Brucella abortus Ornithine Lipids Are Dispensable Outer Membrane Components Devoid of a Marked Pathogen-Associated Molecular Pattern

The brucellae are α-Proteobacteria facultative intracellular parasites that cause an important zoonosis. These bacteria escape early detection by innate immunity, an ability associated to the absence of marked pathogen-associated molecular patterns in the cell envelope lipopolysaccharide, lipoproteins and flagellin. We show here that, in contrast to the outer membrane ornithine lipids (OL) of other Gram negative bacteria, Brucella abortus OL lack a marked pathogen-associated molecular pattern activity. We identified two OL genes (olsB and olsA) and by generating the corresponding mutants found that olsB deficient B. abortus did not synthesize OL or their lyso-OL precursors. Liposomes constructed with B. abortus OL did not trigger IL-6 or TNF-α release by macrophages whereas those constructed with Bordetella pertussis OL and the olsB mutant lipids as carriers were highly active. The OL deficiency in the olsB mutant did not promote proinflammatory responses or generated attenuation in mice. In addition, OL deficiency did not increase sensitivity to polymyxins, normal serum or complement consumption, or alter the permeability to antibiotics and dyes. Taken together, these observations indicate that OL have become dispensable in the extant brucellae and are consistent within the trend observed in α-Proteobacteria animal pathogens to reduce and eventually eliminate the envelope components susceptible of recognition by innate immunity.


Introduction
The members of the genus Brucella are a-2 Proteobacteria that cause brucellosis, an important disease affecting livestock and wild life as well as human beings. These bacteria trigger only low proinflammatory responses in the initial stages of infection and, accordingly, they follow a stealthy behavior that allows them to reach sheltered intracellular niches before effective immunity activation. The outer membranes (OM) of brucellae are of critical importance in this strategy. Whereas most gram-negative have OM molecules bearing the pathogen-associated molecular patterns (PAMP) recognized by innate immunity, at least the Brucella OM lipopolysaccharide (LPS), lipoproteins and flagellin display a reduced PAMP [1,2]. Moreover, smooth (S) brucellae such as B. abortus and B. melitensis have OMs that are unusually resistant to the disrupting action of bactericidal peptides and complement. Thus, periplasmic and internal PAMP-bearing molecules like peptidoglycan or DNA are not readily accessible to pathogen recognition receptors [1,[3][4][5][6][7][8]. The Brucella LPS is clearly implicated in these properties and there is evidence that other lipid molecules also contribute. Brucella OMs contain large amounts of phosphatidylcholine (PC) and blockage of the synthesis of PC with the subsequent replacement by phosphatidylethanolamine (PE) generates attenuation [9,10]. Ornithine lipids (OLs) are present in relatively large amounts in Brucella [11] and, although they have interesting properties in other bacteria, have not been investigated.
It has been reported that Pseudomonas fluorescens grown under conditions that increase OL content becomes more resistant to the polycationic lipopeptide polymyxin B indicating a connection between these amino lipids and the resistance to bactericidal peptides [12,13]. Moreover, OLs of Bordetella pertussis, Flavobacterium meningosepticum and Achromobacter xylosoxidans display antagonistic effects on LPS endotoxicity as well as proinflammatory and inflammatory activity [14][15][16][17][18]. Such an activity in Brucella OL would be in apparent contradiction with the furtive behavior of these bacteria with respect to innate immunity. Therefore, it was of interest to know whether Brucella OLs play a role in the OM stability and resistance to polycations and whether they display a biological activity different from that of other OLs, including the evasion of pathogen recognition receptors.

OLs are OM components of B. abortus
To determine the cellular localization of OLs, we first examined the free-lipids in virulent S B. abortus 2308 Nal R grown in tryptic soy broth to the stationary phase, in the OM fragments released spontaneously during growth [19] and in non-delipidated B. abortus LPS [20]. Thin-layer chromatography of the corresponding chloroform:methanol:water extracts [21] confirmed the presence of OLs in B. abortus and showed their enrichment in the OM fragments (Figure 1 A), thus demonstrating that they are OM components. Although in amounts lower than PE, OLs were also detected in non-delipidated B. abortus LPS suggesting an association in the OM (Figure 1 A). The levels of OLs did not change when the bacteria were cultured in tryptic soy broth or in Brucella Gerhardt's minimal medium (lactate-glutamate-glycerol, mineral salts, vitamins) [22] (Figure 1 A).

