Brucella abortus Encodes an Active Rhomboid Protease: Proteome Response after Rhomboid Gene Deletion

Rhomboids are intramembrane serine proteases highly conserved in the three domains of life. Their key roles in eukaryotes are well understood but their contribution to bacterial physiology is still poorly characterized. Here we demonstrate that Brucella abortus, the etiological agent of the zoonosis called brucellosis, encodes an active rhomboid protease capable of cleaving model heterologous substrates like Drosophila melanogaster Gurken and Providencia stuartii TatA. To address the impact of rhomboid deletion on B. abortus physiology, the proteomes of mutant and parental strains were compared by shotgun proteomics. About 50% of the B. abortus predicted proteome was identified by quantitative proteomics under two experimental conditions and 108 differentially represented proteins were detected. Membrane associated proteins that showed variations in concentration in the mutant were considered as potential rhomboid targets. This class included nitric oxide reductase subunit C NorC (Q2YJT6) and periplasmic protein LptC involved in LPS transport to the outer membrane (Q2YP16). Differences in secretory proteins were also addressed. Differentially represented proteins included a putative lytic murein transglycosylase (Q2YIT4), nitrous-oxide reductase NosZ (Q2YJW2) and high oxygen affinity Cbb3-type cytochrome c oxidase subunit (Q2YM85). Deletion of rhomboid had no obvious effect in B. abortus virulence. However, rhomboid overexpression had a negative impact on growth under static conditions, suggesting an effect on denitrification enzymes and/or high oxygen affinity cytochrome c oxidase required for growth in low oxygen tension conditions.


Introduction
Rhomboid proteases are a superfamily of intramembrane proteins highly widespread and conserved in the three domains of life. Rhomboid active sites are embedded in the membranes, where they cleave their substrates in or close to the transmembrane domains (TMDs). In eukaryotes, they have been implicated in a variety of processes including epidermal growth factor signaling in Drosophila melanogaster [1], lipid metabolism [2], energy production [3], chloroplast development [4], apoptosis regulation [5], endoplasmic reticulum (ER) protein trafficking [6] and surface antigen shedding in the apicomplexan parasites [7].
In bacteria, rhomboid proteases are highly conserved and ubiquitous, suggesting a relevant role in physiology [8,9]. However, the relevance and function of rhomboid proteases in bacteria remain poorly investigated. In Providencia stuartii, the rhomboid protease AarA cleaves an N-terminal fragment of TatA, a membrane-bound component Microorganisms 2022, 10,114 2 of 20 of the twin-arginine translocase secretion apparatus. Processing of TatA activates the translocation process, allowing the export of an unknown quorum sensing signal [10]. In Mycobacteria, rhomboid deletion mutants display altered colony morphology, impaired biofilm formation and increased antibiotic sensitivity [11]. In the archaeon Haloferax volcanii, a rhomboid protease is involved in protein glycosylation of the S-layer [12,13], but the implicated molecular mechanism has not been uncovered yet. The best functionally and structurally-characterized is Escherichia coli rhomboid protease, GlpG [14][15][16]. A recent study links the survival of a pathogenic strain in the mouse gut to the function of GlpG [17], yet the mechanism has not been elucidated. In Gram-positive bacteria, it was postulated that YqgP plays a role in cell division and may be required for glucose export [18]. In addition, it was recently demonstrated that Bacillus subtilis YqgP cleaves the high-affinity magnesium transporter MgtE and that YqgP interacts with the membrane-bound metalloprotease FtsH in order to maintain bacterial magnesium homeostasis [19]. Another recent study identified a new role for Shigella sonnei GlpG and Rhom7, which are involved in quality control of membrane orphan proteins of respiratory complexes in this Gramnegative bacterium [20].
Brucella spp. are intracellular pathogenic bacteria responsible for brucellosis, a chronic and debilitating zoonotic disease. Brucellosis manifestations include sterility and abortion in animals, and non-specific clinical signs like undulating fever, joint pain and fatigue in humans. Consumption of contaminated dairy products or contact with infected animals are the main routes of infection in humans. Brucellosis remains endemic in most of the developing countries with significant impact on economy and on human and animal health [21].
Brucella pathogenesis relies on its ability to replicate in various mammalian cell types [22]. At the onset of the infection, Brucella infects and replicates inside phagocytic cells such as macrophages and dendritic cells. Next, bacteria disseminate to placental trophoblasts (in pregnant females), and to organs of the reproductive tract and the mononuclear phagocyte system to establish a persistent infection in the host [23,24].
The role of membrane proteases in Brucella physiology is poorly characterized. Lon is the most studied membrane-residing protease of Brucella. Several roles have been demonstrated for this protease: it is involved in the modulation of macrophage cytokine response at early infection stages [25]; it functions as a stress response protease [26]; and it has a regulatory role recently identified by comparative RNA-seq analysis [27]. Stress-response proteases have been better characterized. B. abortus HtrA is a periplasmic stress-response protease involved in resistance to hydrogen peroxide and to the antibiotic puromycin [28]. In B. suis, ClpA is involved in bacterial clearance from infected BALB/c mice and in temperature-dependent growth regulation [29]. ClpB has also been characterized as a stress-response protease involved in sensitivity to ethanol stress, high temperature, and acidic pH in B. suis. [30].
To explore the biological role of rhomboid in Brucella, we demonstrate that it is an active protease and then constructed a deletion mutant in B. abortus. In this work, a quantitative high-throughput comparative proteomics approach was applied to understand the importance of rhomboid in B. abortus physiology and to identify putative substrates of this intramembrane protease.

