Molecular Characterization of a Brucella Species Large DNA Fragment Deleted in Brucella abortus Strains: Evidence for a Locus Involved in the Synthesis of a Polysaccharide

Abstract

A Brucella melitensis 16M DNA fragment of 17,119 bp, which contains a large region deleted in B. abortus strains and DNA flanking one side of the deletion, has been characterized. In addition to the previously identified omp31 gene, 14 hypothetical genes have been identified in the B. melitensis fragment, most of them showing homology to genes involved in the synthesis of a polysaccharide. Considering that 10 of the 15 genes are missing in B. abortus and that all the polysaccharides described in the Brucella genus (lipopolysaccharide, native hapten, and polysaccharide B) have been detected in all the species, it seems likely that the genes described here might be part of a cluster for the synthesis of a novel Brucella polysaccharide. Several polysaccharides have been identified as important virulence factors, and the discovery of a novel polysaccharide in the brucellae which is probably not synthesized in B. abortus might be interesting for a better understanding of the pathogenicity and host preference differences observed between the Brucella species. However, the possibility that the genes described in this paper no longer encode the synthesis of a polysaccharide cannot be excluded. Brucellae belong to the alpha-2 subdivision of the class Proteobacteria, which includes other microorganisms living in association with eucaryotic cells, some of them synthesizing extracellular polysaccharides involved in the interaction with the host cell. The genes described in this paper might be a remnant of the common ancestor of the alpha-2 subdivision of the class Proteobacteria, and the brucellae might have lost such extracellular polysaccharide during evolution if it was not necessary for survival or for establishment of the infectious process. Nevertheless, further studies are necessary to identify the entire DNA fragment missing in B. abortus strains and to elucidate the mechanism responsible for such deletion, since only 9,948 bp of the deletion was present in the sequenced B. melitensis DNA fragment.

Microorganisms belonging to the genus Brucella are gram-negative, facultative intracellular bacteria that are able to cause infections in humans and many animal species. Brucellae have been classified, according to differences in pathogenicity and host preference, into six species: Brucella melitensis, B. abortus, B. suis, B. ovis, B. canis, and B. neotomae (24). Several biovars have been also described in some of these species and are currently differentiated by serotyping, phage typing, dye sensitivities, and culture and metabolic properties (4). A high degree of homology has been found within the Brucella genus at the DNA level, which has led to a proposal to classify the genus into just one species (77, 78). However, the preferred-host classification is still used, since the six species and some of their biovars can also be differentiated on the basis of DNA polymorphisms detected by several techniques (1, 11, 12, 20, 21, 27, 29, 30, 39, 42, 53, 54, 58, 80).

In spite of the high degree of DNA homology of the Brucella species, important divergences in DNA should exist between species to explain their differential behavior. Studies leading to the identification of DNA variability between species would be of great interest in our understanding of the differences in pathogenicity and host preference observed in the brucellae.

Small deletions in several genes have been detected in some Brucella species (19, 22, 23, 29, 67), sometimes leading to the lack of production of the encoded protein (67), and the lack of expression of an existing gene has also been reported (28, 42). Moreover, DNA deletions of several sizes and DNA inversions have been shown to exist, by restriction enzyme mapping, in some Brucella species (54), but these DNA regions have not yet been identified. Cloning and sequencing of B. melitensis 16M omp31 (79), a gene coding for a Brucella major outer membrane protein, has allowed us to verify that this gene is missing in B. abortus strains but exists in all the other Brucella species (80). Moreover, DNA bordering this gene is also missing in B. abortus, and the size of the deletion was estimated to be about 10 kb (80). Such a large deletion presumably removes genes in addition to omp31, and their identification might shed more light on the understanding of the pathogenicity and host preference differences between the brucellae. In the present work, we characterized a large DNA fragment missing in B. abortus and assigned functions to the genes involved in this deletion, giving evidence for a gene machinery involved in the synthesis of a novel Brucella polysaccharide.

MATERIALS AND METHODS

Bacterial strains and plasmids.

B. melitensis 16M and B. abortus 544 were obtained from the INRA Brucella Culture Collection, Nouzilly (BCCN), France. The cultures were grown on tryptic soy agar (BioMérieux, Marcy l’Etoile, France) supplemented with 0.1% (wt/vol) yeast extract (Difco Laboratories, Detroit, Mich.). The strains were checked for purity and for species and biovar characterization by standard procedures (4). Escherichia coli JM109 cells bearing the different recombinant plasmids used in this study were cultured overnight on Luria-Bertani medium containing 50 μg of ampicillin ml⁻¹.

The relevant characteristics of the plasmids used in this study are shown in Fig. 1. Plasmid pNV3103 contains a B. melitensis 16M DNA fragment of 17,119 bp that includes a large fragment deleted in B. abortus and adjacent DNA bordering one side of the deletion. This recombinant plasmid was obtained by subcloning, into the NotI site of pGEM-5Zf (Promega, Madison, Wis.), the DNA insert of a phage from a B. melitensis 16M genomic library constructed in λGEM-12 XhoI half-site arms (Promega) as described previously (79). Plasmids pNV3129 and pNV3130 contain the second BamHI fragment of pNV3103 but in the opposite orientation relative to lacZ and were obtained by subcloning the insert of pNV3103 in pGEM-7Zf (Promega). Plasmids pNV3131 and pNV3132 bear the same BamHI-KpnI fragment of pNV3103 cloned in pGEM-7ZF but in the opposite orientation in relation to lacZ. Plasmids pNV3133 and pNV3134 contain the same KpnI-BamHI fragment of pNV3103 cloned in pGEM-7Zf but in the opposite orientation in relation to lacZ.

FIG. 1.

Plasmids and primers used for PCR amplification in this study. pNV3103 contains a 17,119-bp B. melitensis 16M DNA fragment and was obtained by subcloning, into pGEM-5Zf, the insert DNA of a phage from a B. melitensis 16M genomic library as described previously (79). pNV3131, pNV3132, pNV3133, p NV3134, pNV3129, and pNV3130 were obtained by subcloning the pNV3103 insert into pGEM-7Zf. The position of the previously identified omp31 gene (79) is represented in black. Plasmid pNV3138 was obtained by cloning in pGEM-T the B. abortus 544 DNA fragment amplified with primers N3-PCR and N89-P. pNV3140 was obtained by cloning in pGEM-T the B. melitensis 16M DNA fragment amplified with primers N119-P1 and N120-P2. DNA bordering the left side of the B. abortus deletion point and not cloned in the pNV3103 insert is represented as a shaded box in both pNV3138 and pNV3140. Restriction sites: B, BamHI; H, HindIII; K, KpnI. BamHI sites located at both ends of the pNV3103 insert belong to the multiple-cloning site of phage λGEM-12 and not to the B. melitensis 16M DNA insert. Primers are marked by arrows in the 5′ → 3′ direction.

pNV3138 was obtained by cloning into pGEM-T (Promega) the B. abortus 544 DNA fragment PCR amplified with primers N3-PCR and N89-P. pNV3140 bears the B. melitensis 16M DNA fragment amplified with primers N119-P1 and N120-P2 and ligated into pGEM-T.

