Abstract
Brucellaceae are stealthy pathogens with the ability to survive and replicate in the host in the context of a strong immune response. This capacity relies on several virulence factors that are able to modulate the immune system and in their structural components that have low proinflammatory activities. Lipopolysaccharide (LPS), the main component of the outer membrane, is a central virulence factor of Brucella, and it has been well established that it induces a low inflammatory response. We describe here the identification and characterization of a novel periplasmic protein (RomA) conserved in alpha-proteobacteria, which is involved in the homeostasis of the outer membrane. A mutant in this gene showed several phenotypes, such as membrane defects, altered LPS composition, reduced adhesion, and increased virulence and inflammation. We show that RomA is involved in the synthesis of LPS, probably coordinating part of the biosynthetic complex in the periplasm. Its absence alters the normal synthesis of this macromolecule and affects the homeostasis of the outer membrane, resulting in a strain with a hyperinflammatory phenotype. Our results suggest that the proper synthesis of LPS is central to maximize virulence and minimize inflammation.
Brucellaceae are widespread zoonotic intracellular pathogens with the capacity to evade and modulate the immune response of the infected hosts, a hallmark of their infectious processes [1]. Many of these immunomodulatory activities are achieved by a plethora of virulence factors that are able to manipulate the immune system to its benefit, promoting bacterial proliferation and the establishment of the chronic infectious phase [2–7]. Brucella is considered a pathogen with a stealthy strategy, meaning that is able to avoid a strong immune response by “hiding” its pathogen-associated molecular patterns (PAMPs). This strategy is achieved by either downregulating activation of PAMPs or synthesizing structural components with low proinflammatory activities [8]. This last concept raises an interesting question: Are the structural cellular components of Brucella, such as lipopolysaccharide (LPS), silent PAMPs by default, or is the bacterium actively modifying them to be nondetected?
Lipopolysaccharide is the main component of the outer membrane in all Gram-negative bacteria. It is composed of lipid A (the lipidic portion inserted in the membrane), a core oligosaccharide, and the O-antigen, which is the most exposed structure [9]. Lipopolysaccharide is necessary for a wide range of functions, such as protection against harsh environmental conditions, selective permeability, immune protection, and evasion. The synthesis of this macromolecule is achieved through a complex biosynthetic pathway that starts in the cytoplasm, where the precursors are synthesized; continues in the inner membrane and periplasmic space, where the assembly, polymerization, and transport take place; and ends when the complete molecules are inserted in the outer membrane [9]. Many of these processes require multiprotein complexes that coordinate their activities in a spatial and temporal way to effectively synthesize, transport, and insert this complex macromolecule in its final organelle [10]. In Brucella, LPS has been shown to be a central virulence factor necessary for intracellular replication and virulence in mice [8]. Additionally, it has been shown that a structurally complete LPS is needed for efficient immune evasion and that LPS acts as a shield against the innate immune response [1, 11].
We describe the identification in Brucella abortus of a gene encoding a protein with no known function but conserved among almost all α-proteobacteria, which we propose is involved in the homeostasis of the outer membrane. The 84 amino-acid protein has a periplasmic localization, and the protein’s absence results in a strain with several phenotypes: membrane alterations, changes in the LPS composition, defects in the intracellular replication capacity of the bacteria, and deregulated inflammatory response in mice. We hypothesize that the absence of this protein alters the proper synthesis of the LPS, probably modifying the assembly of the biosynthetic complex in the periplasm, which in turn affects outer membrane homeostasis and the modulation of the immune response during the infectious cycle.
METHODS
A complete description of the Methods has been included in the Supplementary Data.
Periplasmic and Cytoplasmic Localization Assay
For the periplasmic localization assays, the B. abortus strains were grown in Tryptic Soy Broth (TSB) for 16–24 hours at 37°C, 2.5 × 1010 bacterial cells were centrifuged for 10 minutes at 3300 × g, and the periplasmic and cytoplasmic fractions were obtained as previously described [12]. For Western blot, an anti-FLAG M2, anti-GroEL (1:2000), and anti–OMP-1 (1:2000) kindly provided by Dr Axel Cloeckaert were used as primary antibodies.