Bacterial Strains and Growth Conditions
Bacterial strains and plasmids used in this study are listed in Table 1. B. abortus strains were inoculated in tryptic soy agar (TSA) or in tryptic soy broth (TSB). Liquid cultures and plates were incubated at 37 • C, ON (200 rpm) or for 72 h, respectively. All work with live B. abortus was performed in a biosafety level 3 laboratory facility. E. coli strains were grown in Luria Broth (LB), at 37 • C ON. Antibiotics were added to the indicated concentrations: 50 µg/mL kanamycin, 5 µg/mL nalidixic acid, 50 µg/mL ampicillin (for B. abortus strains) and 100 µg/mL ampicillin (for E. coli strains).

Construction of B. abortus ∆rhoX
To obtain the rhomboid mutant strain by unmarked gene deletion, the regions flanking the rhomboid gene (bab1_1275) were amplified and ligated using recombinant PCR technique [36]. The primers used for PCR amplification of the 500 bp upstream and downstream regions are Rho_UP_Fw CAGAATTCCGAGAAAGAAGAGGTCGCAG-Rho_compl_Rv cggcaggcgccccttACGGT and Rho_DW_Rv ATGGATCCTTCACAACGTCACCGATTGA-Rho_compl_Fw aaggggcgcctgccgATCCTGCGCCATTTCTGGCT, respectively (complementary sequences in lowercase). DNA fragments amplified by PCR were used in an overlapping PCR. The amplified fragments were ligated into pK18mob-sacB vector [32]. Plasmids were transformed in E. coli S17λpir and transferred to B. abortus 2308 by biparental mating [37]. After selection and counterselection of double recombination events, gene deletion was confirmed by PCR.

Plasmid Constructions
Genetic complementation of the mutant strains was achieved by expression of C-terminal 3xFlag-tagged versions of rhomboid protein from plasmid pBBR1-MCS4-3xFlag (pLF) [34]. DNA fragments coding for rhomboid were amplified using FwRho CAGGATCCCTTCC-GAAAGACAAGAGAGAG and RvRho GCACTAGTTCGCCTCAAAACGGCAT primers containing BamHI and SpeI sites, respectively. The PCR product was digested with BamHI and SpeI, and the resulting fragment cloned into the same sites of pLF to generate in-frame fusions to the 3xFLAG epitope under lac promoter control. The resulting construction, designated pLF_rho, was introduced in B. abortus rhomboid mutant by biparental mating.
Plasmid pLF_rho S153A was obtained by site-directed mutagenesis following the instructions provided by the manufacturers (Quick Change R Site Directed Mutagenesis, Stratagene). Briefly, primers Rho (Ser153/Ala) Fw CGCTGGTGGGTGCGGCGGGCGC-GATTTCG and Rho (Ser153/Ala) Rv CCCGAAATCGCGCCCGCCGCACCCACCAG were used in PCR amplification using pLF_rho as a template. PCR products obtained were transformed into E. coli K12-DH5α F'lq (Kan R ) and colonies were cultured to purify plasmidic DNA. Plasmids were sequenced to detect the point mutation Ser (TCG) to Ala (GCG). The plasmid obtained pLF_rho S153A was transferred to B. abortus by biparental mating.
The plasmid encoding the rhomboid fusion to enhanced yellow fluorescent protein (EYFP) was obtained as follows: rhomboid gene was amplified using primers FwRho CAGGATCCCTTCCGAAAGACAAGAGAGAG and RvRho GCACTAGTTCGCCTCAAAA CGGCAT, containing BamHI and SpeI sites, respectively. The PCR product was digested with BamHI and SpeI, and the resulting fragment cloned into the same sites of pTrc-EYFP [35], yielding pTrc-rho_EYFP. For microscopy analysis, the indicated plasmids were introduced into B. abortus 2308 by biparental mating using the E. coli S17-λpir strain.

Microscopy Analysis
Exponential-phase cultures of B. abortus 2308 expressing rhomboid gene as a fusion to EYFP were placed on a microscope slide layered with a pad of 1% agarose in phosphate buffered saline (PBS) as described in [35]. Samples were examined on an IX81 microscope with an Olympus FV1000 confocal module (60× PLAPO objective, numerical aperture [NA] of 1.42). Images were processed with the Image J program (NIH, Bethesda, MD, USA).

Substrate Preparation for Protease Assay
To prevent possible endogenous rhomboid cleavage, E. coli MG1655 ∆glpEGR::kan with the plasmids encoding the heterologous substrates [10] were used. Cell extracts were prepared according to [12].

