Molecular characterization of Histomonas meleagridis exoproteome with emphasis on protease secretion and parasite-bacteria interaction

The exoproteome of parasitic protists constitutes extracellular proteins that play a fundamental role in host-parasite interactions. Lytic factors, especially secreted proteases, are capable of modulating tissue invasion, thereby aggravating host susceptibility. Despite the important role of exoproteins during infection, the exoproteomic data on Histomonas meleagridis are non-existent. The present study employed traditional 1D-in-gel-zymography (1D-IGZ) and micro-LC-ESI-MS/MS (shotgun proteomics), to investigate H. meleagridis exoproteomes, obtained from a clonal virulent and an attenuated strain. Both strains were maintained as mono-eukaryotic monoxenic cultures with Escherichia coli. We demonstrated active in vitro secretion kinetics of proteases by both parasite strains, with a widespread proteolytic activity ranging from 17 kDa to 120 kDa. Based on protease inhibitor susceptibility assay, the majority of proteases present in both exoproteomes belonged to the family of cysteine proteases and showed stronger activity in the exoproteome of a virulent H. meleagridis. Shotgun proteomics, aided by customized database search, identified 176 proteins including actin, potential moonlighting glycolytic enzymes, lytic molecules such as pore-forming proteins (PFPs) and proteases like cathepsin-L like cysteine protease. To quantify the exoproteomic differences between the virulent and the attenuated H. meleagridis cultures, a sequential window acquisition of all theoretical spectra mass spectrometric (SWATH-MS) approach was applied. Surprisingly, results showed most of the exoproteomic differences to be of bacterial origin, especially targeting metabolism and locomotion. By deciphering such molecular signatures, novel insights into a complex in vitro protozoan- bacteria relationship were elucidated.


Introduction
Histomonas meleagridis is a unicellular microaerophilic flagellate pathogen causing histomonosis (blackhead disease) in gallinaceous birds with a worldwide prevalence. Histomonosis in poultry is of significant importance, as it causes high mortality in turkeys and production losses in chickens [1]. The disease was well controlled in the past with the application of nitroimidazoles and nitrofurans or arsenicals for therapy or prophylaxis. However, due to new legislation in the European Union; the USA and elsewhere in the early 2000s, the use of such therapeutic or prophylactic drugs in food-producing animals became obsolete [2]. This led to the re-emergence of the disease with an increased incidence of outbreaks in poultry flocks, making histomonosis endemic again. Since its' re-emergence, molecular data on this 'new' disease have been accumulating [3][4][5][6][7][8][9]. Molecular investigations of H. meleagridis predominantly relay on in vitro culture, in which the pathogen is isolated from the intestinal content of naturally infected birds and propagated in cell culture media. However, such cultures often constitute an accumulation of diverse poultry-specific caecal microbes including protozoa and bacteria, hampering histomonad specific analysis. To overcome such an impediment, a clonal culture (mono-eukaryotic) of H. meleagridis was established via micromanipulation technique, in which the parasite could be traced back to a single cell [10]. Prolonged in vitro sub-culturing (passaging) resulted in the attenuation of the parasite [11]. Although the establishment of the mono-eukaryotic culture facilitated protozoan specific analysis, the background of coexisting ill-defined xenic bacterial flora complicated advanced molecular studies on H. meleagridis, particularly those encompassing '-omics' approaches. This necessitated the development of monoxenic culture, in which the xenic bacterial background of the mono-eukaryotic culture was exchanged with that of a single bacterial strain, the Escherichia coli DH5α, thereby generating a 'monoxenic-mono-eukaryotic' culture [12]. By exploiting such a defined culture, a recent study on the transcriptome of H. meleagridis disseminated a plethora of genomic information originating from 3356 unique transcripts [8]. Consequently, with the advent of a sizable volume of genomic data, further '-omic' investigations on this protist were expedited. For example, comparative proteome study performed on H. meleagridis unravelled the difference between the native virulent strain and its' mitigated attenuated strain [9].
In the current study, we focused on the role of H. meleagridis derived extracellular proteins (exoproteins/exoproteome), in particular secreted proteases. Exoproteome comprises of extracellular proteins stemming from cellular secretions (classical/non-classical mechanisms), exosome derived proteins, other export proteins and cell lysis [13]. In addition, it includes surface proteins (surfaceome), primarily membrane proteins with their domains exposed towards the extracellular space, which could have been scraped off the membrane due to experimental abrasions. Only the most stable proteins remain in exoproteome for further examination [13], hence their analysis provides invaluable information about the micro-environment of a particular cell type/biological system.