Detergent Sensitivity Assays
For all B. abortus strains, overnight cultures were diluted and seeded onto TSB plates supplemented with detergents as described [13]. Final detergent concentrations were 125 μg/mL of Sarkosyl, 25 μg/mL of Zwittergent 3–16, and 1 g/mL of sodium deoxycholate. Sensitivity was calculated determining the viable colony-forming units (CFUs).
Total Lipid Extraction
For the extraction of total lipids, the Bligh and Dyer method was used [14] on exponentially grown bacteria.
Cristal Violet Staining of Brucella abortus
Serial dilutions of B. abortus cultures were plated in TSB plates and incubated at 37°C until growth was observed. Plates were stained with a crystal violet solution as described [15].
LPS Purification
Lipopolysaccharide was extracted from B. abortus strains using a modification of the phenol-hot water method [16] from 250 mL of stationary cultures. Lipopolysaccharide concentration was determined by the 3-deoxy-D-manno-2-octulosonic acid method, analyzed on 12% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) gels, and stained by silver nitrate.
Gentamicin Protection Assays
Gentamicin protection assays were performed as previously described [17].
Mice Infections
Mice infections were performed as previously described [18].
Analysis of Fluorescent Fusion Proteins
Stationary phase cultures of B. abortus strains were diluted in fresh media and grown until exponential phase was reached. A volume of 3 μL of culture was seeded onto the center of a phosphate-buffered saline 1% agarose pad as previously described [19]. Images were acquired and processed as described above.
RESULTS
Bab1_1280encodes a periplasmic protein necessary for the homeostasis of the outer membrane in Brucella.
In a genetic screen to identify genes from B. abortus coding for secreted or surface-exposed proteins, we isolated Bab1_1280. This gene encodes an 84 amino-acid hypothetical protein conserved in most α-proteobacteria with no known function to date [20] (Supplementary Figure 1). We serendipitously found that an insertion mutation in this gene resulted in a strain with altered membrane properties. Because the gene was identified in a screen for either secreted/periplasmic or surface-exposed proteins, we performed subcellular fractionation assays to further determine its localization. We generated a strain expressing FLAG-tagged RomA from a genomic allele and performed a periplasmic extraction protocol [12]. Figure 1A shows that the FLAG-tagged protein product of Bab1_1280 was detected in the periplasmic/outer membrane fraction as the outer membrane protein 1 (OMP1). Presence of the protein was not due to lysis since no cytoplasmic contamination was observed (GroEL). We additionally lysed the strain and determined membrane association by ultracentrifugation, which indicated that the protein interacts with total membranes (Figure 1B). To determine the degree of this association, total membranes were resuspended in different buffers, recentrifuged, and evaluated if the protein remained associated. Figure 1C shows that only a mild wash with 10 or 50 mM of sodium phosphate was enough to partially loosen the membrane association, a condition that was completely lost with sodium chloride or detergent treatments. These results indicate that the product of Bab1_1280 is a periplasmic protein with a weak association to either the inner or the outer membranes.
The protein product of Bab1_1280 is a membrane-associated periplasmic protein. A, Cell fractions from the merodiploid strain Bab1_1280::3xFLAG were analyzed by Western blot with a monoclonal α-FLAG antibody. Monoclonal anti-Omp1 (α-Omp1) (porin) and anti-GroEL (α-GroEL) (Hsp60 homologue) antibodies were used as outer membrane and cytoplasmic controls, respectively. B, Western blot with α-OMP1 and α-FLAG antibodies on sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) of total membranes prepared from the Bab1_1280::3xFLAG strain. C, Total membranes from the Bab1_1280::3xFLAG strain were resuspended and incubated in different buffer conditions and ultracentrifuged. Pellet and supernantant fractions were analyzed by Western blot with a monoclonal α-FLAG antibody. Abbreviations: GroEL, Hsp60 homologue; HaH2PO4, sodium dihydrogen phosphate; MW, molecular weight; NaCl, sodium chloride; NP40, Nonidet P-40; Omp1, outer membrane protein; SDS, sodium dodecyl sulfate; Sn, supernatant fraction.
