Identification of Immune Effectors Essential to the Control of Primary and Secondary Intranasal Infection with Brucella melitensis in Mice

Brucella (alpha-proteobacteria) are facultative intracellular Gram-negative coccobacilli that infect mammals and cause brucellosis. Human brucellosis is a zoonotic infection transmitted through ingestion, inhalation, or contact of contaminated animal products with the conjunctiva or skin lesions (1). Brucella display strong tissue tropism for the reticuloendothelial and reproductive systems, with a stealthy intracellular lifestyle that limits exposure to protective immune responses (reviewed in Ref. 2). The furtive nature of Brucella is attributed, in large part, to the weak agonist activity of its LPS and active suppression of the innate immune response by immunoregulatory factors such as TIR-containing proteins that inhibit MyD88 signaling pathways. Although rarely fatal, Brucella can cause a devastating multiorgan disease in humans with serious health complications in the absence of prolonged antibiotic treatment (3). Despite significant progress, the incidence of human brucellosis remains very high in endemic areas (4) and is considered to be largely underestimated (5). In addition, Brucella species have been weaponized by several governments and are classified as category B threat agents (6). Because complete eradication of Brucella is unpractical because of its presence in a wide range of wild mammals (7, 8), antibiotic treatment is costly, and patients frequently suffer from resurgence of the bacteria (9), vaccination remains the most rational strategy to confer durable protection on populations living in endemic countries and professionals who are frequently exposed to Brucella. Unfortunately, there is no available human brucellosis vaccine; all commercially available animal vaccines are live vaccines that would cause disease in humans (10, 11).

The first step in the rational selection of candidate vaccines is to identify immune markers associated with protective immunity. The central role played by the MyD88/IL-12 signaling axis (12–14), IFN-γ–producing CD4⁺ T cells (15–17), and specific Abs (17) in the protective immune response against Brucella has been well established. However, those results were obtained primarily through analysis of the splenic immune response following i.p. infection, and they teach us little about mucosal immune defenses constituting the first barrier protecting the host against brucellosis under natural conditions. Human brucellosis is frequently acquired via aerosol exposure (18), and 10–100 bacteria seem sufficient to infect humans in this manner (19). Aerosol or intranasal (i.n.) exposure of mice (20) or macaques (21) to low (10²–10³) doses of virulent B. melitensis 16M caused chronic pulmonary infection, followed by systemic infection of the spleen and liver. Interestingly, the TLRs involved in the control of Brucella following i.n. infection in mice appear different in the lungs, liver, and spleen (22), suggesting that the nature of the protective immune response may depend on the organ. Therefore, caution is warranted when interpreting the significance of host immune effector components identified previously only via analysis of the spleen.

In the current study, we analyzed the protection levels and immune responses elicited in several compartments (lungs, spleen, liver, and blood) at different times after primary and secondary i.n. infection of C57BL/6 mice with virulent B. melitensis 16M. We used a large panel of genetically deficient mice to compare the importance of the IFN-γ–mediated response, the IL-17RA–mediated response, and γ/δ⁺ T, α/β⁺ T, CD4⁺ T, CD8⁺ T, and B cells in the control of primary and secondary Brucella infection.

The procedures used in this study and the handling of the mice complied with current European legislation (directive 86/609/EEC) and the corresponding Belgian law “Arrêté royal relatif à la protection des animaux d'expérience du 6 avril 2010 publié le 14 mai 2010.” The Animal Welfare Committee of the Université de Namur (Namur, Belgium) reviewed and approved the complete protocol (Permit Number 05-558).

Wild-type (wt) C57BL/6 mice were acquired from Harlan (Bicester, U.K.). IFNγR^−/− C57BL/6 mice (23), IL-12p35^−/− C57BL/6 mice (24), and IL12p40^−/− C57BL/6 mice (25) were acquired from Dr. B. Ryffel (University of Orleans). IL17RA^−/− C57BL/6 mice (26) and IL23p19^−/− C57BL/6 mice (27) were acquired from Dr. K. Huygen (Belgian Scientific Institute for Public Health, Bruxelles, Belgium). TAP1^−/− C57BL/6 mice (28) and MHCII^−/− C57BL/6 mice (29) were acquired from Jörg Reimann (University of Ulm, Ulm, Germany). RAG1^−/− C57BL/6 mice (30) were acquired from Dr. S. Goriely (Université Libre de Bruxelles). AID^−/− C57BL/6 mice (31) were acquired from Dr. H. Jacobs (Netherlands Cancer Institute, Amsterdam, the Netherlands). TNF-α^−/− C57BL/6 (32) and IL-6^−/− C57BL/6 mice (33) were acquired from Dr. C. De Trez (Vrije Universiteit Brussel). CD3ε^−/−, TCR-β^−/−, TCR-δ^−/−, IL-1R^−/−, CCR2^−/−, and MuMT^−/− C57BL/6 mice were purchased from The Jackson Laboratory (Bar Harbor, ME). All wt and deficient mice used in this study were bred in the animal facility of the Gosselies campus of the Université Libre de Bruxelles.

We used a wt strain of Brucella melitensis 16M and a strain stably expressing a rapidly maturing variant of the red fluorescent protein DsRed (34), the mCherry protein (mCherry-expressing Brucella melitensis [mCherry-Br]), under the control of the strong Brucella spp. promoter, PsojA. Construction of the mCherry-Br strain was described in detail (35). It was grown in biosafety level III laboratory facilities. Cultures were grown overnight with shaking at 37°C in 2YT medium (Luria-Bertani broth with a double quantity of yeast extract) and washed twice in RPMI 1640 (Life Technologies) (3500 × g, 10 min) before inoculation of the mice.

