Brucella ovis PA and B. melitensis Rev1 were obtained from BCCN (Brucella Culture Collection Nouzilly, maintained at the Institut National de la Recherche Agronomique, Nouzilly, France), and the other B. ovis strains are listed in Table 1. Recombinant plasmids used for mutagenesis of cgs (BOV_RS00 535), bacA (BOV_RS01960), omp10 (BOV_RS10700), omp19 (BOV_RS09115), bepC (BOV_RS04655), and ugpB (BOV_RS13470), and the corresponding single mutants derived from parental B. ovis PA have previously been described (Martín-Martín et al., 2012; Sidhu-Muñoz et al., 2016). B. ovis Δomp31 (BOV_RS12205) and Δomp25 (BOV_RS3460) single nonpolar mutants (Table 1) were obtained from parental B. ovis PA as described below. Multiple mutants (Table 1) were constructed from initial single mutants where one or two of the mentioned genes were additionally deleted. The Δomp31-k and Δomp25c-k mutant strains (omp31 or omp25c replaced by a kanamycin resistance cassette) and the same mutants complemented with the corresponding omp31 or omp25c (BOV_RS00575) wild type genes (Δomp31-k com and Δomp25c-k com strains) were previously obtained (Caro-Hernández et al., 2007).

Brucella ovis strains and the B. melitensis Rev1 attenuated vaccine were cultured in tryptic soy agar or tryptic soy broth (Pronadisa-Laboratorios Conda, Torrejón de Ardoz, Spain) both supplemented with 0.3% yeast extract (Pronadisa-Laboratorios Conda, Torrejón de Ardoz, Spain) and 5% horse serum (Gibco-Life Technologies, Grand Island, NY, United States) (TSA-YE-HS and TSB-YE-HS, respectively). Incubations were performed at 37°C in a 5% CO₂ atmosphere and, in the case of TSB-YE-HS liquid medium, under agitation at 120 rpm. When required, streptomycin (Strep; 50 μg/ml) (Sigma-Aldrich, St. Louis, MO, United States) was added to the culture medium of B. melitensis Rev1 (Strep-resistant strain). Similarly, when necessary for the selection of the recombinant B. ovis strains, kanamycin (50 μg/ml) or 5% sucrose (Sigma-Aldrich, St. Louis, MO, United States) was added. Assays with the B. ovis strains, including Δomp31-k, Δomp25c-k, Δomp31-k com, and Δomp25c-k com strains, were performed in the absence of antibiotics.

Plasmid pGEM-T Easy (Promega, Madison, WI, United States) was used to clone PCR-amplified fragments and pCVD-KanD (Martín-Martín et al., 2012) was the suicide plasmid employed to insert the mutant genes into parental B. ovis PA. They were propagated in Escherichia coli JM109 or CC118 (λpir), respectively, that were incubated at 37°C in Luria Bertani medium supplemented, when required, with ampicillin or kanamycin (50 μg/ml). Their derived recombinant plasmids constructed during this work are mentioned below.

DNA primers (IDT, Leuven, Belgium) used for the construction and verification of the B. ovis Δomp31, Δomp25, and Δomp25c single and multiple nonpolar mutants are listed in Table 2. The additional primers used to check the multiple mutants were previously described (Martín-Martín et al., 2012; Sidhu-Muñoz et al., 2016).

For the construction of the recombinant plasmids used in the mutagenesis process, inactivation of omp31, omp25, and omp25c was performed by in-frame deletion with overlapping PCR (Martín-Martín et al., 2012). Briefly, the 5′- and 3′-ends of each target gene, together with about 300–700 pb upstream or downstream, respectively, were separately amplified by PCR with two pairs of primers (31MUT-F + 31OVL-R and 31OVL-F + 31MUT R for omp31, 25MUTZ-F + 25OVL-R and 25OVL-F + 25MUTZ-R for omp25, 25cdMUT-F + 25cOVL-R and 25cOVL-F + 25cMUT-R) (Table 2). The two amplified fragments were fused, through the overlapping sequences located in the internal primers (primers OVL-F and OVL-R) (Table 2), in a PCR reaction with the two external primers of each fragment (31MUT-F + 31MUT-R for omp31, 25MUTZ-F + 25MUTZ-R for omp25, and 25cdMUT-F + 25cMUT-R for omp25c). The resulting mutant genes were cloned in pGEM-T Easy, verified by DNA sequencing, and subsequently cloned in pCVD-KanD to give pPS31OVL02, pNV25OVL02, and pNV25cOVL02, respectively. Plasmids were introduced in B. ovis PA by electroporation and the selection of bacteria with the corresponding plasmid integrated in the chromosome (intermediate strains), through a single homologous recombination event, was performed on TSA-YE-HS plates with kanamycin. The selected intermediate strains were plated onto TSA-YE-HS supplemented with 5% sucrose to give either the desired mutant strain (wild-type gene replaced by the inactivated gene) or a strain reverting to the parental genotype (Rv). They were differentiated by PCR amplification with the two primers external to each side of the deleted gene (amplified fragment with higher size in Rv strains than in mutant strains) and a second PCR with an external primer and a primer annealing inside the deleted fragment (primers 31-MAT, 25-MAT, and 25c-MAT) (Table 2). The latter PCR reaction produces no amplification in mutant strains.

