INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The genus Brucella is described as constituted by six species: Brucella melitensis, B. abortus, B. suis, B. ovis, B. canis, and B. neotomae, each preferentially infecting an animal host. In addition, several biovars are included in the species B. melitensis, B. abortus, and B. suis. The six Brucella species and their biovars are currently differentiated by pathogenicity and host preference characteristics, serotyping, phage typing, dye sensitivities, and culture and metabolic properties (2). In spite of the high degree of DNA homology that has been found between the six species, which would justify a monospecific genus (53, 54), the classical organization of the genus Brucella has been maintained since it is in accordance with the pathogenicity and host preference characteristics of each species, and several molecular markers, allowing the differentiation between the six species and some of their biovars, have been found (55).

Differences in pathogenicity and host preference found between the Brucella species and biovars could be reflected by differences at the DNA level. However, only small differences have been found between the Brucella species in the genes identified until the present (55). Small deletions in several genes have been detected in some Brucella species (11, 13, 17, 44), sometimes leading to the lack of production of the encoded protein (44), and the lack of expression of an existing gene has also been reported (16, 29). DNA deletions of several sizes and DNA inversions have also been described, but they have not been characterized (35). The first report of a gene absent in one of the Brucella species describes the deletion in B. abortus of omp31 coding for an outer membrane protein (OMP) (58). Further studies revealed that the deletion in B. abortus comprises not only omp31 but also several adjacent genes located at both sides of omp31 (57, 58). Partial characterization of this large DNA region deleted in B. abortus led to the discovery of several genes that might be involved in the synthesis of a polysaccharide not identified in Brucella spp. (57). In the present study, we describe the characterization of the entire B. melitensis 16M large DNA fragment deleted in B. abortus, and we evaluate its occurrence in the other Brucella species and biovars.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains and plasmids. The Brucella strains used in this study were obtained from the INRA Brucella Culture Collection, Nouzilly (BCCN), France. Bacteria were grown on tryptic soy agar (Gibco-BRL/Life Technologies, Eragny, France) supplemented with 0.1% (wt/vol) yeast extract (Difco Laboratories, Detroit, Mich.). For fastidious strains (B. abortus bv. 2 and B. ovis), 5% horse serum (Gibco-BRL) was also added. Species and biovar characterizations were performed by standard procedures (2).

1

1

View larger version (19K): [in this window] [in a new window]   FIG. 1.   B. melitensis 16M DNA inserts of the relevant plasmids used in this study. pNV3132 and pNV3140 were constructed as described previously (57). The insert of pNV31106 is cloned into the NotI site of pGEM-5Zf. The inserts of pNV31109, pNV31111, pNV31112, and pNV31113 are cloned into the SacI site of pGEM-7Zf. Their HindIII and SacI restriction profiles are shown in the B. melitensis 16M DNA fragment drawn in the bottom of the figure, where the limits of the DNA fragment deleted in B. abortus are also shown. The B. melitensis 16M DNA regions previously characterized (57) are patterned. Restriction sites: B, BamHI; E, EcoRI; H, HindIII; K, KpnI; N, NotI; S, SacI; X, XbaI. The framed restriction sites are not present in the B. melitensis 16M DNA, and they belong to the multiple cloning site of -GEM12 or cloning plasmids.

Genomic library screening and Southern blot hybridization. The B. melitensis 16M inserts of pNV3132 and pNV3140 were excised by digestion of the plasmids with BamHI and XbaI and with ApaI and SacI, respectively. Both inserts were purified, with the Geneclean II kit (Bio 101, La Jolla, Calif.), from an agarose gel after electrophoresis and digoxigenin labeled with the DIG DNA Labeling Kit (Roche Diagnostics, Mannheim, Germany) according to the instructions of the manufacturer.