B. abortus OLs are synthesized through a two step pathway
We searched the B. abortus 2308 genome for orthologs of the genes involved in OL synthesis in other a-2 Proteobacteria. In Sinorhizobium meliloti, OlsB acylates the ornithine a amino group with C18:0(3-OH) and OlsA generates the acyloxyacyl group by esterification with C18:0 (Figure 1 B) [23,24]. In Rhizobium tropici, an additional gene (olsC) codes for an oxygenase that generates a 2hydroxy substitution on the ester-linked acyl group [25] (Figure 1 B). We found that ORF BAB1_0147 (annotated as encoding a hypothetical protein) codes for a protein with 55% identity and 66% similarity with OlsB, and that the product of ORF BAB1_2153 (annotated as a phospholipid-glycerol acyltransferase) has 45% identity and 61% similarity with OlsA. Moreover, we located similar ORFs in the genomes of B. melitensis 16M and B. suis 1330 (http://www.ncbi.nlm.nih.gov/genomes/lproks.cgi). ORF BAB1_0147 (henceforth BABolsB) maps as an isolated gene. BAB1_2153 (henceforth BABolsA) is in a possible operon formed by at least BAB1_2151 (putative glycoprotease), BAB1_2152 (putative acetyltransferase) and probably by BAB1_2154 (hypothetical protein; the stop codon of BAB1_2152 and the start codon of BAB1_2153 overlap). Both BABOlsA and BABOlsB contain a sequence [H73 (X)4 D78 and H76 (X)4 D81, respectively] that could correspond to the consensus motif [H(X)4D] of glycerolipid acyltransferases [26]. Moreover, BABOlsA contains two regions, NHTS (amino acids [72][73][74][75] and AEGTT (amino acids 143-147), closely resembling the NHQS and PEGTR motifs highly conserved in lysophosphatidic acid acyltransferases [27]. We also found a DNA sequence with homology to olsC. However, the possible BABolsC carried a frame shift in the same position in all accessible B. abortus, B. melitensis and B. suis genomes so that it corresponds to two ORF (annotated as BMEI0464 and BMEI0465 in B. melitensis 16M). Amino acids 1 to 155 of BMEI0464 have a 87% identity (92% homology) with amino acids 31 to 185 of S. meliloti OlsC, and amino acids 1 to 94 of BMEI0465 are 87% identical (91% homology) to the 187 to 280 stretch of OlsC [28]. Most of the consensus of the OslC-LpxO family of proteins is in BMEI0464 but at the very end of the protein and truncated in the last four amino acids, including the last isoleucine conserved in all OlsC homologues [29]. All these characteristics strongly suggest that the structure of Brucella OLs is similar to those described previously for Rhodospseudomonas sphaeroides ( Figure 1C).
Using the above-described evidence, we constructed internal inframe deletion mutants of the virulent B. abortus 2308 Nal R strain (henceforth BAB-parental) devoid of the consensus motifs following a PCR overlap strategy [30] (Table S1). For BABolsB, we removed amino acids 40 to 229, which resulted in a truncated protein of 107 amino acids (mutant BABDolsB). The deletion in BABolsA (mutant BABDolsA) eliminated amino acids 48 to 245 and resulted in a truncated protein of 69 amino acids. Figure 1 A shows that BABDolsA lacked OLs but synthesized a new ninhydrin-positive component corresponding to the lyso-ornithine lipid (lyso-OL) precursor [31]. In contrast, the only ninhydrin-positive lipid generated by mutant BABDolsB was PE. When we complemented mutant BABDolsB with plasmid pLPI-6 (carrying BABolsB; Table S1), OL synthesis was restored (Figure 1 A). Similarly, BABDolsA complemented with plasmid pYLI-1 (carrying BABolsA) was able to produce OLs (not shown). These results are consistent with a two step pathway in which BABOlsB and BABOlsA act consecutively and where deletion of the former abrogates OL and lyso-OL synthesis (Figure 1 B).
Characterization of mutants BABDolsA and BABDolsB showed no change in colonial morphology, or in catalase, oxidase, and urease activities. They were S according to lysis by B. abortus Sspecific phages, agglutination with anti-A and anti-M monospecific sera, crystal violet exclusion and acriflavine agglutination test.