To determine if the product of Bab1_1280 is involved in maintaining the normal composition of the outer membrane, we measured the resistance of the mutant to the detergents Sarkosyl, Zwittergent 3–16, and sodium deoxycholate. Figure 2A shows that the Bab1_1280 mutant was less resistant to the 3 detergents. To further determine whether this increased sensitivity to detergents is related to the hydrophobicity of the membrane, we performed an N-phenyl-1-naftilamine (NPN) incorporation assay. As can be observed in Figure 2B, the hydrophobic probe was incorporated more efficiently in the mutant, which indicates that the outer membrane is more permeable to hydrophobic molecules. These alterations in the outer membrane properties of the mutant could be the consequence of differences in the phospholipids or fatty-acid profile. One and 2-dimensional thin layer chromatography using carbon 14–labeled total bacteria or extracted periplasms (phospholipids of the outer membrane) showed no differences between strains (Figure 2C and Supplemetary Figure 2). These results indicate that the alterations observed with the mutant in terms of detergent sensitivity as well as membrane hydrophobicity are not the result of a differential phospholipid composition but are probably due to differences in other components of the outer membrane. The localization of the protein, together with the defects in the membrane properties of the mutant strain, suggest that RomA might be playing a role in maintaining a degree of homeostasis of the periplasm and/or outer membrane. For these reasons, we renamed Bab1_1280 as romA, for regulator of outer membrane.
![The absence of Bab1_1280 causes pleiotropic membrane defects. A, Detergent sensitivity assays. Stationary phase cultures of either the wild-type 2308 or the ΔBab1_1280 strains were grown at 37°C and diluted to optical density (OD)600 = 1, and serial dilutions were plated in solid media containing different detergents and incubated at 37°C for colony-forming unit determination. The figure shows significant differences between the means of both strains for Sarkosyl (*P < .01), Zwittergent (*P = .006), and sodium deoxycholate (*P = .001). B, N-phenyl-1-naphtylamine incorporation assay. Stationary-phase cultures of either the wild-type 2308 or the ΔBab1_1280 strains were grown at 37°C, diluted to OD600 = 0.1, and grown at 37°C until exponential phase was reached. To measure baseline fluorescence before adding N-phenyl-1-naftilamine (NPN) (10 μM final concentration), 7.5 × 108 cells (resuspended in 250 μL of phosphate-buffered saline) were used per well in 96-well black plates, and measurements were made every 18 seconds for 5 minutes. Relative fluorescence units were calculated by dividing each value by the mean obtained for the baseline for each strain (*P < .001). In all cases (A–C), error bars are standard deviations, and P values were calculated by unpaired t tests. C, Two-dimensional thin layer chromatography (2D-TLC) of total phospholipids. Total [14C]acetate-labeled lipids were extracted from cultures in the presence of choline and analyzed by 2D-TLC and autoradiography. Lipids spots corresponding to cardiolipin, phosphatidylglycerol, ornithine lipid, phosphatidylethanolamine, and phosphatidylcholine are indicated. D, Crystal violet