PCR and DNA sequencing. Insert DNA from plasmids pNV31112 and pNV31113 (Fig. 1) was unidirectionally digested with exonuclease III by using the Erase-a-Base system (Promega) as specified by the manufacturer. A series of plasmids differing in ca. 400 bp was obtained for each initial plasmid and used to determine the entire sequence of both strands of the 13,112-bp insert. Plasmid DNA was obtained and purified by using the Wizard Plus SV Minipreps System (Promega) according to the instructions of the manufacturer. Purified plasmid DNA was sequenced by primer-directed dideoxy sequencing (45) with an ABI PRISM 377 DNA sequencer (Perkin-Elmer, Foster City, Calif.) and the forward pUC19 primer. Some specific Brucella DNA primers were also used to complete the sequence. The same procedure was followed for sequencing both strands of the 8,418-bp insert of pNV31109 and pNV31111 (Fig. 1).

DNA and protein analysis. Search for open reading frames (ORFs) and putative genes in the DNA sequences was performed with the DNAStrider 1.2 program (32) and the GeneMark.fbf prediction program (47) (Website, http://dixie.biology.gatech.edu/GeneMark/fbf.cgi), respectively. Searches for DNA and protein homologies were performed with the FASTA program (39) (FASTA Website, http://www.infobiogen.fr/services/analyseq/cgi-bin/fasta_in.pl; version 2, June 2000). Cellular location and PROSITE motifs of the predicted proteins were determined with the PSORT (37) (Website, http://psort.nibb.ac.jp/form.html) and MOTIF programs (30) (Website, http://www.motif.genome.ad.jp/), respectively.

Nucleotide sequence accession number. The B. melitensis 16M nucleotide sequence determined here has been deposited in the GenBank/EMBL/DDBJ databases under accession number AF076290.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

DNA sequence of the B. melitensis 16M DNA fragments cloned in pNV31112/pNV31113 and pNV31109/pNV31111. A B. melitensis 16M DNA fragment of 17,119 bp, cloned in pGEM-7Zf giving recombinant plasmid pNV3103, was sequenced in a previous study (57). This fragment contained 9,948 bp corresponding to the right side of a large deletion detected in B. abortus strains (57). However, the left end of the B. abortus deleted fragment could not be determined. In order to completely characterize the B. abortus deletion, two B. melitensis 16M DNA fragments were used as probes to screen a B. melitensis 16M genomic library. The first B. melitensis 16M DNA fragment, which was cloned in pGEM-7Zf resulting in recombinant pNV3132 plasmid, is composed of 3,971 bp located at the left end of the previously sequenced pNV3103 insert (57) (Fig. 1). The second B. melitensis 16M DNA fragment, which was cloned in pGEM-T resulting in recombinant pNV3140 plasmid, corresponds to 438 bp known to be located adjacent to the left end of the fragment deleted in B. abortus (57) (Fig. 1). Screening of a B. melitensis 16M genomic library by using the pNV3132 and pNV3140 inserts as digoxigenin-labeled probes allowed us to select two recombinant phages containing overlapping inserts hybridizing with pNV3132 and pNV3140, respectively. Subcloning of the B. melitensis 16M DNA insert of these phages, as described in Materials and Methods, allowed us to obtain plasmids pNV31109/pNV31111 and pNV31112/pNV31113 (Fig. 1), which were used to determine the nucleotide sequence of the left end of the fragment missing in B. abortus.

View larger version (6K): [in this window] [in a new window]   FIG. 2.   Nucleotide sequence of the B. melitensis 16M DNA flanking the 25-kb fragment deleted in B. abortus (A) and of the corresponding region in B. abortus 544 (B). The B. abortus sequence was determined in a previous work (57). The direct repeats of four nucleotides found at both sides of the B. melitensis 16M DNA fragment deleted in B. abortus are shaded. The point where the deletion occurred is marked with an arrow in B. abortus, and the DNA flanking the right side of the deletion is underlined in B. melitensis and B. abortus.

Putative genes and encoded proteins identified in the B. melitensis 16M DNA fragments cloned in pNV31112/pNV31113 and pNV31109/pNV31111. Sequencing of the B. melitensis 16M DNA cloned in plasmids pNV31112/pNV31113 and pNV31109/pNV31111 has allowed us to identify 17 new putative genes potentially encoding proteins (Fig. 3; Tables 1 and 2). Three of these genes (bme16, bme17, and bme1) had been partially sequenced previously in B. melitensis 16M and/or B. abortus 544 (57). Putative genes were identified with GeneMark.fbf and sequence comparison, with FASTA, of the encoded proteins with other proteins in the database.

