Global proteomic profiling of Yersinia ruckeri strains

Yersinia ruckeri is the causative agent of enteric redmouth disease (ERM) of salmonids. There is little information regarding the proteomics of Y. ruckeri. Herein, we perform whole protein identification and quantification of biotype 1 and biotype 2 strains of Y. ruckeri grown under standard culture conditions using a shotgun proteomic approach. Proteins were extracted, digested and peptides were separated by a nano liquid chromatography system and analyzed with a high-resolution hybrid triple quadrupole time of flight mass spectrometer coupled via a nano ESI interface. SWATH-MS technology and sophisticated statistical analyses were used to identify proteome differences among virulent and avirulent strains. GO annotation, subcellular localization, virulence proteins and antibiotic resistance ontology were predicted using bioinformatic tools. A total of 1395 proteins were identified in the whole cell of Y. ruckeri. These included proteases, chaperones, cell division proteins, outer membrane proteins, lipoproteins, receptors, ion binding proteins, transporters and catalytic proteins. In virulent strains, a total of 16 proteins were upregulated including anti-sigma regulatory factor, arginine deiminase, phosphate-binding protein PstS and superoxide dismutase Cu–Zu. Additionally, several virulence proteins were predicted such as Clp and Lon pro-teases, TolB, PPIases, PstS, PhoP and LuxR family transcriptional regulators. These putative virulence proteins might be used for development of novel targets for treatment of ERM in fish. Our study represents one of the first global proteomic reference profiles of Y. ruckeri and this data can be accessed via ProteomeXchange with identifier PXD005439. These proteomic profiles elucidate proteomic mechanisms, pathogenicity, host-interactions, antibiotic resistance ontology and localization of Y. ruckeri proteins. Electronic supplementary material The online version of this article (doi:10.1186/s13567-017-0460-3) contains supplementary material, which is available to authorized users.


Introduction
Enteric redmouth disease (ERM) is one of the most important bacterial diseases of salmonids and causes significant economic losses in the aquaculture industry worldwide. ERM can affect fish from all age classes and appears as a more chronic condition in older and larger fish. The disease is caused by Yersinia ruckeri, a Gramnegative rod-shaped enterobacterium [1,2]. Y. ruckeri enters the fish via the secondary gill lamellae and from there spreads to the blood and internal organs [3]. Clinical signs of the disease include exophthalmia, darkening of the skin in addition to subcutaneous hemorrhages in and around the mouth and throat. The spleen is often enlarged and can be almost black in color and the lower intestine can become reddened and filled with an opaque, yellowish fluid [1,2]. Focal areas of necrosis can be present in the organs (spleen, kidney and liver). Degenerated renal tubules, glomerular nephritis and a marked increase in melano-macrophages may be observed in the kidney of infected fish [1,2,4]. Several virulence factors of Y. ruckeri have been identified such as extra-cellular products and Yrp1. Extra-cellular products have been shown to reproduce the clinical signs of the disease [5]. The 47 kDa metalloprotease Yrp1 is necessary for virulence and degrades fibronectin, actin and myosin of the fish [6].
Strains of Y. ruckeri have been categorized into two biotypes: biotype 1 strains are motile and lipase positive, while biotype 2 strains are negative for these phenotypes [2,7]. Previously, the majority of epizootic outbreaks in salmonids were caused by biotype 1 strains which could be easily controlled by vaccination with a bacterin vaccine [5]. Nevertheless, biotype 2 strains have recently emerged and have been responsible for outbreaks in Open Access *Correspondence: Gokhlesh.Kumar@vetmeduni.ac.at 1 Clinical Division of Fish Medicine, University of Veterinary Medicine, Veterinärplatz 1, 1210 Vienna, Austria Full list of author information is available at the end of the article both naive and vaccinated fish, thereby suggesting that biotype 2 strains may be less sensitive to the traditional ERM vaccine which is made from a biotype 1 strain [8,9]. This relationship between vaccine failure and emergence of biotype 2 has led to the hypothesis that the loss of the flagellum is essential for resistance to immersion vaccination [9,10]. However, bivalent or biotype 2 vaccines provide good protection against the biotype 2 strains [2,11].
Whole genome sequences of Y. ruckeri strains have been annotated and can now be used for comparative genomic analysis of strains and other research purposes [12]. Global proteomic identification and comparative analysis of Y. ruckeri strains are required to create a proteomic map, understanding proteomic biology, proteomic changes and proteomic differences between strains. Little is known about the proteomics of Y. ruckeri. Outer membrane protein and whole cell protein patterns of Y. ruckeri isolates were described using SDS-PAGE and 2D-PAGE [11,13,14]. Reference proteome maps of many bacteria including Y. pestis have been created, and this work is leading to an understanding of the virulence mechanisms and the regulatory networks used by pathogenic bacteria [15]. However, for fish pathogens, in-depth proteomic analysis is not yet well established.
In our previous study, we compared two culture conditions of Y. ruckeri strains and focused only on proteins expressed in response to iron-limited culture conditions [16]. In this study, we identified, quantified and analyzed the global proteomic profiles of Y. ruckeri strains grown under standard culture conditions using a shotgun proteomic approach. Furthermore, we predicted virulence proteins and antibiotic resistance ontology in the proteome of Y. ruckeri.