In Vitro Protease Assay
B. abortus membranes were prepared as described in [35]. For in vitro protease assay, substrate preparations (100 µg) were incubated for 18 h with B. abortus membrane fractions (100 µg) and samples were further processed as described in [12].

Protease Assays in B. abortus
Substrate cleavage assays were performed under aerobic conditions. B. abortus strains harboring pBBR2-gurken or pBBR2-tatA plasmids were grown ON at 37 • C (250 RPM). Bacteria were then harvested for SDS-PAGE and Western blotting. Complemented strains carried simultaneously two plasmids (derived from pBBR1-MCS2 and pBBR1-MCS4) that were maintained with different antibiotic resistance.

SDS-PAGE and Western Blot
Samples were boiled for 5 min before loading on SDS-PAGE gels. Proteins were transferred to nitrocellulose membranes and blots probed with anti-His antibody (Invitrogen). IR Dye secondary antibodies were used for detection on the Odyssey Infrared Imaging System.

SDS-PAGE
Thirty µg of each sample were loaded onto polyacrylamide 10% (w/v) gels. The gels were run at RT (20 mA) until samples were concentrated into one protein band in the separation gel. Proteins were visualized with a Coomassie brilliant blue (CBB-G250) stain [49].

In-Gel Tryptic Digestion
Excision of proteins from gel bands, then cutting into pieces of about 1 × 1 mm size, de-staining of gel was carried out as described [50]. Protein reduction and alkylation was done according to [51]. Tryptic peptide digest was obtained following procedure in [52]. Upon drying in SpeedVac, peptides were stored at −20 • C. For mass spectrometry, the peptides were re-dissolved in 20 µL 0.1% formic acid in LC-MS/MS grade water, with 2% (v/v) acetonitrile (buffer A). For analysis, 8 µL of peptides were injected.

Protein Identification and Quantification
Protein identification was performed by MaxQuant version 1.5.3.175 with Andromeda [53] and quantified with the LFQ algorithm [54] embedded in MaxQuant searching against the complete proteome database of B. abortus 2308 (Uniprot taxon identifier 359391). After quantification, intensities from the LFQ normalization were filtered and compared with Perseus (v1.5.8.5) [55], removing rows with fewer than two values under either condition. Search parameters were fully specific cleavage by trypsin with no more than two missed cleavages: carbamidomethylation of cysteine (static), oxidation of methionine (dynamic), glutamine (Gln) to pyro-glutamate (pyro-glu) at protein N-terminus. A protein was considered as significantly regulated at a p value ≤ 0.05 (Student's t-test). Mass spectrometry raw data are included in Supplementary Table S1.

Automated Growth Curves
Overnight TSB cultures of the different strains were diluted (1/100) in TSB to the same initial OD 600 . For shaking experiments, 100 µL of diluted cultures were plated at least by triplicate in 96-well microplates while 250 µL of diluted cultures were plated for static experiments, in order to obtain a low-oxygenated condition. Growth curve experiments were performed using a TECAN Infinite Sunrise microplate reader (Thermo). OD600 nm was measured every 30 min at 37 • C.

Identification of a Rhomboid Protease in B. abortus
To examine the occurrence of rhomboid in Brucella, the annotated sequenced genomes were manually verified using the BLAST program, available on the NCBI website. The in silico analysis showed that 292 annotated Brucella genomes encode a rhomboid-like protein. A unique genome organization at the rhomboid locus (bab1_1275 in B. abortus 2308) across the genus is observed, with many of the rhomboid surrounding genes conserved. Typically, flanking genes encode a cystathione β-synthase (CBS) domain-containing protein (bab1_1274 in B. abortus 2308) and a conserved hypothetical protein containing a PAS domain (bab1_1277 in B. abortus 2308). bab1_1274 and bab1_1275 are separated by 171 bp and they are predicted to be in the same operon (the probability is 0.628, according to the tool described in methods). Rhomboid proteases are highly conserved across the genus, except for a few amino acid substitutions in B. suis, B. vulpis, B. canis and B. inopinata ( Figure 1).

Localization of Rhomboid Protease in B. abortus
Four different topological classes for rhomboid-like proteins have been proposed [56]. The basic class consists of a six-transmembrane domain core. It is found in E. coli GlpG and some eukaryotic rhomboids such as Saccharomyces cerevisiae Rbd2 (YPL246C) [57][58][59]. The next class has an extra TMD fused to the C terminus and a variable N-terminal domain [1]. Drosophila Rhomboid-1 and many bacterial rhomboids have a 6 + 1 TMD structure. The third class contains a globular domain inserted into first loop and some modifications in the active site. Finally, the PARL-subfamily (Presenilin-Associated-Rhomboid-Like) has a predicted extra TMD fused to the N terminus of the rhomboid core, thus the catalytic residues are inserted in TMD5 and TMD7, instead of TMD4 and TMD6 in other rhomboids.
Based on several bioinformatic tools (HMMTOP, Phobius and TMHMM), B. abortus rhomboid is predicted to comprise seven TMDs (Figures 1 and 2A). According to the topological classification, it belongs to the fourth class (1 + 6 TMD) together with the PARL subfamily of mitochondrial rhomboid proteases. The first sixteen to twenty-seven amino acids of the protein (depending on the species) are predicted to be in the extracellular space/periplasm, whereas the C-terminus is predicted to be cytoplasmic. The serine (GXSG) and histidine residues typical of the catalytic dyad of this protease family [60] are conserved in TMD5 and TMD7, respectively (Figures 1 and 2A).  Figure 1).