Pathogens are known to release a wide array of toxins and virulence factors including proteases into their cellular environment, which in turn can aid the development of a disease [13]. The role of exoproteins has been addressed in the context of pathological processes for a number of parasitic diseases, including trichomonosis [14,15]. Exoproteins present themselves at the interface milieu between the host and the pathogen, therefore being fundamental molecules that facilitate host-pathogen interactions [13]. It can be hypothesized that the establishment of histomonosis within the host relies on multifactorial host-parasite interactions, involving both contact-dependent and contact-independent mechanisms. Exploring virulent factors expressed in the H. meleagridis exoproteome constitutes a fundamental step towards better understanding of molecular mechanisms underlying the pathogenesis of histomonosis.
In the present study we dissected the H. meleagridis exoproteome, in order to mine for virulent factors and to obtain a comprehensive molecular outline. By applying traditional gelbased zymography and the high resolution micro-LC-ESI-MS/MS, we demonstrated the active secretion of lytic factors such as proteases by the parasite. In addition, we obtained a plethora of molecular data on extracellular environment, which at the same time, reflected the H. meleagridis-bacteria relationship.

Extraction of parasite exoproteome
Both, the virulent and the attenuated H. meleagridis cultures, were maintained in sterile T-75 cm 2 closed cap tissue culture flasks (Sarstedt, Wiener Neudorf, Austria) at 40˚C for 50-56h representing the logarithmic phase of growth [16]. Following this, a protocol employing serum-free culture was executed. Briefly, the H. meleagridis cultures were transferred into sterile 50mL falcon tubes (Sarstedt, Wiener Neudorf, Austria) and centrifuged at 200x g for 5 min, the supernatant was discarded and the parasite pellet was gently washed using pre-warmed serum-free culture media. This step was repeated five times. Following this, the parasite cultures were pulled down to a 30mL serum-free stock aliquot, which was used to adjust the parasite concentration in the serum-free cultures. Inoculums of 30mL (1x10 6 parasites/mL) serum-free cultures were prepared using pre-warmed serum-free culture media and 2mL overnight E. coli DH5α culture incubated in the same serum-free medium. The inoculums were incubated in sterile T-25 cm 2 closed cap tissue culture flasks (Sarstedt, Wiener Neudorf, Austria) for different duration (0h, 2h, 4h, 6h, 12h, and 24h). In addition, 3 replicates of the 6h culture were prepared to be utilized in the protease inhibitor susceptibility assay and for shotgun proteomics analysis. During the incubation periods, the cell integrity was monitored under a light microscope. Cell numbers were counted using Neubauer counting chamber and cell viability was determined by trypan blue exclusion test. Throughout the experimental procedure, cells were treated gently to minimize cell lysis. Cell-free filtrates were prepared from the 30mL inoculums by differential centrifugation: 200x g for 5 min to remove parasites followed by 2600x g for 10min to remove bacteria. The supernatant was filtered through 0.22μm cellulose acetate filters (Millipore, VWR) to obtain the cell-free filtrate containing the exoproteome from the culture. For the protease inhibitor susceptibility assay and shotgun proteomics, different protease inhibitors were added at this step with concentrations mentioned below. Afterwards, the cell-free filtrate was concentrated from its' initial 30mL volume to~100μL. This enrichment of the sample was achieved by performing ultrafiltration, using two types of filters. Firstly, Centriprep Ultracel YM-3; 3000 NMWL (Millipore, Vienna, Austria) was used to reduce the volume from 30mL to 1mL and in the second step Amicon Ultra 0.5mL Ultracel 3K; 3000 NMWL filter (Millipore, Vienna, Austria) was implemented to obtain a final volume of~100μL. Both types of filters were used following manufacturer's instructions. Finally, the protein concentration of each sample was measured using Bradford Protein Assay (Fermentas, Thermo Scientific).
Acquisition of proteins for identification runs. Peptides were separated on an Eksigent NanoLC 425 system using a microflow pump module (Sciex, Framingham, MA, USA). Sample pre-concentration and desalting were accomplished with a 5 mm YMC-Triart C18 precolumn (500 μm inner diameter, 3 μm particle size, and 12 nm pore size) (YMC, Dinslaken, Germany). For sample loading and desalting, ultra pure LC-MS grade H2O with 0.1% formic acid (FA) was used as a mobile phase with a flow rate of 10 μl/min. Separation of peptides was performed on a 15 cm YMC-Triart C18 column (300 μm inner diameter, 3 μm particle size, and 12 nm pore size; YMC, Dinslaken, Germany) with a flow rate of 5 μl/min. The gradient started with 3% B (ACN with 0.1% FA) and increased in two steps to 25% B (68 min) and 35% (73min). It was followed by a washing step with 80% B. Mobile Phase A consisted of ultra pure H2O with 0.1% FA. For mass spectrometric analysis the LC was directly coupled to a high resolution quadrupole time of flight mass spectrometer (Triple TOF 5600+; Sciex, Framingham, MA, USA). For information dependent data acquisition (IDA runs,) MS1 spectra were collected in the range of 400-1250 m/z for 250 ms. The 40 most intense precursors with charge state 2-4, which exceeded 150 counts per second, were selected for fragmentation. MS2 spectra were collected in the range 200-1500 m/z for 50 ms. Precursor ions were dynamically excluded from reselection for 13 s. The HPLC system was operated by Eksigent Control Software version 4.2 (Sciex, Framingham, MA, USA) and the MS by Analyst Software 1.7.1 (Sciex, Framingham, MA, USA).