For multiple gene mutagenesis, single mutants listed in Table 1 were subjected to a second mutagenesis round with a recombinant plasmid containing another inactivated gene. A third round of mutagenesis was conducted on some selected double mutants to inactivate a third gene. The selection of mutant strains was performed with specific PCRs targeting each inactivated locus.

Growth characteristics of the mutant strains in TSA-YE-HS plates and TSB-YE-HS liquid medium were compared to those of parental B. ovis PA. Numbers of CFU/ml corresponding to bacterial suspensions in PBS with optical density scores at 600 nm (OD₆₀₀) of 0.2 were determined for each mutant by triplicate plating on TSA-YE-HS of the properly diluted suspensions. The initial suspensions were prepared from bacteria cultured in TSA-YE-HS plates for 44 h. Colony size was photographed 5 days after plating and colonies enumerated 8 days after plating. Growth curves were established for triplicate bacterial suspensions in TSB-YE-HS medium (30 ml) with initial OD₆₀₀ readings of 0.05 that were incubated at 37°C under agitation (120 rpm) and a 5% CO₂ atmosphere. OD₆₀₀ scores were measured through a 120-h period, and CFU/ml numbers were evaluated at the beginning of the experiment (t0), and after 24, 52, and 77 h of incubation (t24, t52, and t77, respectively), by plating the properly diluted cultures on TSA-YE-HS.

The autoagglutination assay was performed as described previously (Caro-Hernández et al., 2007; Martín-Martín et al., 2012) by measuring the evolution, over 48 h of static incubation, of the OD₆₀₀ values of bacterial suspensions with initial readings of 0.8 (100% OD₆₀₀) in TSB-YE-HS. Susceptibility to 1 mg/ml of polymyxin B (Sigma-Aldrich, St. Louis, MO, United States) and 0.1 mg/ml of sodium deoxycholate (Sigma-Aldrich, St. Louis, MO, United States) in PBS was expressed as the percentage of survival after 80 min of exposure. It was determined by comparison of the numbers of CFU in untreated (incubation in PBS, 100% survival) and treated bacterial suspensions (Caro-Hernández et al., 2007; Martín-Martín et al., 2011). The results were expressed as means ± standard deviation (SD) of three assays.

Reactivity of B. ovis mutants with MAbs specific for cell-envelope antigens (Table 3) was measured by indirect enzyme-linked immunosorbent assay (iELISA) as previously described (Soler-Lloréns et al., 2014). MAbs specific for Brucella peptidoglycan (PG), R-LPS, S-LPS, Omp2b, Omp10, Omp16, Omp19, Omp25, Omp31, or periplasmic BP26 were used (Cloeckaert et al., 1990, 1991, 1996a,b; Vizcaíno et al., 2001b; Seco-Mediavilla et al., 2003). Briefly, 96-well plates were coated overnight at room temperature with sonicated (mild homogenization for 10 s at 40% intensity in a Sonic Dismembrator model 120, Thermo Fisher Scientific, Waltham, Ma, United States) bacterial suspensions in PBS (OD₆₀₀ = 1), which were prepared from cultures in TSA-YE-HS plates. MAbs (hybridoma supernatant) diluted 1/2 and a goat anti-mouse IgG-horseradish peroxidase conjugate (Bio-Rad, Hercules, CA, United States) diluted 1:9000 were used as primary and secondary antibodies, respectively. Antigen–antibody binding was revealed by incubation for 20 min with TMB as substrate for peroxidase and subsequent addition of a 1 M HCl-based stop solution (Interchim, Montluçon, France). The results were expressed as means ± SD of the values recorded at 450 nm (OD₄₅₀) in a Labsystems Multiskan Ascent microplate reader (Thermo Fisher Scientific) for three repeats by MAb and strain.

Additionally, sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS–PAGE) and immunoblot were also performed and carried out as previously described (Vizcaíno et al., 2001b; Martín-Martín et al., 2009). Briefly, bacterial suspensions concentrated to OD₆₀₀ values of 20 were prepared in H₂O with the different B. ovis strains. They were submitted to SDS–PAGE on a Protean II xi cell (Bio-Rad, Hercules, CA, United States) and either stained with Coomassie blue (Bio-Rad, Hercules, CA, United States) or transferred to a nitrocellulose membrane in a semidry electroblotter (Amersham, GE Healthcare, Little Chalfont, United Kingdom). Prestained protein marker VI (Applichem-Panreac, Barcelona, Spain) was used as protein standard. Nitrocellulose strips were saturated with skim milk and then incubated with sera obtained, as described before (Martín-Martín et al., 2009), by immunization of rabbits with Omp31b, Omp25c, Omp25d, and Omp22 purified recombinant proteins. Binding of the secondary antibody – a goat anti-rabbit IgG-peroxidase conjugate (Sigma-Aldrich, St. Louis, MO, United States) – was detected with a 4-chloro-1-naphthol substrate solution.