Culture conditions
The culture conditions and growth yield of Y. ruckeri strains have been previously described [16]. Briefly, a single colony of each strain was used to inoculate duplicate 5 mL tryptic soy broth cultures. Duplicate starter cultures of each strain (OD 600 0.10) were then used to inoculate 25 mL tryptic soy broth cultures and grown overnight at 22 °C until the late log phase. The yield of CSF007-82, 7959-11 and YRNC-10 strains (OD 600 1.62) were similar to each other but the yield of SP-05 strain was slightly lower (OD 600 1.32) compared to the other three strains [16]. Cells were harvested and washed three times with sterile phosphate buffered saline containing bacterial protease inhibitor cocktail.

Protein extraction and digestion
The protein extraction procedures used have been previously described [16]. Briefly, bacterial cells were resuspended in denaturing lysis buffer (7 M urea, 2 M thiourea, 4% 3-[(3-cholamidopropyl)dimethyl-ammonio]-1-propane sulfonate and 1% dithiothreitol) containing bacterial protease inhibitor cocktail. Cells were then sonicated on ice and cellular debris removed by centrifugation. Protein digestion was performed using the standard two-step insolution digestion protocol for Trypsin/LysC mix according to the user manual (Promega) and digested samples were acidified.

Nano LC-MS/MS analysis
Tryptic peptides were separated by a nano liquid chromatography system (Dionex Ultimate 3000 RSLC) and analyzed with a high-resolution hybrid triple quadrupole time of flight mass spectrometer (TripleTOF 5600+, Sciex) coupled via a nano-ESI interface. Preconcentration and desalting of samples were accomplished with a 5 mm Acclaim PepMap µ-Precolumn (Dionex). Details of the LC-MS/MS procedure were described previously [16]. Briefly, 370 ng of digested protein were used per injection and peptide separation was performed on a 25 cm Acclaim PepMap C18 column with a flow rate of 300 nL/min. The gradient started with 4% mobile phase B (80% acetonitrile with 0.1% formic acid) and increased to 35% B over 120 min. MS1 survey scans were collected in the range of 400-1500 mass-to-charge ratio (m/z). The 25 most intense precursors with charge state 2-4, which exceeded 100 counts per second, were selected for fragmentation for 250 ms. MS2 product ion scans were collected in the range of 100-1800 m/z for 110 ms. Precursor ions were dynamically excluded from reselection for 12 s.
For quantitative measurements, data independent sequential window acquisition of all theoretical spectra (SWATH) technology based on MS2 quantification was used [19,20]. Peptides from biological and technical replicates were fragmented in 35 fixed fragmentation windows of 20 Dalton (Da) in the range of 400-1100 Da with an accumulation time of 50 ms in TOF MS mode and 80 ms in product ion mode. The nano-HPLC system was operated by Chromeleon 6.8 (Dionex) and the MS by Analyst Software 1.6 (Sciex).