DNA preparation.

For Brucella genomic DNA extraction, the strains were cultured for 24 h at 37°C on tryptic soy agar-yeast extract slopes and harvested, in 3 ml of sterile distilled water, by centrifugation at 2,000 × g for 10 min. The pellet was suspended in 567 μl of TE/sodium buffer (50 mM Tris, 50 mM EDTA, 100 mM NaCl [pH 8.0]). Then, 30 μl of 10% (wt/vol) sodium dodecyl sulfate (SDS) solution and 3 μl of 2% (wt/vol) proteinase K solution were added, and the mixture was kept at 37°C for 1 h. The lysed cell suspension was extracted twice with phenol-chloroform, and nucleic acids were precipitated by gently mixing the aqueous phase with 2 volumes of cold ethanol. The precipitate was dissolved in 100 μl of TE (10 mM Tris, 1 mM EDTA [pH 8.0]). The amount of DNA was measured by electrophoresis of an aliquot of each sample through 0.8% agarose gels and comparison with standard DNA solutions.

Plasmid DNA was extracted from recombinant E. coli JM109 cells by standard procedures (66).

DNA sequencing.

Insert DNA from plasmids pNV3129, pNV3130, pNV3131, pNV3132, pNV3133, and pNV3134 (obtained by subcloning the pNV3103 insert) (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 approximately 400 bp was obtained for each initial plasmid and used to determine the entire sequence of the 17,119-bp insert of B. melitensis 16M contained in pNV3103. pNV3138 and pNV3140 inserts were sequenced without exonuclease digestion. Plasmid DNA was obtained and purified by using the Wizard Plus SV minipreps system (Promega) as specified by the manufacturer.

Purified plasmid DNA was sequenced by primer-directed dideoxy sequencing (68) with an ABI PRISM 377 DNA sequencer (Perkin-Elmer, Foster City, Calif.) and the forward pUC19 primer. In some cases, specific Brucella DNA primers or the reverse pUC19 primer were also used.

DNA amplification and cloning of PCR-amplified DNA.

Primers for PCR amplification of B. melitensis 16M or B. abortus 544 DNA fragments were selected according to the pNV3103 or pNV3138 deduced sequence. The following primers were used (Fig. 1): 31R, 5′-TCTGCGTGATGAAATGCTGG-3′; 31R-P, 5′-CCAGCATTTCATCACGCA-3′; N3-PCR, 5′-ACTGGTTTCATTCCCGCC-3′; N95-P, 5′-GCGATAGTTCACCGTTGT-3′; N49-PCR, 5′-AGAGCAGGCGTTCCACAC-3′; N89-P, 5′-ATCAAGCCTGCGGGACAT-3′; N119-P1, 5′-TTGCTGGTCTTGCGGTGT-3′; N120-P1, 5′-CCGTGCCGATTTTTATGG-3′; and N120-P2, 5′-AAGCCTTTTCGGATGAGC-3′. Amplification reaction mixtures were prepared in volumes of 50 μl containing 1× PCR buffer (MAD-GEN, Valencia, Spain), 1.5 mM MgCl₂, 200 μM each deoxynucleoside triphosphate, 1 μM each primer, 200 ng of genomic DNA, and 2.5 U of DNA polymerase Super-Therm (MAD-GEN). The temperature cycling for the amplification was performed in a PTC-100 thermocycler (MJ Research Inc., Watertown, Mass.) as follows: the first cycle was 94°C for 5 min (denaturation); the next 35 cycles were 58°C for 1 min (annealing), 72°C for 2 min (extension), and 94°C for 1 min (denaturation); the last cycle was 58°C for 1 min (annealing) and 72°C for 10 min (extension). The size of the amplified DNA was determined by electrophoresis on 0.8% agarose gels and comparison with DNA molecular weight standards (Boehringer, Mannheim, Germany). When desired, the amplified DNA was cloned into plasmid pGEM-T (Promega) as specified by the manufacturer.

DNA and protein analysis.

DNA sequences were analyzed for putative coding regions by using the DNAStrider 1.2 program (51). Searches for DNA and protein homologies were performed with the FASTA program (26a, 59). Cellular location and motifs of the predicted proteins were determined with the PSORT (57, 60a) and MOTIF programs (55a), respectively. Multiple alignments were performed with CLUSTAL W 1.74 (23a, 76).

Nucleotide sequence accession number.

The nucleotide sequences of the B. melitensis 16M inserts of pNV3103 and pNV3140 and of the B. abortus 544 insert of pNV3138 have been submitted to the DDBJ/EMBL/GenBank databases under accession no. AF076290, AF076289, and AF076288, respectively.

RESULTS

DNA sequence of the pNV3103 insert DNA.

Plasmid pNV3103, which had been obtained previously, bears a B. melitensis 16M DNA insert of about 17 kb that was shown to contain a large DNA fragment, including the omp31 gene (missing in B. abortus strains) and flanking DNA (80). The entire nucleotide sequence of both strands of the pNV3103 insert was determined by sequencing, as described in Materials and Methods, the insert DNA from plasmids pNV3131, pNV3132, pNV3133, pNV3134, pNV3129, and pNV3130, obtained by subcloning the pNV3103 B. melitensis 16M insert (Fig. 1). The B. melitensis insert of pNV3103 was shown to be 17,119 bp (Fig. 2), and analysis of the nucleotide sequence with DNA Strider and FASTA has allowed us to identify 15 open reading frames (ORFs) that potentially encode proteins (Fig. 2; Tables 1 and 2).

FIG. 2.

ORF distribution of the B. melitensis 16M DNA fragment cloned in plasmid pNV3103. The ORFs deleted in B. abortus strains are shaded, except that corresponding to the previously identified omp31 (79), which is represented in black. The vertical line shows the limit of the right side of the deletion in B. abortus located between nucleotides 9948 and 9949. Arrows show the direction of ORF transcription. The extent of the B. melitensis 16M plasmids inserts used for sequencing is also marked.

TABLE 1.