Localization of Rhomboid Protease in B. abortus
Four different topological classes for rhomboid-like proteins have been proposed [56]. The basic class consists of a six-transmembrane domain core. It is found in E. coli GlpG and some eukaryotic rhomboids such as Saccharomyces cerevisiae Rbd2 (YPL246C) We further predicted the 3D structural model of rhomboid with the help of I-TASSER server. The crystallized synthetic protein E. coli GlpG (pdb 2iC8) [61] was ranked as the best template with a normalized Z-score of 2.14 for the alignment and 72% coverage. The C-score, which gives the confidence for the quality of the predicted models, is −0.83, and the TM-score is 0.61 ± 0.14 [45]). The predicted 3D structure ( Figure 2B) contains seven transmembrane segments supporting B. abortus rhomboid as a transmembrane protein homologous to other proteins of the rhomboid family.
in other rhomboids.
Based on several bioinformatic tools (HMMTOP, Phobius and TMHMM), B. abortus rhomboid is predicted to comprise seven TMDs (Figures 1 and 2A). According to the topological classification, it belongs to the fourth class (1 + 6 TMD) together with the PARL subfamily of mitochondrial rhomboid proteases. The first sixteen to twenty-seven amino acids of the protein (depending on the species) are predicted to be in the extracellular space/periplasm, whereas the C-terminus is predicted to be cytoplasmic. The serine (GXSG) and histidine residues typical of the catalytic dyad of this protease family [60] are conserved in TMD5 and TMD7, respectively (Figures 1 and 2A).  To ascertain rhomboid intramembrane localization, a B. abortus strain expressing rhomboid as a fusion to yellow fluorescent protein (EYFP) was obtained. As shown in Figure 2C, the fusion protein is localized to the bacterial membrane. Supporting this result, Western blot analysis of total membrane preparations of a strain expressing rhomboid as a fusion to 3xFLAG epitope, revealed a membrane association of the fusion protein, indicating that it is either an inner or an outer membrane-associated protein ( Figure 2D). Further periplasmic extraction experiments showed that the protein is not associated to the periplasm or outer membrane, confirming its association to the inner membrane ( Figure 2D).

B. abortus Rhomboid (RhoX) Is an Active Protease
In the absence of a known substrate, the identification of active rhomboid proteases has been possible due to the capability of rhomboid proteases from evolutionarily divergent organisms to recognize and cleave a common set of substrates [1,14,[62][63][64][65]. To determine if B. abortus rhomboid (RhoX hereafter) is an active protease, we tested the in vitro proteolysis of D. melanogaster Gurken and P. stuartti TatA, two model substrates for rhomboid proteases. These substrate analogues are chimeric proteins that consist of E. coli maltose binding protein (MBP) with a signal peptide (SP) at the N-terminus, followed by the TMD of Gurken or TatA protein, and a thioredoxin (Trx) domain and His tag at the C-terminus [14] ( Figure 3). These model substrates was expressed in E. coli MG1655 ∆glpEGR::kan, and cell extracts were incubated with solubilized membranes from B. abortus wild type (2308) and a deletion mutant in rhomboid gene (∆rhoX). As shown in Figure 3A,B, the model substrates were cleaved in the presence of B. abortus wild type membranes, releasing añ 18 kDa-His tagged C-terminal fragment, whereas the proteolytic activity was undetectable in the presence of B. abortus ∆rhoX membranes. The rhomboid activity was restored in membranes of the B. abortus ∆rhoX (prhoX) complemented strain and only traces of activity were detected in membranes of B. abortus ∆rhoX (prhoX S153A), complemented with a RhoX protein with a substitution of one of the residues of the catalytic dyad.
boid as a fusion to 3xFLAG epitope, revealed a membrane association of the fusion protein, indicating that it is either an inner or an outer membrane-associated protein ( Figure  2D). Further periplasmic extraction experiments showed that the protein is not associated to the periplasm or outer membrane, confirming its association to the inner membrane ( Figure 2D).