View larger version (21K): [in this window] [in a new window]   FIG. 3.   ORF distribution of the B. melitensis 16M DNA fragment deleted in B. abortus and flanking DNA. The patterned area shows the DNA deleted in B. abortus. Genes identified in a previous study (57) are marked with an asterisk. The extension of bme19 is shown as determined by sequence homology of the encoded protein with other proteins in the database. However, a stop codon (marked with a dashed line) is detected in the reference strains of B. abortus and the three biovars of B. melitensis, but it is not detected in the other Brucella reference strains (see Results).

(i) bme18, bme19, and bme16 The three ORFs are located outside the 25-kb DNA fragment deleted in B. abortus and would be transcribed from the direct strand (Table 1; Fig. 3). The first ORF of 459 bp (bme18) would encode a highly hydrophobic (data not shown) putative protein of 152 amino acids showing 42.6 and 43.4% identity with the C-terminal ends of Pseudomonas aeruginosa PA5030 and E. coli YnfM, respectively, that have been defined as probable transport proteins (Fig. 3; Tables 1 and 2). Identity with only the C-terminal end of both proteins suggests that the entire bme18 sequence is not cloned in plasmids pNV31112/pNV31113.

(ii) bme17, bme20, and bme21 The transcription of ORFs bme17, bme20, and bme21 would occur from the reverse strand (Table 1; Fig. 3). ORF bme17 has been partially characterized previously in B. abortus 544 (57). The left end of the B. abortus deletion is located inside this ORF (Fig. 3) that would encode in B. melitensis 16M a protein of 218 amino acids with a probable cytoplasmic location (Table 1). Bme17 displayed 43.7% identity with a putative protein of Streptomyces coelicolor (protein SCF62.20) of unknown function and lower homology degrees with proteins of other bacteria also with an unknown function (Table 2). Although the function of SCF62.20 from Streptomyces coelicolor has not been defined, the corresponding gene is located in a cluster encoding proteins that seem related to the synthesis of a polysaccharide (accession number AL121855), and some other genes from this cluster also show homology with the Brucella genes identified here (see below).

(iii) bme22 The bme22 gene, which would be transcribed from the direct strand, was predicted to code for a protein of 184 amino acids probably located in the cytoplasm (Table 1; Fig. 3). Search for homology with other proteins in the database revealed that Bme22 from B. melitensis displayed ca. 45% identity with nucleotide-sugar epimerases from several bacteria (Table 2), among them StrX and Y4gL from Streptomyces glaucescens and Rhizobium sp., respectively. StrX from Streptomyces glaucescens is thought to act as an NDP-4-ketohexose-3,5-epimerase involved in the synthesis of the amino-glycoside 5'-hydroxystreptomycin (7) and the Y4gL protein from Rhizobium sp., defined as a dTDP-4-dehydrorhamnose-3,5-epimerase, is predicted to be involved in the synthesis of a polysaccharide (18).

(iv) bme23, bme24, bme25, bme26, and bme27 The five genes bme23, bme24, bme25, bme26, and bme27 would be transcribed from the reverse strand (Table 1; Fig. 3), the end of bme23 overlapping by 38 nucleotides the end of bme22 that would be transcribed from the direct strand. Bme23, a probable cytoplasmic protein, would be made up of 437 amino acids (Table 1) and showed 51.7% identity with SCF62.22 from Streptomyces coelicolor (Table 2), a protein of unknown function but encoded by a gene located in a cluster containing several genes that seem related to the synthesis of a polysaccharide (accession number AL121855). Bme23 also showed 32.8% identity with Cj1295 of Campylobacter jejuni. The function of this protein is also unknown, but its gene is located close to other C. jejuni genes that might be involved in the synthesis of a polysaccharide (i.e., putative sugar-nucleotide epimerase-dehydratase, aminotransferase, or acetyltransferase) (accession number AL139078) (Table 2).