OLs are not required for Brucella OM resistance to bactericidal peptides and complement Due to the OL abundance in Brucella and their zwitterionic nature, it has been proposed that they play a relevant role in the stabilization of negative charges of LPS and, therefore, in the stability of the OM [32]. To test this, we first examined the sensitivity to bactericidal peptides. We found no significant differences in the minimal inhibitory concentrations of polymyxin B and colistin on BAB-parental, BABDolsA and BABDolsB strains. Since bactericidal peptides are also OM permeabilizing agents, we probed the mutants with polymyxin B plus lysozyme under hypotonic conditions in comparison with Escherichia coli. This treatment was effective in killing E. coli but had no action on mutants BABDolsB or BABDolsA or on BAB-parental ( Figure 2). Moreover, EDTA alone or combined with polymyxin B did not promote lysozyme uptake, proving that divalent cations were not taking over the hypothetical role of OLs in OM stabilization ( Figure 2). Finally, sensitivity to a set of antibiotics (penicillin, doxycycline, clarithromycin, erythromycin and rifampicin) that penetrate the OM by hydrophilic or hydrophobic pathways, or of dyes like thionine blue, fuchsine and safranin remained unchanged (not shown). All these results indicate that OLs are neither necessary to stabilize the OM of B. abortus against bactericidal peptides nor influence its permeability.
B. abortus mutants altered in PC synthesis, with a truncated LPS or an upset OM protein profile are sensitive to killing by nonimmune serum [33][34][35]. However, OL deficiency did not have a similar effect because we observed only a small and not significant (p.0.05) increase in serum sensitivity in mutant BABDolsB (Figure 3 A). Although small differences in antibody-independent complement activation were suggested when large amounts of bacteria were used (Figure 3 B), the differences were not statistically significant (p.0.05). Finally, we also tested the surface hydrophobicity of the mutants and the exposure of major OM proteins. The partition coefficient of the wild type BAB-parental  The absence of OLs does not influence proinflammatory responses to B. abortus or the ability to multiply intracellularly and generate chronic infections Since B. abortus induces a negligible proinflammatory response in mice at early times of infection [1], we examined whether the lack of OLs could alter this property by examining several markers. First, we tested the generation of fibrin D-dimer that would result from the activation of the blood clothing cascade caused by endotoxic microorganisms. For this, we infected Swiss Webster mice 10 6 CFU/mouse of BAB-parental or mutant BABDolsB (controls received 0.1 ml of 100 mM phosphate buffered saline (pH 7.2) [PBS]) and examined the serum 24 h later. For BAB-parental, the results confirmed previous reports [1] on the absence of any increase in fibrin D-dimers above the positive threshold (0.5 mg/ml). Similarly, infection with the OLdeficient mutant did not have a significant effect on fibrin D-dimer generation. Then, we determined the number of leukocytes in the peritoneal fluid and blood of mice injected intraperitoneally with 10 6 CFU of BAB-parental or BABDolsB, 10 5 CFU of S. typhimurium or 0.1 ml of PBS. Whereas the latter induced a leukocyte peritoneal recruitment and a progressive reduction in blood leukocyte numbers at later times, these linked effects were not observed in BAB-parental or BABDolsB mutant infected mice ( Figure 4). Similarly, lymphocyte, neutrophil and monocyte numbers in blood and peritoneal fluid did not reveal significant differences in Brucella infected mice ( Figure 5). We also measured TNF-a, IL-6, IL-10 and IL-12p40 in the blood of the same animals, and found that the normalized levels of these cytokines were similar in BAB-parental and BABDolsB infections ( Figure 6). These levels were similar to and much lower than those reported for B. abortus 2308 and S. typhimurium, respectively [1].
To complement the above-described studies, we tested the ability of the BABDolsB OL-deficient mutant to multiply in cells and to generate chronic infections. First, we infected bone marrow derived macrophages, RAW 264.7 macrophages and HeLa cells and monitored the intracellular survival of bacteria [36]. Figure 7 shows that the BABDolsB mutant retained the ability to multiply intracellularly in all these cells. Then, we inoculated BALB/c mice intraperitoneally with 5610 4 CFU/mouse of BABDolsB or of BABparental. Two, 6, and 12 weeks later, the bacteria in the spleens were counted, the identity of the isolates confirmed by PCR, and the spleen weights recorded. The results showed that the parental were practically identical throughout the experiment (average log 10 CFU 6 SD values for BABDolsB and BAB-parental, respectively, were: week 2, 6.0960. 24  Consistent with the experiments in vitro, these observations suggest that the absence of OLs does not affect the OM stability in vivo and, accordingly, that OLs seem not to hamper the release of PAMP-bearing molecules and the establishment of chronic infections. Moreover, since BABDolsB did not promote a lower or higher proinflammatory response than BAB-parental, the results suggest that these lipids are neither detected by the pathogen recognition receptors nor antagonists in the recognition of other OM molecules during brucellosis.