Infection assays of murine macrophage-like J774.A1 cells (DSMZ ACC170) and epithelial HeLa (ATCC CCL-2TM) cells were performed as described previously (Sidhu-Muñoz et al., 2016). Briefly, 2 × 10⁴ J774.A1 macrophages or 1.5 × 10⁴ HeLa cells/well were seeded on 96-well sterile microplates and incubated for 24 h at 37°C under a 5% CO₂ atmosphere. After incubation for 2 h with the B. ovis strains (4 × 10⁶ CFU/well for macrophages or 8 × 10⁶ CFU/well for HeLa cells) and killing of extracellular bacteria by incubation with gentamycin for 1 h, intracellular bacteria were enumerated in three wells per strain after lysis of the phagocytes with H₂O [t0 post-infection (p.i.)] (Sidhu-Muñoz et al., 2016). The remaining wells were maintained in the presence of gentamycin and intracellular bacteria were similarly determined at 20 (t20) and 44 h (t44) p.i. The results were expressed as means ± SD of the log CFU/well at each selected p.i. time point (t0, t20, and t44) and are representative of at least two experiments.

Female 6-week-old BALB/c mice (Charles River Laboratories, Chatillon-sur-Chalaronne, France), received 1 week previously, were used. They were randomly distributed into experimental groups and kept with water and food ad libitum in the animal experimentation facilities of the University of Salamanca (registration number PAE SA-001) or the Unidad de Producción y Sanidad Animal, Instituto Agroalimentario de Aragón-IA2 (CITA-Universidad de Zaragoza) (registration number ES502970012005).

Procedures with mice were designed according to Spanish and European legislation regarding the use of animals in research (RD 53/2013 and directive 2010/63/UE). Microbiological practices and animal experimentation were approved by the Biosecurity and Bioethics Committees of the University of Salamanca, and authorized by the competent authority of Junta de Castilla y León, Spain. The Animal Welfare Committee of the CITA (Spain) also reviewed and approved the protocols.

Preliminary evaluation of virulence in mice was performed by intraperitoneal inoculation of 10⁶ or 10⁸ CFU/mouse depending on the previous information about each single mutant (Martín-Martín et al., 2012; Sidhu-Muñoz et al., 2016). One mouse was used per strain, dose and time of analysis and splenic colonization was evaluated 3 or 7 weeks (W3 and W7) p.i. as described previously (Sancho et al., 2014). These time points in parental B. ovis PA correspond to the acute and chronic phase of infection, respectively (Caro-Hernández et al., 2007; Martín-Martín et al., 2012; Sidhu-Muñoz et al., 2016).

Spleen colonization of the selected mutants – according to the results of the preliminary analysis – was evaluated at W3 and W7 p.i. in mice inoculated intraperitoneally with 10⁶ CFU, and at W1, W3, W5, W7, and W11 p.i. in mice inoculated intraperitoneally with 10⁸ CFU. Five mice per strain and time point were used. The results were expressed as means ± SD (n = 5) of the log of CFU/spleen for each strain and time point. The identity of the recovered colonies was checked by PCR. Antibodies specific for B. ovis PA were determined by iELISA (Sancho et al., 2014) in sera obtained from submandibular blood from the same mice. Briefly, a suspension in PBS of heat-inactivated B. ovis PA whole cells (OD₆₀₀ = 1) was used as the coating antigen of 96-well plates and a goat antimouse IgG-peroxidase conjugate (Sigma–Aldrich, St. Louis, MO, United States) was used as the secondary antibody. OD₄₀₅ readings were recorded on a Multiskan Go Microplate Reader (Thermo Fisher Scientific) after 30 min incubation at room temperature with the substrate solution constituted by 1 mM 2,2′-azino-di-(3-3-ethylbenzothiazoline-sulfonic acid) (ABTS; Sigma-Aldrich, St. Louis, MO, United States) and 2 mM H₂O₂ (Sigma-Aldrich, St. Louis, MO, United States) in 0.1 M citrate, pH 4.2. Antibody titers in serum were defined as the inverse of the highest serum dilution scoring an OD₄₀₅ value twice as high as that obtained with the blank (mean OD₄₀₅ of six wells in which serum was replaced by dilution buffer). The results were represented as means ± SD of the log of the titers obtained with five mice analyzed individually.

BALB/c mice were inoculated intraperitoneally with PBS, 10⁵ CFU of the classical vaccine B. melitensis Rev1, or 10⁸ CFU of the attenuated mutant B. ovis Δomp10ΔugpBΔomp31. Seven weeks later, corresponding with the time point where the B. ovis vaccine was cleared from spleen, five mice per group were either challenged with 2 × 10⁵ CFU of B. ovis PA, or processed for the evaluation of the antibody and cellular immune responses specific for B. ovis PA. Determination of virulent B. ovis PA in spleen was evaluated 3 weeks after experimental challenge (Sancho et al., 2014). The CFU number of virulent B. ovis PA in mice vaccinated with B. melitensis Rev 1 was obtained by subtracting the values obtained in TSA-YE-HS-Strep medium from those obtained in the same medium without antibiotic. Results were expressed as means ± SD (n = 5) of the log CFU/spleen of B. ovis PA for each vaccination group.