(v) bme28 The bme28 gene would be transcribed from the direct strand and would code for a protein of 413 amino acids that would be probably located in the cytoplasmic membrane (Table 1; Fig. 3). Significant homology was only found with SCF62.25 from S. coelicolor (Table 2), a protein with unassigned function but whose gene is located in a cluster of genes that seem related to the synthesis of a polysaccharide (accession number AL121855).

(vi) bme29 and bme30 Genes bme29 and bme30 would be transcribed from the reverse strand (Table 1; Fig. 3). The bme29 gene would code for a cytoplasmic protein of 337 amino acids (Table 1) giving the highest percentages of homology with glycosyl transferases LgtD and LgtA from Neisseria spp. (Table 2) intervening in the biosynthesis of lipooligosaccharide (5, 14, 26). The amino acid sequence EX₇E, which is found in many glycosyltransferases (3, 46), was also found in Bme29.

(vii) bme31 and bme1 Transcription of bme31 and bme1 would take place from the direct strand (Table 1; Fig. 3). The bme31 gene would code for a protein of 373 amino acids containing multiple transmembrane-spanning segments that would probably be located in the cytoplasmic membrane (Table 1). The PSORT program predicted an N-terminal signal sequence cleavable by signal peptidase at amino acid 21 that would release a protein of 352 amino acids and 37.6 kDa. Bme31 showed ca. 35% identity with proteins defined as undecaprenyl phosphate-N-acetylglucosamine transferase (UndP-GlcNAc transferase) from several bacteria (Table 2). UndP-GlcNAc transferase has been shown to be involved in the synthesis of bacterial polysaccharides by catalyzing the addition of the first sugar of a polysaccharide unit (GlcNAc from a UDP-GlcNAc sugar nucleotide donor substrate) to a UndP acceptor (1, 38, 61). UndP-GlcNAc transferases are predicted to be integral membrane proteins (61), as is also the case for Bme31.

Occurrence in the genus Brucella of the 25-kb DNA fragment deleted from B. abortus HindIII-digested DNA from the reference strains of all of the Brucella species and biovars was tested in Southern blot hybridization against digoxigenin-labeled pNV31112 and pNV31106 probes in order to investigate the occurrence in the Brucella genus of the 25-kb DNA fragment deleted from B. abortus.

View larger version (31K): [in this window] [in a new window]   FIG. 4.   Southern blot hybridization of Brucella spp. DNA HindIII restriction fragments with pNV31112 (A) and pNV31106 (B) digoxigenin-labeled probes. Lanes: molecular mass markers (lanes 1 and 2), B. melitensis 16M (bv. 1) (lane 3), B. melitensis 63/9 (bv. 2) (lane 4), B. melitensis Ether (bv. 3) (lane 5), B. suis 1330 (bv. 1) (lane 6), B. suis Thomsen (bv. 2) (lane 7), B. suis 686 (bv. 3) (lane 8), B. suis 40 (bv. 4) (lane 9), B. suis 513 (bv. 5) (lane 10), B. ovis 63/290 (lane 11), B. canis RM6/66 (lane 12), B. neotomae 5K33 (lane 13), B. abortus 544 (bv. 1) (lane 14), B. abortus 86/8/59 (bv. 2) (lane 15), B. abortus Tulya (bv. 3) (lane 16), B. abortus 292 (bv. 4) (lane 17), B. abortus B3196 (bv. 5) (lane 18), B. abortus 870 (bv. 6) (lane 19), and B. abortus C68 (bv. 9) (lane 20).

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The nucleotide sequence of a B. melitensis 16M large DNA fragment of 25,064 bp (Fig. 3), known to be deleted in B. abortus (57, 58) and partially sequenced previously (57), has been entirely determined. A short direct repeat of 4 bp (GTGC) was found in B. melitensis 16M at each side of the fragment deleted in B. abortus (Fig. 2), the deletion removing one of this 4-bp repeat in B. abortus (Fig. 2). Short direct repeats have been reported to intervene in the generation of large deletions by a mechanism of illegitimate recombination involving strand slippage during replication that removes one of the direct repeats (40, 59, 60). It has been shown, in bacterial model systems, that deletion frequencies are increased by reducing the distance between the direct repeats (10), by increasing the direct repeat lengths, and by the presence of inverted repeats flanking the direct repeats (40). No inverted repeats were found flanking the 4-bp direct repeats located at both sides of the B. melitensis 16M DNA fragment deleted in B. abortus. However, several regions able to form hairpin structures have been detected in the deleted fragment (data not shown) that might have contributed to the excision of the 25-kb fragment in B. abortus. This deletion would have a low formation frequency since the length of the direct repeats is very short, the distance between them is very long, and no inverted repeats are found close to the direct repeats. Moreover, the deletion has been detected only in B. abortus and probably occurred before the differentiation of this species in its biovars, all of them lacking the 25-kb DNA fragment (Fig. 4).