staining. Stationary-phase cultures of either the wild-type 2308 or the ΔBab1_1280 strains were grown at 37°C and diluted to OD600 = 1, and serial dilutions were plated in solid media and incubated at 37°C until colonies were observed. The crystal violet staining was performed as described. Abbreviations: CFU, colony-forming unit; CL, cardiolipin; DOC, sodium deoxycholate; OL, ornithine lipid; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PG, phosphatidylglycerol; RFU, relative fluorescence unit; Sarkosyl, N-lauroylsarcosine; Zwittergent, Zwittergent 3–16.](https://oup.silverchair-cdn.com/oup/backfile/Content_public/Journal/jid/217/8/10.1093_infdis_jiy002/3/m_jiy00202.jpeg?Expires=1747925302&Signature=e6wuzxR4vpZ8toq5eiWwf6JwEFNVbvpr5P0dfCcGnbfpNv5N1UkKTPWJzqn8moVOcxy-BB7NWUXs7UhgtX~UQr1L6aVPe0KD--LHgmr5Kks-fJsw9iQmV4lYM0-tHxFzND54WPyRYC2qSxF0gKdzXaKitAygerjWM8pfP6k8fw3BapDA3Rpv7kBUZeltIDp8LSibrioxmEpnhYPP91ksVhfOgAG8uCSRnpCkp0PZ~9aZdnlYkGQVrWHxCx01eXXLlc3qWKB3UOacy1BmGC2E7VAkJveOWMlzuqW6oeOLKMsQg9FQYfyj40QfSBfoe08lWaFbblAFIFHO4JQBl3jvQw__&Key-Pair-Id=APKAIE5G5CRDK6RD3PGA)
The absence of Bab1_1280 causes pleiotropic membrane defects. A, Detergent sensitivity assays. Stationary phase cultures of either the wild-type 2308 or the ΔBab1_1280 strains were grown at 37°C and diluted to optical density (OD)600 = 1, and serial dilutions were plated in solid media containing different detergents and incubated at 37°C for colony-forming unit determination. The figure shows significant differences between the means of both strains for Sarkosyl (*P < .01), Zwittergent (*P = .006), and sodium deoxycholate (*P = .001). B, N-phenyl-1-naphtylamine incorporation assay. Stationary-phase cultures of either the wild-type 2308 or the ΔBab1_1280 strains were grown at 37°C, diluted to OD600 = 0.1, and grown at 37°C until exponential phase was reached. To measure baseline fluorescence before adding N-phenyl-1-naftilamine (NPN) (10 μM final concentration), 7.5 × 108 cells (resuspended in 250 μL of phosphate-buffered saline) were used per well in 96-well black plates, and measurements were made every 18 seconds for 5 minutes. Relative fluorescence units were calculated by dividing each value by the mean obtained for the baseline for each strain (*P < .001). In all cases (A–C), error bars are standard deviations, and P values were calculated by unpaired t tests. C, Two-dimensional thin layer chromatography (2D-TLC) of total phospholipids. Total [14C]acetate-labeled lipids were extracted from cultures in the presence of choline and analyzed by 2D-TLC and autoradiography. Lipids spots corresponding to cardiolipin, phosphatidylglycerol, ornithine lipid, phosphatidylethanolamine, and phosphatidylcholine are indicated. D, Crystal violet staining. Stationary-phase cultures of either the wild-type 2308 or the ΔBab1_1280 strains were grown at 37°C and diluted to OD600 = 1, and serial dilutions were plated in solid media and incubated at 37°C until colonies were observed. The crystal violet staining was performed as described. Abbreviations: CFU, colony-forming unit; CL, cardiolipin; DOC, sodium deoxycholate; OL, ornithine lipid; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PG, phosphatidylglycerol; RFU, relative fluorescence unit; Sarkosyl, N-lauroylsarcosine; Zwittergent, Zwittergent 3–16.
