Corynebacterium pseudotuberculosis biovar equi strain 258 was isolated from a horse in Belgium; this strain was cultivated under routine conditions in brain–heart infusion broth (BHI-HiMedia Laboratories Pvt. Ltd., India) at 37°C. When necessary, 1.5% agar was added to the medium for solid culture. For extracellular proteomic analyses, 258_equi was grown in a chemically defined medium (CDM) [(Na₂HPO₄_7H₂O (12.93 g/L), KH₂PO₄ (2.55 g/L), NH₄Cl (1 g/L), MgSO₄_7H₂O (0.20 g/L), CaCl₂ (0.02 g/L), and 0.05% (v/v) Tween 80] with 4% (v/v) MEM Vitamins Solution (Invitrogen, Gaithersburg, MD, USA), 1% (v/v) MEM Amino Acids Solution (Invitrogen), 1% (v/v) MEM Non-Essential Amino Acids Solution (Invitrogen), and 1.2% (w/v) glucose at 37°C (Moura-Costa et al., 2002).

The infection parameters were performed according to Moraes et al. (2014) and Ribeiro et al. (2014). In this study, female BALB/c mice between 6- and 8-weeks-old were utilized; they were provided by the Animal Care Facility at the Biological Sciences Institute at the Federal University of Minas Gerais and were handled in accordance with the CEUA guidelines of the UFMG Ethics Committee on Animal Testing (Permit Number: CETEA 103/2011). For the bacterial passage assay, three mice were infected via intraperitoneal injection with 10⁶ colony forming units (CFU) of strain 258_equi. Thirty-six hours after infection, the animals were sacrificed, and the spleen was aseptically removed for recovering the bacteria. Each spleen was individually macerated in a sterile saline solution (0.9% NaCl₂) and seeded onto BHI agar plates for incubation at 37°C for 48 h. Subsequently, one bacterial colony of each BHI plate was isolated and cultured in BHI broth at 37°C with shaking (180 rpm) until the OD₆₀₀ = 0.8. Three different stock cultures were generated and stored at −80°C in BHI broth and 10% glycerol. The recovered bacteria are referred to as Recovered (Rc), and bacteria with no previous host contact were used as the Control (Ct). For bacterial virulence assays, bacteria from the three individual frozen stocks of Rc and the Ct condition were centrifuged at 5,000 × g for 5 min and washed twice in saline solution, followed by resuspension in saline solution. Three groups of five mice were infected with bacteria from the Rc or Ct condition via intraperitoneal injection of a suspension containing 10⁶ or 10⁵ CFU. The animals' survival rates were calculated and represented in GraphPad Prism v.5.0 (GraphPad Software, San Diego, CA, USA) using the Kaplan-Meier survival function.

For proteomic analysis, three independent control and recovered colonies from the three individual frozen stocks were grown in CDM to an OD600 = 0.8. The cultures were then centrifuged for 20 min at 2,700 × g. The supernatants were filtered using 0.22-μm filters, 30% (w/v) ammonium sulfate was added to the samples, and the pH of the mixtures was adjusted to 4.0. Next, 20 mL/L N-butanol was added to each sample. The samples were centrifuged for 10 min at 1,350 × g and 4°C. The interfacial precipitate was collected and resuspended in 1 mL of 20 mM Tris-HCl, pH 7.2 (Paule et al., 2004). Proteins were quantified using the Bradford assay. For label-free proteomic analysis, the protein extract was concentrated using a spin column with a 10 kDa threshold (Millipore, Billerica, MA, USA). The protein was denatured (0.1% RapiGEST SF at 60°C for 15 min) (Waters, Milford, CA, USA), reduced (10 mM DTT), alkylated (10 mM iodoacetamide) and enzymatically digested with trypsin (Promega, Sequencing Grade Modified Trypsin, Madison, WI, USA). Glycogen phosphorylase (Waters Corporation, SwissProt P00489) was added to the digests to a final concentration of 20 fmol/μl as an internal standard for normalization prior to each replicate injection. The digestion process was stopped by adding 10 μL of 5% TFA (Fluka, Buchs, Germany) (Silva et al., 2006).