Twenty-one putative genes identified in the B. melitensis 16M DNA are completely deleted in B. abortus (Fig. 3), two other genes located in each end of the deleted fragment (bme10 and bme17) are partially removed from this species, and the corresponding proteins might not be synthesized in B. abortus. Most of these 23 genes would code for proteins showing a significant degree of identity with proteins involved in the synthesis of polysaccharides in other bacteria (57) (Table 2), suggesting that they might lead to the synthesis of a Brucella polysaccharide. Two main pathways have been described for the biosynthesis of polysaccharides in bacteria. In both mechanisms the synthesis of the polysaccharide subunit starts with the linkage of a sugar nucleotide, usually UDP-GlcNAc, to UndP that is anchored to the inner membrane. The first pathway, known as the Wzy (polymerase)-dependent pathway, has been shown to be involved in the synthesis of many lipopolysaccharide (LPS) O-chains, mainly in heteropolymeric O-chains (those made of repeating units of different sugars) (61, 62). In the Wzy-dependent pathway, the repeat units of the polysaccharide are synthesized, at the cytoplasmic face of the inner membrane, by the action of glycosyltransferases that transfer a sugar from a nucleotide-sugar complex. The repeat units are exported across the cytoplasmic membrane by Wzx, a multiple membrane-spanning protein, and polymerized at the periplasmic face by the Wzy polymerase (61, 62). The second main mechanism for the biosynthesis of bacterial polysaccharides, known as the Wzy-independent pathway, has been observed in homopolymeric polysaccharides (made of repeating units of the same sugar). The polysaccharide is synthesized, in the cytosolic face of the inner membrane, by the sequential addition of sugars by glycosyltransferases, and then it is transported across the cytoplasmic membrane by an ATP-binding cassette (ABC) transporter (61, 62). In the case of LPS molecules, the O-polysaccharide chains synthesized by the Wzy-dependent or Wzy-independent pathway would then be ligated to the lipid A core complex by the WaaL ligase (61, 62). The mechanism of transport of LPS and polysaccharides not attached to lipid A core, such as capsular polysaccharides and exopolysaccharides (EPS), across the outer membrane of gram-negative bacteria remains almost unexplored. However, some periplasmic proteins and OMPs have been suggested to be involved in this process in some capsular polysaccharides and EPS (8, 19, 43, 49, 50, 63).

According to homologies with other proteins in the database, Bme31 might be the enzyme catalyzing the transfer of the first sugar of a polysaccharide unit (probably GlcNAc that would be transferred from a UDP-GlcNAc sugar-nucleotide donor substrate) to UndP attached to the cytoplasmic membrane. Six putative glycosyltransferases (Bme1, Bme6, Bme7, Bme8, Bme27, and Bme29), enzymes that add a sugar unit from a nucleotide-sugar substrate to a polysaccharide growing chain, were identified in the 25-kb DNA fragment deleted in B. abortus (57) (Table 2). A specific sugar transferase is required for each different linkage between the sugars of a polysaccharide unit, thus suggesting that if the genes identified in the DNA fragment deleted in B. abortus lead to the synthesis of a polysaccharide, it would be made up of at least six sugar units, in addition to the initiator GlcNAc.