For i.n. infection, mice were anesthetized with a mixture of xylazine (9 mg/kg) and ketamine (36 mg/kg) in PBS before being inoculated i.n. with 2 × 10⁴ CFU wt or mCherry-expressing B. melitensis in 30 μl PBS. Control animals were inoculated with the same volume of PBS. The infectious doses were validated by plating serial dilutions of the inoculums. At the selected time postinfection (p.i.), mice were sacrificed by cervical dislocation. Immediately after sacrifice, spleen, liver, and lung cells were collected for bacterial count, flow cytometry, and/or microscopic analyses.

The i.p. infection technique used as a comparison with our i.n. model was described previously (35). Briefly, 2 × 10⁴ CFU wt or mCherry-expressing B. melitensis in 500 μl of PBS were injected i.p. without anesthesia.

Antibiotic treatment was administered to immunized and control mice for 2 wk. The oral treatment was a combination of rifampicin (12 mg/kg) and streptomycin (450 mg/kg) (adapted from Ref. 17) prepared fresh daily and given in the drinking water. To ensure that the antibiotic treatment was effective, some mice from each group were sacrificed 1 wk prior to the challenge, and the CFU counts were evaluated in the spleen.

Neutralizing Abs against IL-17A (clone 17F3) or the recommended isotype control MOPC-21 (Bio X Cell) were administrated by i.p. injection (200 μg/200 μl) 1 h before the i.n. Brucella challenge and at 2, 6, and 12 d after the challenge. In addition, to saturate the lungs, mice received 50 μg/20 μl Abs against IL-17A or the isotype control by i.n. inoculation immediately after the i.n. Brucella challenge.

Spleens, livers, and lungs were crushed and transferred to PBS/0.1% Triton X-100 (Sigma-Aldrich). We performed successive serial dilutions in RPMI 1640 to obtain the most accurate bacterial count and plated them onto 2YT medium. The CFU were counted after 5 d of culture at 37°C. For bacterial counts in the blood, 75 μl blood was collected from the tail into heparinized capillary tubes at selected time points and diluted in PBS/0.1% Triton X-100 (Sigma-Aldrich). Serial dilutions were prepared in RPMI 1640 and plated onto 2YT medium. The CFU were counted after 5 d of culture at 37°C.

As described previously (17), spleens were harvested, cut into small pieces, and incubated for 30 min at 37°C with a mix of DNase I fraction IX (Sigma-Aldrich) (100 μg/ml) and 1.6 mg/ml collagenase (400 M and l U/ml). Spleen cells were washed, filtered, and incubated with saturating doses of purified 2.4G2 (anti-mouse FcR; American Type Culture Collection) in 200 μl PBS, 0.2% BSA, and 0.02% NaN3 (FACS buffer) for 20 min at 4°C to prevent Ab binding to the FcR. Various fluorescent mAb combinations in FACS buffer were used to stain 3–5 × 10⁶ cells. We acquired the following mAbs from BD Biosciences: FITC-coupled 145-2C11 (anti-CD3ε), FITC-coupled M1/70 (anti-CD11b), FITC-coupled 1A8 (anti-Ly6G), PE-coupled GL3 (anti– TCR-γ/δ), PE-coupled RM4-5 (anti-CD4), PE-coupled 53-6.7 (anti-CD8α), PE-coupled PK136 (anti-NK1.1), biotin-conjugated RB6-8C5 (anti-Gr1), streptavidin PE, allophycocyanin-coupled 1-A/1-E (anti-MHC class II [MHCII]), and allophycocyanin-coupled BM8 (anti-F4/80). The cells were analyzed on a FACSCalibur cytofluorometer. Dead cells and debris were eliminated from the analysis according to size and scatter.

For intracellular staining, spleen cells were incubated after DNase-collagenase treatment for 4 h in RPMI 1640, 10% FCS with 1 μl/ml Golgi Stop (BD Pharmingen) at 37°C, under 5% CO₂. After washing with FACS buffer and staining for cell surface markers, cells were fixed in PBS/1% PFA for 15–20 min at 4°C. They were permeabilized for 30 min using a saponin-based buffer (10× Perm/Wash in FACS buffer; BD Pharmingen) and stained with allophycocyanin-coupled XMG1.2 (anti–IFN-γ; BD Biosciences). After final fixation in PBS/1% PFA, cells were analyzed on a FACSCalibur cytofluorometer. No signal was detectable with isotype controls.

Control and immunized mice were sacrificed, and lung and spleen cells were harvested. Cells were centrifuged and cultured in RPMI 1640 supplemented with 10% FCS, 1% l-glutamine, 1% nonessential amino acids, 1% pyruvate sodium, and 0.1% gentamicin in six-well plates with 10⁷ cells/well in a volume of 2 ml. For stimulation, a concentration of 2 × 10⁷ bacteria/ml heat-killed B. melitensis (HKBr) or 0.5 μg/ml Con A (Sigma-Aldrich) was used. Cells were incubated for 7 h at 37°C, 5% CO₂. After adding 1 μl/ml Golgi Stop (BD Pharmingen), the incubation was continued for an additional 13 h at 37°C, 5% CO₂. Cells were washed and stained for flow cytometric analysis, as described above.