B. abortus Rhomboid (RhoX) Is an Active Protease
In the absence of a known substrate, the identification of active rhomboid proteases has been possible due to the capability of rhomboid proteases from evolutionarily divergent organisms to recognize and cleave a common set of substrates [1,14,[62][63][64][65]. To determine if B. abortus rhomboid (RhoX hereafter) is an active protease, we tested the in vitro proteolysis of D. melanogaster Gurken and P. stuartti TatA, two model substrates for rhomboid proteases. These substrate analogues are chimeric proteins that consist of E. coli maltose binding protein (MBP) with a signal peptide (SP) at the N-terminus, followed by the TMD of Gurken or TatA protein, and a thioredoxin (Trx) domain and His tag at the Cterminus [14] (Figure 3). These model substrates was expressed in E. coli MG1655 ΔglpEGR::kan, and cell extracts were incubated with solubilized membranes from B. abortus wild type (2308) and a deletion mutant in rhomboid gene (ΔrhoX). As shown in Figure  3A,B, the model substrates were cleaved in the presence of B. abortus wild type membranes, releasing an ~18 kDa-His tagged C-terminal fragment, whereas the proteolytic activity was undetectable in the presence of B. abortus ΔrhoX membranes. The rhomboid activity was restored in membranes of the B. abortus ΔrhoX (prhoX) complemented strain and only traces of activity were detected in membranes of B. abortus ΔrhoX (prhoX S153A), complemented with a RhoX protein with a substitution of one of the residues of the catalytic dyad. were incubated with the solubilized membranes. After incubation, samples were TCA-precipitated and used in SDS-PAGE and Western blot with anti-His antibody. The bands corresponding to unprocessed and processed substrates are pointed out by arrow heads. Migration of molecular mass markers (kDa) is indicated in each blot. On the left, a schematic representation of the chimeric protein substrates is shown. 2308: wild type; ∆rhoX: rhomboid deletion mutant; ∆rhoX + prhoX: rhomboid deletion mutant complemented with wild type rhomboid; ∆rhoX + prhoXS153A: rhomboid deletion mutant complemented with rhomboid mutant S153A.
It was previously described that some rhomboid proteases are inhibited by the general serine protease inhibitor 3,4-dichloroisocoumarin (3,4-DCI) [1]. To test if B. abortus rhomboid activity is impaired by this inhibitor, we performed the in vitro assay in the presence of DCI with two rhomboid chimeric substrates: Gurken TMD and P. stuartii TatA TMD.
Rhomboid activity towards Gurken and TatA chimeric substrates was completely abolished in the presence of 400 mM DCI ( Figure 3B).
To analyze if rhomboid proteolytic activity also occurs in vivo, we expressed two rhomboid chimeric substrates, Gurken and TatA, under constitutive promoters in B. abortus, and whole cell extracts were analyzed by Western blot. As shown in Figure 4, both substrates were cleaved when expressed in B. abortus 2308 but not in B. abortus ∆rhoX background. As shown in vitro, B. abortus ∆rhoX (prhoX) complementation was able to restore the proteolytic activity in vivo. These results demonstrate that RhoX is active towards the model substrates in vivo.
It was previously described that some rhomboid proteases are inhibited by th eral serine protease inhibitor 3,4-dichloroisocoumarin (3,4-DCI) [1]. To test if B. a rhomboid activity is impaired by this inhibitor, we performed the in vitro assay presence of DCI with two rhomboid chimeric substrates: Gurken TMD and P. stuarti TMD. Rhomboid activity towards Gurken and TatA chimeric substrates was comp abolished in the presence of 400 mM DCI ( Figure 3B).
To analyze if rhomboid proteolytic activity also occurs in vivo, we expresse rhomboid chimeric substrates, Gurken and TatA, under constitutive promoters in B tus, and whole cell extracts were analyzed by Western blot. As shown in Figure 4 substrates were cleaved when expressed in B. abortus 2308 but not in B. abortus Δ background. As shown in vitro, B. abortus ΔrhoX (prhoX) complementation was a restore the proteolytic activity in vivo. These results demonstrate that RhoX is acti wards the model substrates in vivo.

Identification of Putative RhoX Substrates by Label-Free Quantitative Proteomics
Having demonstrated that RhoX is an active protease, we sought to identify pu substrates by a label-free quantitative proteomics approach. We also intended to ad metabolic and signaling pathways altered in response to rhomboid deletion. To ac The bands corresponding to processed substrates are pointed out by an arrowhead. A schematic representation of the chimeric protein substrates is shown on the right.

Identification of Putative RhoX Substrates by Label-Free Quantitative Proteomics
Having demonstrated that RhoX is an active protease, we sought to identify putative substrates by a label-free quantitative proteomics approach. We also intended to address metabolic and signaling pathways altered in response to rhomboid deletion. To achieve this, the differential proteome map between the B. abortus wild type (2308) and rhomboid deletion mutant (∆rhoX) was obtained. Two growth conditions were used to compare membrane, cytoplasmic and supernatant fractions of wild type and rhomboid mutant. Both strains were grown to early stationary phase (OD 600 1.8-1.9) in TSB rich medium (for membrane and cytoplasmic proteins) or in a defined minimal medium (supplemented RPMI) (for membrane, cytoplasmic and supernatant proteins) ( Figure 5A). Proteins obtained from triplicate cultures of both strains were subjected to SDS-PAGE, digested with trypsin, and then analyzed by nano-LC-MS/MS as described in the Experimental Section.
Out of 3023 total proteins from the predicted B. abortus proteome, 1554 (51.41%) were identified from bacteria grown in TSB and 1391 (46.01%) from bacteria grown in RPMI, including both strains and all cell fractions. Venn diagrams indicating the number of identified proteins in each of the cellular fractions are shown in Figure 5B (left panel TSBgrown bacteria and right panel RPMI-grown bacteria). The protease RhoX was not detected in membrane fractions of B. abortus 2308. Indeed, integral membrane proteins containing several TMDs and/or hydrophobic stretches are barely detected by MS technology, as they are underrepresented in trypsin digestion products [66]. However, they can be identified with trypsin/chymotrypsin digestion [13].