Data analysis
Database searches of raw files of data dependent acquisition were carried out with Protein Pilot Software version 5.0 (Sciex). UniProt database (Released 10_2016) was restricted to Y. ruckeri. Mass tolerance in MS mode was set with 0.05 and 0.1 Da in MS/MS mode for the rapid recalibration search as well as 0.0011 Da in MS and 0.01 Da in MS/MS mode for the final search. The following sample parameters were applied: trypsin digestion, cysteine alkylation set to iodoacetamide and the search effort set was to rapid identification. False discovery rate analysis was performed using the integrated tools in Pro-teinPilot. The global false discovery rate (FDR) was set to < 1% on the protein level, peptide level as well as spectra level. Information dependent data acquisition identification results were used to create the SWATH ion library with the MS/MS (ALL) with SWATH Acquisition Micro-App 2.0 in PeakView 2.2 (both Sciex). Peptides were chosen based on a FDR rate < 1%, excluding shared and modified peptides. Up to six peptides per protein and up to 6 transitions per peptide were used. MarkerView 1.2.1 (Sciex) was used for calculation of peak areas of SWATH samples after retention time alignment and normalization using total area sums. The resulting protein lists were then used for visualization of data after principal component analysis (PCA) in form of loading plots and score plots to get a first impression of the overall data structure and to assess variability between technical and biological replicates.
Differentially expressed proteins were determined by statistical analysis in R programming language [21]. Raw peak areas after normalization to total area sums were log 2 -transformed to approach a normal distribution. On a logarithmic scale, technical replicates were aggregated by arithmetic mean before application of statistical tests. This procedure is equivalent to the application of a hierarchical model in the subsequent ANOVA, as the same number of technical replicates was measured per biological replicate. Differential expression of proteins in each strain was assessed using one-way ANOVA for each protein. To adjust for multiple testing, the method of Benjamini and Hochberg [22] was used to control the FDR. Differences were considered significant if adjusted p-values were smaller than the significance level of α = 0.001. For those proteins, Tukey's honest significant difference method was applied as post hoc test to assess the significance of the pairwise comparisons. Protein expression was considered differential if the adjusted p-value was below α and the absolute fold change was at least three (fold change < −3 or > +3).

GO annotation and prediction of virulent proteins
Venn diagrams were used to show the differences between protein lists originating from different strains [23]. Gene ontology annotation of all identified proteins was classified using the software tool for researching annotations of proteins [24]. Subcellular localization of proteins was predicted by PSORTb version 3.0 [25]. Virulence proteins were predicted by a method based on bi-layer cascade Support Vector Machine using Virulent-Pred [26].

Antibiotic resistance ontology and their validation
Antibiotic resistance ontology was identified using a comprehensive antibiotic resistance database [27]. The antibiotic resistance phenotypes predicted by in silico analysis were validated using the disc diffusion technique and minimal inhibitory concentration (MIC) determination [28,29]. The antimicrobial commercial Oxoid discs (µg disc/mL, Thermo Scientific): gentamicin (10 µg), polymyxin B (300 UI), erythromycin (15 µg), rifampin (5 µg), novobiocin (5 µg) and mupirocin (5 µg) were applied to inoculated Mueller-Hinton agar (Thermo Scientific) in triplicate. In parallel, MIC ranges for the same antibiotics were determined using microtiter plates and solutions of antibiotics prepared from powders of known potencies (Sigma-Aldrich). All plates were incubated for 48 h at 22 °C. The diameter of the inhibition halo of antimicrobial discs and lowest concentration of antibiotic that inhibited visible growth of bacteria were defined and categorized as susceptible or resistant (Additional file 1) as previously using standard methods [28,29].

Protein identification
A total of 1395 proteins in the whole cell of Y. ruckeri were identified (Additional file 2). The number of proteins identified in each strain was 1193 for SP-05, 1263 for CSF007-82, 1244 for 7959-11 and 1208 for YRNC-10. The list of identified proteins in each strain is given in Additional file 3. Forty-six proteins in SP-05, 43 proteins in CSF007-82, 31 proteins in 7959-11 and 13 proteins in YRNC-10 were uniquely identified ( Figure 1). PCA score plots of all strains suggested that strain SP-05 differs from the other three strains (CSF007-82, 7959-11 and YRNC-10) but the latter three strains showed minor proteomic differences (Additional file 4). The list of uniquely identified proteins in each strain is given in Additional file 5.