B. abortus OL do not stimulate cytokine secretion in murine macrophages
To test whether Brucella OLs carry a PAMP, we used the BABparental and BABDolsB lipids, since the presence of PC in Brucella allows obtaining stable liposomes. As a positive control, we extracted the lipids of B. pertussis (containing OL but not PC) and generated B. pertussis OL-liposomes using the OL-free lipids of BABDolsB as carriers. After verification of the liposome composition (Figure 8 A), we stimulated RAW 264.7 macrophages. While the B. pertussis-BABDolsB liposomes notably stimulated TNF-a and IL-6 secretion, the BAB-parental or BABDolsB liposomes were inactive (Figure 8 B). These results clearly show that Brucella OL, in contrast to those of B. pertussis, do not bear a marked PAMP. In addition, they demonstrate that Brucella phospholipids do not inhibit PAMP recognition.

Discussion
The OM of most gram negative bacteria hinders the penetration of harmful molecules, and the LPS plays a key role in this important property. The LPS inner sections (core oligosaccharide and lipid A) are rich in acidic sugars and orthophosphate, and these negatively charged groups bind divalent cations and polyamines that bridge the LPS molecules and hamper the partition of hydrophobic permeants into the OM [37,38]. However, this supramolecular arrangement makes OM sensitive to divalent cation chelators like EDTA and to bactericidal peptides. Moreover, this set of properties is connected to the PAMP of a variety of OM molecules, primarily the LPS. Interestingly, Brucella OM is comparatively permeable to hydrophobic compounds and resistant to those agents. At least in part, these properties are accounted for by a low number of negatively  charged groups which are limited to two 2-keto-3-doxyoctulosonic acid residues and the lipid A phosphates [39][40][41][42][43]. Moreover, it was proposed that the Brucella OLs could shield those negatively charged groups by virtue of their positively charged amino groups, as postulated for P. fluorescens [44]. Indeed, the results of this and previous works in other bacteria [45][46][47] support that Brucella OLs are in fact N-(acyloxyacyl)-ornithine OM structural elements with a free amino group that should be positively charged at neutral and acidic pH. Such an OL role would represent an advantage for a pathogen because the hydrophobic anchorage should make OLs more resistant than divalent cations to displacement by bactericidal peptides and proteins. However, since OL deficiency did not increase the sensitivity of Brucella cells to polymyxins or the permeability to lysozyme (a cationic peptide), this hypothesis was disproved. Furthermore, any possible defect not detected by the in vitro methods seems not to be relevant in vivo, at least in cells and  mice. One possibility is that OLs are in the inner leaflet of the OM and, therefore, not in contact with the polar moiety of the LPS. In fact, our results suggest that PE (also with a free amino group) is the more important LPS associated-lipid in B. abortus.
The interactions of OLs with innate immunity have been analyzed in some c and b Proteobacteria. The OLs from B. pertussis and A. xylosoxidans stimulate IL-1, TNF-a and prostaglandin E2 synthesis in macrophages [14,[48][49][50]. Moreover, F. meningosepticum OLs are mitogenic for B-lymphocytes and exhibit adjuvant activity in C3H/HeJ mice, suggesting that receptors other than TLR4 are involved in OL recognition [14,51,52]. Consistent with these observations, OLs of B. pertussis presented in BABDolsB liposomes triggered cytokine release. Therefore, the fact that this effect was significantly lower when similar liposomes carried B. abortus OLs demonstrates that innate immunity fails to efficiently recognize these Brucella amino lipids. This is a well-known property of Brucella LPS and other Brucella putative PAMP bearing molecules such as flagella and lipoproteins and, therefore, our results extend this ability to another OM element [1,53]. It is known that the acyl groups in LPS, other glycolipids and synthetic aminolipids or lipopeptides modulate the inflammatory activity [54][55][56] and, indeed, the acyl chains of Brucella OL differ from those of other bacteria in length and, in some cases, in the presence of a hydroxyl group (Figure 1 C) [57][58][59][60][61]. In some bacteria, OL have esterlinked fatty acyl groups with a hydroxyl group at the 2-position and this hydroxyl group may affect the biological activity [62]. However, this hydroxylation is a post-synthesis modification that requires OlsC, which is inactive in Brucella according to our genomic analysis. All these data strongly suggest that the reduction of Brucella OL PAMP is connected to the acyl chains, longer in a-Proteobacteria than in other phylogenetic groups (Figure 1 C). Indeed, these results add more weight to the hypothesis that the hydrophobic moieties of Brucella OM elements are critical in avoiding recognition of this pathogen by innate immunity [1,8,63].
Here, we also presented evidence that Brucella OLs are dispensable elements. Taking into account that Brucella OLs are quantitatively important lipids, it is surprising that they are dispensable. However, the a-Proteobacteria show a marked tendency to live in tight interactions with eukaryotic cells [64] and there is evidence that this ability is associated to a reduction in envelope molecule PAMPs. Bartonella possess a low endotoxic LPS, a reduced number of OM lipoproteins and flagellins that are not recognized by TLR5 [65][66][67]. Similarly, Rickettsia carry a non canonical LPS and a reduced number of lipoproteins [68], and Wolbachia do not posses genes to synthesize LPS, flagella or fimbriae, have a low number of lipoproteins and an unusual peptidoglycan [69]. Finally, the genomes of Ehrlichia and Anaplasma contain a low number of lipoprotein genes and do not have the genetic machinery to synthesize LPS, peptidoglycan or flagellin [70,71]. Accordingly, we hypothesize that free-living Brucella ancestors carried OLs that were useful OM structural elements in their environment but lost their importance once the major OM target of immunity (i.e. the LPS) became adapted to the host. This adaptation was marked by the modification in LPS PAMP connected features (i.e., acyl chains and charged groups) and had as a result that LPS no longer needed stabilizing agents in the OM. Within this perspective, the degeneracy of the olsC homologue is in keeping with the hypothesis that B. abortus OL represent dispensable ancestral structures that could be eventually eliminated during the evolutionary process.