Acquisition of SWATH data. The LC parameters employed were identical to the IDA runs described above. Separation of peptides was performed on a 15 cm YMC-Triart C18 column (300 μm inner diameter, 3 μm particle size, and 12 nm pore size; YMC, Dinslaken, Germany) with a flow rate of 5 μl/min. The gradient started with 3% B (ACN with 0.1% FA) and increased in two steps to 25% B (38 min) and 35% (43min). For information independent data acquisition (SWATH runs) MS1 spectra were collected in the range of 400-1250 m/z with an accumulation time of 50 ms. Product ion spectra were collected in 70 windows in the range of 400-1250 m/z with a width of 4.8 to 323.9 Da depending on the density of precursor masses in the mass segment. For each window ions were accumulated for 50 ms. Parameters for database search were: trypsin digestion, cysteine alkylation set to iodoacetamide, search effort set to rapid ID. False discovery rate analysis (FDR) was performed using the integrated tools in ProteinPilot Software version 5.0 (Sciex, Framingham, MA, USA). Global false discovery rate was set to <1% on protein level. For quantification of proteins, IDA 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, USA). Peptides were chosen based on a FDR rate <1%, excluding shared and modified peptides. Up to 6 peptides per protein and up to 6 transitions per peptide were used. MarkerView 1.2.1 (Sciex, USA) was used for calculation of peak areas of SWATH samples after retention time alignment and normalization using total area sums.
Statistical evaluation using R programming language. Statistical evaluation was performed using R programming language. cRAP proteins and proteins quantified with just one peptide were removed from Marker View raw protein list before further processing. Raw peak areas after normalization to total area sums were log2-transformed to approach a normal distribution. On a logarithmic scale, technical replicates were aggregated by arithmetic mean before application of statistical tests. Differential expression of proteins was assessed using two-tailed t-Test for independent samples for each protein. To adjust for multiple testing, the method of Benjamini and Hochberg was used to control the FDR. Protein expression was considered differential if the adjusted p-value was below α = 0.05 and the absolute fold change was at least two (fold change < −2 or > +2).

Bioinformatics analysis
Gene ontology, Biological Process (BP) clustering. Using pre-existing annotations from the H. meleagridis transcriptome [8] the identified proteins were distributed into 12 categories according to their biological process.
Prediction of secreted proteins. In order to detect structural features indicative for protein secretion by detecting N-terminal signal peptides predicting a possible secretion, the translated amino acid sequences of the 176 H. meleagridis exoproteins were submitted to Sig-nalP 4.1 server [18] using default parameters and organism group: eukaryotes. The SecretomeP 2.0 server [19] employing default parameters and organism group: Gram-negative bacteria, was used to predict protein secretion via non-classical pathways independent of signal peptides. Additionally, the E. coli exoproteins identified by SWATH analysis to be differentially expressed were also subjected to analysis via SignalP 4.1 and SecretomeP 2.0 server.
Clustering of proteins quantified by SWATH-MS analysis. Proteins quantified to be differentially expressed between the virulent and the attenuated H. meleagridis exoproteomes were clustered according to their biological process. A color coded cytoscape [20] network was generated displaying the n-fold over-expression of proteins from each H. meleagridis culture.

Growth of H. meleagridis under serum-deprived condition
In order to assure cell integrity, H. meleagridis cells were monitored under serum-free conditions. Under such conditions, the cell count of virulent H. meleagridis remained constant up to 6h, which was followed by a steady decline. This indicated that upon the medium exchange, cells stopped to divide but the progressive cell death started only after 6 hours of incubation in medium without serum (Fig 1). Only viable cells were observed, suggesting bursting of dead cells. The attenuated H. meleagridis displayed a slight increase in cell number within first 2 hours of incubation, which was followed by a plateau until 6 hours and steady decline in cell count after 6 hours in medium without serum (Fig 1). Since in both cultures, trophozoites started to die after 6h of incubation in serum-free medium, this time point was selected for detailed analysis of exoproteomes.