GO annotation and subcellular localization of proteins
The identified proteins were associated with cellular process, metabolic process, regulation, localization and response to stimulus (Figure 2A). Proteins were localized in the cytoplasm, plasma membrane, ribosome, macromolecular complex, nucleus, chromosome and others ( Figure 2B). Proteins involved in catalytic activity and binding were the most abundant among those identified proteins, 51 and 39%, respectively ( Figure 2C). The identified proteins were predicted in the cytoplasmic space (67%), unknown (16%), cytoplasmic membrane (8%), periplasmic space (6%), outer membrane (2%) and extracellular space (1%) (Figure 3). The unknown group included proteins with multiple subcellular and unknown localizations.
We also predicted antibiotic resistance ontology in 12 antibiotic classes (Table 3) in the proteome of Y. ruckeri, which contains 14 proteins such as bacterial regulatory protein (cyclic AMP receptor protein), membrane fusion protein of the resistance-nodulation-division (RND) family multidrug efflux pump, bifunctional polymyxin resistance protein ArnA and RND efflux system inner membrane transporter CmeB.

Discussion
Here we identify global proteomic reference profiles of Y. ruckeri strains (PXD005439) grown under standard culture conditions. These global proteomic profiles help us to understand the physiology, protein biology, virulence factors, host-interactions, localization and antibiotic resistance of Y. ruckeri. The total number of proteins identified was 1395 in Y. ruckeri (Additional file 2). These included proteases, chaperones, cell division proteins, outer membrane proteins, chromosome partitioning proteins and transporters. Proteins have been classified into different functional categories such as biological process and molecular function ( Figure 2) and this information will be useful for further studies in the direction of extracellular (flagellin and flagellar hook-associated protein), interaction with cells (invasin and manganese ABC transporter, periplasmic-binding protein SitA), antioxidant (thioredoxin reductase and glutathione amide-dependent peroxidase) and molecular transducer (methyl-accepting chemotaxis protein) activity. The identification of predicted virulence proteins (Table 2; Additional file 7) and antibiotic resistance ontology (Table 3) contributes to our understanding of this pathogen and will aid in the rational design of novel treatment strategies for ERM disease.
Biotype 2 strains showed minor proteomic differences among each other (Additional file 4). However, the Austrian biotype 1 strain (SP-05) showed major proteomic differences when compared to the USA biotype 1 strain (CSF007-82) and biotype 2 strains (7959-11 and YRNC-10). These major differences may be due to the slightly lower yield and growth rate of SP-05 strain compared to the other three strains (CSF007-82, 7959-11 and YRNC-10) or its avirulent nature toward the fish.
Sixteen upregulated proteins were identified in virulent Y. ruckeri strains using a sophisticated statistical analysis (avirulent SP-05 strain versus virulent strains). We found strong upregulation of bacterioferritin (5.7-to 6.8-fold) and DNA protection during starvation protein (3.2-to 3.8-fold) in Y. ruckeri strains. However, iron dependent proteins (bacterioferritin and iron-sulfur cluster assembly scaffold protein IscU) were downregulated (−3-fold) in Y. ruckeri strains in response to iron-limited culture conditions [16]. The phosphate-binding protein PstS is a high affinity phosphate binding protein of the Pst transport system and has been shown to be involved in pathogenesis, invasion and biofilm formation of many bacteria [30]. Superoxide dismutase Cu-Zn is an important for oxidative stress and has been shown to contribute to the pathogenicity of many bacteria [31]. Arginine deiminase protects bacterial cells against the damaging effects of acidic environments and enhances the ability of cells to survive in acidic extracellular conditions [32]. We observed strong upregulation of phosphate-binding protein PstS (> 3-fold), superoxide dismutase Cu-Zn (3.3-to 3.4-fold) and arginine deiminase (5.2-to 5.8-fold) in Y. ruckeri strains. Based on the results of the present study, it appears that upregulated proteins (avirulent strain versus virulent strains) such as PstS, SOD-Cu-Zn and arginine deiminase may be involved in the establishment of disease inside the host and the survival of Y. ruckeri during the infection process. Several proteases such as HtrA, Lon, carboxy-terminal, signal peptidase I, La Type II, HslUV, pyrrolidonecarboxylate, FtsH, protease III, Clp, protease 4, putative protease, peptidase B, T and M37 were identified. These proteases were serine, threonine, cysteine, metalloproteinase and ATP-dependent type proteases, and belonged to the C15, M16, M17, M20B, M23, S16, S26, S41A, S49, U32, AAA ATPase and Clp families including PDZ domains. Proteases play critical roles in the invasion of host tissues, contribute to virulence and damage host tissue during infection [33]. The Yrp1 protease of Y. ruckeri has been implicated in the hydrolysis of different matrix and muscle proteins of fish and vaccination with Yrp1 elicits a strong protection against the development of  enteric redmouth disease [6]. Additionally, the Clp and Lon pro-teases have been shown to have a role in the regulation of the type III secretion systems (T3SS) in various bacterial pathogens. The T3SS forms a needle-like structure in several Gram negative bacteria that allows direct transfer of bacterial virulence factors into the cytoplasm of host cells. The T3SS has been linked to flagellum biosynthesis [34]. We also identified flagellar biosynthesis proteins (FliC, FliG, FliH and FliN), flagellar hook proteins (FlgD, FlgE and FlgK), flagellar brake protein