Spleens and livers were fixed for 4 h at room temperature (RT) in 2% paraformaldehyde (pH 7.4), washed in PBS, and incubated overnight at 4°C in a 20% PBS-sucrose solution. Lungs were submitted to the same treatment at RT under a vacuum. Tissues were embedded in Tissue-Tek OCT compound (Sakura) and frozen in liquid nitrogen, and cryostat sections (5 μm) were prepared. For staining, tissue sections were rehydrated in PBS and incubated in a PBS solution containing 1% blocking reagent (Boehringer) for 20 min before incubation overnight in the same type of solution containing any of the following mAbs or reagents: DAPI nucleic acid stain Alexa Fluor 350 or 488 phalloidin (Molecular Probes) to visualize the structure of the organ and Alexa Fluor 647–coupled BM8 (anti-F4/80; Abcam), Alexa Fluor 647–coupled M1/70 (anti-CD11b; Abcam), Alexa Fluor 647–coupled N418 (anti-CD11c; BioLegend), biotin-coupled RB6-8C5 (anti-Gr1; eBioscience), or purified M-19 (anti-NOS2; Santa Cruz Biotechnology) to stain the cells of interest. Incubation with a secondary mAb, Alexa Fluor 488–coupled goat anti-rabbit IgG (Molecular Probes), for 2 h was necessary for the anti-NOS2 mAb. Biotinylated Abs were incubated for 2 h with Alexa Fluor 350 streptavidin. Detection of Brucella Ag was performed by incubation overnight with rabbit polyclonal Abs against HKBr (35), followed by incubation for 2 h with Alexa Fluor 488–coupled goat anti-rabbit IgG (Molecular Probes). Slides were mounted in Fluoro-Gel medium (Electron Microscopy Sciences, Hatfield, PA). Labeled tissue sections were visualized with an Axiovert M200 inverted microscope (Zeiss, Jena, Germany) equipped with a high-resolution monochrome camera (AxioCam HR; Zeiss). Images (1384 × 1036 pixels, 0.16 μm/pixel) were acquired sequentially for each fluorochrome with A-Plan 10×/0.25 N.A. and LD Plan-NeoFluar 63×/0.75 N.A. dry objectives and recorded as eight-bit gray-level *.zvi files. At least three slides/organ were analyzed from three animals, and the results are representative of two independent experiments.

Specific murine IgM, IgA, IgG1, IgG2a, and IgG3 isotypes were determined by ELISA. Polystyrene plates (Nunc; 269620) were coated with HKBr (107 CFU/ml). After incubation overnight at 4°C, plates were blocked for 2 h at RT with 200 μl PBS–3.65% casein. The plates were incubated overnight at 4°C with 50 μl serial dilutions of the serum in PBS–3.5% casein. Sera from unimmunized mice were used as the negative control. After four washes with PBS, isotype-specific goat anti-mouse HRP conjugates were added (50 μl/well) at the appropriate dilutions (hIgM from Sigma-Aldrich; hIgA from Invitrogen; LO-MG1-13 HRPO, LO-MG2a-9 HRPO, and LO-MG3-13 HRPO from LO-IMEX). Plates were incubated for 2 h at RT and washed four times in PBS before adding 100 μl substrate solution (BD OptEiA) to each well. After 10 min of incubation at RT in the dark, the enzyme reaction was stopped by adding 25 μl/well 2 N H₂SO₄, and absorbance was measured at 450 nm.

We used a (Wilcoxon) Mann–Whitney U test provided by GraphPad Prism software to analyze our results. Each group of deficient mice was compared with wt mice. The p values were considered to represent a significant difference.

The course of B. melitensis 16M in wt C57BL/6 mice following i.n. infection was analyzed previously by several research groups (22, 36, 37). However, because the infectious dose, the origin of the mice, and the conditions in the animal facility could influence the course of the infection, we began by establishing the kinetics of infection under our laboratory conditions. Because lungs naturally contain commensal bacteria, we used a mCherry-Br strain of B. melitensis 16M displaying kanamycin resistance to evaluate the Brucella CFU count during the course of infection.

Following i.n. inoculation with 2 × 10⁴ CFU of B. melitensis, wt C57BL/6 mice were sacrificed at 1, 3, 6, 12, 20, 28, and 50 d p.i., and CFU were counted in the lungs, spleen, and liver (Fig. 1A). We observed that bacteria persisted at the same level in the lungs for ∼12 d and then were progressively eliminated. At 28 d p.i., only rare mice displayed detectable CFU in the lungs. The dissemination of Brucella from the lungs to the spleen and liver took ∼3 d. From 6 d p.i., the spleen and liver of the majority of mice displayed detectable CFU counts, which peaked at 12–28 d p.i. Infection persisted in the spleen but decreased below the level of detection in the liver of the majority of mice. Compared with i.p. infection with the same dose of bacteria, a striking characteristic of i.n. infection was the slow dissemination from the primary site of infection to the spleen and liver (38), as well as the absence of detectable bacteria in the bloodstream (Fig. 1B). Because our detection limit was 10 CFU/ml, this result suggests that a very small number of bacteria travel simultaneously in the bloodstream.

After i.n. administration of 2 × 10⁴ CFU, flow cytometry analysis of infected lungs did not reveal a significant modification in the frequency or number of neutrophils (CD11b⁺ Ly-6G⁺ F4/80⁻), alveolar macrophages (CD11c⁺ Ly-6G⁻ F4/80⁺), and interstitial macrophages (CD11c⁻ Ly-6G⁻ F4/80⁺) until 12 d p.i. (data not shown). Intraperitoneal Brucella infection induces dense and easily detectable granulomas in the spleen and liver, which primarily contain inducible NO synthase (iNOS)/NOS2⁺ monocytes and inflammatory dendritic cells surrounded by granulocytes (35). All of these cells express high levels of CD11b. Thus, we used iNOS and CD11b markers to detect the presence of aggregates of activated myeloid cells in the lung tissue. We observed (Fig. 2A, data not shown) that lungs of infected mice do not display a greater number of CD11b⁺ aggregates compared with lungs of control naive mice. In addition, the few CD11b⁺ aggregates present in the lungs of infected mice are negative for iNOS or the presence of Brucella Ags (data not shown). Thus, in agreement with reports by other investigators (39), we conclude that, although nonnegligible CFU levels are reached, Brucella lung infection does not lead to granuloma formation during any phase of infection. This suggests that the early control of Brucella growth in mucosal lung tissue does not induce or require an inflammatory reaction. To test this hypothesis, we compared the course of Brucella infection in wt, IL-1R^−/−, IL-6^−/−, TNF-α^−/−, and CCR2^−/− [deficient for monocyte recruitment (40)] C57BL/6 mice (Fig. 3). Mice deficient for IL-1R, IL-6, and TNF-α or the CCR2-dependent recruitment of monocytes displayed CFU counts in the lungs similar to those of wt mice during the first 5 d of infection, confirming that the inflammation process does not seem to be implicated in the early control of Brucella infection in the lungs. At 12 and 28 d p.i., only TNF-α^−/− mice displayed significantly higher CFU counts in the organs. IL-1R deficiency had a slightly significant impact on the spleen only at 28 d p.i.