Proteins Differentially Represented in ∆rhoX Mutant
Using the criteria described in M&M, a total of 82 and 36 proteins were differentially represented between 2308 parental strain and ∆rhoX mutant strain grown in TSB or RPMI media, respectively ( Figure 5C,E). Supplementary Table S2 shows the proteins which evidenced significant increase, organized by growth media, strain and cellular fraction. Ten proteins were differentially represented both in TSB and in RPMI between wild type and ∆rhoX ( Table 2). Out of 108 differential proteins, 22 were predicted to contain a signal peptide and 8 contain at least one transmembrane domain. 48.23% of the proteins that exhibited variation in their concentration were detected in membrane and secretome fractions, in accordance with RhoX exerting its role within the membrane. Accordingly, about 50% of the proteins were predicted to be in the inner membrane, outer membrane, periplasm or to have unknown localization as predicted by Psortb ( Figure 5D,F). Volcano plots ( Figure 6A) were generated to visualize protein abundance changes obtained in these experiments, by plotting −log10 (p-value) vs. log2 (fold-change) for the compared samples. Of the proteins meeting statistically significant criteria (p value ≤ 0.05), those with the lower p values (higher values of −log10 (p-value) were highlighted in the figures ( Figure 6A). In the cytoplasmic fraction of TSB-grown bacteria, four proteins exhibited lower p values: Q2YJW2 (Nitrous-oxide reductase NosZ); Q2YP89 (Fumarylacetoacetate hydrolase); Q2YMQ4 (Periplasmic binding protein/LacI transcriptional regulator) and Q2YRV5 (immunogenic 31 KDa TRAP transporter solute receptor, TAXI family protein). In the membrane fraction of TSB-grown bacteria Q2YNA5 (DNA topoisomerase 4 subunit A), Q2YJ39 (ABC transporter: AAA ATPase) and Q2YIK8 (Heme transporter BhuA) evidenced the most significant difference in amount. In RPMI-grown bacteria, proteins meeting this criteria were Q2YLF3 (4-hydroxyproline epimerase) and Q2YLX8 (Ribosephosphate pyrophosphokinase) in the secretome, Q2YP92 (Oxidoreductase, N-terminal) and Q2YLR4 (diaminopimelate epimerase) in the cytoplasm, and Q2YIM6 (PntA, NAD(P) transhydrogenase, alpha subunit) and Q2YRS4 (OmpW family protein) in the membrane. The differential proteins belonged to several functional classes ( Figure 6B,C) and proteins involved in catalytic activity and binding were the most represented, suggesting that these processes are affected by RhoX.

Proteins Differentially Represented in the Membrane Fraction: Potential Rhomboid Targets
Rhomboid activity may expose or release a fragment of the substrate, leading to changes in size, stability and/or location of the cleaved substrate. Most characterized substrates are single TMDs, type I or III proteins (N-terminus facing the extracellular side with or without a SP, respectively). Thus, single TMD proteins exhibiting accumulation or decrease can be considered putative targets of rhomboid. A total of 51 proteins were differentially represented in membrane fractions of TSB-and RPMI-grown bacteria (Table S2), comprising 47.22% of the total altered proteins. Six proteins identified as differential are predicted to have at least 1 TMD and seven contain signal peptides that likely target them to the cell surface. Proteins without predicted TMDs and/or signal peptides may be associated to the membrane via interactions with lipids or other proteins and might represent indirect targets of rhomboids. However, it cannot be ruled out that some of these proteins are contaminants from cytoplasmic origin.
Q2YJT6 and Q2YP16 are integral membrane proteins (with single TMDs) that accumulated in the wild type membrane fraction. Q2YJT6 is nitric oxide reductase subunit C (NorC) and Q2YP16 is periplasmic protein LptC involved in LPS transport to the outer membrane.

Proteins Differentially Represented in the Secretome
Fourteen proteins were differentially represented in the secretome (Table S2). Out of these, seven are predicted secretory proteins, two are expected to be membranes associated via 1 TMD, and the remaining are likely soluble proteins. Vesicle transport (outer membrane vesicles) or cell lysis might explain the presence of proteins without a predicted SP in supernatants. Soluble proteins, with or without an SP, differentially represented may be a consequence of an indirect effect of the rhomboid deletion. Two integral membrane proteins with a single TMD were underrepresented in ∆rhoX culture media supernatants (Table S2). Q2YQA6 is a periplasmic protein, named EipA, involved in membrane integrity [67], and Q2YM85 is Cbb3-type cytochrome c oxidase subunit.