Characteristics of the proteins deduced from the putative genes found in the B. melitensis 16M DNA insert of plasmid pNV3103

Protein	Start nucleotide (putative start codon)	End nucleotide	Putative RBS sequence (position)a	Strand of transcription	No. of amino acids	Size of protein (kDa)	Cellularb localization
Bme1c		618		Direct
Bme2	615 (ATG)	1232	GGAG (12)	Direct	205	22.0	P
Bme3	1310 (ATG)	2491	GGAG (4)	Direct	393	44.7	CM
Bme4	2503 (ATG)	3117	AAGG (9)	Direct	204	22.0	CM
Bme5d	3514 (ATG)	4222	GAGG (7)	Direct	240	25.3	OM
Bme6	5431 (ATG)	4307	AGGAGG (9)	Reverse	374	40.5	CM
Bme7	6663 (ATG)	5428	GGA (9)	Reverse	411	44.6	C
Bme8	7934 (ATG)	6684	AGGA (10)	Reverse	416	46.4	C
Bme9	8198 (TTG)	9268	GGAGG (9)	Direct	356	40.6	C
Bme10	9252 (ATG)	9992	GGAGGT (9)	Direct	246	26.5	C
Bme11	11583 (GTG)	10336	GGA (5)	Reverse	415	45.0	P
Bme12	13782 (GTG)	11593	AAGGAG (8)	Reverse	729	79.0	CM
Bme13	14586 (ATG)	13819	GAGG (9)	Reverse	255	28.2	CM
Bme14	14949 (ATG)	15707	GGT (7)	Direct	252	28.5	CM
Bme15c		16084		Reverse

^aPosition of the putative RBS expressed as nucleotides upstream of the start codon. 

^bThe most probable cellular localization of each protein was determined with the PSORT program (60a). C, cytoplasm; CM, cytoplasmic membrane; OM, outer membrane; P, periplasm. 

^cOnly DNA encoding the C-terminal end of Bme1 and Bme15 was cloned in pNV3103. Therefore, the start position, RBS, number of amino acids, size, and cellular localization are not given for these proteins. 

^dBme5 corresponds to the previously identified outer membrane protein Omp31 (79). 

TABLE 2.

Most representative homologies of the B. melitensis 16M hypothetical proteins encoded by pNV3103 to other proteins in the databases

pNV3103 protein	No. of aaa	Similar proteins and source	No. of aa	Putative function	% Identity (aa overlap)	Expectation value	Accession no.
Bme1b	205	KdtX, Serratia marcescens	257	Glycosyltransferase	28.2 (173)	1.2 × 10−2	U52844
		HI0635, Haemophilus influenzae	254	Unknown	26.1 (172)	1.4 × 10−1	U32748
		LgtF, Neisseria meningitidis	252	β-1,4-Glucosyltransferase	26.1 (176)	3.0	U58765
Bme2	205	ORF C, Streptomyces coelicolor	203	Inhibition of plasmid maintenance	27.3 (176)	9.9 × 10−2	X15866
		HI1701, Haemophilus influenzae	247	Unknown	31.3 (150)	1.0 × 10−1	U32843
		GdmH, Staphylococcus gallinarum	330	Epidermin secretion	25.3 (162)	1.2 × 10−1	U61158
		EpiH, Staphylococcus epidermidis	330	Epidermin secretion	26.3 (148)	1.6 × 10−1	U77778
Bme3	393	Hypothetical, Escherichia coli	208	Unknown	24.7 (166)	1.3	AE000270
		SmpX, Synechococcus spp.	269	Channel protein	24.2 (149)	1.6	D43774
Bme4	204	Galactoside acetyltransferase, Escherichia coli	182	Galactoside acetyltransferase	39.7 (156)	2.3 × 10−23	D90843
		NodL, Rhizobium leguminosarum	190	Acetyltransferase	31.3 (180)	6.5 × 10−11	Y00548
		NodL, Rhizobium meliloti	183	Acetyltransferase	28.9 (180)	1.6 × 10−9	X61083
		MJ1064, Methanococcus jannaschii	214	Galactoside acetyltransferase	27.1 (140)	1.9 × 10−7	U67549
Bme5c	240	Omp31, Brucella melitensis 16M	240	Unknown	100 (240)	0	U39453
		Omp25, Brucella suis	213	Unknown	34.7 (245)	3.7 × 10−6	U39397
		Omp25, Brucella neotomae	213	Unknown	34.7 (245)	3.7 × 10−6	U39359
Bme6	374	Hypothetical, Synechocystis spp.	386	Unknown	26.6 (391)	1.1 × 10−13	D90901
		MTH370, Methanobacterium thermoautotrophicum	395	LPS biosynthesis, RfbU related	30.5 (364)	7.8 × 10−13	AE000822
		MTH173, Methanobacterium thermoautotrophicum	382	LPS biosynthesis, RfbU related	27.0 (322)	9.5 × 10−12	AE000805
Bme7	411	MTH450, Methanobacterium thermoautotrophicum	411	LPS biosynthesis, RfbU related	33.3 (213)	4.8 × 10−7	AE000829
		YefL, Escherichia coli	406	Unknown	32.5 (243)	9.8 × 10−7	D90842
		BplH, Bordetella pertussis	390	Glycosyltransferase	29.1 (237)	3.0 × 10−6	X90711
		AmsK, Erwinia amylovora	407	EPS synthesis	31.8 (198)	1.3 × 10−5	X77921
Bme8	416	Hypothetical, Synechocystis spp.	424	Unknown	29.1 (409)	0	D90901
		Cap5L, Staphylococcus aureusd	401	Glycosyltransferase	23.3 (390)	5.5 × 10−21	U81973
		YefD, Escherichia coli	407	Glycosyltransferase	22.2 (388)	1.0 × 10−10	D90843
		WbpJ, Pseudomonas aeruginosa	413	Glycosyltransferase	23.7 (367)	1.7 × 10−10	U50396
Bme9	356	NoeL, Rhizobium spp.	351	GDP–d-mannose-4,6-dehydratase	68.1 (339)	0	AE000064
		RfbD, Vibrio cholerae MO45e	348	GDP–d-mannose dehydratase	60.9 (345)	0	U24571
		Gmd, Yersinia enterocolitica	372	GDP-mannose dehydratase	60.6 (350)	0	U46859
		HP0044, Helicobacter pylori	381	GDP–d-mannose dehydratase	55.5 (371)	0	AE000526
Bme10	246	NolK, Rhizobium spp.	314	Nucleotide sugar epimerase/dehydrogenase	58.1 (210)	0	AE000064
		WbcJ, Yersinia enterocolitica	321	Epimerase-reductase	53.4 (204)	0	U46859
		WcaG, Escherichia coli	321	Epimerase-reductase	51.2 (207)	0	U38473
		YefB, Escherichia coli	321	Nucleotide sugar epimerase/dehydrogenase	51.2 (207)	0	D90843
Bme11	415	ExoF, Rhizobium meliloti	421	Biosynthesis of succinoglycan	25.2 (409)	1.5 × 10−18	L05588
		PrsE, Rhizobium leguminosarum	435	Protein secretion	25.7 (249)	1.8 × 10−7	Y12758
		PrsE, Rhizobium meliloti	439	Protein secretion	24.7 (287)	2.6 × 10−7	U89163
		GumB, Synechocystis spp.	504	Unknown	28.5 (172)	9.9 × 10−5	D90904
Bme12	729	GumC, Xanthomonas campestris	449	Biosynthesis of xanthan gum	25.2 (449)	2.6 × 10−17	U22511
		ExoP, Rhizobium meliloti	786	Biosynthesis of succinoglycan	23.7 (739)	1.2 × 10−16	L20758
		AceD, Acetobacter xylinum	637	EPS export	23.1 (628)	1.8 × 10−14	X94981
Bme13	255	RegX, Rhodobacter sphaeroides	212	Response regulator	35.3 (215)	2.5 × 10−20	U80081
		FlhR, Paracoccus denitrificans	236	Unknown	30.2 (212)	5.9 × 10−9	AJ223460
		GlpR, Pseudomonas aeruginosa	221	Glycerol regulatory protein	26.3 (209)	7.5 × 10−9	M60805
		EpsR, Pseudomonas solanacearum	236	Negative regulator of EPS synthesis	25.9 (212)	3.5 × 10−8	M61197
Bme14	252	Fnr, Rhodobacter sphaeroides	248	Anaerobic regulatory protein	26.0 (196)	3.6 × 10−10	Z49746
		FixK, Bradyrhizobium japonicum	237	Transcriptional regulator	25.5 (212)	8.7 × 10−10	M86805
		AadR, Rhodopseudomonas palustris	239	Transcriptional regulator	25.7 (187)	8.8 × 10−10	M92426
		FnrA, Pseudomonas stutzeri	244	Transcriptional regulator	25.4 (209)	2.3 × 10−9	Z26044
Bme15b	344	HI0148, Haemophilus influenzae	379	Unknown	36.6 (344)	0	U32700
		YjhT, Escherichia coli	404	Unknown	33.0 (342)	7.4 × 10−19	AE000501