Ethics Statement
All animals were handled and sacrificed according to the approval and guidelines established by the ''Comité Institucional para el Cuido y Uso de los Animales'' of the Universidad de Costa Rica, and in agreement with the corresponding law ''Ley de Bienestar de los Animales No 7451'' of Costa Rica (http://www. micit.go.cr/index.php/docman/doc_details/101-ley-no-7451-leyde-bienestar-de-los-animales.html).
BALB/c mice (Charles River, Elbeuf, France) were accommodated in the animal building of the CITA of Aragón (ID registration number ES-502970012005) in cages with water and food ad libitum and under biosafety containment conditions, for 2 weeks before the start and all along the experiment. The animal handling and procedures were in accordance with the current European legislation (directive 86/609/EEC) supervised by the Animal Welfare Committee of the institution (protocol number R102/2007).

Bacterial strain and growth conditions
Bacteria were grown in tryptic soy broth or agar either plain or supplemented with kanamycin (Km) at 50 mg/ml, or/and nalidixic (Nal) at 25 mg/ml, or/and gentamicin (Gm) at 20 mg/ml, or/and chloramphenicol at 20 mg/ml (all from Sigma), or/and 5% sucrose. Where indicated, the defined medium of Gerhardt [72] was used. All strains were stored in skim milk at 280uC. The origin of the B. abortus and S. typhimurium strains is described in previous works [1,73,74]. B. pertussis is a clinical isolate kindly provided by G. Martínez de Tejada.