IgG titers in serum were analyzed, as described above, in five mice per group 7 weeks after vaccination by using heat-inactivated B. ovis PA whole cells as the coating antigen in a iELISA test. Additionally, IgG isotypes were determined under the same conditions but using goat anti-mouse IgG₁-, IgG_2a-, or IgG_2b-peroxidase conjugates (Santa Cruz Biotechnology, Dallas, TX, United States). The same mice were processed as previously described (Sancho et al., 2014) to evaluate the cytokine response of splenocytes to a second stimulus with B. ovis PA. Briefly, spleen cells from immunized mice were cultured in 24-well sterile plates and stimulated by exposure to heat-inactivated (1 h at 65°C) B. ovis PA whole cells (10⁷ CFU/well), 10 μg/ml of the mitogen concanavalin A (Sigma-Aldrich, St. Louis, MO, United States) as a positive control of cell proliferation, or culture medium as a negative control. After 72 h of incubation, the culture supernatants were harvested to evaluate the levels of interferon-γ (IFN-γ), tumor necrosis factor-α (TNF-α), IL-10, and IL-12(p40). They were determined by sandwich ELISA with OptEIA^TM Mouse Sets specific to each cytokine, as instructed by the manufacturer (BD Biosciences, San Diego, CA, United States). Two wells were used for each experimental condition and mouse. The results for each vaccination group were expressed as means ± SD of the cytokine amount (ng/well) in the supernatants of splenocytes obtained from five individual mice. The results obtained with the positive and negative controls (concanavalin A and RPMI as stimulating agents, respectively) were as expected and are not shown.

Statistical comparisons between means were performed using analysis of variance. The significance of the differences (P < 0.05) between the experimental groups was determined with the post hoc Fisher’s protected least significant differences (PLSD) test. To simplify the figures and tables, no ranking of P-values has been established and all significant differences are marked as P < 0.05.

The B. ovis mutants (Table 1) were genotypically characterized as described in the section “Materials and Methods.” Two PCR reactions were settled for each inactivated gene, one reaction with external primers to each side of the deleted locus (amplified fragment of lower size in mutant strains than in parental strain), and a second reaction with an external primer to the deleted loci and a primer annealing inside the deleted region (no amplification in mutant strains). All single and multiple mutants gave the expected results for each inactivated gene (data not shown).

Although not all combinations of single genes shown in Table 1 were tempted in multiple mutants, two combinations were specially assayed without success: a double mutant in lipoproteins Omp10 and Omp19 and a double mutant in major Omp25 and Omp31 proteins. Despite multiple attempts, the intermediate strains for these double mutants always reverted to the single mutant initial genotype. Similarly, two independent omp25c intermediate strains always reverted to the parental genotype and, consequently, the non-polar Δomp25c single mutant was not obtained. However, the non-polar deletion of omp25c, with plasmid pNV25cOVL02, was successful in the Δomp25Δomp25c double mutant and in the Δomp10Δomp31Δomp25c and Δomp31Δomp10Δomp25c triple mutants. The Δomp25c-k mutant, where omp25c was replaced by a kanamycin resistance cassette (Caro-Hernández et al., 2007), was used in this work for comparisons with the non-polar Δomp25c double and triple mutants.

The numbers of CFU in bacterial suspensions with OD₆₀₀ values of 0.2 were determined for each mutant in TSA-YE-HS plates. Bacterial counts obtained for Δomp25, ΔbepC, ΔugpB, and ΔbacA single mutants and for ΔbepCΔugpB double mutant did not significantly differ from those observed with the parental strain. On the contrary, Δomp31, Δomp25c-k, Δomp10, Δomp19, Δcgs, and the remaining multiple mutants showed, in different degrees, lower CFU values than B. ovis PA (Table 1). The Δomp31Δcgs double mutant and its derived triple mutants, the Δomp25Δcgs mutant and the Δomp10Δomp31Δomp25c and Δomp31Δomp10Δomp25c triple mutants (bearing the same three genes inactivated but in a different order) showed the lowest CFU/ml values in TSA-YE-HS plates (Table 1). The numbers of CFU/ml obtained for each mutant were used for the calculation of the bacterial doses in further experiments.