^aaa, amino acid. 

^bbme1 and bme15 were not entirely cloned in pNV3103. Therefore, the number of amino acids shown in the table corresponds to the C-terminal domain cloned in pNV3103. 

^cBme5 corresponds to the previously identified Omp31 (79). 

^dSame level of homology to Cap8L from Staphylococcus aureus. 

^eBme9 also showed homology to other RfbD proteins from Vibrio cholerae. 

(i) bme1, bme2, bme3, and bme4.

The first ORF (bme1) is located between nucleotides 1 and 618 (Fig. 2; Table 1). The deduced amino acid sequence had 28.2% identity to the C-terminal end of the kdtX product from Serratia marcescens (Table 2), which has been defined as a putative glycosyltransferase involved in core lipopolysaccharide (LPS) biosynthesis (40). A 26.1% identity was also observed to the C-terminal end of LgtF from Neisseria meningitidis (Table 2), which is a glycosyltransferase involved in the inner-core biosynthesis of lipooligosaccharide (45). It is probable that the entire bme1 gene is not present in the pNV3103 insert, since no putative ribosome binding site (RBS) has been found near an ATG codon and since homology to KdtX and LgtF covers only the C-terminal end of these proteins. The characteristic motif EX₇E, found toward the C-terminal end of many glycosyltransferases (5, 70), was also found in Bme1, which suggests that the probable function of this protein is as a glycosyltransferase involved in the synthesis of a polysaccharide.

bme2 starts at position 615, slightly overlapping the end of bme1, and ends at nucleotide 1232 (Fig. 2; Table 1). The most probable location of the deduced protein was determined to be the periplasm, and a putative signal peptidase cleavage site was detected at amino acid 50. Homology was found to the proteins encoded by orfC and HI1701 of Streptomyces coelicolor and Haemophilus influenzae, respectively (Table 2). The protein encoded by orfC is involved in the inhibition of extrachromosomal maintenance of the Streptomyces plasmid SLP1 (37). GdmH and EpiH, from Staphylococcus gallinarum and Staphylococcus epidermidis, respectively, also had homology to Bme2 (Table 2). Both proteins seem to be implicated in the secretion of the lantibiotic epidermin (60).

bme3, located between nucleotides 1310 and 2491, would encode a highly hydrophobic cytoplasmic membrane protein (Fig. 2; Table 1). Only the N-terminal end of the deduced protein showed homology (about 24% identity) to a hypothetical protein of E. coli of unknown function and to SmpX from Synechococcus spp. (Table 2), which has been defined as a pore-forming channel protein belonging to a family of homologous intrinsic membrane proteins involved in a variety of transport processes (46). The hydrophobicity profile of Bme3 was very similar to that of Wzx proteins (Fig. 3), reported to be involved in the transfer of polysaccharides across the cytoplasmic membrane and also described as multiple membrane-spanning proteins (73). The Wzx proteins of different polysaccharide biosynthesis clusters have little similarity at the amino acid sequence level but are predicted to have structural homology (49, 73).

FIG. 3.

Comparison of the hydrophobicity profiles of Bme3 and Wzx homologous proteins, determined by the method of Kyte and Doolittle (47) with a window length of 11 amino acid residues. (A) Bme3. (B) Wzx of the E. coli K-12 colanic acid cluster (accession no. U38473). (C) Wzx of the P. aeruginosa B-band LPS O-chain cluster (accession no. U50396). (D) Wzx of the Y. enterocolitica O:8 LPS O-chain cluster (accession no. U46859).

bme4 is located between nucleotides 2503 and 3117 (Fig. 2; Table 1) and would code for a putative protein which displays significant homology to several acetyltransferases (Table 2), enzymes involved in the acetylation of several substrates. The highest homology was observed to the galactoside acetyltransferase from E. coli K-12 (44) and NodL from Rhizobium leguminosarum. The latter protein acetylates various substrates, such as lipooligosaccharides, chitin fragments, and N-acetylglucosamine, and appeared to be located in the cytoplasm (10). Analysis of the amino acid sequence of the hypothetical Bme4 protein with the PSORT program identified the cytoplasmic membrane as the most probable location of the protein, rather than the cytoplasm as described for the NodL acetyltransferase. However, other acetyltransferases, including the acetyltransferase intervening in the synthesis of the Rhizobium meliloti exopolysaccharide (EPS) succinoglycan, have been described as cytoplasmic membrane proteins (36).

No regions of dyad symmetry have been found in the intergenic spaces between the four genes, which suggests that they might be transcribed as a single unit.

(ii) bme5.

bme5 corresponds to the previously identified omp31 gene (79), coding for an immunogenic major outer membrane protein in all Brucella species except in B. abortus, where omp31 was shown to be absent (80). The gene extended from nucleotides 3514 to 4236 (Fig. 2; Table 1) (79). A putative cleavage site for signal peptidase is located between amino acids 19 and 20 of the deduced protein (79).

Although no clear regions of dyad symmetry have been found following the translation termination codon of bme4, omp31 is probably not cotranscribed with bme4. Expression of omp31 has been detected with an anti-Omp31 monoclonal antibody in recombinant E. coli bearing a plasmid containing a PCR product covering omp31 and part of the intergenic region between bme4 and omp31. Expression of omp31 has been detected independently of the orientation of the PCR product in relation to the lac promoter (data not shown), which suggests that omp31 is transcribed from its own promoter.