Bacteriological procedures, antibiotic sensitivity and cell surface characterization
The mutants were characterized according to standard Brucella typing procedures [75]: colonial morphology after 3 days of incubation at 37uC, crystal violet exclusion, catalase, oxidase, urease, acriflavine agglutination, sensitivity to Tb, Wb, Iz and R/ C phages, agglutination with anti-A and anti-M monospecific sera, CO 2 and serum dependence, and susceptibility to thionine blue, fuchsine, and safranin. Moreover, the minimal inhibitory concentrations of polymyxin B, colistin, penicillin, doxycycline, clarithromycin, erythromycin and rifampicin were determined in Müller-Hinton medium by standard procedures.
Surface hydrophobicity was analyzed as described by Kupfer and Zusman [76]. Bacteria were grown in tryptic soy broth until stationary phase, incubated in 0.5% sodium azide at 37uC overnight, collected by centrifugation (70006g, 10 min, 4uC), washed twice with a solution of K 2 HPO 4 .3H 2 0 97mM, KH 2 PO 4 53mM, urea 21mM, MgSO 4 .7H 2 O 0.8mM and resuspended to an OD 470 = 1.0. A volume of 1.5 ml of this bacterial suspension was mixed with 0.5ml of n-hexadecane and incubated during 10 min at 37uC. After shaking for 40 seconds, the tubes were settled to allow phases separation to occur. The partition was calculated as (12OD 470 of water phase)/OD 470 of the water phase. Western-blot analysis with monoclonal antibodies to the major Omps was carried out as described before [77].

DNA manipulations, construction of mutants and complementation
Plasmid and chromosomal DNA were extracted with Qiaprep spin Miniprep (Qiagen GmbH, Hilden, Germany), and Ultraclean Microbial DNA Isolation kit (Mo Bio Laboratories) respectively. When needed, DNA was purified from agarose gels using Qiack Gel extraction kit (Qiagen). DNA sequencing analysis was performed by the Servicio de Secuenciación de DNA del Centro de Investigación Médica Aplicada (Navarra, Spain). Primers were synthesized by Sigma-Genosys Ltd.. Searches for DNA and protein homologies were carried out using the NCBI (National Center for Biotechnology Information (http://www.ncbi.nlm.nih. gov) and the EMBL-European Bioinformatics Institute server (http://www.ebi.ac.UK/ebi_home.html). In addition, sequence data were obtained from The Institute for Genomic Research at http://www.tigr.org. Genomic sequences of B. melitensis 16M, B. abortus and B. suis were analyzed using the database of the L'Unité de Recherche en Biologie Moléculaire (URBM, Namur, Belgium) (http://www.serine.urbm.fundp.ac.be/,seqbruce/GENOMES/ Brucella_melitensis).
For constructing the BABDolsB mutant, oligonucleotides olsB-F1 (59 -CTTCTGTCATCGTCGCGTAG-39) and olsB-R2 (59-GATGCGTCCCAGAATGATG-39) were used to amplify a 270-bp fragment including codons 1 to 39 of the olsB ORF, as well as 153 bp upstream of the olsB first putative start codon, and oligonucleotides olsB-F3 (59-CATCATTCTGGGACGCATCC-CAAAGGAAGCGATCAACAA-39) and olsB-R4 (59-TTAAAA-CCGGAACCGCTCTA-39) were used to amplify a 294-bp fragment including codons 230 to 297 of the olsB ORF and 87-bp downstream of the olsB stop codon. Both fragments were ligated by overlapping PCR using oligonucleotides olsB-F1 and olsB-R4 for amplification, and the complementary regions between olsB-R2 and olsB-F3 for overlapping. The resulting fragment, containing the olsB deletion allele, was cloned into pCR2.1 (Invitrogen), to generate plasmid pIRI-2, sequenced to ensure the maintenance of the reading frame, and subcloned into the BamHI and the XhoI sites of the suicide plasmid pJQ200KS. The mutator plasmid (pLRI-7) was introduced in BAB-parental by conjugation [78]. Integration of the suicide vector was selected by Nal and Gm resistance, and the excision (generating the mutant strain by allelic exchange) was selected by Nal and sucrose resistance and Gm sensitivity. The resulting colonies were screened by PCR with primers olsB-F1 and olsB-R4 which amplify a fragment of 564 bp in the mutant and a fragment of 1134 bp in the parental strain. The mutation resulted in the loss of the consensus amino acids responsible for the enzymatic activity.
The BABDolsA mutant was constructed using primers olsA-F1 (59-GATTGCGCAGGATACCATCT -39) and olsA-R2 (59-AACAATGCGGTGGAAGAAAT-39) to amplify a 498-bp fragment including codons 1 to 47 of the olsA ORF, as well as 357-bp upstream of the olsA start codon and primers olsA-F3 (59ATTTCTTCCACCGCATTGTTACGATGGAAAATCGG-GTGAG-39) and olsA-R4 (59-CAGCGCGGAATAGAGTTTTT-39) to amplify a 372-bp including codons 246 to 268 of the olsA ORF and 303 bp downstream of the olsA stop codon. Both fragments were ligated by overlapping PCR using primers olsA-F1 and olsA-R4 and the fragment obtained, containing the deletion allele, was cloned into pCR2.1 to generate pYLI-2, sequenced to confirm that the reading frame had been maintained, and subcloned in pJQ200KS to produce the mutator plasmid pYLI-3. This plasmid was introduced in BAB-parental and the deletion mutant generated by allelic exchange was selected by Nal and sucrose resistance and Gm sensitivity and by PCR using oligonucleotides olsA-F1 and olsA-R4 which amplify a fragment of 870 bp in the deletion strain and a fragment of 1464 bp in the parental strain. The mutation generated results in the loss of 73.8% of the olsA ORF.
For complementation, plasmids carrying olsB and olsA were constructed using the Gateway cloning Technology (Invitrogen). Gene olsB was amplified from BAB-parental using primers olsB-F13 (59 GGGGACAAGTTTGTACAAAAAAGCAGGCTT-CATGACAGCACTGCTTGGAATGG 39) and gene olsB-R14 (59 GGGGACCACTTTGTACAAGAAAGCTGGGTC CTAG-ACAAAGCGGTTTGCTTC 39), that contain the attB sequences (underlined), and cloned into vector pDONR221 (Invitrogen) to generate pLPI-5. The ORF was subsequently cloned in pRH001 [79], able to multiply in Brucella, to produce the complementation plasmid pLPI-6. Since the sequence of olsA from B. abortus and B. melitensis is 99% identical, the clone carrying olsA was extracted from the B. melitensis ORFEOMA [80] and the ORF subcloned into plasmid pRH001 [81] to produce plasmid pYLI-1. To complement the olsB mutation, plasmid pLPI-6 was introduced into the BABDolsB mutant by mating with E. coli S17 lpir and the conjugants harboring this plasmid (designated as BABDolsB pLPI-6) were selected by plating the mating mixture onto tryptic soy agar-Nal-Km plates. The olsA mutation was complemented following the same protocol by introducing plasmid pYLI-1 into the BABDolsA mutant.