Table 2 Lists of important virulence proteins of Yersinia ruckeri
Proteins were predicted by a method based on bi-layer cascade support vector machine using VirulentPred.

Protein Function Cascade of SVMs and PSI-BLAST, score
Gene expression modulator/haemolysin expression modulating protein Haemolysin expression 1.0520 HtrA protease Serine-type endopeptidase activity 0.5097 Outer membrane stress sensor protease DegQ serine protease Serine-type endopeptidase activity 1.0012 Anti-sigma regulatory factor Protein serine/threonine kinase activity 0.9363 Beta-barrel assembly-enhancing protease Chaperone and a metalloprotease 1.0339  YcgR, flagellar motor protein MotB and pilus assembly protein PilW in Y. ruckeri. FliC and FliH flagellar proteins have been linked with pathogenesis in the fish pathogen, Edwardsiella tarda [35]. Additionally, the Y. ruckeri flagellin protein has been shown to elicit a robust innate immune response and protect fish against biotype 1 and biotype 2 Y. ruckeri strains [36]. More research on the role of proteases and T3SS in Y. ruckeri virulence is needed to more fully understand the pathogenicity of Y. ruckeri.
We also identified other important virulence proteins such as the UvrY response regulator, peptidyl-prolyl cistrans isomerase (PPIases), TolB, PhoP and LuxR family transcriptional regulators. UvrY is a response regulator of the BarA-UvrY two-component system and has been shown to be involved in the pathogenesis of Y. ruckeri, probably through its regulation of both the invasion of epithelial cells and protection against oxidative stress induced by immune cells [37]. PPIases are FKBP domaincontaining ubiquitous folding proteins and have been reported as virulence factors in several bacterial pathogens [38]. Upregulation of FKBP-type peptidyl-prolyl cis-trans isomerases has been observed in iron-starved biotype 2 Y. ruckeri strains [16], which may be involved in virulence of Y. ruckeri. PhoP is part of a two component system and is important for bacterial survival and replication in macrophages [39]. TolB is the periplasmic component of the Tol-Pal system and is important for antibiotic resistance and pathogenicity in Gram negative pathogens and has been suggested as a suitable candidate for the development of novel drugs against Pseudomonas aeruginosa [40]. The LuxR transcriptional regulator is a key player in quorum sensing and affects survival, virulence, antibiotic biosynthesis and biofilm formation of bacteria [41].
A number of chaperone proteins (CbpA, ClpB, DnaK, DnaJ, fimC, HscA, HscB, HtpG, skp, SurA, ProQ), an acid stress chaperone HdeB, universal stress protein E, cold shock (CspC and CspE) and a phage shock protein were identified in Y. ruckeri. Bacterial pathogens produce a number of chaperone proteins for survival during changing environments and stress conditions [42]. Some chaperone proteins have also been implicated in bacterial virulence [43]. DnaK chaperone protein plays a role in protein folding and interacts with ClpB in reactivating proteins which have become aggregated after heat shock [44]. The DnaK/DnaJ chaperone machinery and ClpB have been shown to be involved in the invasion of epithelial cells and survival within macrophages of the host, leading to systemic infection of Salmonella enterica and Francisella tularensis in mice [43,45]. Upregulation of ClpB, HtpG and universal stress protein A have been observed in Flavobacterium psychrophilum during in vivo growth in fish and were suggested to play an important role in the pathogenesis of F. psychrophilum [46]. Based on these data, we suggest that some chaperone proteins may be important for in vivo survival and pathogenesis of Y. ruckeri.
