, all of the wbk region (wbkA, gmd, per, wzm, wzt, wbkB, and wbkC). Interestingly, these O-polysaccharide genes were found to be highly conserved not only in the classical S Brucella species and biovars but also in B. ovis and B. canis, the two species that lack the O-polysaccharide [14]. Therefore, an implication of these observations is that the R phenotype of B. ovis and B. canis cannot be explained by the absence of any of those seven wbk genes. More recently, the wbk region has been extended to include wbkE, manA O-Ag , manB O-Ag , manC O-Ag , wbkF, and wkdD [12]. The present study includes an analysis of some of these genes and the results not only show the existence of specific markers but, more important, they also improve our understanding of the genetics-structure relationship in Brucella LPS. Concerning the O-polysaccharide, the results are relevant to interpret the variations in O-polysaccharide linkages of S Brucella and add further weight to our previous finding (12) that the putative mannose genes in wbk are not essential for perosamine synthesis. Furthermore, they help to explain the differences existing between S and R Brucella species.
Despite extensive transposon mutagenesis searches, only four putative glycosyltransferase genes have been implicated in N-formylperosamine polymerization in Brucella: wbkA, wbkE, wboA and wboB. As mentioned above, wbkA is conserved in classical Brucella species [14], and the results reported here show that wboA, wboB and wbkE are simi-  [20]. In keeping with this, it has been observed that strain RB51 (a wboA mutant of the A-dominant B. abortus 2308 S strain [21]) generates small amounts of atypical M-type polysaccharides [22]. All this evidence suggests that, rather than the presence of a α A surprising feature of the wbk is the presence of genes that are not essential for O-polysaccharide synthesis. Godfroid et al. [13] analyzed the functions of the ORFs between BMEI1404 (wbkA, encoding a putative mannosyltransferase [perosaminyltransferase since mannose and perosamine are related]) and BMEI1418 (wbkC, encoding a putative formyltransferase) and found that disruption of ORF BMEI1417 (wbkB) generated no R phenotype. Later, it was found that the genome of B. melitensis contains three putative mannose synthesis genes (ORFs BMEI1394 to BMEI1396) adjacent to wbkA. Because mannose is the direct precursor of perosamine and O-polysaccharide genes usually cluster together, Monreal et al. [23] proposed the names of manA O-Ag , manB O-Ag , manC O-Ag for BMEI1394 to BMEI1396, and their assignment to wbk is supported by the finding by González et al. [12] that disruption of ORF BME1393 (wbkE) blocks O-polysaccharide synthesis. The latter authors provided proof that at least manB O-Ag , is dispensable for perosamine synthesis but also pointed out that the existence of manB core -manCcore (ORFs BMEII0900 and BMEII0899) preclude to rule out any role for the wbk putative mannose synthesis genes since there could be internal complementation [12]. All these results are fully consistent with the observation that, although manB O-Ag is disrupted by IS711 in B. pinnipedialis and B. ceti, these two species keep the S phenotype. The wbk region has features suggestive of horizontal acquisition [14] whereas manB core (and manC core ) are Brucella older genes necessary for the synthesis of the LPS core oligosaccharide [23,24]. Accordingly, a drift to dysfunction of the wbk man genes may have been made possible by the redundancy created after horizontal acquisition of wbk, and the similarity in this regard between B. ceti and B. pinnipedialis suggests a common ancestor.
The results of this research also shed additional light on the genetic basis behind the R phenotype of B. ovis and B. canis. Previous work has shown a large deletion in B. ovis that encompasses wboA and wboB [16,17]. The present work confirms the absence of these two putative perosaminyltraneferase genes in B. ovis, an absence that can account by itself for the lack of O-polysaccharide in this species [12,25]. To this evidence, the present work adds the nucleotide deletion detected in B. ovis wbkF. Indeed, the frame-shift thus created predicts a very modified protein. Presumably, WbkF is involved in catalyzing the transfer of an acetylated aminosugar to undecaprenylphosphate, thus priming this carrier for O-chain polymerization. The N-terminal region of the E. coli WbkF homologue was found to be necessary for this function [26] and, therefore, it seems likely that the frame-shift in B. ovis wbkF produces a non-functional protein, thus explaining in part the R phenotype of this species. Other changes detected in several B. ovis LPS genes do not have this dramatic effect. As discussed above, the man wbk genes are dispensable and, therefore, the nucleotide substitution and frame shift detected in B. ovis manA O-Ag do not contribute to the R phenotype. Since disruption of manB core generates a deep R-LPS [24,24], the presence of two more nucleotides in the sequence of B. ovis manB core was interesting. However, this deletion modified only the C-terminal sequence (5 last amino-acids) of the protein making unlikely a change severe enough to contribute to the R phenotype. In support of this interpretation, B. ovis R-LPS is not deeply truncated like that of manB core mutants. Moreover, the same two nucleotide addition was detected in B. suis, and it is known that a functional manB core is required for the synthesis of S-LPS in this species [27].
A DNA deletion of 351 bp. including 3' end of wbkF and 3' end of wbkD was detected in B. canis, which might have occurred by a slipped mispairing mechanism (a direct repeat sequence of 7 bp «GGATCAT» is present at both sides of the deleted sequence in the other Brucella species ( Figure 5). It is clear that this deletion has profound consequences in the synthesis of LPS. We have discussed above the essential role of wbkF in O-polysaccharide synthesis, and wbkD seems involved in the synthesis of quinovosamine, a sugar that is possibly linking the Brucella O-polysaccharide to the R-LPS [12]. This double mutation clearly explains the R phenotype of B. canis and is consistent with the absence of quinovosamine in this species [28].

Conclusion
The analyses carried out suggest new hypothesis to study the genetics of Brucella O-polysaccharide serotypes and provide evidence on both the dispensability of some wbk genes which is consistent with their horizontal acquisition. They also confirm the essential role of wbkD and wbkF in O-polysaccharide synthesis and, at the same time, contribute to understand the R phenotype of B. ovis and B. canis. Finally, they provide several biovar and species specific markers that can be used to design the corresponding molecular typing tools.

Brucella strains
The strains (Table 1) were maintained freeze-dried in the INRA Brucella Culture Collection, Nouzilly (BCCN), France. For routine use, they were grown on tryptic soy agar (Difco)-0.1% (w/v) yeast extract (Difco). Fastidious strains (B. abortus biovar 2 and B. ovis) were grown on the same medium supplemented with 5% sterile horse serum (Gibco BRL). All strains were checked for purity, and species and biovar confirmed by standard procedures [2].

DNA preparation
Bacteria were cultured at 37°C for 24 h, suspended in 3 ml sterile distilled water, harvested (2000 × g, 10 minutes) and resuspended in 567 μl of 50 mM Tris, 50 mM EDTA, 100 mM NaCl (pH 8.0). Then, 30 μl of 10% (w/v) SDS and 3 μl of 2% (w/v) proteinase K were added, the mixture was held at 37°C for 1 h and extracted twice with phenolchloroform. Nucleic acids in the aqueous phase were precipitated with two volumes of cold ethanol, dissolved in 100 μl of 10 mM Tris, 1 mM EDTA (pH 8.0) and the amount of DNA estimated by electrophoresis on 0.8% agarose gels using appropriate DNA solutions as the standards.