In addition, the size of the colonies in TSA-YE-HS was monitored over time and recorded 5 days after inoculation (see some representative results in Figure 1A). Colony size of the Δomp25, ΔbacA, ΔbepC, ΔugpB, Δomp25c-k, ΔbepCΔugpB, and Δomp25Δomp25c mutants was undistinguishable from that of the parental strain (see B. ovis PA and the Δomp25Δomp25c mutant in Figure 1A). Growth of the Δomp10 and Δomp19 mutants was slightly delayed when compared to the parental strain but the initial differences in colony size were not apparent after 5 days of incubation (data not shown). Colonies of the Δomp31 and Δcgs single mutants appeared hardly visible 72 h after inoculation, 24 h later than those of the B. ovis PA parental strain, and were smaller than those of the parental strain after 5 days (see Δomp31 in Figure 1A). A similar or higher delay, depending on the strain, was detected with colonies of their multiple mutants. Despite the triple gene deletion, the Δomp10ΔugpBΔomp31 mutant showed a colony growth pattern undistinguishable from that of the single Δomp31 mutant (Figure 1A). The most important growth defects were observed with the Δomp31Δcgs mutant and its derived triple mutants (with the additional deletion of omp10 or omp19), whose colonies started to be detected 96 h after inoculation and showed the smallest size after 5 days of incubation (Figure 1A). Colonies of the Δomp10Δomp31Δomp25c and Δomp31Δomp10Δomp25c triple mutants were also detected after 96 h of incubation but, at day 5 post-inoculation, they were bigger (see colonies of the Δomp10Δomp31Δomp25c mutant that are representative for both mutants) than those of B. ovis Δomp31Δcgs (Figure 1A).

The Δomp31Δcgs, Δomp10Δomp31Δomp25c, and Δomp10ΔugpBΔomp31 mutants (selected according to the results obtained in the evaluation of virulence described below), the Δomp25Δomp25c mutant (selected because it lacks two major OMPs in B. ovis PA) and their corresponding single mutants were also analyzed regarding growth in TSB-YE-HS liquid medium (Figures 1B,C). Cultures were adjusted to OD₆₀₀ values of 0.05 and incubated at 37°C under agitation (120 rpm) and a 5% CO₂ atmosphere. As expected, according to the correlation OD₆₀₀-CFU/ml in solid medium previously determined, values of CFU/ml at t0 for mutants with inactivated omp31 or cgs were lower than those of B. ovis PA (P < 0.05), the Δomp31Δcgs and Δomp10Δomp31Δomp25c mutants presenting the lowest values (P < 0.05) (Figure 1C). However, the CFU/ml values of the Δomp31 and Δomp10ΔugpBΔomp31 mutants at the beginning of the stationary phase (approximately t52) were similar to those of B. ovis PA (P > 0.05), which correlates with the higher OD₆₀₀ scores observed with these mutants by this time (P < 0.05). Although the Δomp31Δcgs and Δomp10Δomp31Δomp25c mutants, those showing the most important growth defects in solid medium (Figure 1A), also had higher OD₆₀₀ values at this moment (P < 0.05), their CFU/ml counts never reached the maximum level of B. ovis PA (detected at t52) (Figures 1B,C). The Δcgs, Δomp25, Δomp25c-k, and Δomp25Δomp25c mutants had slightly higher OD₆₀₀ scores in the stationary phase, when compared to the parental strain, but not important differences in the CFU/ml values (Figures 1B,C and data not shown). The Δomp10 and ΔugpB mutants behaved similarly to the parental strain (data not shown).

For a preliminary assay of virulence, aiming to select the most relevant mutants, mice were inoculated with 10⁶ CFU of the parental strain or its derived mutants and bacteria were enumerated in spleen at W3 and W7 p.i. The multiple cgs mutants were excluded from this first analysis, since the single Δcgs mutant showed poor spleen colonization in a previous study (Martín-Martín et al., 2012). Only those mutants that, when compared to the parental strain, showed bacterial counts lower than 1.5 logarithmic units were considered as probably attenuated and they were later inoculated in a dose of 10⁸ CFU/mouse. Uniquely the Δcgs, Δomp10ΔugpBΔomp31, Δomp10Δomp31Δomp25c, and Δomp31Δomp10Δomp25c mutants met this requirement and, together with the Δcgs multiple mutants, were evaluated with the increased dose (except B. ovis Δomp31Δomp10Δomp25c that bears the same mutations as the Δomp10Δomp31Δomp25c but in a different order). All these mutants were present in spleen at W3 p.i. but the CFU were between 1 and 2.5 log units lower than those obtained with the parental strain (Table 1). All cgs mutants and B. ovis Δomp10Δomp31Δomp25c had similar splenic counts at W7 p.i., which were not drastically different from those observed at W3 p.i., while the Δomp10ΔugpBΔomp31 mutant was not detected in spleen at W7 p.i. (Table 1).

According to these results, the Δomp31Δcgs, Δomp10 ΔugpBΔomp31, and Δomp10Δomp31Δomp25c mutants were retained for a detailed evaluation of virulence and the selection of vaccine candidates. First, to statistically verify the previous results, the spleen colonization was evaluated in five mice per group at W3 and W7 after intraperitoneal infection with 10⁶ CFU (Figure 2A). The spleen weight (Figure 2B) and the levels in serum of IgG reacting against B. ovis PA whole cells (Figure 2C) were also determined. As expected, spleen infection of B. ovis PA was noteworthy at both sampling points – with mean values ranging between log 6 and 7 of CFU/spleen – while the three mutants presented a poor colonization under these conditions (Figure 2A). Spleen infection of the mutants was undetectable in several mice per group, although two out five mice inoculated with the Δomp10Δomp31Δomp25c mutant gave values of about log 5 CFU/spleen at W7 p.i. (Figure 2A). Spleen weight and IgG titers in serum were also significantly higher (P < 0.05) in mice inoculated with the parental strain (Figures 2B,C).