Oxygen-Limiting Conditions and Rhomboid Protease
To explore the role of RhoX in B. abortus, we subjected ∆rhoX to multiple phenotypic assays including growth in rich/minimal media under aerobic or anaerobic conditions, resistance to detergents or to oxidative stress, intracellular replication in phagocytic and nonphagocytic cells, and virulence in mice. Deletion of rhomboid had no effect in these assays.
It has been recently described that S. sonnei rhomboid proteases GlpG and Rhom7 prevent protein aggregation of orphan proteins in the inner membrane [20]. They specifically target components of respiratory complexes. With that in mind, we noticed that many of the differentially represented proteins between B. abortus 2308 and ∆rhoX are associated to denitrification and oxidative phosphorylation complexes: Q2YJW2 nitrous-oxide reductase NosZ, Q2YJU8 copper-containing nitrite reductase, Q2YJT6 nitric oxide reductase subunit C (NorC), Q2YJY3 nitrate reductase subunit beta (NarH), Q2YJY4 nitrate reductase subunit alpha (NarG) and Q2YM85 (high oxygen affinity) Cbb3-type cytochrome c oxidase unit.
To investigate the biological role of RhoX we first addressed nitrate respiration in anaerobic conditions, but no differences could be detected among wild type, ∆rhoX and the complemented strains. Then, we evaluated bacterial replication under static or shake conditions in rich medium TSB. All strains behaved similarly under shake conditions (Figure 7 left panel). However, rhomboid mutant strain overexpressing wild type RhoX exhibited growth impairment under the static condition. Initial growth was similar, but after a few hours a metabolic change became evident as a diauxic curve. After adaptation, growth rate recovered to wild type levels, describing a linear performance as oxygen transference rate became the limiting factor (Figure 7 right panel). In contrast, complementation of the mutant with the mutated version of rhomboid (S153A) had no effect in the replication curve, indicating that adaptation to low oxygen conditions is affected by overexpression of a functional rhomboid protease. ure 7 left panel). However, rhomboid mutant strain overexpressing wild type RhoX exhibited growth impairment under the static condition. Initial growth was similar, but after a few hours a metabolic change became evident as a diauxic curve. After adaptation, growth rate recovered to wild type levels, describing a linear performance as oxygen transference rate became the limiting factor (Figure 7 right panel). In contrast, complementation of the mutant with the mutated version of rhomboid (S153A) had no effect in the replication curve, indicating that adaptation to low oxygen conditions is affected by overexpression of a functional rhomboid protease.

Discussion
Here we identified and characterized B. abortus rhomboid protease. We found that it is catalytically active and cleaves model heterologous substrates like D. melanogaster Gurken and P. stuartii TatA in vitro and in vivo. This activity was abolished in the deletion mutant and restored after complementation with a wild type gene. Supporting these results, protease activity could be inhibited by (3,, a general inhibitor of serine proteases including rhomboids [68,69].
Rhomboid proteases typically contain a six-transmembrane domain core with or without an extra TMD [8]. RhoX has an additional N-terminal TMD like mitochondrial PARL-like rhomboids. This might be related to the fact that mitochondria have an alphaproteobacterial ancestry [70]. In members of this subfamily, the GxSx and histidine of the catalytic dyad are shifted to TMD5 and TMD7, respectively. This topology differs from most of bacterial rhomboids, and it is not clear whether this difference infers evolutionary or functional significance, such as substrate recognition or regulation of proteolytic activity. Immunofluorescence and subcellular fractionation studies indicate that RhoX is an inner membrane protease. Cellular localization of the endogenous protein is yet to be done to confirm the localization determined for the EYFP and FLAG-tagged proteins.
Over the years, the search for rhomboid substrates in bacteria has remained challenging. A recent study has shed some light about the potential reasons [20]. The study shows that the substrates are only expressed during growth of S. sonnei in specific conditions and