Three independent biological replicates of each experimental condition were digested, as described above, for MS^E analysis. Qualitative and quantitative nanoUPLC tandem nanoESI-HDMS^E (Nano Electrospray High Definition Mass Spectrometry) experiments were performed using a 1 h reversed-phase gradient from 7 to 40% (v/v) acetonitrile (with 0.1% v/v formic acid) at 500 nL.min⁻¹ using a nanoACQUITY UPLC 2D RPxRP Technology system (Gilar et al., 2005). All analyses were performed using nano-electrospray ionization in the positive ion mode (nanoESI (+)) and a NanoLockSpray (Waters, Manchester, UK) ionization source. The mass spectrometer was calibrated with an MS/MS spectrum of human [Glu1]-Fibrinopeptide B (Glu-Fib) solution (100 fmol.mL−1) delivered through the reference sprayer of the NanoLockSpray source. The double-charged ion ([M + 2H]²⁺ = 785.8426) was used for initial single-point calibration, and MS/MS fragment ions of Glu-Fib were used to obtain the final instrument calibration. Multiplexed data-independent (DIA) scanning with additional specificity and selectivity for non-linear “T-wave” ion mobility (HDMS^E) experiments were performed using a Synapt G2-S HDMS mass spectrometer (Waters, Manchester, UK), which was constructed to automatically switch between the application of standard MS (3 eV) and elevated collision energies HDMS^E (19–45 eV) to the transfer “T-wave” CID (collision-induced dissociation) cell with argon gas.

The proteins were identified, and quantitative data were packaged using dedicated algorithms (Silva et al., 2005; Geromanos et al., 2009) and searching against a database with default parameters to account for ions (Li et al., 2009). The databases used were reversed “on-the-fly” during the database queries and appended to the original database to assess the false positive rate during identification. For proper spectra processing and database searching conditions, the ProteinLynx Global SERVER v.2.5.2 (PLGS) with Identity^E and Expression^E informatics v.2.5.2 (Waters, Manchester, UK) was used. UniProtKB (release 2013_01) with manually reviewed annotations was used, and the search conditions were based on taxonomy (Corynebacterium pseudotuberculosis). The maximum allowed missed cleavages by trypsin was up to 1, and various modifications, including carbamidomethyl (C), N-terminal acetyl, phosphoryl (STY) and oxidation (M), were allowed. A peptide mass tolerance value of 10 ppm was used. The search threshold to accept each spectrum was the default value in the program with a false discovery rate value of 4% (Curty et al., 2014). For protein quantitation, PLGS v2.5.2 software was used with the Identity^E algorithm using Hi3 methodology and glycogen phosphorylase (muscle form; P00489) peptides were used as internal standards. The collected proteins were organized by the PLGS Expression^E tool algorithm into a statistically significant list (p-value ≤ 0.05) that corresponded to higher or lower regulation ratios between the different groups. The calculation of the log ratio and the confidence interval was based on a Gaussian distribution model, which allows for the possibility of an uncertain peptide assignment, an incorrect assignment of data to a cluster or interference. The confidence interval of 95% was used, and the probability distribution of the measured value of a log2 ratio more than a 1.2 was more symmetric than that obtained for the direct ratio, making the results interpretations more meaningful (Levin et al., 2011). For comparing pairs of experimental groups, proteins with a differential expression log₂ ratio greater than or equal to 1.2 between the two conditions were considered for higher or lower abundance level determination (Levin et al., 2011).

The proteins identified in both conditions were analyzed using the following prediction tools: SurfG+ v1.0 (Barinov et al., 2009) was used to predict subcellular localization, SecretomeP 2.0 server was used to predict proteins exported from non-classical systems (positive prediction scores greater than 0.5; Bendtsen et al., 2005a), TatP was used to predict proteins with twin-arginine signal peptides (Bendtsen et al., 2005b) and the PIPs software was used to predict the proteins in pathogenicity islands (Soares et al., 2012). Gene ontology (GO) functional annotations were generated using the COG data base (Tatusov et al., 2001). Pathway enrichment analysis of significant proteins was carried out using the Kyoto encyclopedia of genes and genomes (KEGG) database. A protein-protein interaction network was generated using Cytoscape version 2.8.3 (Shannon et al., 2003) with a spring-embedded layout.