In drastic contrast to the lungs, colonization of the spleen (Fig. 2B) and liver (Fig. 2C) was accompanied by CD11b⁺ iNOS⁺ granuloma formation. Rabbit polyclonal Abs against HKBr confirm the presence of Brucella Ags inside granulomas (Fig. 2D). Unfortunately, the low frequency of Brucella in organs does not allow for direct visualization of the bacteria in tissue sections using the mCherry-expressing strain. Interestingly, the location of granulomas in the spleen after i.n. infection was clearly different from their location after i.p. infection. As we reported previously (35), following i.p. infection, granulomas are mainly located in the red pulp during the first days of infection and are found in the white pulp only after 12 d. After i.n. infection, granulomas appear immediately and only inside the marginal zone and white pulp (Fig. 2C), suggesting that the localization of granulomas in the red pulp could be a consequence of the rapid and abrupt infection of the spleen following i.p. injection.

The central role of the TLR4/TLR9/MyD88/IL-12 axis in promoting IFN-γ–producing CD4⁺ T cells to control Brucella growth in the spleen following i.p. infection is well established (14, 16, 35). By comparing the course of i.p. infection in the spleen in wt, IL-12p35^−/−, IL-23p19^−/−, and IL-12p40^−/− (deficient for both IL-12 and IL-23) mice, we observed that IL-12p40^−/− mice display higher susceptibility than IL-12p35^−/− mice, which suggests a weakly compensatory role for IL-23 (affecting the IL-17A–mediated response) in the absence of IL-12p35 (affecting the IFN-γ–mediated response) in the spleen (16). In our i.n. infection model, we analyzed the impact of this deficiency on the control of Brucella growth in the lungs, spleen, and liver (Fig. 4A). IL-23p19^−/− mice displayed CFU counts similar to wt mice at all time points and in all organs tested. IL-12p35^−/− and IL-12p40^−/− mice displayed enhanced CFU counts in the spleen throughout the entire course of infection. In drastic contrast to the spleen, IL-12p35 deficiency only had a transient impact on CFU counts in the lungs and liver. At 50 d p.i., these mice eliminated Brucella like wt mice. Only IL-12p40 deficiency leads to long-term uncontrolled persistence in the lungs and liver at 50 d p.i. These results demonstrate that the immune requirements for the control of Brucella infection are organ-specific and suggest that the IL-23–dependent IL-17A–mediated response may compensate for the partial absence of an IFN-γ–mediated response to control Brucella in the lungs and liver.

Then we compared the impact of direct IFN-γR and IL-17RA deficiencies on the course of i.n. Brucella infection. As reported previously (41–43), the complete absence of functional IFN-γ signaling leads to high CFU counts in all tested organs (Fig. 4B) and 100% mortality after 35 d p.i. (Fig. 5A). By comparison, the impact of IL-17RA deficiency appears to be limited to higher CFU counts in the lungs during the early stages of infection (Fig. 5), without any associated mortality (data not shown). Interestingly, the impact of IL-17RA deficiency is already pronounced at 5 d p.i. At that point in time, IFN-γR signaling deficiency has no detectable impact in the lungs, suggesting that the IL-17RA–mediated response starts before the IFN-γR–mediated response in the lungs and could constitute the first line of defense against Brucella infection.

The fatal course of Brucella infection in all mice displaying a deficiency in IFN-γ signaling [IFN-γ^−/− (41, 43), IRF1^−/− (42) mice] supports the idea that IFN-γ plays a pivotal role in mediating a protective immune response. However, the absence of mortality in RAG1^−/− (16), MyD88^−/− (14), and IL-12p35^−/− (16) mice is surprising because T cells are described as the main producer of IFN-γ during chronic Brucella infection (14, 16), and MyD88 (14) and IL-12p35 (16) have been closely implicated in IFN-γ induction. This shows that mortality results only from the complete absence of IFN-γ and is not proportional to its level. We observed that CFU levels in organs were nearly similar in CD3ε^−/− and IFN-γR^−/− mice (Fig. 5B), but only IFN-γR^−/− mice died (Fig. 5A), suggesting that the lack of control of Brucella growth in the organs is not the only reason for IFN-γR deficiency–associated mortality.