The same parameters described above were evaluated at several time points over a 11-week period in mice inoculated with 10⁸ CFU of each strain (Figures 2D–F), the dose usually employed for evaluation of attenuated strains as vaccines (Sancho et al., 2014; Soler-Lloréns et al., 2014; Silva et al., 2015b). The parental strain showed high levels of spleen infection through the entire experiment with mean log CFU/spleen values of 5.99 at W1 p.i. and a peak at W3 p.i. of 7.64. B. ovis PA counts decreased thereafter but were still high at the end of the experiment (mean log CFU/spleen values of 5.01 at W11) (Figure 2D). The Δomp31Δcgs and Δomp10Δomp31Δomp25c mutants displayed lower levels of splenic infection at W1, with mean CFU/spleen values 1–1.5 log lower than those of the parental strain (P < 0.05). However, although their maximum CFU/spleen scores (about log 6 at W3–W5 p.i.) did never reach those of B. ovis PA (log 7.64 at W3 p.i.), they persisted in spleen with similar values to those of the parental strain at W11 p.i. (P > 0.05) (Figure 2D). The spleen colonization of the Δomp10ΔugpBΔomp31 mutant did not show statistically significant differences with the parental strain at W1 p.i., but the bacterial counts did not increase at W3 p.i. and decreased thereafter until the detection limit of infection at W7 p.i. (P < 0.05) (Figure 2D).

At W1 p.i., the spleen weight in mice infected with 10⁸ CFU of the parental strain (0.102 ± 0.030 g) was analogous to that currently obtained with mice inoculated with PBS and that was also observed in the PBS group used to determine the production of cytokines by splenocytes in the experiment described below (0.096 ± 0.005 g) (P > 0.05). However, a prominent splenomegaly was detected at W3 p.i. (0.478 ± 0.020 g/spleen) which, although decreased after W3 p.i., persisted until the end of the experiment (0.205 ± 0.046 g/spleen) (P < 0.05) (Figure 2E). On the contrary, the B. ovis mutants did never reach mean values of 0.200 g/spleen (Figure 2E). At W1 p.i., all groups of mice had low levels of IgG reactive with whole cells of B. ovis PA. An important increase of the antibody response was detected at W3 p.i. in all groups (P < 0.05), except in mice vaccinated with the Δomp31Δcgs mutant (P > 0.05). At W5 p.i. and thereafter, mice inoculated with the parental strain or the Δomp31Δcgs and Δomp10Δomp31Δomp25c mutants displayed IgG titers of about log 4, while mice inoculated with B. ovis Δomp10ΔugpBΔomp31 gave titers in the order of 0.5 logarithmic units higher (P < 0.05) than those detected in the B. ovis PA group (Figure 2F).

In addition to the three selected mutants, other relevant mutants were included in the remaining studies by their interest to increase the knowledge about the B. ovis cell envelope. Several tests related to the OM properties of Brucella spp. (i.e., autoagglutination and resistance to polymyxin B and sodium deoxycholate) (Martín-Martín et al., 2011; Vizcaíno and Cloeckaert, 2012) were performed. In the autoagglutination assay, most of the single mutants behaved as the B. ovis PA parental strain (P > 0.05), remaining in suspension until the end of the experiment (Figure 3 and data not shown). The exceptions were B. ovis Δomp31 – with %OD₆₀₀ values of about 30 and 10% after 24 and 48 h of static incubation, respectively – and the Δcgs and Δomp25c-k mutants, whose %OD₆₀₀ values gradually decreased until 70% at the end of the experiment (P < 0.05 when compared to the parental strain) (Figure 3). The omp31 multiple mutants behaved similarly to the single mutant, although the Δomp10Δomp31Δomp25c and Δomp31Δomp10Δomp25c mutants agglutinated more quickly (they were almost completely settled after 10 h) (P < 0.05) (Figure 3). B. ovis Δomp25Δcgs behaved like the single Δcgs mutant while B. ovis Δomp25Δomp25c did not show differences with the parental strain (P > 0.05) (Figure 3).

Brucella ovis ΔbacA and the B. ovis ΔbacAΔomp31 double mutant were more resistant than B. ovis PA to polymyxin B exposure for 80 min (Table 4). On the contrary, B. ovis Δomp25Δomp25c, and the Δomp31 single and multiple mutants – except the previously mentioned ΔbacAΔomp31 mutant – showed higher susceptibility to polymyxin B than the parental strain (Table 4). The lowest survival percentages were obtained with the Δomp10Δomp31Δomp25c and Δomp31Δomp10Δomp25c mutants (less than 10% survival). Only the Δomp31Δcgs double mutant and its derived Δomp31ΔcgsΔomp10 and Δomp31ΔcgsΔomp19 triple mutants were more susceptible to sodium deoxycholate than the parental strain (about 30% survival versus 86% survival, respectively) (Table 4).