In the first in vivo assay, BALB/c mice were infected with 10⁶ CFU of bacteria that had no previous host contact and with bacteria that were recovered from mouse spleens. We observed that all of the infected animals under both the control (Ct) and recovered (Rc) conditions died within 48 h of infection (Figure 1A). These results reveal the virulence potential of 258_equi; however, in this assay, we did not observe differences in the virulence potential between the control and recovered condition. Next, the Ct and Rc condition were analyzed using a new survival assay in BALB/c mice, but a 10⁵ CFU infection dose was used (Figure 1B). In this assay, we observed altered virulence in the Rc condition; the mice began dying in the first 10 days post-infection (40% decrease in survival rate) and mortality reached 100% in less than 20 days post-infection. For the Ct condition, while mice died during the first 10 days of infection, this early stage only resulted in 20% mortality, and mortality did not reach 100% until 23 days post-infection. Finally, we detected abscess formation in the internal organs (kidney and liver) of all animals infected with either the Ct or Rc condition in the assays using 10⁵ CFU (data not shown). These results show that passage in a murine host affects the virulence potential of 258_equi.

After passage of 258_equi in BALB/c mice, we detected changes in its virulence potential. To assess whether this change is reflected in its proteome, considering the importance of exported proteins in bacterial infection (Hilbi et al., 2012), we used a TPP/LC-MS^E approach (Pacheco et al., 2011) to compare the extracellular proteome of the control and recovered conditions. From our proteomic analysis, a total of 113 non-redundant 258_equi proteins were detected with high confidence and were identified in at least two of the three biological replicates of the two conditions tested, with an average of 17 peptides per protein and an FDR of 1%. The peptides were identified with a normal distribution of 10 ppm error for the total identified peptides (Supplementary Figure 1A). In addition, only the source fragments of peptides with a charge state of at least [M + 2H]²⁺ and the absence of decoys were considered to increase data quality.

The absolute quantitation of proteins present within a complex protein mixture is extremely important for understating physiological adaptations in response to biological demands promoted by environmental changes (Mallick and Kuster, 2010). To estimate the absolute abundance of identified proteins in the 258_equi exoproteome, we utilized the Hi3 method (Silva et al., 2006) where the average MS signal responses for the three most intense tryptic peptides for each protein were determined, including those of the internal standard protein glycogen phosphorylase (muscle form; P00489). All samples were normalized prior to injection using “scouting runs,” and the stoichiometry between the intensity and molarity proportion prior to the replicate runs per condition were considered. From this analysis a dynamic range of protein abundance was generated spanning three orders of magnitude (Supplementary Figure 1B). Lysozyme M1 was the most abundant protein detected. This protein, which is related to bacterial virulence, is localized in the pathogenic island Cp258PiCp02. Lysozyme M1 was also detected in a membrane shaving of a field isolate of C. pseudotuberculosis biovar ovis (Rees et al., 2015). Other proteins, such as hydrolase domain containing protein, trehalose corynomycolyl transferase B, which is involved in the cell wall synthesis, and FtsX, a protein related to division cellular, were among the most abundant proteins. All of the identified proteins on the protein abundance scale are listed in Supplementary Table 1.

For evaluating the relative differences between the core exoproteome of the Ct and Rc conditions, we used label-free quantification (Silva et al., 2005, 2014; Pacheco et al., 2011). In agreement with the PLGS analyses, 105 proteins between the Rc and Ct conditions presented significant statistical values (p < 0.05) using the Expression^E algorithm tool (Supplementary Table 2). Differential expression was considered for proteins that were significantly different (p < 0.05) and had log₂ ratios equal to or greater than a factor of 1.2, as described by Levin et al. (2011). Based on this analysis, 39 proteins were induced and 16 were down-regulated in the Rc condition (Table 1). In addition, we detected proteins exclusive to the proteome of each condition; cytochrome c nitrate reductase small subunit NrfH was detected only in the Ct condition. In only the Rc condition, two multidrug resistance proteins, the cytochrome oxidase assembly, a thioredoxin-related protein and three proteins with unknown function were identified (Supplementary Table 3).