In vitro secretion kinetics of proteases by H. meleagridis
The kinetics of proteases actively released by virulent and attenuated H. meleagridis cells into the culture media was evaluated from exoproteomes after incubating both strains in serum-free conditions for 0h, 2h, 4h, 6h, 12h, and 24h. Analysis of both exoproteomes by 1D-IGZ revealed at least six prominent protease activities corresponding to the molecular masses (MM) of approximately 17kDa, 22 kDa, 30 kDa, 36 kDa, 68 kDa and 120kDa (Fig 2). The proteases displayed a time-dependent accumulation in the exoproteomes, with their activity appearing at 2h and reaching a maximum 12h onwards (Fig 2), which indicated an active secretion. Comparison of the zymograms between the virulent and the attenuated H. meleagridis revealed a stronger protease activity of low MM proteases (< 40kDa) in the former at any time-point except 0 hours (Fig 2A and 2C). The pattern of high MM proteases did not differ between two strains. Examination of the 1D SDS-PAGE gels revealed a gradual build-up of proteins in both strains (Fig 2B and 2D).

Protease inhibitor susceptibility assay
The mechanistic class of proteases in the H. meleagridis exoproteome was characterized based on their susceptibility to different protease inhibitors. The exoproteomes from both, virulent and attenuated H. meleagridis cells, were treated with the protease inhibitors: TLCK, E-64, PMSF and EDTA and subjected to 1D-IGZ analysis. The susceptibility of six protease bands at MM of~17kDa, 22 kDa, 30 kDa, 36 kDa, 68 kDa and 120kDa was most evident (Fig 3). A dose-dependent manner of protease inhibition was observed, with the effect of TLCK being most profound, followed by E-64, PMSF and EDTA. In addition, strain dependant variations in the minimal concentration of inhibitor needed to suppress the protease activity were noticed. For TLCK, a potent dual inhibitor of cysteine and serine proteases, a concentration of 135μM was sufficient to subdue the majority of the protease activity in the attenuated H meleagridis exoproteome (Fig 3A). In contrast, a higher concentration of 540μM was required to achieve similar effect in the virulent exoproteome ( Fig 3A). The inhibitor E-64, which selectively inhibits cysteine proteases, showed a similar inhibitory effect than TLCK. Again, much lower concentration (70μM) was sufficient to subdue the majority of the protease activity in the attenuated exoproteome, in contrast to its' virulent counterpart that needed 270 μM E-64 for suppression of protease activity (Fig 3B). To determine the distribution of serine proteases, inhibitor PMSF was applied. In general, the effect of PMSF on the exoproteome of either H. meleagridis strain was almost non-existent ( Fig 3C). Exceptional to this is the proteolytic activity at~68kDa, which exhibited susceptibility to the PMSF in the attenuated exoproteome at 500 and 1000μM. In contrary to this, even with the highest concentration applied (1000μM) the virulent exoproteome was not influenced by PMSF. The distribution of metalloproteases was assessed by addition of EDTA, however its' inhibitory effect remained largely inert to exoproteomes of both strains (Fig 3D).

Shotgun proteomics to identify proteins in the H. meleagridis exoproteome
Following ESI-MS/MS, the resulting peptides were analysed and mapped to the H. meleagridis reference transcriptome (3356 non-redundant transcripts) [8] identifying a total of 176 proteins with at least two peptides displaying confident ion scores, at false discovery rate (FDR) of <1% (S1, S2 and S3 Tables). The 176 proteins were grouped into 12 categories based on the gene ontology, biological process (BP) (Fig 4). The 176 proteins identified in the exoproteome represented 5.2% of the H. meleagridis transcriptome. The SignalP algorithm detected 17 proteins possessing the N-terminus signal peptide, whereas SecretomeP 2.0 algorithm predicted 37 proteins to be secreted independently of signal peptides via non-classical pathways (S1 Table). In addition to H. meleagridis proteins, the analysis identified 647 E. coli proteins (S2 and S3 Tables).

Differentially expressed proteins quantified by SWATH-MS analysis
Exoproteomes from three biological replicates of virulent and attenuated H. meleagridis cultures were compared by SWATH-MS to determine the difference in protein expression. The SWATH data from two injections for each replicate were analysed by ProteinPilot Software version 5.0 (Sciex, USA). Data analysis (p-value < 0.05; fold change < −2 or > +2) resulted in the identification of 35 differentially expressed proteins (Fig 5). Out of the 35 proteins, two proteins, a C2 domain containing protein and a RNA binding protein, were mapped exclusively to the virulent H. meleagridis ( Table 1). The remaining 33 proteins were mapped to E. coli, with 17 proteins over-expressed in the exoproteome of the virulent (Table 1) and 18 in the attenuated culture ( Table 2). The 17 E. coli proteins identified as over-expressed in the virulent H. meleagridis background included proteins involved in bacterial motility, carbohydrate membrane transport and carbohydrate metabolism (Table 1), whereas the 18 E. coli proteins over-expressed in the attenuated exoproteome can be related to the stress response toward the parasite ( Table 2).