A region of dyad symmetry that might function as rho-independent transcription terminator was detected 18 bp downstream the stop codon of the gene.

(iii) bme6, bme7, and bme8.

The three hypothetical genes bme6 to bme8 would be transcribed from the reverse strand according to the schema shown in Fig. 2 and are probably cotranscribed, since no regions of dyad symmetry have been found in the intergenic spaces. The dyad symmetry region downstream of omp31 might also function as the transcription terminator of this hypothetical operon.

bme6 is located between nucleotides 4307 (end) and 5431 (start) and would produce a 374-residue protein, whose most probable location is the cytoplasmic membrane (Fig. 2; Table 1). The highest homology level was found to a Methanobacterium thermoautotrophicum protein, which has been defined as an LPS biosynthesis RfbU-related protein (72) (Table 2). RfbU has mannosyltransferase activity in Salmonella enterica (50). The protein also showed homology to a wide number of mannosyltransferases of different microorganisms (data not shown), enzymes that act by transferring mannose for the synthesis of polysaccharide chains.

bme7 extends between nucleotides 5428 (end) and 6663 (start) and would encode a protein that is probably located in the cytoplasm (Fig. 2; Table 1). The C-terminal domain of Bme7 showed homology (Table 2) to the C-terminal end of several proteins that are supposed to be glycosyltransferases (e.g., E. coli YefL and Bordetella pertusis BplH, which is thought to be a glycosyltransferase required for LPS biosynthesis [2]).

bme8 would encode a protein of 416 amino acids located in the bacterial cytoplasm. The ORF covers the pNV3103 insert region contained between nucleotides 6684 (end) and 7934 (start) (Fig. 2; Table 1). A search for homologies revealed that Bme8 had significant homology (Table 2) to a wide number of proteins, from several microorganisms, that are supposed to function as glycosyltransferases (e.g., Cap5L and Cap8L from Staphylococcus aureus, involved in the synthesis of the type 5 and 8 capsular polysaccharides, respectively [69]; YefD from E. coli; and WbpJ from Pseudomonas aeruginosa, involved in the synthesis of LPS [16]).

As described for several glycosyltransferases (5, 70), Bme6, Bme7, and Bme8 exhibit the characteristic motif EX₇E toward the C-terminal end of the protein, which accounts for the probable function of these proteins as glycosyltransferases.

(iv) bme9 and bme10.

No regions of dyad symmetry have been found between bme9 and bme10, suggesting that they might be transcribed from the same promoter as a single unit.

Bme9 has a high level of homology to GDP-mannose-4,6-dehydratases, which function as sugar oxidoreductases. The highest homology levels were found to NoeL from Rhizobium spp. (68.1% identity in a 339-amino-acid overlap) (31) and RfbD from Vibrio cholerae (60.9% identity in a 345-amino-acid overlap) (Table 2). According to the nucleotide sequence of pNV3103, the first ATG codon of bme9 is located at position 8549 and the end of the ORF is at nucleotide 9268. However, no characteristic sequence for an RBS was found upstream of the ATG codon, and homology to GDP-mannose dehydratases occurred up to the TTG codon at position 8198. A sequence (GGAGG) that might function as an RBS was found 9 bp upstream of this TTG codon. We propose that bme9 starts at this TTG codon, a codon that has been proposed as the start of translation in other genes (2, 60). If this is the case, the deduced protein would be composed of 356 amino acids and would probably be located in the cytoplasm (Table 1). The region of Bme9 extending from amino acids 66 to 118 showed 71.1% identity to a short identified fragment (53 amino acids) of an RfbD-homologous protein from B. melitensis (accession no. AF043475) and B. abortus (accession no. AF021920), involved in the synthesis of the LPS O chain (3) (Fig. 4).

FIG. 4.

Multiple alignment of amino acids 66 to 118 of Bme9 and the 53 amino acids corresponding to the published partial coding sequence of one GDP–d-mannose dehydratase from B. melitensis (accession no. AF043475) and B. abortus (accession no. AF021920 [3]). Alignment was performed with CLUSTAL W 1.74 (23a). Identical amino acids are shown by asterisks. Conserved and semiconserved substitutions are shown by colons and periods, respectively.

The bme10 start codon is located at position 9252, slightly overlapping the end of bme9, and the end of the ORF is located at nucleotide 9992 (Fig. 2; Table 1). The putative protein would probably have a cytoplasmic location (Table 1). Homology (58.1% identity) was found between Bme10 and NolK from Rhizobium spp. (Table 2), which is proposed to be a NAD-dependent nucleotide sugar epimerase-dehydrogenase required for the biosynthesis of lipochito-oligosaccharidic Nod factors, involved in the nodulation process of the Rhizobium-legume symbiosis (31). Bme10 also showed 53.4 and 51.2% identity to WbcJ from Yersinia enterocolitica serotype O:8 and WcaG from E. coli, respectively (Table 2). Both proteins have been described as sugar epimerase-reductases involved in the conversion of GDP-4-keto-6-deoxy-mannose to GDP-fucose, one of the sugars of the Y. enterocolitica O:8 LPS O antigen (85) and the E. coli EPS colanic acid (73). Although the bme10 ORF ended at nucleotide 9992, homology of Bme10 was found to similar proteins after the stop codon with a change in the reading frame. Close to the stop codon of this other reading frame, located at position 10231, was found a region of dyad symmetry that might act as the transcription terminator of the probably cotranscribed bme9 and bme10. It is possible that the ORF originally finished at this position but that a single nucleotide has been removed during evolution, thereby changing the length of the gene. Additional studies are necessary to elucidate whether this probable reduction of the gene size has altered the function of the encoded protein.

(v) bme11, bme12, and bme13.

The three genes bme11 to bme13 would be transcribed from the reverse strand of the pNV3103 insert, and they might be transcribed as a single unit, since no regions of dyad symmetry have been found in the intergenic spaces. The dyad-symmetry domain found downstream of bme10 might also function as transcription terminator of this three-gene operon transcribed from the reverse strand.

Bme11 shows 25% identity to ExoF from Rhizobium meliloti (Table 2), which is involved in the biosynthesis of the EPS succinoglycan and is reported to be located in the periplasm but probably anchored to the cytoplasmic membrane through its amino terminus (63). ExoF was initially thought to be required for the addition of the first sugar to the lipid carrier of succinoglycan (63). However, although no homology to other proteins in the database was reported for ExoF when the sequence was published (56), later comparisons revealed homology of ExoF to GumB from Xanthomonas campestris, postulated to be involved in the polymerization of the repeating unit of xanthan gum, and KpsD from E. coli, needed for translocation of the capsular polysaccharide (35). The possible role of ExoF in polymerization or translocation of succinoglycan in R. meliloti was then taken into account (35). Local homology was also detected between Bme11 and GumB from X. campestris and between Bme11 and PrsE from R. leguminosarum and R. meliloti, a protein that has been implicated in the secretion of proteins required for the production of low-molecular-weight succinoglycan (83). The first ATG codon of bme11 was located at position 11238 of the pNV3103 insert, 355 bp downstream of the stop codon of bme12. No region of dyad symmetry has been found between the two genes, and homology to exoF extended up to close to a GTG codon located at nucleotide 11583, 10 bp downstream of the bme12 stop codon. A putative RBS (GGA) was detected 5 bp upstream of this GTG. We propose that this GTG codon is the initiation codon, although the real start codon is difficult to establish, since other GTG codons are present between the first GTG codon and the first ATG of the ORF. Accordingly, the ORF would extend from nucleotides 10336 (end) to 11583 (start) and would encode a protein of 415 amino acids with a putative signal peptidase cleavage site at amino acid 26 and the most probable localization in the periplasm (Fig. 2; Table 1).