Extracellular proteins produced by prokaryotic organisms are a subset of proteins present in the extracellular milieu, which is composed of both proteins with signal peptides that are actively secreted by classical secretion systems and proteins without signal sequences that are exported by non-classical secretion systems (Bendtsen et al., 2005a; Desvaux et al., 2009). To identify proteins that contain signal peptides and to determine their subcellular localizations, we utilized the SurfG tool (Barinov et al., 2009), which enables the classification of proteins within the following categories: cytoplasmic (CYT), membrane (MEM), potentially surface-exposed (PSE) and secreted (SEC) (Figure 2A and Supplementary Tables 2, 3). Of the total proteins identified, 66% (n = 74) presented positive predictions for signal peptides (Figure 2B). This group was composed of predicted proteins in the SEC and PSE categories. When these results were compared with the in silico data of the 258_equi genome, we had identified approximately 65% and 16% of the proteins predicted to be SEC and PSE, respectively. The proteins that did not present positive predictions for signal peptides were analyzed by SecretomeP (Bendtsen et al., 2005a) to identify proteins that could eventually be exported by non-classical secretion systems. According this analysis, nine proteins were predicted to be PSE, seven proteins were predicted to be MEM and two proteins were predicted to be CYT as they presented High SecP scores above 0.5 (Supplementary Tables 2, 3), suggesting that these proteins might be exported by a non-classical secretion system. Taken together, 86% of the 258_equi exoproteome was composed of extracellular proteins (Figure 2B). In addition, our proteomic analysis detected 17 proteins predicted to be lipoproteins (Supplementary Table 2).

To evaluate the functional characteristics of the 258_equi exoproteome, we performed a Clusters of Orthologous Groups analysis (Tatusov et al., 2001). According GO analysis, the proteins were organized by clusters of orthologous groups (Figure 3A). When we evaluated each functional category, we observed that the majority of the proteins detected as induced in the Rc condition were predicted as “function unknown” and “general function only” (Table 1 and Figure 3B). These results represent a lack of knowledge regarding a protein set that might play an important role in the pathogenic process of 258_equi, and therefore, more studies are necessary to investigate the true roles of these proteins in the virulence of this strain. When we evaluated proteins with known or predicted functions, the majority of those that were more abundant in the Rc condition were related to cellular processes and signaling (Figures 3A,B). According to in silico data of the 258_equi genome, this pathogen has five iron uptake systems (Supplementary Figure 2; Soares et al., 2013a). Interestingly, in our proteomic analysis, we identified components of each of the 5 systems as more abundant in the Rc condition, including CiuA, FhuD, FagC, HmuT, HmuV, and HtaA (Table 1), suggesting that iron uptake pathways may play an important role in the pathogenesis of C. pseudotuberculosis. Moreover, we detected several proteins related to bacterial pathogenesis that contribute to processes of adherence, intracellular growth and evasion of the immune system (Table 1 and Supplementary Tables 2, 3).

To identify the most relevant biological pathways of the proteins differentially expressed between the Ct and Rc conditions, we performed a KEGG enrichment analysis. Enrichment results revealed eight biological pathways with significant differences (p < 0.05). The proteins that were induced in the Rc condition are in pathways such as ABC transporters, bacterial secretion systems, peptidoglycan biosynthesis and protein export (Figure 3C). This finding confirms that secretion and secreted proteins are key elements in C. pseudotuberculosis virulence and adaptation, as suggested by previous reports that identified several secreted proteins as potential virulence factors in C. pseudotuberculosis (Pacheco et al., 2011; Silva et al., 2013; Rees et al., 2015). Most proteins perform their function in a context of networks by interacting with other proteins (Schleker et al., 2012). To evaluate the 258_equi exoproteome at the network level, we performed a protein-protein interaction analysis of the differentially expressed proteins using the Cytoscape tool. After Cytoscape analysis, the 258_equi exoproteome network was composed of 87 proteins (Figure 4). In the PPI-network, we observed enrichment clusters in heme biosynthesis and ABC transporters related to iron uptake, peptidoglycan biosynthesis and antibiotic biosynthesis. In addition, we observed that some clusters were formed by unknown proteins, which shows that these proteins may play an important role in the virulence of C. pseudotuberculosis.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.