Discussion
The present study focused on investigating the exoproteomes of H. meleagridis, derived from virulent and attenuated cells cultivated in serum-free conditions. Since serum-free conditions were a necessary step to prevent the unspecific results from serum proteins [21], the cellular integrity of H. meleagridis was monitored for any adverse effects arising from such conditions [22]. Despite showing no active growth under serum-free conditions [23], slight differences between the growth of two H. meleagridis cultures were noticed. In contrast to virulent histomonads, the cell count of attenuated parasites slightly increased within first two hours of incubation. This could be due to the fact that at the start of the serum-free experiment, cells were taken from the mid logarithmic phase and just continued their programmed cell division. Attenuated H. meleagridis cells are well adapted to in vitro growth conditions in complete   [16]. It is therefore plausible that attenuated histomonads needed slightly longer to completely cease their growth. This observation is coherent with the recent study reporting in vitro phenotypic differences between the two strains, where drastic changes in temperature and pH levels that can account as 'stress factors', resulted in pronounced effects in the virulent H. meleagridis culture [16].

Species-Contig ID/Accession
The exoproteomes, used in the experiments with protease inhibitors and for proteomic analysis, were extracted from cultures incubated for 6h. At this time point, drastic cell death could not be detected for both strains and sufficient protein for analysis could be harvested. Therefore, it can be anticipated that the majority of proteins originated from an active secretion, as demonstrated by the in vitro secretion kinetics of proteases by H. meleagridis. Furthermore, since discrimination between proteins sequestered inside extracellular vesicles from their soluble counterparts was not performed, a sizable proportion of the proteins derived from exosomes should be expected.
The biological role of proteases has been extensively discussed in the context of parasitic protozoa, including their role in pathogenesis [24][25][26]. Applying 1D-IGZ investigations, a Table 2 wide spread distribution of proteases was detected ranging from low to high MM, similar to results obtained for Trichomonas gallinae and Trichomonas vaginalis [27,28]. It can be predicted that these molecules play an important role in the H. meleagridis micro-environment, since their secretion was noticed already after 2h of incubation. The existence of these lytic factors in the exoproteome of H. meleagridis supports the notion of nutrient acquisition by attacking the co-cultivating E. coli.

Species-Contig ID/Accession
Inhibitory effects of TLCK and E-64 were the most profound, suggesting that the majority of H. meleagridis extracellular proteolysis is carried out by cysteine proteases. This is not surprising given the fact that transcripts encoding for cysteine proteases were the most over-represented protease family in the H. meleagridis transcriptome [8]. The results of the TLCK and E-64 inhibition tests demonstrate that the activity of cysteine proteases within the virulent H meleagridis exoproteome displays higher robustness than the one within the attenuated exoproteome. Higher concentrations of TLCK and E-64 were required to subdue the protease activity of virulent exoproteome, showing reduced sensitivity to the inhibitors in contrast to its' attenuated counterpart. Cysteine proteases have been implicated in virulence of a number of pathogens, including parasitic protozoa such as trichomonads [29][30][31]. For instance, from the study on T. gallinae [27] one could easily emphasize the association of cysteine proteases with the pathogenicty, as the addition of TLCK and E-64 averted the cytotoxic effect of the parasites' exoproteome on LMH cells. Mass spectrometric investigations of the T. gallinae proteases revealed the action of a particular clan of cysteine protease to be the major cytotoxic molecule; the C1 family cathepsin L-like cysteine protease [27]. Likewise, research on the human pathogen T. vaginalis highlights the role of its' cysteine proteases in the context of pathogenesis [28,32]. Particularly cathepsins have been demonstrated to facilitate the parasites' adherence to host cells and induce cellular damage [33][34][35].