To reconcile this enigma, we examined the lungs, spleen, and liver of wt, IL-12p35^−/−, CD3ε^−/−, and IFN-γR^−/− mice 35 d after i.n. infection by in situ fluorescent microscopy. We observed that only IFN-γR^−/− infected mice displayed a dramatic recruitment of Gr1⁺ cells in tissues (Fig. 6A for lungs, Supplemental Fig. 1 for comparison of the lungs, spleen, and liver). Flow cytometry analysis identified Gr1⁺ cells in the spleen as neutrophils, as demonstrated by their typical cell surface phenotype: Ly-6G⁺ F4/80⁻ MHC-II⁻ CD11b⁺ (Supplemental Fig. 2). This dramatic neutrophilia was not correlated with the observation of a higher frequency of IL-17A⁺ cells by flow cytometry in the spleen of IFN-γR^−/− infected mice (data not shown), suggesting that a compensatory IL-17A–mediated response is not implicated in mortality. Interestingly, the IFN-γR deficiency strongly affected the cellular composition of the granulomas in the liver (Fig. 6B) and spleen (data not shown). We showed previously (35) that granulomas induced by Brucella infection are primarily composed of iNOS⁺ monocytes (CD11b⁺ CD11c⁻ F4/80^low) and inflammatory dendritic cells (CD11b⁺ CD11c⁺ F4/80^high) with few neutrophils (CD11b⁺ Ly-6G⁺/GR1⁺). We observed a similar cellular composition of the granulomas in the liver (Fig. 6B) and spleen (data not shown) after i.n. infection in wt and IL-12p35^−/− mice, with only a partial reduction in iNOS staining in IL-12p35^−/− mice. In contrast, IFN-γR^−/− infected mice displayed large granulomas dominated by iNOS⁻ CD11b⁺ Gr1⁺ CD11c⁻ F4/80⁻ cells. These neutrophil-rich granulomas did not seem able to restrict Brucella dissemination because they were correlated with high CFU counts in the blood from all IFN-γR^−/− infected mice (Fig. 6C). Note that a kinetic analysis showed that Brucella becomes detectable in the blood of IFN-γR^−/− mice only starting at 20 d p.i. (data not shown). No bacteria are detected during the course of infection in wt mice (Fig. 1B) and IL-12p35^−/− infected mice (data not shown), and only a small fraction of infected CD3^−/− mice display a small number of bacteria in the blood (Fig. 6C). As a whole, these results demonstrate that the high mortality observed in IFN-γR^−/− mice following B. melitensis infection is associated with dramatic neutrophilia leading to multiple organ failure and a major alteration in granuloma formation associated with the persistence of bacteria in the blood.

To determine the lymphocyte populations implicated in the control of primary i.n. Brucella infection, we compared the course of infection in various strains of C57BL/6 mice genetically deficient for key genes affecting specific lymphocyte subsets.

The absence of α/β⁺ or γ/δ⁺ T lymphocytes leads to a lack of Brucella control in the lungs (Fig. 7A). TCR-β deficiency affects early Brucella control in the lungs and late control in all organs. In contrast, TCR-δ deficiency strongly impairs early Brucella growth in the lungs but has only a transient impact on spleen and liver bacterial loads at 12 d p.i., with no impact on late control at 28 and 50 d p.i. Comparison of TAP1 and MHCII deficiencies (Fig. 7B) suggests that CD8⁺ α/β⁺ T lymphocytes only participate in the early control of Brucella in the lungs and for a short period of time in the spleen at 28 d p.i. In striking contrast, MHCII deficiency affecting CD4⁺ α/β⁺ T lymphocytes fully abrogates Brucella control in all organs from 12 to 50 d p.i., thus confirming the crucial role played by CD4⁺ α/β⁺ T lymphocytes in the protective immune response against chronic Brucella infection (Fig. 7B). Thus, although γ/δ⁺ T cells and CD8⁺ α/β⁺ T cells constitute a first line of defense in mucosal lung tissue, CD4⁺ α/β⁺ T cells appear to be involved in the later control of infection in all organs.

Comparison of the cell surface phenotype of IFN-γ–producing cells in the spleen of wt mice following i.p. and i.n. Brucella infection showed that CD4⁺ α/β⁺ T lymphocytes are the dominant IFN-γ–producing cells at all times during infection (Fig. 8). In contrast to i.p. infection, we did not observe an early peak in IFN-γ–producing γ/δ⁺ T lymphocytes, NK, or CD8⁺ α/β⁺ T lymphocytes during i.n. infection, suggesting that early activation of these cells during i.p. infection could be a consequence of the rapid and massive dissemination of Brucella in the spleen.

As we observed previously in an i.p. model (16), the course of Brucella infection appears to be similar in the lungs, spleen, and liver of i.n. infected wt, AID^−/− (deficient for activation-induced deaminase-mediated isotype switching), and MuMT^−/− (deficient for B lymphocytes) C57BL/6 mice at all time points tested (Fig. 9). This demonstrates that Abs do not play a crucial role in the control of primary Brucella infection in our i.n. infectious model.

To gain further insight into the secondary immune response against Brucella, wt mice received i.n. administrations of 30 μl of PBS alone (control mice) or 2 × 10⁴ CFU of the wt (nonfluorescent) strain of B. melitensis. At 28 d p.i., both groups were treated daily with a combination of antibiotics (see 2Materials and Methods) for 2 wk to eliminate the Brucella. As shown previously (17), this treatment led to the complete elimination of Brucella in all tested organs of the infected mice. After a rest period of 2 wk, control mice and previously infected mice were challenged with an mCherry-Br strain of B. melitensis. Kinetic analysis of the CFU counts in the lungs, spleen, and liver (Fig. 10A) showed that the previously infected animals displayed a greater ability to control Brucella infection. The CFU counts in the lungs had already decreased by >1 log at 6 d. In contrast to primary infected mice, a majority of the secondary infected mice did not display bacteria in the spleen at 12 and 28 d p.i. It is interesting to note that i.n. infection protects against i.n. and i.p. challenges (Fig. 10B). On the whole, these results demonstrate that Brucella infection induces development in the lung of a protective memory that limits or impairs the dissemination of bacteria from the lung to the spleen and liver.