RomA is Implicated in the Biosynthesis of LPS
The fact that the ΔromA mutant showed altered membrane properties but no differences in the phospholipid composition raised the possibility that it might have a defective LPS. We performed a crystal violet (CV) staining to determine whether the strain has a complete LPS (smooth strain, excludes the staining) or it lacks the assembled O-antigen (rough strain, includes the staining). Surprisingly, the ΔromA (ΔBab1_1280) strain showed a higher degree of exclusion of CV (Figure 2D), strongly suggesting a modified LPS, which was further confirmed by Western blot on whole bacteria. Figure 3A shows that the mutant strain exhibited an LPS pattern that seemed to have not only a higher antigenic load but also higher molecular weight forms. To further advance its characterization, we performed a Western blot on whole bacteria using anti–O-antigen (α-S-LPS) and antirough (α-R-LPS) monoclonal antibodies. As can be observed in Figure 3B and 3C, the ΔromA mutant showed higher levels of smooth LPS and lower levels of the rough LPS, which indicates that the strain might have an altered equilibrium of the S-LPS/R-LPS ratio in the membrane. To confirm that this alteration is present in the outer membrane and it is not a consequence of an accumulation of the S-LPS in the inner membrane, we performed a periplasmic/outer membrane extraction and analyzed the LPS. As can be observed in Figure 3B (right panel), the same pattern of S-LPS/R-LPS was observed in the outer membrane with the mutant, confirming that this strain exhibits a higher percentage of S-LPS. An additional characteristic that we observed while analyzing the gels of the ΔromA LPS was that it seemed to have longer O-antigen chains, although an alternative explanation could be that it was the result of a higher concentration of smooth LPS and not a chain length issue. To distinguish between these 2 possibilities, we characterized purified LPS from both strains by Western blot with the α-S-LPS and α-R-LPS monoclonal antibodies and silver staining. In Figure 3D it can be seen that purified LPS showed the same pattern observed either with whole cells or periplasmic extractions. These preparations were used to chemically characterize the O-antigen and the core (see Methods). Monosaccharide analysis by high-performance anion-exchange chromatography coupled with pulsed amperometric detection (HPAEC-PAD) of the oligosaccharides released from the LPSs showed that the ΔromA strain LPS presents a higher ratio of perosamines (Rha4N, present in the O-antigen) to N-acetyl-glucosamines (GlcN, present in the core), indicating that the mutant has a longer O-antigen (Figure 4A and Supplementary Figure 3).
RomA is required for lipopolysaccharide (LPS) homeostasis. A, Whole cell extracts of the wild-type 2308 and ΔromA strains were analyzed by Western blot using a rabbit polyclonal α-Brucella antibody. Monoclonal α-GroEL, α-Omp1, α-Omp16, and α-Omp19 antibodies were used for loading controls. B, Whole cells and periplasmic fractions were analyzed by Western blot using an α-Smooth LPS (S-LPS) or an α-Rough LPS (R-LPS). C, Densitometry of gels in part B. D, Lipopolylsaccharide was extracted from either the wild-type 2308 or the ΔromA strains and analyzed by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and Western blot using the α-S-LPS and α-R-LPS antibodies (left panel), or SDS-PAGE and LPS silver staining (right panel). Abbreviations: α-GroEL, anti-GroEL; α-Omp1, anti-outer membrane protein 1; α-Omp16, anti-outer membrane protein 16; α-Omp19, anti-outer membrane protein 19; MW, molecular weight; R-LPA, rough lipopolysaccharide; S-LPS, smooth lipopolysaccharide;
RomA is required for controlling the O-antigen length. A, Aminosugar analysis by High-Performance Anion-Exchange Chromatography coupled with Pulsed Amperometric Detection (HPAEC-PAD) of the oligosaccharides released after acid hydrolysis of the lipopolylsaccharide (LPS) of the wild-type 2308 or the ΔromA strains. Comparison of the perosamine/glucosamine (Rha4N/GlcN) stoichiometry ratio in both strains. B, Lipopolysaccharide was purified from the wild-type 2308 ΔromA mutant strains, and the O-antigen was released by acid hydrolysis. The released oligosaccharides were analyzed by mass spectrometry (MS) in the positive ion mode, which showed that the mutant strain has up to 25 perosamine subunits (Rha4N) per O-antigen chain (m/z = 4835.5) in comparison with the 14 subunits of the wild-type strain (m/z = 3100.0). An inset of the wild-type strain in the range of m/z = 4200–5000 shows the absence of the peak corresponding to an O-antigen with 25 subunits of perosamine. Abbreviation: a.u., arbitrary units; Glc, glucose; GlcN, glucosamine; Fo, formyl groups; Kdo, keto-deoxyoctulosonate; Man, mannose; Quin, quinosamine; Rha4N, perosamine.