Discussion
Here we identified and characterized B. abortus rhomboid protease. We found that it is catalytically active and cleaves model heterologous substrates like D. melanogaster Gurken and P. stuartii TatA in vitro and in vivo. This activity was abolished in the deletion mutant and restored after complementation with a wild type gene. Supporting these results, protease activity could be inhibited by (3,4-DCI), a general inhibitor of serine proteases including rhomboids [68,69].
Rhomboid proteases typically contain a six-transmembrane domain core with or without an extra TMD [8]. RhoX has an additional N-terminal TMD like mitochondrial PARL-like rhomboids. This might be related to the fact that mitochondria have an alphaproteobacterial ancestry [70]. In members of this subfamily, the GxSx and histidine of the catalytic dyad are shifted to TMD5 and TMD7, respectively. This topology differs from most of bacterial rhomboids, and it is not clear whether this difference infers evolutionary or functional significance, such as substrate recognition or regulation of proteolytic activity. Immunofluorescence and subcellular fractionation studies indicate that RhoX is an inner membrane protease. Cellular localization of the endogenous protein is yet to be done to confirm the localization determined for the EYFP and FLAG-tagged proteins.
Over the years, the search for rhomboid substrates in bacteria has remained challenging. A recent study has shed some light about the potential reasons [20]. The study shows that the substrates are only expressed during growth of S. sonnei in specific conditions and that cleavage can be impeded by the presence of partner proteins. More importantly, GlpG and Rhom7 rhomboid proteases selectively target non-functional proteins, hence their absence might not lead to noticeable phenotypes. Similar results were obtained after deletion of B. abortus rhomboid. None of the phenotypes evaluated were statistically different between wild type and ∆rhoX.
As a step towards defining the function of RhoX, we used a label-free quantitative proteomic approach. We analyzed differentially represented proteins in membrane, cytoplasmic and secreted fractions in wild type and ∆rhoX strains. This approach has been successful in identifying rhomboid substrates in H. volcanii [13], Trichomonas vaginalis [71] and S. sonnei [20], among others.
About 50% of the predicted proteome was identified and 108 differentially represented proteins were detected. Of those showing the most significant variation (highlighted in volcano plots), three were predicted to contain signal peptide (Nitrous-oxide reductase NosZ, a 31 KDa immunogenic protein and OmpW family protein) and six were predicted to be either membrane-associated (Nitrous-oxide reductase (NosZ), ATP/GTP-binding site motif A (P-loop):ABC transporter: AAA ATPase, Heme transporter BhuA, NAD(P) transhydrogenase subunit alpha and OmpW family) or periplasmic (Periplasmic binding protein/LacI transcriptional regulator) proteins. The role of some of these proteins has been described: NosZ is the last reductase of the denitrification pathway and catalyzes N 2 O reduction to N 2 ; OmpW family of small outer membrane proteins is widely distributed in Gram negative bacteria and has been associated to host-bacterial interactions in Pseudomonas aeruginosa and in many Vibrio species [72]. It could also participate in the transport of small molecules across the outer membrane [73,74]. Heme transporter BhuA is required to maintain a chronic spleen infection in BALB/c mice experimentally infected with B. abortus [75].
A high proportion of differentially represented proteins belonged to the membrane fraction of both TSB an RPMI-grown bacteria. Among these, of special interest are proteins with SP or single TMDs: transporters of branched amino acids (Leu/Ile/Val), a 31 kDa immunogenic protein, nitric oxide reductase subunit C (NorC), periplasmic protein LptC, secretion protein HlyD and OmpW family protein. The 31 kDa immunogenic protein is also known as BCSP-31. It is a Brucella surface antigen largely used in diagnostics of brucellosis [76]. NorC is a membrane-anchored c-type cytochrome catalyzing the reduction of NO to N 2 O. LptC is a component of the LPS transport machinery [77]. HlyD is the membrane fusion protein (MFP) component of type I secretion systems [78].
Since rhomboid substrates identified so far are integral membrane proteins, differentially represented proteins identified in cytoplasmic fractions are probably a result of yet unidentified gene expression and/or proteolytic processes regulated by rhomboids. Yet, it cannot be ruled out that some of these proteins are contaminants from membranous origin. Differential proteins with predicted SP or TMDs that may have been released to the cytoplasm because of rhomboid activity were identified. The list includes Q2YLG0 Leu/Ile/Val-binding protein homolog 2 (SP); Q2YRV5 31 kDa immunogenic protein (SP); Q2YJW2 Nitrous-oxide reductase (NosZ) (SP); Q2YJU8 Copper-containing nitrite reductase (SP); Q2YM85 Cbb3-type cytochrome c oxidase unit (1 TMD); Q2YMF9 small uncharacterized protein (SP); Q2YPP2 DUF1775 domain-containing protein (SP) and Q2YRS4 OmpW family protein (SP).
In the secretome, all differential proteins with SP or single TMD were underrepresented in ∆rhoX: a putative lytic murein transglycosylase, OmpW family protein, nitrous-oxide reductase NosZ and Cbb3-type cytochrome c oxidase subunit involved in the oxidative phosphorylation pathway.
Finally, rhomboid in B. abortus might target multiprotein membrane complexes participating in adaptation to low oxygen conditions by a yet unidentified mechanism, as evidenced by overexpression of wild type rhomboid in static cultures. We assume rhomboid activity might be tightly regulated and this phenotype is only observed after gene overexpression. A few years ago, a study demonstrated that in respiration at low oxygen tension, the two-component systems PrrBA and NtrYX coordinately regulate the expression of denitrification and high-affinity cytochrome oxidase genes [79]. Further studies are needed to determine if denitrification enzymes like NarG, NarH, NosZ and NorC, and high-affinity cyto-chrome cbb3 oxidase identified by the proteomics approach are rhomboid substrates. Also of interest is to determine whether B. abortus rhomboid targets orphan proteins to mediate quality control of respiratory protein complexes such as already described in S. sonnei [20].