The first ATG codon of bme12 was located at position 13701, but not clear region for an RBS was found near this ATG codon. However, an RBS sequence (AAGGAG) was detected 8 bp upstream of a GTG codon located at nucleotide 13782, which was considered to be the start codon of bme12. Accordingly, bme12 would extend from nucleotides 11593 (end) to 13782 (start) and would code for a protein of 729 amino acids that is probably located in the cytoplasmic membrane (Fig. 2; Table 1). A search for homology to other proteins in the database revealed that Bme12 is 25.2% identical (449 amino acid overlap) to GumC from X. campestris, a protein of 449 amino acids required for the production of the EPS xanthan gum. A 23.7% identity, covering the entire protein, was detected to ExoP from R. meliloti, reported to be implicated in succinoglycan chain length determination and export (8). AceD from Acetobacter xylinum, involved in EPS export, also showed homology to Bme12 (Table 2).

bme13 extended from nucleotides 13819 (end) to 14586 (start) and would encode a protein with the most probable location in the cytoplasmic membrane (Fig. 2; Table 1). A search for protein motifs revealed a 28-amino-acid region, located toward the C terminus of Bme13 (amino acids 201 to 228), characteristic for bacterial regulator proteins of the LuxR family (43, 74) (data not shown). Bme13 showed homology to proteins which are members of the environmentally responsive two-component regulatory systems (Table 2). Interestingly, two of the homologous proteins, GlpR from Pseudomonas aeruginosa and EpsR from Pseudomonas solanacearum, have been defined as regulator proteins involved in the biosynthesis of EPSs (18, 71).

(vi) bme14.

bme14 would be transcribed from the direct strand of pNV3103 and extended from nucleotides 14949 to 15707, encoding a putative protein probably located in the cytoplasmic membrane (Fig. 2; Table 1). Several regions of dyad symmetry have been found between bme14 and bme15, the gene located downstream of bme14 that would be transcribed from the reverse strand, and that might function as transcription terminators of both genes. Bme14 showed homology to several proteins included in a family of transcriptional regulators responding under low-oxygen concentrations (e.g., FnrL from Rhodobacter sphaeroides [84] and FixK from Bradyrhizobium japonicum [6]) (Table 2).

(vii) bme15.

bme15, which would be transcribed from the reverse strand of pNV3103, seems not to be entirely cloned in pNV3103, since no RBS sequence was found near a start codon and homology was observed only to the C-terminal end of other proteins in the database (Table 2). The highest level of homology was found to two proteins of unknown function (HI0148 from H. influenzae and YjhT from E. coli) (Table 2). The partial ORF extended from nucleotide 16084 (end) to the end of the pNV3103 insert (Fig. 2; Table 1).

Identification of the B. abortus deletion endpoints in the B. melitensis insert of pNV3103.

According to previous results (80), it was thought that the DNA fragment deleted in B. abortus strains would extend from a point close to the left end of the pNV3103 B. melitensis insert to a point located between the fifth and the sixth HindIII sites of the insert (Fig. 1). Therefore, in an attempt to identify the endpoints of the B. abortus deletion, PCR amplification of B. abortus DNA was performed with a primer covering a region near the left end of the pNV3103 insert (primer 31R or N3-PCR) and another one covering a region of the HindIII fragment mentioned above (N95-P, N49-PCR, or N89-P) (Fig. 1). Successful amplification was obtained only with primers N3-PCR and N89-P, which amplified a B. abortus DNA fragment of 1,854 bp that was cloned in plasmid pGEM-T to give plasmid pNV3138 (Fig. 1). Surprisingly, analysis of the nucleotide sequence of this fragment revealed that the last 799 bp is also present in the pNV3103 insert, corresponding to the 799 nucleotides located upstream of primer N89-P, while the first 1,055 bp was not cloned in pNV3103 (Fig. 5). Therefore, PCR amplification occurred through specific hybridization of primer N89-P and nonspecific hybridization of primer N3-PCR. According to previous work, it was thought that the entire DNA fragment missing in B. abortus was cloned in pNV3103 (80). However, sequencing of the pNV3138 insert made this hypothesis unlikely. Two hypotheses were possible, at this point, to explain this unexpected result. The first is that the B. melitensis DNA fragment cloned in pNV3103 does not contain the left side of the DNA deleted in B. abortus, and the second is that the DNA missing in B. abortus has been replaced by another DNA fragment absent in B. melitensis. To clarify this point, B. melitensis DNA was PCR amplified by using primers N119-P1 and N120-P2 (Fig. 1), which were located in the region of the B. abortus pNV3138 insert that was not cloned in the B. melitensis pNV3103 insert. A fragment of 438 bp was obtained, cloned into pGEM-T to give pNV3140 (Fig. 1), and sequenced, revealing that this fragment corresponds to the fragment located between both primers in B. abortus (Fig. 5). Therefore, DNA bordering the left side of the fragment missing in B. abortus was also present in B. melitensis, in agreement with the hypothesis of a large deletion in B. abortus that was not entirely cloned in the B. melitensis DNA insert of pNV3103. In an attempt to identify the DNA fragment absent in B. abortus that was not cloned in pNV3103, PCR amplification of B. melitensis DNA was performed with primers N120-P1 and 31R-P (Fig. 1), with unsuccessful results. In our PCR reaction conditions, fragments of about 2 kb have been successfully amplified but amplification of fragments of 3 kb with specific primers was not obtained (data not shown). Therefore, it seems likely that the DNA deleted in B. abortus strains and not cloned in pNV3103 is large size and would be greater than 2 kb, since no amplification of B. melitensis DNA was obtained with primers N120-P1 and 31R-P. Although the left side of the B. abortus deletion has not been identified, the limit of its right side has been determined to be located between nucleotides 9948 and 9949, near the end of bme10, of the pNV3103 B. melitensis insert (Fig. 2 and 5). Therefore, the DNA deleted in B. abortus accounts for more than 9,948 nucleotides, and bme1 to bme9 and most of bme10 are missing in this species.

FIG. 5.