A number of cell division proteins (BolA, DedD, DamX, FtsA, FtsE, FtsH, FtsP, FtsZ, ZapA, ZapB and ZapD), chromosome partitioning proteins (ParA, ParB, MukB and MukE) and biosynthesis proteins (iscR, MraZ, basR/pmrA, IF-1, IF-3, S2-S21, L1-L6, RsmA-RsmC and RsmG-RsmI) were identified. The FtsZ and ParA proteins have been identified as potential drug targets against clinically important bacterial pathogens [47]. Protein synthesis (transcriptional and translational) proteins have been targeted for inhibition of bacterial pathogens [48]. However, cell division and chromosome partitioning proteins may act as new drug targets for Y. ruckeri. Additionally, we predicted 12 antibiotic resistance classes (Table 3) in the Y. ruckeri proteome, particularly for cys regulon transcriptional activator CysB, bifunctional polymyxin resistance protein ArnA, copper-sensing two-component system response regulator CpxR and isoleucine-tRNA ligase. We observed intermediate susceptibility of aminoglycoside (gentamicin, MIC = 4-8 μg/mL) and polymyxin B (MIC = 4 μg/mL) antibiotics against Y. ruckeri strains. Similar results were previously reported in French Y. ruckeri isolates with aminoglycoside (gentamicin) [49] and greatest variation (MIC = 2-512 μg/ mL) in antibiotic sensitivity of polymyxin B was reported among Y. ruckeri strains [50]. These higher MIC values suggest that Y. ruckeri strains may harbor acquired or intrinsic resistance mechanisms to aminoglycosides and polymyxin B. Additionally, our Y. ruckeri strains were highly resistant to erythromycin (MIC = 1024 μg/mL) and rifampin (MIC = 32 μg/mL), consistent with observations by Calvez et al. [49] and Stock et al. [51], who found Y. ruckeri strains to be resistant to erythromycin (MIC = 32-64 μg/mL) and rifampin (MIC = 8-16 μg/ mL). Erythromycin and novobiocin discs did not show inhibition zone against the Chinese Y. ruckeri strain H01 [52]. Similarly, we did not observe any inhibition zone of novobiocin and mupirocin discs against the Y. ruckeri strains examined. Inherent resistance to erythromycin and rifampin has been described for the other Yersinia species (Y. enterocolitica, Y. mollaretii and Y. aldovae) [51]. Our results support these findings and suggest that Y. ruckeri strains might also be resistant to novobiocin and mupirocin. Moreover, two efflux pumps of the RND family were identified. This family is widespread among Gram negative bacteria and, in Enterobacteriaceae such as E. coli, contributes to the intrinsic resistance against several antibiotics, including macrolide and novobiocin [53]. This is consistent with our present results that found Y. ruckeri to be resistant to both antibiotics. Finally, it is important to note that the antimicrobial agents used to validate the results of our antibiotic resistance ontology are generally not approved for use in aquaculture. Y. ruckeri strains are susceptible to commonly applied antimicrobial agents such as florfenicol and oxytetracycline to treat fish diseases [49].
In conclusion, our study provides the first global proteomic profiles of Y. ruckeri and this work will provide a better understanding of the physiology, proteomic biology, proteomic changes, virulence mechanisms and localization of Y. ruckeri proteins. The most commonly expressed proteins such as SOD-Cu-Zn and PstS might be useful to develop a single vaccination protocol or single drug therapy for both biotype 1 and biotype 2 strains. Additionally, proteins associated with virulence and antigenicity such as Clp and Lon pro-teases, TolB, PPIases, PhoP and LuxR family transcriptional regulators may be used for the construction of novel vaccines for yersiniosis in fish. The comprehensive data set generated in this study will serve as a reference proteome for future studies such as protein-protein interaction and network analysis.

Data deposition
Shotgun proteomics data have been deposited in the Pro-teomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository [58] with the dataset identifier PXD005439.