Eight hypothetical proteins that might be involved in the synthesis of nucleotide-sugar complexes or in the chemical modification of sugars have also been identified: Bme9, Bme10, Bme22, and Bme24 show homology with epimerases and dehydratases from other polysaccharide clusters (57) (Table 2); Bme4 was identified as a putative acetyltransferase (57), Bme26 was identified as a putative aminotransferase, and Bme21 was identified as a putative methyltransferase; Bme20 might act as a glucose-1-phosphate cytidylyltransferase catalyzing the production of the sugar-nucleotide CDP-D-glucose from CTP and D-glucose-1-phosphate; and Bme9 showed 64.0% identity in 356 amino acids overlapping with Gmd from B. melitensis 16M (data not shown), a putative GDP-mannose dehydratase involved in biosynthesis of the LPS O-chain (24). However, the highest level of identity (68.1% identity) was obtained with NoeL (57), a GDP-mannose dehydratase from Rhizobium sp. (18).

According to its hydrophobicity profile, Bme3 was previously proposed to be equivalent to Wzx (57), a multiple membrane-spanning protein that promotes the export of polysaccharides across the cytoplasmic membrane in the Wzy-dependent pathway (15, 62). However, the multiple membrane-spanning hypothetical protein Bme30 identified in this work showed homology with Wzx from E. coli and serovar Typhimurium and with ExoT from R. meliloti, a protein that seems to be involved in the export of EPS (23). Therefore, Bme30 seems to be a better candidate than Bme3 for the transport of the hypothetical Brucella polysaccharide across the cytoplasmic membrane. Wzy, the polymerase in the Wzy-dependent pathway, is also predicted to be located in the cytoplasmic membrane and to contain multiple membrane-spanning segments (62). B. melitensis 16M Bme3, showing a hydrophobicity profile similar to that of Wzy from other bacteria (data not shown), is the sole hypothetical protein without a clear assigned function that seems to fulfill this requirement. Thus, Bme3 might be the Wzy polymerase for the synthesis of the Brucella polysaccharide.

Another protein intervening in the synthesis of bacterial polysaccharides by the Wzy-dependent pathway is Wzz, which is involved in the chain length determination (62). The Wzz functional homologous proteins are ca. 42 kDa, contain two highly conserved potential transmembrane domains in the N- and C-terminal regions, and have been mainly associated with the biosynthesis of bacterial O-polysaccharides and some capsular polysaccharides (62). Several proteins of ca. 80 kDa mainly involved in the biosynthesis of EPS have been identified as Wzz-like proteins although, in general, their function has not been experimentally established. These proteins are larger than the Wzz functional homologues and contain a C-terminal cytoplasmic domain that Wzz lacks (25, 62). One of the Wzz-like proteins is ExoP from Rhizobium meliloti, which is thought to intervene in the polymerization and/or export process of the succinoglycan octasaccharide units (6, 42), although it has also been suggested that ExoP must play another critical role in the biosynthesis of this EPS (25). The B. melitensis 16M bme12 gene would code for a protein of 79 kDa (57), predicted to be located in the cytoplasmic membrane with two membrane-spanning domains in the N- and C-terminal ends of the protein, which shows homology with ExoP from R. meliloti (57) and a similar hydrophobicity profile (data not shown). Therefore, Bme12 might be involved in the regulation of polysaccharide length and/or export of the polysaccharide.

Considering the homology of several putative genes located in the B. melitensis 16M 25-kb DNA fragment deleted in B. abortus with genes involved in the synthesis of several bacterial EPS, the possibility that these genes lead to the synthesis of an EPS must be taken into account. This possibility would be reinforced by the identification of proteins involved in the transport of the polysaccharide across the periplasm and the outer membrane. Genes encoding periplasmic proteins and OMPs have been identified in several clusters directing the synthesis of several bacterial EPS and capsular polysaccharides and are thought to be involved in the export of the polysaccharide across the periplasm and the outer membrane (8, 19, 43, 49, 50, 63). Three proteins predicted to be located in the periplasmic spaceBme2, Bme11, and Bme25would be encoded in the 25-kb DNA fragment of B. melitensis 16M deleted in B. abortus (57) (Table 1). Bme11 showed homology with R. meliloti ExoF (57) involved in the synthesis of the EPS succinoglycan. ExoF shows homology with the periplasmic E. coli KpsD protein, which is involved in the biosynthesis of the K5 capsule (43, 63), and was suggested to play a role in the export of the rhizobial EPS succinoglycan (22). Accordingly, periplasmic Bme11 might be involved in the last steps of export of the hypothetical B. melitensis polysaccharide. Bme5 that corresponds to Omp31 (56, 57), a B. melitensis 16M major immunogenic OMP, might be involved in the export of the polysaccharide across the outer membrane, as was shown for other bacterial OMPs. This is the case for the OMP Wza encoded in the cluster responsible for the synthesis of the EPS colanic acid in E. coli and Salmonella enterica (43, 49, 50), the OMP AmsH of Erwinia amylovora involved in the synthesis of EPS (8), and the Neisseria meningitidis OMP CtrA involved in capsule expression (19).