Furthermore, when the oligosaccharides released after mild acid hydrolysis of the corresponding LPSs were characterized by Matrix-Assisted Laser Desorption/Ionization Time-Of-Flight mass spectrometry (MALDI-TOF m.s) in the positive mode, differences in the high mass range were detected. Thus, the spectrum corresponding to the wild-type oligosaccharide presented the highest mass signals at m/z 3077.9 and m/z 3100.0 (ΔNa) (calculated [calc.] m/z 3099.2977; C122H210N15Na2O73), corresponding to a structure bearing KdoGlcQuinMan2Rha4N14Fo4Na2. In accordance, signal at m/z 2858.2 (calc. m/z 2857.2574; C114H199N15NaO66) corresponds to the structure GlcQuinMan2Rha4N14Fo4Na. On the other side, the spectrum of the mutant strain showed not only the signals described above but also signals at m/z 4349.3 (calc. m/z 4349.8865; C173H299N23NaO101), attributed to a structure bearing KdoGlcQuinMan2Rha4N22Fo8Na, and m/z 4325.3 (calc. m/z 4325.8630; C173H296N23Na2O99), attributed to anhKdoGlcQuinMan2Rha4N22Fo7Na2. Furthermore, a signal at m/z 4835.5 (calc. m/z 4835.0851; C193H331N26Na2O111) corresponds to the latter with 3 additional Rham4N and 2 Fo groups (Figure 4B). The fact that the ratio of Rha4N to GlcN in the mutant is increased 5 times and the O-antigen is almost 2-folds longer confirms that the mutant has an altered smooth-to-rough ratio. Despite the O-antigen being longer, the core showed no differences (data not shown).
Altogether these results indicate that RomA affects membrane properties and composition and is required for proper LPS assembly.
RomA Is Involved in the Virulence Process and Its Inactivation Profoundly Alters the Inflammatory Response
The changes in the LPS profile in the ΔromA mutant strain led us to evaluate the potential role of this gene in the virulence of B. abortus. Figure 5A shows that the ΔromA strain exhibited a significant defect in the intracellular survival capacity in murine bone marrow–derived macrophages during the initial stages of infection (4 and 24 hours after infection) but was able to replicate, and at 48 hours after infection, we did not observe any differences with the wild-type parental strain (P = .002 and P = .005). This intracellular replication pattern was similar when J774 A.1 cells were used (Supplementary Figure 4). Because of the altered membrane of the ΔromA mutant, we further analyzed whether these early effects were due to a defect in the adhesion of the bacteria to the cells or in their reduced capacity to exclude the lysosomal marker Lamp-1 during the intracellular trafficking. As can be seen in Supplementary Figures 5 and 6, the mutant showed a statistically defect in the adhesion to J774 A.1 cells as well a reduced capacity to exclude Lamp-1 at 24 hours after infection. These results indicate that the altered membrane structure of the ΔromA mutant probably impacts on several steps of the interaction of the bacterium with the host cells (adhesion/invasion as well as in the intracellular trafficking).
The results obtained in vitro encouraged us to evaluate the role this gene might play during pathogenesis in the mouse model. Surprisingly, the ΔromA mutant showed a dramatic increase in the number of bacteria in the spleens of intraperitoneally infected mice at 14 days after infection, which correlated with an increased splenomegaly (Figure 5B–D). An interesting observation was that the complemented strain exhibited less bacterial load and splenomegaly than the wild-type strain, which led us hypothesize that it might also have an altered LPS. Supplementary Figure 7 shows that the ΔromA complemented strain, expressing RomA from a plasmid, had an LPS with less assembled O-antigen (partly rough).
RomA is important for the intracellular cycle, and its absence results in a hiperinflammatory strain. A, Intracellular multiplication of wild-type 2308, ΔromA, and ΔromA complemented strains in bone marrow–derived macrophages (BMDMs). *P = .005 and **P = .002. Error bars are standard deviations, and P values were calculated by unpaired t test. B, Spleens extracted from the infected mice were homogenized for colony-forming unit determination by direct plating 15 days after infection. *P < .0001. Error bars are standard errors of the mean, and P values were calculated by unpaired t test. C, Spleen weight before homogenization. ***P = .0001 and ****P < .0001. Error bars are standard errors of the mean. D, Comparative sizes of spleens of mice intraperitoneally infected with 1 × 105 colony-forming units per animal of the wild-type 2308, ΔromA, and ΔromA complemented strains 15 days after infection. Abbreviation: CFU, colony-forming unit.