Nucleotide sequence of the B. abortus 544 DNA fragment amplified with primers N3-PCR and N89-P and cloned in pNV3138. The primers are underlined. The primer N3-PCR sequence is shown in small capitals because it is likely that it hybridized nonspecifically, and it is not included in the sequence submitted to the database. The position delimiting both sides of the B. abortus deletion is shown by an open circle. The B. melitensis DNA fragment amplified with primers N119-P1 and N120-P2 is shown in bold type. Potential coding regions are marked and translated into amino acids, although bme10 and bme17 are probably not translated in B. abortus since both genes seem to be interrupted by the deletion. bme10 and bme11 correspond to partial sequences of the same genes found in the B. melitensis insert of pNV3103. Regions of dyad symmetry that might function as rho-independent transcription terminators are shown by inverted arrows.

In addition to the ends of bme10 and bme11, two other potential coding regions, designated bme16 and bme17, were found in the B. abortus insert of pNV3138 (Fig. 5). The hypothetical bme16 would be transcribed from the direct strand and would extend from nucleotide 19 (the region corresponding to primer N3-PCR was not taken into account since this primer probably hybridized nonspecifically with the B. abortus DNA) to nucleotide 267 (Fig. 5). The deduced amino acid sequence showed homology (the highest homology was 52.7% identity in 74 amino acid overlap) to several uracil permeases (e.g., accession no. AE000202, X76083, U32802, and M59757) (data not shown), enzymes intervening in the de novo pyrimidine nucleotide biosynthesis (34, 61). Since no RBS was found near an ATG codon and homology covered only the C-terminal end of the uracil permeases, it is likely that bme16 has not been entirely cloned in pNV3138. Two regions of dyad symmetry that might act as rho-independent transcription terminators were found downstream of bme16 (Fig. 5). In addition to bme10, the deletion in B. abortus strains appears to interrupt another hypothetical gene, bme17, whose putative product showed homology to an E. coli protein of unknown function (accession no. U73857) just covering the C terminus (25% identity in a 160-amino-acid overlap) (Fig. 5). Deletion in B. abortus has probably led to the loss of expression of both bme10 and bme17. Sequencing of the B. melitensis DNA corresponding to the left side of the B. abortus deletion will be necessary to determine the entire sequence of the hypothetical bme17.

DISCUSSION

In a previous study, a large DNA fragment, containing the gene encoding the Brucella Omp31 and DNA bordering both sides of this gene, was found to be missing in B. abortus strains but not in the other Brucella species (80). According to Southern blot hybridization and PCR amplification results, the size of this fragment was suggested to be about 10 kb, and it was thought that the entire DNA sequence absent in B. abortus was contained in the B. melitensis insert of plasmid pNV3103 (80). In the present work, we analyzed the nucleotide sequence of the pNV3103 insert, searching for the limit of the DNA missing in B. abortus and the genes involved in this deletion.

A search for the ORFs of the B. melitensis pNV3103 DNA insert and homology of the deduced proteins to other proteins in the database has allowed us to identify 14 putative new genes, in addition to omp31 (Fig. 2; Tables 1 and 2). Most of the proteins encoded by these genes exhibited a significant degree of homology to proteins involved in the biosynthesis process of several polysaccharides from different bacteria (Table 2), which point to the possibility of the implication of this group of genes in the synthesis of a Brucella polysaccharide.

Four proteins (Bme1, Bme6, Bme7, and Bme8) showed a significant degree of homology to several sugar transferases (Table 2), enzymes that act by transferring the sugar from one nucleotide-sugar complex to the growing chain of a polysaccharide. In the synthesis of a polysaccharide, a sugar transferase is needed for each different linkage between sugars (62), which would mean that the hypothetical Brucella polysaccharide would be composed of at least four sugar units. Bme4 shows homology to acetyltransferases and might act by acetylating one or more sugars of the polysaccharide. Bme9 and Bme10 are homologous to GDP–d-mannose dehydratases and epimerase-reductases, respectively, which might be involved in the modification of some of the sugars before their incorporation into the growing polysaccharide chain. A partial amino acid sequence for B. melitensis and B. abortus proteins homologous to GDP–d-mannose dehydratases has been found in the database (accession no. AF043475 and AF021920, respectively). The two partial sequences were identical, but homology to the equivalent fragment of Bme9 reached only 71.1% identity (Fig. 4). The GDP–d-mannose dehydratase from B. abortus is involved in the synthesis of the smooth-LPS (S-LPS) O chain (3). Bme9 might be involved in the synthesis of a different polysaccharide, since genes for the synthesis of each individual polysaccharide are usually arranged in independent clusters (62).

Biosynthesis of polysaccharides occurs through two different mechanisms. The Wzy-dependent pathway involves Wzx, which would transfer the repeat polysaccharide units across the cytoplasmic membrane to the periplasmic face, a polymerase (Wzy) of the polysaccharide units, and Wzz, which would regulate the polysaccharide length (81). The second pathway requires the ATP binding and transmembrane components of an ATP binding cassette transporter, the polymerase is not needed, and regulation of the polysaccharide length is independent of Wzz (81). Homologous proteins with bacterial ATP binding cassette transporters have not been found among the hypothetical proteins supposed to be encoded by the B. melitensis DNA fragment cloned in pNV3103. Although no homologous proteins with Wzy or Wzx have been identified among the pNV3103 hypothetical proteins, it must be noted that Wzy and Wzx of different polysaccharide biosynthesis clusters have little similarity at the amino acid sequence level (49, 73). Wzx, involved in the transfer of polysaccharides across the cytoplasmic membrane, has been described as a multiple membrane-spanning protein (73). Bme3 is predicted to be a highly hydrophobic protein located in the cytoplasmic membrane, with multiple membrane-spanning segments, that showed a hydrophobicity profile very similar to other Wzx-like proteins (Fig. 3), and it might have a Wzx function, allowing the translocation of the polysaccharide to the periplasm. Moreover, the N-terminal end of Bme3 showed homology to SmpX from Synechococcus spp. (accession no. D43774), which has been defined as a pore-forming channel protein belonging to a family of homologous intrinsic membrane proteins involved in a variety of transport processes (46). Polymerization of the polysaccharide subunits might be mediated by Bme11 and/or Bme12, which might also participate in the transport of the polysaccharide. Both proteins showed homology to other proteins that have been implicated in the polymerization or export of several EPSs (see Results), which points to the possibility that the genes identified in the B. melitensis insert of pNV3103 are part of a pathway for the synthesis of an EPS. This possibility is reinforced by the presence of an outer membrane protein, Bme5, among the proteins found to be encoded in this DNA fragment. Bme5 corresponds to the previously identified Omp31 (79), a protein that is located in the outer membrane of Brucella spp. and for which a possible role as porin has been suggested (79). Bme5 might be involved in the secretion of a hypothetical EPS, as has been suggested for other outer membrane proteins (32, 73). Bme2, probably located in the periplasm, shows only local homology to proteins involved in the secretion of the lantibiotic epidermin (60) and might also contribute to the secretion of an EPS.