Lipid analysis
Total lipids were extracted as described by Bligh and Dyer [82], and analyzed on silica gel 60 high-performance thin layer chromatography plates (Merck Chemicals), the plates were prewashed by solvent migration with chloroform-methanol-water (140:60:10, vol/vol), dried thoroughly. Then samples were applied and chromatography performed in the same mixture of solvents [83]. Plates were developed with 0.2% ninhydrin in acetone and heating at 120uC for 5 min or by charring with 15% sulfuric acid in ethanol at 180uC. L-b,c-dipalmitoyl-a cephalin (Sigma-Aldrich) was used as a standard.

Sensitivity to non immune serum and complement consumption
Exponentially growing bacteria were adjusted to 10 4 CFU/ml and mixed with fresh sheep normal serum (45 ml of cells plus 90 ml of serum per well) in microtiter type plates in duplicate. After incubation for 90 min at 37uC with gentle stirring, brain heart infusion broth (200 ml/well) was added, mixed and 100 ml aliquots plated out by triplicate. The results were expressed as the % survival with respect to the CFU in the inocula. Complement consumption was estimated as the reduction of the hemolytic activity of rabbit serum complement incubated with live bacteria, and S. typhimurium SL1344 was used as a positive control [1].

Infections and bacterial counts in cells
Bone marrow cells were isolated from femurs of 6-10-week-old C57Bl/6 female mice and differentiated into macrophages [84]. Infections were performed at a multiplicity of infection of 50:1 by centrifuging bacteria onto macrophages at 400 g for 10 min at 4uC, followed by a15 min incubation at 37uC under a 5% CO 2 atmosphere. Macrophages were extensively washed with Dulbeccos's modified Eagle's medium to remove extracellular bacteria and incubated for an additional 90 min in medium supplemented with 50 mg/ml gentamicin to kill extracellular bacteria. Thereafter, the antibiotic concentration was decreased to 10 mg/ml. At each time point, samples were washed three times with 100 mM PBS before processing. Alternatively, murine RAW 264.7 macrophages (Raw 264.7; American Type Culture Collection No. TIB-71) or Hela cells (American Type Culture Collection No. CCL-2) were used in a similar protocol. To monitor Brucella intracellular survival, infected cells were lysed with 0.1% (vol/vol) Triton X-100 in H 2 O after PBS washing and serial dilutions of lysates were rapidly plated onto TSB agar plates to enumerate CFU. The attenuated B. abortus virB mutant (Table S1) was used as a control.