The increase in the spleen size was the result of an enhanced inflammatory response. Figure 6A–D shows that the mutant induced a higher inflammatory response in comparison with the wild type, measured by the production and circulation of 2 proinflammatory cytokines (tumor necrosis factor α and interferon γ). This was also confirmed by histological observation of the spleens, which showed a higher cellularity of the red pulp with more macrophages and neutrophils, as well as more pronounced granulomatous lesions in the mutant (Supplementary Figure 8). Additionally, Supplementary Figure 9 shows that the ΔromA mutant triggered an increased inflammatory response as early as 48 hours after infection, indicating that the strain per se is more proinflammatory and that the robust inflammation it induced was not the consequence of an increased bacterial proliferation. To further analyze the kinetics of the infection in mice, we determined spleen colonization of the mutant in comparison with the wild-type parental strain at 7 and 42 days after infection. We did not observe differences at these 2 time points (Supplementary Figure 10), indicating that the increased bacterial load is a phenomenon restricted to the acute phase of the infection and it does not persist during the chronic phase. At both time points we observed enhanced splenomegaly (not shown).
The mutant ΔromA has an increased inflammatory response. A–D, Inflammatory cytokines determined in the infected animals at 15 days after infection. Interferon gamma and tumor necrosis factor alpha levels were measured in spleens (A and C) and sera (B and D) by enzyme-linked immunosorbent assay as described. In all cases, significant differences were found between the means of the wild-type and mutant strains and between the means of the mutant and the complemented strain. No differences were found between the wild-type and complemented strain. A, *P = .03 and ***P = .009; B, *P = .02 and **P = .001; C, **P = .004 in both groups; D, **P = .001 in both cases. Error bars are standard errors of the mean, and P values were calculated using ordinary 1-way analysis of variance multiple comparisons between groups. Abbreviations: IFN- γ, interferon γ; TNF-α, tumor necrosis factor α.
Absence of RomA Alters the Positioning of LptD, an LPS Biosynthetic Protein in the Periplasmic Space
The localization of RomA as well as the membrane-related phenotypes found in the ΔromA mutant suggest that this protein might be involved in the organization/localization of LPS biosynthetic complexes in the periplasmic space and/or outer membrane. To evaluate whether this was the case, we constructed a fusion of the gene lptD that codifies for a protein involved in the transport of the LPS to the outer membrane [10, 21] with super-folder GFP (sfGFP) [22] and evaluated its localization in the wild-type and ΔromA strains. As can be observed in Figure 7A, LptD showed an altered distribution in the mutant in comparison with the parental strain. More specifically, whereas most of the wild-type cells showed a single localization spot, the mutant cells had a more homogenous distribution. This was not observed with the inner membrane protein responsible for flipping the O-antigen lipid intermediate to the periplasm, RfbD [9] (Figure 7B), indicating that only some of the LPS biosynthesis proteins have an aberrant distribution in the mutant.
RomA participates in the positioning of LptD, a lipopolysaccharide (LPS) transport protein. A, The outer membrane LPS transporter LptD was expressed with a C-terminal sfGFP from a pBBR4 plasmid in the Brucella abortus 2308 and ΔromA strains. Bacteria were grown at 37°C until exponential phase, placed on agarose pads, and analyzed by fluorescence confocal microscopy. B, The inner membrane O-antigen flippase RfbD was expressed with a C-terminal EYFP from pTRC-EYFP in the B. abortus 2308 and ΔromA strains. Bacteria were grown at 37°C until exponential phase, placed on agarose pads, and analyzed by fluorescence confocal microscopy. Abbreviations: EYFP, yellow-fluorescent protein; LptD, LPS-transport D; RfbD, flippase; sfGFP, super-folder green-fluorescent protein.