Two genes, bme13 and bme14, encoding proteins homologous to regulatory proteins have been identified in the pNV3103 insert. It is difficult to assign them a regulatory function in the synthesis of the hypothetical polysaccharide, but several clusters for the synthesis of other bacterial polysaccharides are controlled by the action of regulatory proteins (18, 38, 74). Although bme15 was not entirely cloned in pNV3103, Bme15 exhibited a significant level of homology to two proteins of unknown function (Table 2), and it is difficult to determine whether it might be involved in the synthesis of a polysaccharide or have an unrelated function.

The number of genes in polysaccharide clusters is normally between 6 and 19, depending on the complexity of the polysaccharide (62). Most of the 15 hypothetical genes found in the pNV3103 B. melitensis DNA insert are supposed to be involved in the synthesis of a polysaccharide, but sequencing of adjacent DNA to both sides of the pNV3103 insert will be necessary to determine whether there are more genes that might be implicated in the synthesis of the hypothetical polysaccharide.

Although it was thought that the pNV3103 insert of B. melitensis contained the entire fragment missing in B. abortus strains (80), results presented in this paper demonstrate that the pNV3103 insert does not contain this entire DNA fragment. In spite of this, the exact point delimiting one side of the DNA absent in B. abortus has been determined to be located between nucleotides 9948 and 9949 of the pNV3103 insert, close to the end of bme10. Therefore, the fragment missing in B. abortus is at least 9,948 bp and covers 10 of the 15 genes identified in the pNV3103 insert (from bme1 to bme10). According to the PCR results presented in this paper, it seems likely that more than 2 kb of the B. abortus deletion remains unknown, and further studies are necessary to completely identify the DNA missing in B. abortus and to clarify the mechanism responsible for this deletion.

If genes contained in the pNV3103 insert are members of a cluster that efficiently synthesize a polysaccharide, it seems likely that the polysaccharide would not be present in B. abortus strains, since 10 of the 15 genes of the pNV3103 insert are deleted in B. abortus. Several polysaccharides have been identified in the genus Brucella. S-LPS has been widely studied since it is an important diagnostic and protective antigen and it is considered as a virulence factor in infections caused by the brucellae. Two S-LPS types, A and M, have been found to exist in smooth Brucella spp., with differences residing in the O chain of the S-LPS (13, 52, 82). The O chain of type A S-LPS is composed of a homopolymer of α-1,2-linked residues of 4,6-dideoxy-4-formamido-d-mannopyranose, and the O chain of type M S-LPS is composed of pentasaccharide repeats of the same sugar but with four α-1,2-linked residues and one α-1,3-linked residue (13, 14, 17). The brucellae have been classified into A>M, M>A, or A⁺M⁺ according to the distribution of both S-LPS types, and strains A>M and M>A have been shown to exist in all the species which can exist in smooth phase. Therefore, it seems unlikely that the genes contained in the pNV3103 insert are involved in the synthesis of the Brucella O-polysaccharide chain, since no differences have been found between the O chains of the different species. Moreover, the DNA fragment cloned in pNV3103 also seems to exist in rough Brucella strains, since omp31 (bme5) has been amplified from both smooth and rough strains (80). It is also unlikely that these genes are involved in the synthesis of the native hapten, a polysaccharide with the same structure of the S-LPS O chain and that also exists in B. abortus (86). Another polysaccharide, polysaccharide B, a polymer of cyclic β-1,2-glucan, has been found in the genus Brucella, but it has been shown to exist in both B. melitensis and B. abortus (15).

Therefore, if the group of genes identified in the B. melitensis insert of pNV3103 are part of the gene machinery leading to the synthesis of a polysaccharide, it is probable that this polysaccharide has not yet been described in the brucellae. The Brucella S-LPS has been described as an important virulence factor (3, 64, 65), and capsular polysaccharides and EPS have also been described as virulence factors in other microorganisms. Identification of a novel Brucella polysaccharide might be of great interest for a better understanding of the differences in pathogenicity and host preference observed between the brucellae. In an attempt to identify a polysaccharide present in B. melitensis but absent in B. abortus, SDS-polyacrylamide gel electrophoresis (PAGE) followed by periodate-silver staining for polysaccharides has been performed with B. melitensis 16M and B. abortus 544 cells grown in solid medium and with the supernatant recovered after centrifugation of the resuspended cells. However, these studies proved unsuccessful, since no differences in the polysaccharide profile were found between these organisms (data not shown). LPS and capsular polysaccharides, including LPS of Brucella spp. (26, 33), are currently detected by this technique (5, 9, 85). However, while periodate-silver staining after SDS-PAGE is a good method for detection of LPSs, other polysaccharides not attached to a lipid moiety are unable to migrate in SDS-PAGE, as has been shown for the Brucella native hapten (7). The possibility cannot be excluded that B. melitensis produces a polysaccharide not detectable by the technique we used or that the polysaccharide is synthesized only when contact with the host is established and is not detected under the conditions usually employed in the laboratory. Another possibility that should be taken into account is that the genes found in B. melitensis no longer lead to the synthesis of a polysaccharide. Brucellae are members of the alpha-2 subdivision of the class Proteobacteria and are genetically close to bacteria of the same group, which are associated with eucaryotic cells (25, 55). Bacteria from the genus Rhizobium and Agrobacterium, living in association with plant cells, produce EPS which is important for the interaction with the host eucaryotic cell (48, 75). The common ancestor of the alpha-2 subdivision of the class Proteobacteria has been suggested to have EPS, and the genes discovered in B. melitensis might be a remnant from this common ancestor. Modifications, during the course of evolution, in one or more of the genes involved in the synthesis of this polysaccharide might have led to the loss of this polysaccharide, which would not be necessary for survival or the establishment of the infection process. The alpha-2 subdivision of the class Proteobacteria contains several motile members, and recently, although the brucellae are defined as nonmotile, genes encoding homologs of several flagellum-related proteins have been found in B. abortus (41). One of the hypothesis made by Halling to explain this fact was that the brucellae may have lost motility during evolution because it is not essential and may even be detrimental for their cycle as animal pathogens (41).

Although small deletions in several genes have been identified in some Brucella species (19, 22, 23, 29, 67) and unidentified DNA deletions of several sizes exist in some brucellae (54, 80), this is the first report describing the characterization of a large DNA fragment deleted in one of the Brucella species. In spite of the high level of DNA homology reported for this genus (77), important divergences between the brucellae should exist at the DNA level that are able to explain their differences in pathogenicity and host preference. Here we have found several genes that seem to be implicated in the synthesis of a polysaccharide not yet described in the genus Brucella. Although this polysaccharide has not been detected, the possibility cannot be excluded that it was not detected under the conditions we used or that it is synthesized only in contact with the host. Discovery of a novel Brucella polysaccharide, probably not synthesized in B. abortus, would be of great interest since it might play an important role in virulence and might explain some of the differences in pathogenicity and host preference among the brucellae.