Characterization of the lymphocyte populations and Th cell polarization necessary to control secondary mucosal Brucella infection constitutes the first step in understanding the underlying mechanisms of the protective immune response and developing a rational strategy to select a protective candidate vaccine against brucellosis. Previous studies by our group using an i.p. B. melitensis infectious model with the same B. melitensis strain (17) showed that genetic deficiencies leading to a lack of CD4⁺ T cells (MHCII^−/− mice), B cells (MuMT mice), or IL-12 (IL-12p35^−/− mice) completely impaired the ability of mice to develop a protective secondary immune response limiting infection of the spleen.

wt, CD3ε^−/−, TCR-β^−/−, TCR-δ^−/−, TAP1^−/−, MHCII^−/−, MuMT^−/−, IL-12p35^−/−, IL-17RA^−/− and IL-23p19^−/− mice infected with wt B. melitensis 28 d previously and treated with antibiotics as described previously were challenged with mCherry–B. melitensis. Naive wt mice were used as the internal control. Fig. 11 shows the CFU counts in the spleen at 28 d postchallenge. We observed that only the spleen of CD3ε ^−/− and TCR-β^−/− mice displayed a lack of control of Brucella growth compared with wt mice. They were also the only mice to present persisting infection in the lungs following the Brucella challenge (data not shown). In striking contrast with an i.p. infection model (17), deficiencies in MHCII, IL-12p35, or B cells do not significantly impair the development of protective immune memory.

The negligible role of B lymphocytes in the control of an i.n. secondary challenge is not due to an impaired ability of i.n. infection–induced specific Abs to control Brucella. This is demonstrated by the fact that i.p.– and i.n.–infected mice displayed specific circulating Abs against Brucella at 28 d p.i. (Supplemental Fig. 3A) and were equally able to control the bacteria counts in the blood after i.p. challenge with 5 × 10⁷ CFU of B. melitensis (Supplemental Fig. 3B).

The efficient control of secondary Brucella infection observed in IL-12p35^−/− mice challenges the idea that the IFN-γ–mediated response is crucial to Brucella control. To determine whether the IL-17A–mediated response compensates for the lack of the IL-12–dependent IFN-γ–mediated response in IL-12p35^−/− immunized mice, we tested the impact of IL-17A neutralization in these mice. After i.n. challenge, wt and IL-12p35^−/− immunized mice were treated with neutralizing Abs against IL-17A, as indicated in 2Materials and Methods. We observed (Fig. 11B) that this treatment abolishes the protection in IL-12p35^−/− challenged mice but not in wt mice. A similar treatment with the isotype control had no detectable impact (data not shown). Taken together, these results suggest that the IL-17A–mediated response can compensate for the reduced IFN-γ–mediated response in IL12p35^−/− mice to control a Brucella challenge. Unfortunately, a similar experiment would not be feasible in IFN-γR^−/− mice because the complete absence of the IFN-γ signaling pathway leads to the death of all Brucella-infected mice around 35–40 d p.i., even with oral antibiotic treatment (described in 2Materials and Methods section) starting at 20 d and reinforced by i.p. injection of the antibiotic (data not shown).

Finally, to identify the IFN-γ– and IL-17A–producing cells in the lung and spleen during the memory response, we performed a kinetic analysis of the cytokines by flow cytometry in challenged wt mice. In the absence of in vitro restimulation, we did not detect IFN-γ and IL-17A in the lung and spleen cell populations from challenged mice at any time point tested (data not shown), which suggests that the levels of these cytokines are very low in vivo. To solve this problem, we stimulated the lung and spleen cells from control and immunized wt mice overnight in vitro with HKBr or 0.5 μg/ml of Con A (Fig. 12A). HKBr stimulation induces IFN-γ⁺ cells in the spleen of immunized mice but not control mice. These cells are α/β⁺ T cells (Fig. 12B) and mainly CD4⁺ (data not shown). Similar stimulation of lung cells did not lead to detectable IFN-γ production. Unspecific Con A stimulation is required to induce IFN-γ⁺ cells in lung cells. These cells also appear to be α/β⁺ T cells (Fig. 12B) and mainly CD4⁺ (data not shown). Lung cells from immunized mice display a higher frequency of IFN-γ⁺ cells upon Con A stimulation compared with control mice, suggesting that Brucella infection induces specific CD4⁺ α/β⁺ T cells in the lung. Unfortunately, under the same experimental conditions, we did not detect IL-17A–producing cells in response to HKBr or Con A, suggesting that the frequency of these cells is extremely low or that these cells are lost during the flow cytometry protocol.

A comparison of immune effector mechanisms essential to the control of Brucella infection in i.p. and i.n. infection models is shown in Fig. 13.

A growing number of clinically relevant infectious diseases are caused by pathogen persistence in the host and require long-lasting and costly therapies (44). Chronic infection is generally the consequence of the ability of a pathogen to escape the adaptive immune response and generate a microenvironment in which it can survive. Therefore, it is crucial to neutralize such pathogens before they disseminate and colonize the host. The major point of entry for many pathogens is the gastrointestinal or respiratory tract, or the genital mucosa (reviewed in Ref. 45), suggesting that mucosal protective immune responses must be induced to counteract pathogen dissemination. However, the ways by which mucosal immunization could generate protective frontline immunity against pathogens remain largely undefined. Brucella establishes a stealthy infection in mice and constitutes an interesting experimental model that can be used to analyze the relationships between hosts and intracellular stealth pathogens (46, 47). Although the aerosol route is a common cause of infection in nature (1), a relatively limited number of Brucella studies (20, 22, 36, 37, 48, 49) investigated the protective immune response following respiratory infection. In this study, we analyzed the progression of B. melitensis in the lungs, spleen, and liver of C57BL/6 mice, as well as the nature of the protective immune response following primary and secondary i.n. infection, and compared them with our results (16, 17, 35, 38) obtained in the classical i.p. infection model.

In contrast to i.p. infection (35, 38), i.n. infection does not lead to the rapid dissemination of Brucella in the spleen and liver or to detectable and prolonged persistence of Brucella in the blood. This could explain why Brucella-induced granulomas in the spleen immediately appeared primarily in the white pulp in this i.n. model and not in the red pulp as previously described following i.p. infection (35). We hypothesize that i.p. infection saturates the marginal zone filter of the spleen consisting of marginal zone macrophages and dendritic cells, thus allowing for the infection of red pulp F4/80⁺ macrophages. These observations suggest that the route of infection could potentially alter the tissue localization of the bacteria and the phenotype of the primary infected cells.