The shotgun proteomics analysis of the H. meleagridis exoproteome identified three cysteine proteases represented by C1 family cathepsins and C13 family legumains. The majority of the cysteine proteases identified here had a low MM range (30-40 kDa region), with two cathepsins and one legumain protease displaying a theoretical MM of 34.88 kDa, 35.65 kDa and 47.25 kDa, respectively. Correlating this data to the 1D-IGZ results, it seems that the majority if not all of the proteolytic activities up to~40 kDa in the H. meleagridis exoproteome, stem from the cysteine proteases. Discrepancy of few kDa between the theoretical and actual MM as observed in here is not unusual and discrepancies of up to 13 kDa have been reported [36]. Proteins often undergo chemical changes due to posttranslational modifications such as glycosylation and ubiquitination/sumoylation [37], which can shift their mobility pattern during electrophoresis. Alternatively, the reported H. meleagridis transcripts may not be of full length.
The H. meleagridis exoproteome remained predominantly inert to PMSF, suggesting a sparse distribution of serine proteases. In agreement with this, the shotgun proteomic approach identified only a single serine protease in the exoproteome with one confident peptide (theoretical MM:18.68kDa, S2 Table). Interestingly, the only protease activity which displayed some susceptibility to PMSF during 1D-IGZ analysis was the protease at~68kDa in the exoproteome of attenuated H. meleagridis. The activity of this protease was more robust in the exoproteome of virulent H. meleagridis in which it persisted even with the addition of 1000μM PMSF. As shotgun proteomics analysis was negative for higher MM serine proteases, the effects of PMSF might be due to the presence of other class of proteases. In addition to serine proteases, PMSF also inhibits clan C1 family peptidases to certain extent, the only class of cysteine protease susceptible to this inhibitor [38]. Based on the sensitivity of the~68kDa protease band to TLCK and E-64, it seems very likely that this is a cathepsin, which is also susceptible to higher concentrations of PMSF.
The shotgun proteomic approach identified metalloproteases belonging to the M20 family T-like metallopeptidase, M24 family aminopeptidase and M49 family metalloproteases. However, during the 1D-IGZ analysis, the H. meleagridis exoproteome remained overall inert to EDTA, suggesting negligible presence of metalloproteases. Such a discrepancy may arise from amino acid sequence homology analysis, which identified several of these proteins as metalloproteases. However, they might also represent non-peptidase homologues, which carry out metal binding activity instead of actual proteolysis [39]. An alternative explanation might be that the metalloproteases identified here are inactive zymogens, which require specific modifications under appropriate conditions to become active.
Unlike the differences in proteolytic intensities between the virulent and the attenuated H. meleagridis exoproteome noticed by 1D-IGZ analysis, the applied SWATH-MS approach could not quantify any differences in protease expressions between the two strains. It appears that the lack of protease inhibitors in protease secretion kinetics assays, as opposed to the 1D-IGZ inhibitor assays and the mass spectrometric analysis, actually demonstrated the potency of proteases in the virulent strain rather than its' expression levels. This is corroborated by the results of protease inhibitor susceptibility assay, in which the lowest effective concentration of all tested inhibitors was much higher for the virulent as opposed to the attenuated exoproteome.
The shotgun proteomic approach identified 176 proteins with at least two confident peptides, which accounts for 5.2% of the H. meleagridis transcriptome. By utilizing the pre-existing annotations from the H. meleagridis transcriptome database [8], the proteins were classified into twelve categories based on their biological process. In silico analysis by SignalP and Secre-tomeP algorithms, showed that proteins secreted by non-classical pathways are dominant in the exoproteome, although the very small proportion of proteins from a sporadic cell lysis cannot be excluded. The actual results suggest that H. meleagridis predominantly exports proteins by means of non-classical pathways in contrast to traditional signal peptide based motifs, which might prove to be advantageous to the parasite. Proteins not dictated by premeditated export signal peptide motifs might actually carry out two unrelated functions, within the cytoplasm and in the extracellular milieu, demonstrating a moonlighting property [19]. This could facilitate H. meleagridis survival and/or virulence by maximizing its' proteomic functions.