The reactivity of the B. ovis mutants with MAbs raised against Brucella S-LPS, R-LPS, PG, major and minor OMPs, and periplasmic BP26 was analyzed by iELISA (Figure 4 and data not shown). As expected, the MAb-specific against S-LPS did not react with any mutant or with the parental strain (data not shown) and omp31, omp25, omp10, or omp19 single and multiple mutants did not react with the respective MAbs (see Figure 4 for representative results). No relevant differences were observed between the parental strain and the mutants regarding reactivity with MAbs specific against Omp31 (except with mutants lacking omp31), Omp10 (except with mutants lacking omp10), Omp16, Omp2b, PG, and R-LPS. On the contrary, all omp31 mutants showed a stronger reaction with MAb C09 (specific against Omp25) and omp19 mutants showed a stronger reaction with MAb B02, specific against BP26 (Figure 4, upper panels and data not shown). Accordingly, several omp31 mutants were tested in iELISA with three MAbs specific against Omp25, while omp19 mutants were tested with four anti-BP26 MAbs. All the Δomp31 and Δomp19 mutants showed a stronger reactivity with the anti-Omp25 and anti-BP26 MAbs, respectively (Figures 4G,H), which confirmed the results obtained in the previous analysis.

The protein profile of the most relevant strains was evaluated by SDS–PAGE followed by Coomassie blue straining (Figure 5A). Proteins were also transferred to nitrocellulose to assess the reactivity with sera raised against the proteins of the Omp25/Omp31 family. A serum against Omp25c was used to detect both Omp25c and Omp25 since these proteins display cross-reacting epitopes (Martín-Martín et al., 2009), while the detection of Omp31 was performed with a serum raised against Omp31b (an OMP absent in B. ovis) that strongly cross-react with Omp31 (Martín-Martín et al., 2009). Omp25d and Omp22 were detected by reactivity with their respective anti-sera (Martín-Martín et al., 2009).

A multiple band pattern, dependent on the electrophoretic conditions, is frequently observed in SDS–PAGE for Omp25, Omp25c, and Omp31 (Cloeckaert et al., 1990; Martín-Martín et al., 2009). Two protein bands corresponding to Omp25c were detected by immunoblot in parental B. ovis PA and were absent in all omp25c mutants (Figure 5B). Additionally, the intensity of the upper band in omp31 mutants was higher than that observed with the parental strain (Figure 5B). Absence of the lower band of Omp25c in the Δomp25c mutants was also evident in the SDS–PAGE gel (Figure 5A, lanes Δomp25c-k, Δomp25Domp25c, and Δomp10Δomp31Δomp25c) while the absence of the upper band was only apparent in the Δomp10Δomp31Δomp25c triple mutant (Figure 5A). This result can be explained by the fact that this mutant also lacks Omp31, which is a major OMP that exhibits an electrophoretic mobility similar to that of Omp25c (Figure 5A) and, consequently, masks the absence of the upper Omp25c band in the other Δomp25c mutants.

Omp25 was detected between the two Omp25c bands except in the omp25 mutants. Absence of Omp25 in these mutants was evidenced in both the SDS–PAGE gel (Figure 5A) and the immunostained nitrocellulose (Figure 5B) (see Δomp25, Δomp25Δcgs, and Δomp25Δomp25c lanes in comparison with PA lanes). With both techniques, the Omp25 band of the Δomp31 mutants was more intense than in B. ovis PA. Omp31 was detected in immunoblot and with the same intensity in all the strains, except those lacking the encoding gene (Figure 5C). Absence of Omp31 in these latter mutants was also evident in the SDS–PAGE gel (Figure 5A). The Omp22 band was revealed in B. ovis PA and the analyzed mutants, while Omp25d – that has only been observed in a complemented mutant overexpressing the protein (Martín-Martín et al., 2009) – was not detected (data not shown).

The intracellular behavior of B. ovis PA and the most relevant mutants was evaluated in J774.A1 murine macrophages and HeLa cells. No significant differences between strains (P > 0.05) were observed regarding internalization in J774.A1 macrophages (Figure 6, upper panels). However, although the intracellular bacterial numbers decreased at t20 with all the strains, the reduction was more pronounced with strains Δomp25Δomp25c (Figure 6A), Δomp25Δcgs, Δomp31Δcgs, and its derived triple mutants (Figure 6C), and with B. ovis Δomp10ΔugpBΔomp31 and Δomp10Δomp31Δomp25c (Figure 6B) (P < 0.05). After this moment, all strains were able to replicate, reaching intracellular CFU numbers at t44 about 1 log unit higher than those detected at t20 for each strain (P < 0.05) (Figure 6, upper panels).