The cytokines IL-1, IL-6, and TNF-α are sometimes referred to as the inflammatory triad because they are key mediators of local and systemic inflammatory responses. Impaired IL-1 (50, 51), IL-6 (52–54), and TNF-α (55–57) signaling causes enhanced susceptibility of mice to infection with various pathogens. Surprisingly, in our experimental model, we observed that early Brucella control in the lungs is not affected by IL-1R, IL-6, or TNF-α deficiency. Furthermore, during the late phase of control, only TNF-α deficiency strongly affected CFU counts of Brucella in the organs. In keeping with this observation, Brucella was recently shown to induce weak IL-1β, IL-6, and TNF-α production in lung alveolar macrophages compared with peritoneal macrophages (58). The chemokine receptor CCR2 has been implicated in the recruitment of inflammatory monocytes during infection (40). CCR2 deficiency impaired development of a protective immune response in several infectious models (40, 59, 60). We observed that CCR2 deficiency had no impact on the course of Brucella infection in the lungs, liver, or spleen at all time points tested. We also did not detect recruitment of inflammatory cells in the lungs during Brucella infection by flow cytometry (data not shown) or modification of the cellular organization, such as the formation of granulomas, by microscopy during the period of infection. These results suggest that control of Brucella growth in the lungs does not involve the development of a strong local inflammatory response or the recruitment of inflammatory monocytes.

Initially, the IL-17RA–mediated response was primarily associated with the host defense against extracellular pathogens (reviewed in Ref. 61). However, a growing number of recent studies suggest that this response is also involved in the control of intracellular microorganisms, such as Listeria monocytogenes, Salmonella enterica, and Mycobacterium tuberculosis (reviewed in Refs. 62, 63). In the lungs, IL-17A was reported to act primarily on nonhematopoietic cells, such as epithelial cells, endothelial cells, and fibroblasts (reviewed in Ref. 63), to promote the production of numerous proteins with antimicrobial properties and chemokines that attract neutrophils. Although first identified in CD4⁺ T cells, IL-17A can also be produced by CD8⁺ T cells and innate cells, such as γ/δ⁺ T cells, invariant NKT cells, neutrophils, and innate lymphoid cells (63). There have been few studies of the function of IL-17A–mediated responses in immunity to Brucella organisms. In a classical i.p. infection model, we reported that IL-23p19 deficiency, which is known to reduce the IL-17A production response (64), does not significantly affect the control of Brucella in the spleen but can partially compensate for IL-12p35 deficiency (16). In the present work, we observed that, following i.n. infection, IL-17RA and IL-23p19 deficiency only had minor effects on CFU counts in the spleen and liver throughout infection. In contrast, IL-17RA^−/− mice displayed an important lack of control in the lungs during the early stages of infection, until 12 d p.i. Interestingly, during the same period, IL-23p19 and IFN-γR deficiency had no or only minor effects on CFU counts of Brucella in the lungs. These results suggest that the IL-17RA–mediated response constitutes the first line of defense against Brucella in the lungs. IL-17A production is generally considered strictly dependent on IL-23, but reduced IL-17A production can still be detected in IL-23p19^−/− mice (27). The absence of an effect of IL-23p19 deficiency in our model suggests that low counts of IL-17A could be sufficient to mediate Brucella control in the lung. The IFN-γ–mediated response seems to develop late in the lungs, and an IL-17RA–mediated response may compensate for reduced IFN-γ production due to IL-12p35 deficiency, as suggested by the fact that wt and IL-12p35^−/− mice, but not IL-12p40^−/− mice, control the infection at later stages in the lungs. A similar compensatory phenomenon is observed in the liver. Unfortunately, presumably due to the stealthy nature of Brucella infection, we failed to detect IL-17A⁺ cells in the lungs by flow cytometry and, thus, to identify their source during primary or secondary infection. However, we observed that TCR-δ and TAP1 deficiency, but not MHCII deficiency, also affected early Brucella control in the lungs, suggesting that these cells could be a potential source of IL-17A in our model. In keeping with this observation, IL-17 production by γ/δ T cells was reported in a Brucella model (65) and in several other pulmonary infectious models (66–68).

Studies carried out in i.p. infectious models clearly established the crucial role of IFN-γ–secreting CD4⁺ T cells in the primary control of Brucella infection (16, 35, 41–43). We confirmed this key role in our i.n. model. From 12 d p.i., IFN-γR and MHCII deficiency led to a dramatic lack of Brucella control in the lungs, spleen, and liver. Moreover, we demonstrated that mortality associated with a deficiency in IFN-γ signaling is linked to uncontrolled Brucella growth in tissues and probably to severe neutrophilia, which causes multiple organ failure. We failed to reduce Brucella-induced mortality in IFN-γR^−/− mice by repeated administration of depleting Abs, such as 1A8 or GR1 specific to neutrophils (data not shown). Most likely, this can be attributed to the chronic nature of Brucella infection, which leads to continuous recruitment of new neutrophils. This neutrophilia is not associated with a compensatory IL-17A–mediated response, because we did not detect a higher frequency of IL-17A⁺ cells in the spleen of infected IFN-γR^−/− mice by flow cytometry compared with infected wt mice (data not shown). Likewise, cutaneous and musculoskeletal inflammation observed following i.p. Brucella infection in IFN-γ^−/− mice is independent of the Th17 response (69). In an Ehrlichia muris infection model (70), IFN-γ was shown to control infection by directly promoting the differentiation of myeloid cells into monocytes. In the absence of IFN-γ signaling, E. muris infection only promotes the expansion of neutrophils, which do not play a protective role during Brucella infection (71) and are generally associated with massive tissue damage, particularly in the lungs (reviewed in Ref. 72). In our i.n. infectious model, the absence of IFN-γR was associated with a drastic alteration in the granuloma composition, which appeared to primarily contain neutrophils, and a lack of tissue confinement of Brucella, as demonstrated by the high levels of bacteria observed in the blood of IFN-γR^−/− mice during chronic infection but not during the early phases of infection. In summary, our results strongly suggest that the protective role of IFN-γ against Brucella infection is primarily limited to the chronic phase of infection and may even have been overestimated based on the exceptional mortality associated with IFN-γ deficiency (41, 43). IFN-γ is clearly crucial to late Brucella control in all organs, but is not sufficient, because we observed, in an i.p. infection model (16), that a similar frequency of IFN-γ–producing CD8⁺ T cells does not replace IFN-γ–producing CD4⁺ T cells to control Brucella growth in the spleen following primary infection.