DISCUSSION
In this study we have identified a novel gene (romA) in Brucella and conserved in almost all α-proteobacteria that codes for a small periplasmic protein with no known function. A mutant in romA is pleiotropic and displays several phenotypes, all related with an altered periplasm and/or outer membrane. The sensitivity of the mutant to several detergents and the fact that its membrane is more permeable to hydrophobic compounds indicated an altered outer membrane. Lipopolysaccharide analysis showed that the mutant strain has several modifications. It has an altered smooth-to-rough ratio, which results in a strain with more assembled LPS compared with the wild-type strain. Additionally, the LPS has an O-antigen with a higher degree of polymerization, with at least twice the amount of perosamines but substituted with formyl residues. These modifications have several implications and raise interesting questions for future studies. To our knowledge this is the first report of a mutant that has an altered smooth/rough LPS ratio, strongly suggesting that Brucella (and probably other members of this group) controls this equilibrium. It is tempting to speculate that a certain level of incomplete LPS is necessary to assemble or expose other outer membrane components that could be affected if the O-antigen is present in all LPS molecules. If this hypothesis is correct, the amount of smooth-to-rough LPS could be determined by a compromise between 2 needs: to protect against harmful conditions encountered in the environment and to allow the assembly and positioning of a set of proteins or supramolecular structures necessary for motility, virulence, attachment, and protein secretion, among other actions. A similar hypothesis has been postulated for the length of the O-antigen and the type III secretion system in Shigella [23]. The authors proposed that the length of the O-antigen is determined by 2 opposing necessities: its protective properties and the efficiency of the type III injectisome. Currently we do not have a molecular explanation for why the ΔromA mutant has a longer O-antigen but speculate that the stoichiometry of the LPS biosynthetic machinery is probably altered and that this affects the synthesis. In this regard, we have shown that LptD, involved in the transport of LPS to the outer membrane [9], but not the flippase located in the inner membrane, showed a mislocalization in the mutant, strongly suggesting that RomA participates in the organization/assembly of this machinery in the periplasm and/or outer membrane.
One possible hypothesis is that the modified LPS in the romA mutant probably disturbs the homeostasis of the outer membrane, and this impacts the virulence of the bacterium in several ways. For example, even though the mutant was less infective in the cellular model of infection, it triggered an exacerbated inflammatory response that actually increased the bacterial load in the spleens of infected mice during the acute phase of the infectious process. This inability to tune down the inflammatory response, a hallmark of the Brucella infection [1], is probably the result of a combination of factors, both structural and functional, although we cannot completely rule out at this stage that the modified LPS is the only component that could account for all of the phenotypes. It has been described that a mutant in the fliC gene in Brucella that codes for a component of a flagellar-like structure induced increased splenomegaly and showed higher bacterial loads in the spleens as well as more tissue damage, similar to that observed with the ΔromA mutant [24]. Because the flagellum is assembled in the outer membrane and because we have observed that the expression of fliC is not affected (not shown), it could be speculated that the altered LPS affects the assembly of this structure and this results in a similar phenotype to the fliC null mutant.
The fact that the mutant induces severe inflammation during the acute phase of the infectious process indicates that Brucella has active mechanisms to modify its cellular structure to tune down the immune response. This is not trivial because it has been suggested that the default cellular structure of Brucella is mainly responsible for its stealthy strategy and it implies that the bacterium could actually modulate up and down the inflammatory response depending on the phase of the infectious cycle or its needs, modifying its membrane composition or structure. It could be speculated that the inflammatory balance and the necessity of the bacteria to assemble membrane structures needed for virulence must be tightly equilibrated to establish a successful chronic infection. In this view, Brucella should be able to finely counterbalance these 2 needs to be a successful pathogen.
Supplementary Data
Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.
Notes
Acknowledgments. We thank members of the J.E.U. laboratory for useful discussions and Thomas Bernhardt from Harvard University for kindly providing the superfolder GFP coding plasmid. C. C., S. G. A., A. S. C., K. A. P., J. M. S., J. C. and J. E. U. are members of the National Research Council of Argentina (CONICET).
Financial support. This work was supported by grants PICT-PRH08-160 to C. C., PICT-PRH08-230 and PICT-1028-2014 to J. E. U., and grants from the University of San Martín to C. C. and J. E. U. E.V. was supported by a fellowship of the National Research Council of Argentina (CONICET).
Potential conflicts of interest. All authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.