OL stimulation of macrophages
For macrophage stimulation, OL were included in liposomes. Total free-lipid extracts of BAB-parental, BABDolsB or a 1:1 mixture of B. pertussis and BABDolsB free-lipids were evaporated under a nitrogen stream, resuspended thoroughly in 250 ml of 10mM HEPES to a final concentration of 5 mg/ml, and the composition verified by thin-layer chromatography. Phase contrast microscopy and fluorescein entrapment controls demonstrated that Brucella total lipids formed liposomes that were stable for at least 24 h without addition of exogenous lipids. The murine RAW 264.7 macrophages used in these experiments were grown at 37uC under 5% CO 2 in Dulbecco's medium supplemented with 10% fetal bovine serum, 2.5% sodium bicarbonate, 1% glutamine (all from Gibco), penicillin (100 units/ml) and streptomycin (100 mg/ ml). Macrophages (5610 5 cells) were treated with 20 mg/mL or 100 mg/mL of liposomes of BAB-parental, BABDolsB or B. pertussis-BABDolsB in 10% fetal calf serum-Dulbecco's Modified Eagle's Medium. After 90 min of incubation, fresh 10% fetal calf serum-Dulbeccos's modified Eagle's medium was added to obtain a final concentration of 10 mg/ml and 50 mg/ml of liposomes, and the amount of TNF-a and IL-6 in the supernatants was assessed by ELISA (eBioscience) after 6 and 24 h.

Proinflammatory responses in mice
Swiss CD1 mice of 18 to 20 g were used. Fibrin D-dimers were determined from the plasma of mice 24 h after intraperitoneal infection with 0.1 ml of bacteria (1610 6 UFC/mouse) in pyrogenfree PBS (pH 7.2). Fibrin D-dimers were assessed by the semiquantitative D-Di testH latex agglutination assay (Diagnostica Stago). The levels of IL-6, IL-10, IL-12p40 and TNF-a were estimated by ELISA (eBioscience) in the sera of mice (n = 5) infected intraperitoneally with 10 6 CFU of BABDolsB or BABparental. For leukocyte counts, mice were intraperitoneally infected with 1610 6 CFU of BABDolsB or BAB-parental or 10 5 CFU of S. typhimurium SL1344 in pyrogen-free sterile PBS (pH 7.2), or with pyrogen-free sterile PBS as a control. Blood was collected from the retrorbital plexus at different time points in tubes with 2mg/ml of EDTA. Then, 5 ml of ice cold PBS were injected in the peritoneal cavity, and 3.0 to 4.5 ml of fluids were collected with a syringe. After centrifugation, the peritoneal cells were resuspended in 0.2 ml of PBS, and total leukocytes, neutrophils, monocytes and lymphocytes were counted. The number of cells recruited in the peritoneum was corrected according to the volume of fluid collected from each animal [1].

Virulence and splenomegaly in mice
Groups of 30 mice (seven-week-old female BALB/c mice [Charles River, Elbeuf, France]) were inoculated intraperitoneally with 5610 4 CFU/mouse of BABDolsB or BAB-parental in 0.1 mL of PBS, and the spleen weights and number of viable bacteria in spleens were determined in five mice at 2, 6, and 12 weeks post-inoculation. Experimental procedures (i.e. preparation and administration of inocula, retrospective assessment of the exact inoculating doses, and determination of the number of CFU/ spleen) were performed as described previously [85]. The identity of the spleen isolates was confirmed by PCR amplification at each selected post-inoculation point time. Spleen weights were expressed as the mean and SD (n = 5) of grams/spleen and infections as mean 6 SD (n = 5) of log 10 CFU/spleen at each selected postinoculation point-time, previous logarithmic normalization of individual data. Statistical comparisons between means were performed by ANOVA and the Fisher's Protected Least Significant Differences test [86].