The identification of actin and several gylcolytic enzymes, such as enolase, fructose bisphosphate aldolase (FBA), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), and triosephosphate isomerase (TPI), in the H. meleagridis exoproteome is particularly intriguing. Common traits of all these proteins are their alternative localization on the parasite's surface and the capacity to bind plasminogen [40]. The recent comparative proteome studies of H. meleagridis, identified actin, enolase, FBA and GAPDH as up regulated by virulent histomonads [9,41]. However, this only reflected spots localized on positions that differed to the bulk of both proteins, indicating possible post-translational modification(s) and/or truncation(s). Both studies suggested that such modified actin, enolase, GAPDH and FBA molecules might be components of alternative mechanism (moonlighting) unrelated to cytoskeleton and carbohydrate metabolism. Similar observations, involving the up regulation in the virulent strain and alternative positions on 2D-proteome map, were also reported for T. vaginalis [42,43] Binding of plasminogen to these parasites' surface exposed molecules, is a crucial step in the activation of plasminogen to plasmin, a serine protease [44]. Using this mechanism, the parasite interacts with the fibrinolytic system of the host, which is involved in degradation of fibrin and extracellular matrices [44]. The presence of actin, enolase, GAPDH, FBA and TPI molecules in the exoproteome of H. meleagridis could therefore be related to their tentative plasminogen-binding role. Considering that the dissemination of H. meleagridis from caecum to liver and other organs occurs via blood, a mechanism involving the interaction with host's fibrinolytic system might be of great importance. Beside potential plasminogen binding proteins, another group of moonlighting proteins, such as adhesins AP-33, AP-51, and AP-65, was identified in the H. meleagridis exoproteome. Classically localized in hydrogenosomes as metabolic enzymes, these proteins have been reported to be surface attached, mediating cytoadherence to host cells [45]. An earlier study, reported only the H. meleagridis AP-33 and its' dual localization, both in the hydrogenosomes and on the cell surface, whereas AP-65 could only be detected within the hydrogenosome [7]. The results presented here, suggest that besides of being intracellular, AP-33, AP-51 and AP-65 are also localized on the cell surface or secreted by H. meleagridis, which could reinforce their moonlighting property.
Data mining also revealed the presence of several lytic molecules in the H. meleagridis exoproteome particularly surfactant B/saposin-like proteins, zinc dependent phospholipase C and N-acetylmuramoyl-L-alanine amidase. Most of these lytic molecules operate by attacking target cell membrane and compromising cellular integrity [46][47][48][49]. In the present context, these molecules are most likely exploited by H. meleagridis to launch assaults against co-incubating bacteria, a function they might display in vivo against host cells.
The SWATH-MS approach was applied to obtain a quantitative proteome expression between the virulent and the attenuated H. meleagridis cultures. The majority of the differentially expressed proteins were of E. coli origin whose detection was not surprising due to the monoxenic co-cultivation with E. coli DH5α. The applied statistical parameters filtered only two H. meleagridis proteins as differentially expressed, a C2 domain containing protein and a RNA-binding protein, both over-expressed in the virulent exoproteome. Both proteins have a wide spectrum of functions [50,51]. The C2 domain is a membrane binding motif found in a broad and diverse set of eukaryotic proteins [50]. The majority of the C2 domains function as phospholipid binding moieties relying on calcium (Ca 2+ ), hence sometimes referred to as a Ca 2+ dependent lipid binding domains [52]. Lipid-binding domains (LBDs) can engage in a wide set of molecular functions, such as membrane anchoring of the cytoskeleton [53,54]. The identification of such a protein in the exoproteome stems most likely from experimental conditions, which introduced a removal of an extracellular domain of such membrane bound protein. Its' up-regulation in the virulent H. meleagridis, might indicate a possible cytoadherence mechanism that involves the interaction of the parasite's C2 motif protein and host cells.
The RNA-binding proteins engage in the formation of ribonucleoprotein complexes by binding to sequence/structural motifs in the RNA, which can alter the fate of that RNA molecule [51]. The presence of parasite derived extracellular RNA (exRNA) recognition motif can thus have far-reaching implications. Such proteins can serve as 'traps' targeting host cell exRNA, which upon binding can aggravate host susceptibility. It is noteworthy to point that, exRNAs particularly in extracellular vesicles are an important source of cellular communication both in protists and multi-cellular organisms [55,56].Therefore one can hypothesize on the implication of H. meleagridis derived exRNA-binding motif in the disruption of the host cell signalling.
It is intriguing to consider why only two H. meleagridis proteins were detected as significantly over-expressed. One possible explanation would be that the serum-free conditions provoked a stress effect and by that possibly reduced the overall metabolism of both H. meleagridis strains which lead to a minimization of excreted proteins. Additionally, since the attenuated H. meleagridis strain originated from the virulent by prolonged in vitro cultivation, both strains are genetically very similar [11]. The recent H. meleagridis proteome analysis study somewhat argues for the first hypothesis, since the use of identical H. meleagridis strains and the SWATH MS approach identified 42 differentially expressed H. meleagridis proteins [41]. However, beside the fact that in this study the medium for cultivation of parasites contained serum, the analysed protein fraction was different and contained endogenous proteins. Taken together, it seems that even though the serum-free stress conditions have to be considered, the bulk of variations between the two H. meleagridis strains in in vitro conditions is endogenous The majority of differentially expressed proteins identified in the present study were of E. coli origin. This demonstrates the rapid change in the prokaryote's gene expression which is dependent on the parasite's phenotype. At the same time, it indicates a strong interaction between H. meleagridis and bacteria. This relationship is particularly thought-provoking, as the bacteria are not only required for proliferation of H. meleagridis in vitro; but their presence in the intestines of host birds is essential for successful infection [57]. Furthermore, different bacterial strains mediate diverse stimulatory effects on H. meleagridis growth in vitro [12]. A recent comparative proteomics study supports the idea of interdependence and complex interactions between H. meleagridis and bacteria [9]. However, the nature of the H. meleagridis -prokaryote relationship is still unresolved.