To identify potential protective immune markers to select candidate vaccines against brucellosis, we previously attempted to characterize the cells and signaling pathways involved in the generation of a protective immune memory response following i.p. priming with live B. melitensis (17). We showed that IL-12p35, MHCII, and B cell deficiency completely impair the control of secondary Brucella infection in the spleen. Using a similar approach in the i.n. infection model, we analyzed the course of secondary B. melitensis infection. Secondary infections appear to be primarily controlled in the lung, because the majority of immunized mice displayed a 10-fold reduction in CFU counts in the lungs at 6 d p.i. compared with primary infected mice, and 60% presented no detectable bacteria in the spleen or liver at 12 d. Using genetically deficient mice, we observed that CD3 and TCR-β deficiency, but not TCR-δ or B cell deficiency, completely abrogates protection in the spleen, thus demonstrating that α/β⁺ T cells constitute the core of the adaptive immune system implicated in protective immunity against Brucella. Because the selective absence of CD4⁺ α/β⁺ T cells (in MHCII^−/− mice) or CD8⁺ α/β⁺ T cells (in TAP1^−/− mice) does not affect the efficacy of the secondary protective immune response, we conclude that these two populations are equally able to mediate protective immune responses against an i.n. Brucella challenge. Thus, although Brucella display immune escape mechanisms that affect CD8⁺ T cell activity (73, 74), memory CD8⁺ T cells seem to be able to compensate for the absence of memory CD4⁺ T cells and, thus, to control secondary i.n. infection with Brucella. We also observed that IL-12p35, IL-17RA, and IL-23p19 deficiency did not significantly affect protective immunity against secondary i.n. infection. However, protection in IL-12p35^−/− mice is completely abolished by injecting neutralizing IL-17A–specific Abs, which suggests that the IL-17A–mediated response is able to compensate for the reduced IFN-γ–mediated response in IL-12p35^−/− mice. In keeping with this observation, live oral brucellosis vaccines were reported to stimulate IFN-γ–mediated and IL-17A–mediated responses in mice (75). Oral immunization with attenuated ΔznuA and RB51 strains of B. melitensis induces protection in IFNγ^−/− BALB/c mice against an i.n. challenge with a virulent B. melitensis strain. Neutralization of IL-17A in IFNγ^−/− mice reduces the protection conferred by the RB51, but not by the ΔznuA, vaccine strain. Thus, the control of secondary i.n. Brucella infection seems to be characterized by a redundancy of protective immune mechanisms and, consequently, a great robustness to selective deficiencies of IL-17A–mediated and IFN-γ–mediated response or T cell subsets (γ/δ⁺ T cells, CD4⁺ α/β⁺ T cells, CD8⁺ α/β⁺ T cells).

We observed previously (38) that, following i.p. inoculation, Brucella display a brief extracellular phase in the blood and then are able to invade erythrocytes and to persist for several weeks in the bloodstream. Circulating specific Abs present in immunized mice are able to neutralize Brucella before erythrocyte invasion and, thus, to reduce Brucella dissemination to organs (38). Surprisingly, we observed that protective immunity against an i.n. challenge is not significantly affected by a B cell deficiency. We hypothesize that, following i.n. infection, Brucella disseminate from lung tissue inside myeloid phagocytic cells and, thus, avoid neutralizing Abs, a result that is in keeping with the observation that Brucella are rapidly found in dendritic cells and alveolar macrophages in lung-draining lymph nodes following primary i.n. infection (39). Thus, the control of i.n. Brucella infection appears to be primarily dependent on cellular immunity and independent of humoral immunity.

As a whole, our work highlights the importance of analyzing the protective immune response in experimental models using natural routes of infection. By comparing i.p. and i.n. infection models, we have shown that the location of bacteria in organs and the effector immune mechanisms required to control primary and secondary infection are strongly dependent on the route of inoculation (summarized in Fig. 13). Our results challenge the idea that the Th1 response mediated by CD4⁺ T cells is indispensable to the development of a protective memory response against Brucella infection. In our i.n. infection model, we found that CD8⁺ T cells may compensate for the absence of CD4⁺ T cells and control secondary Brucella infection. Similarly, a reduced IFN-γ–mediated response can be compensated for by an IL-17A–mediated response. The precise effector mechanisms implicated in these responses remain to be clarified, but our work opens up new areas for future investigations. These findings could improve our ability to develop protective vaccines or therapeutic treatments against brucellosis or other infectious diseases caused by intracellular bacteria. Because cellular immunity only develops following the administration of live bacteria (17), live attenuated vaccines or inactivated, yet metabolically active, bacteria seem to remain the easiest and most potent tools for the production of candidate protective vaccines against brucellosis.

We thank Dr. S. Goriely and Dr. Heinz Jacobs for providing RAG1^−/− C57BL/6 mice and AID^−/− C57BL/6 mice, respectively.

The authors have no financial conflicts of interest.