The B. melitensis periplasmic Bme2 and Bme25 proteins did not show clear homology with other proteins in the database (57) (Table 2), and it is difficult to determine their hypothetical functions. Taking into account the probable periplasmic locations of these proteins, if they are involved in the synthesis of a Brucella spp. polysaccharide they might act in the last steps of the pathway after the export of the polysaccharide across the cytoplasmic membrane. Bme17, Bme23, and Bme28 showed homology with three Streptomyces coelicolor proteins of unknown function but whose genes are located close to other genes that seem involved in the synthesis of a polysaccharide (accession number AL121855), and they might also be related to the synthesis of a Brucella polysaccharide.

Several clusters for the biosynthesis of bacterial polysaccharides are controlled by the action of regulatory proteins (9, 27, 51). Bme13 and Bme14, which would be encoded by genes flanking the right side of the B. melitensis 16M DNA fragment deleted in B. abortus (Fig. 3), showed homology with regulatory proteins (57) and might have a regulatory role in the synthesis of the hypothetical Brucella polysaccharide. However, they might be unrelated to this process and be involved in the regulation of another process. Finally, putative B. melitensis 16M proteins encoded by genes flanking the left side of the large DNA fragment deleted from B. abortus (Bme18, Bme19, and Bme16) (Fig. 3; Table 2) seem to be unrelated to the synthesis of a polysaccharide.

Considering that the 25-kb DNA fragment deleted in B. abortus has been detected in all of the other Brucella species (Fig. 4), if the genes that have been identified lead to the synthesis of a polysaccharide, that polysaccharide would be present in all of the Brucella species with the exception of B. abortus. Since the polysaccharides identified in the genus Brucella have been detected in all of the species including B. abortus, the genes described here and in a previous study (57) might intervene in the synthesis of a polysaccharide not identified until now. In spite of the fact that most of these genes are homologous to genes involved in the synthesis of other bacterial polysaccharides and that Omp31 (Bme5) has been shown to be synthesized in all of the Brucella species except B. abortus (58) and to be located in the Brucella outer membrane (12), the possibility that they no longer lead to the synthesis of a polysaccharide must be taken into account. Genes encoding homologs of several flagellum-related proteins have been found in B. abortus that, as in the other Brucella species, is a nonmotile bacterium, and it was hypothesized that brucellae may have lost motility during evolution because it is not essential or even detrimental for its cycle as animal pathogens (28). The genus Brucella is a member of the alpha-2 subdivision of the class Proteobacteria that contains other members that live in close association with eucaryotic cells, such as Rhizobium and Agrobacterium spp., both of which synthesize an EPS important for the interaction with the host plant cells. The common ancestor of the alpha-2 subdivision of the class Proteobacteria has been suggested to have EPS (36), and the genes found in the B. melitensis 16M DNA fragment of 25 kb deleted in B. abortus might be a remnant of this common ancestor.

Deletions that enhance the virulence of bacterial pathogens have been described and suggested to complement gene acquisition, such as pathogenicity islands, in the evolution of bacterial pathogens, even allowing a broadening of the host range (34). The deletion that occurred in B. abortus might have contributed to an evolution of Brucella spp., allowing the adaptation to survive in a new animal host. Further studies are necessary to determine whether a polysaccharide not identified until present is synthesized in the genus Brucella. Discovery of a novel polysaccharide, probably not synthesized in B. abortus, would be of great interest since polysaccharides are virulence factors in Brucella spp. and other microorganisms and this might also explain some of the differences in pathogenicity and host preference among the brucellae.