In HeLa cells, the Δomp25c mutants (Δomp25c-k, Δomp25Δomp25c, Δomp10Δomp31Δomp25c, and Δomp31 Δomp10Δomp25c) showed a deficient internalization, with intracellular CFU at t0 in the order of 1–1.5 log units lower than those determined for B. ovis PA (P < 0.05). On the contrary, all mutants bearing the omp31 deletion, except those with the Δomp25c genotype (B. ovis Δomp10Δomp31Δomp25c and Δomp31Δomp10Δomp25c), showed an increased internalization in HeLa cells, with CFU at t0 about 1 log higher than those of the parental strain (P < 0.05) (Figure 6, lower panels). All strains suffered a reduction (P < 0.05) of intracellular CFU a t20 while at t44 the bacterial numbers were, in general, similar to those obtained at t20 (Figure 6, lower panels).

Searching for an explanation to the internalization differences observed in HeLa cells, the Δomp25c-k mutant complemented with the wild-type gene and the Δomp31-k mutant together with its complemented strain (Caro-Hernández et al., 2007) were also analyzed regarding internalization in HeLa cells (CFU determined at t0) and their behavior in immunoblot with sera reacting with Omp31 and Omp25c (Figure 7). Complementation of the Δomp25c-k mutant restored its ability to enter HeLa cells like the parental strain (Figure 7A) and the production of Omp25c (Figures 6B,C). The complemented Δomp31-k mutant recovered the ability to produce Omp31, but its level was lower than that of the parental strain (Figures 7B,D). This fact was concomitant with a higher intensity of Omp25c bands, when compared to B. ovis PA that, however, did not reach the intensity observed with the Δomp31 and Δomp31-k mutants (Figure 7C). Internalization of the complemented Δomp31-k mutant in HeLa cells was similar to that of the Δomp31 mutant and significantly higher than that observed with the parental strain (Figure 7A).

The usefulness as vaccine of B. ovis Δomp10ΔugpBΔomp31 against B. ovis PA was evaluated in mice and compared to that of the B. melitensis Rev1 heterologous vaccine. Spleen colonization of the challenge strain was evaluated 3 weeks after infection (Figure 8F) and the humoral (Figure 8E) and cellular (Figures 8A–D) immune response was evaluated at the time point selected for the challenge with the virulent strain (W7 post-vaccination).

Weight of spleens obtained at W7 post-vaccination from mice inoculated with the Δomp10ΔugpBΔomp31 mutant (0.104 ± 0.012 g) did not show statistically significant differences with that of mice inoculated with PBS (0.096 ± 0.005 g), while splenomegaly (P < 0.05) was detected in mice vaccinated with B. melitensis Rev1 (0.136 ± 0.011 g) (data not shown). Additionally, B. melitensis Rev1 was detected in spleen (log CFU/spleen values of 3.66 ± 0.07), while the Δomp10ΔugpBΔomp31 mutant was cleared by this time (data not shown).

Splenocytes were stimulated with B. ovis PA whole cells to evaluate, by specific ELISA tests, the production of cytokines. Splenocytes of mice inoculated with PBS secreted limited amounts of IFN-γ, while about 30 and 60 times higher levels (P < 0.05) were detected in the groups vaccinated with the Δomp10ΔugpBΔomp31 mutant and B. melitensis Rev1, respectively (Figure 8A). In a lesser extent, splenocytes from vaccinated mice also produced more IL-10 (P < 0.05) than those obtained from mice of the PBS group (mean values of 14.7, 9.9, and 1.4 ng/well for the B. melitensis Rev1, B. ovis Δomp10ΔugpBΔomp31, and PBS groups, respectively) (Figure 8C). Regarding the production of TNF-α, splenocytes from mice vaccinated with the two Brucella attenuated strains secreted about double amounts (mean values of 3 ng/well) than those obtained from the unvaccinated control group (P < 0.05) (Figure 8B). No statistically significant differences were observed between groups concerning the production of IL-12(p40) (Figure 8D).

At the time of challenge, mice inoculated with PBS 7 weeks earlier showed low titers (close to the detection limit of log 1.7) of serum antibodies of the IgG₁, IgG_2a, and IgG_2b subclasses able to react with B. ovis PA whole cells in iELISA (Figure 8E). On the contrary, mice vaccinated with B. ovis Δomp10ΔugpBΔomp31 had titers of the three IgG subclasses (ranging from log 3.99 ± 0.16 for IgG_2b to log 4.65 ± 0.13 for IgG₁) that were about 1.5 log units higher (P < 0.05) than those detected in mice vaccinated with the B. melitensis Rev1 heterologous vaccine. Within each vaccination group, no relevant differences were observed between the three IgG subclasses (Figure 8E).

The homologous B. ovis Δomp10ΔugpBΔomp31 vaccine was not detected in spleen 3 weeks after the experimental challenge while it prevented spleen colonization of virulent B. ovis PA (CFU/spleen counts were under the detection limit) (Figure 8F). When compared to control mice inoculated with PBS, vaccination with heterologous B. melitensis Rev1 also drastically reduced the spleen colonization of B. ovis PA (Figure 8F), but both the vaccine strain and the challenge strain were detected in some mice.