The E. coli proteins over-expressed in the H. meleagridis virulent background were involved in the flagellar motility system or carbohydrate uptake. It seems that the low glucose environment of Medium 199 induced the uptake mechanisms of carbohydrate sources, such as maltodextrin or D-allose. The MalE protein was also shown to be important in glucose transport under glucose-limited conditions [58]. Aside from this, a considerable number of proteins involved in flagellar motility were up-regulated, which presumably facilitated bacteria in accessing a more favourable environment [59]. The medium that was used in the present study contains low glucose and is supplemented with rice starch as an additional carbohydrate source. Since E. coli is unable to use rice starch due to the absence of endogenous amylolytic enzymes [60], glucose remains the only possible carbohydrate source. In contrast, H. meleagridis is able to hydrolyse starch molecules [8], and could potentially contribute to the increase of extracellular glucose. However, it is likely that the serum-free medium caused the virulent H. meleagridis to severely minimize/or stop metabolizing, since even in optimal conditions, this strain of the parasite is not well adapted to in vitro growth and shows much lower metabolic rates as compared to the attenuated parasite [9,16].
In contrast to the virulent culture, E. coli proteins up regulated during co-cultivation with attenuated H. meleagridis indicate a stress response toward the parasite. This particularly concerns ATP-dependent Clp protease (ClpX), hydrogen peroxide and cadmium resistance periplasmic protein (YgiW) and CsbD family protein, which were all described to be associated with stress response [61][62][63]. As adapted to the in vitro growth, the attenuated H. meleagridis seems to cope more successfully with the serum-free conditions than its' virulent counterpart [9,16]. It is possible that attenuated H. meleagridis phagocytised co-cultivating E. coli more effectively than the virulent parasite, which in turn resulted in the up-regulation of E. coli stress proteins. In addition, the up-regulation of the tryptophanase (TnaA) together with proteins involved in the formation of type -1 fimbriae (FimA and FimC), detected in the attenuated H. meleagridis exoproteome, can be related to the interaction with the parasite. Type-1 fimbriae are most likely involved in the direct contact with the parasite, since they enable the bacteria to bind specific receptor structures [64]. However, the up-regulation of the TnaA protein might indicate a potential defence mechanism of E. coli towards phagocytosis by the parasite. The tryptophanase is an enzyme responsible for the inter-conversion between tryptophan and indole [65]. The more favoured product of the reaction, the indole, has been proposed to act as an extracellular signalling molecule influencing a variety of biological processes, even playing role in inter-species communication signals [66]. In Vibrio cholerae, active tryptophanase and the presence of indole are related to grazing resistance against the phagocytic eukaryote Dictyostelium discoideum [67]. Whether a similar mechanism is employed by E. coli-histomonad interaction has to be investigated in more detail. However, considering that a reduction of E. coli during co-cultivation with H. meleagridis was not observed [12], it is possible that the prokaryote employs some protective mechanism to protect from phagocytosis by the parasite. Finally, the strikingly high rate of overexpression of E. coli AdhP alcohol dehydrogenase in the exoproteome of attenuated H. meleagridis is especially intriguing. Unlike the more common alcohol dehydrogenase AdhE, the AdhP is an ethanol-inducible dehydrogenase whose expression is not linked to anaerobic conditions [68]. Its' up-regulation in the attenuated H. meleagridis culture indicates an increased concentration of ethanol in the media, whose likely contributor is the parasite itself. This situation resembles low-iron-dependent changes described for Tritrichomonas foetus, in which the pyruvate metabolism is shifted to the cytoplasm resulting in the production of ethanol and some glycerol [69]. The existence hydrogenosomal iron-sulphur proteins, such as PFOR, ferredoxin and iron-hydrogease in H. meleagridis, indicates the requirement for iron supplementation in order to maintain normal hydrogenosomal metabolism and energy production [8]. In in vitro conditions the serum represents an important iron source that is absent in the current experiment. Therefore, it is likely that due to the lack of iron, the majority of hydrogenosomal metabolism is inhibited and pyruvate is utilized in the cytoplasm, resulting in the production and subsequent secretion of the ethanol [69]. The E. coli AdhP enzyme, in turn, converts ethanol to acetaldehyde which can again be used by the protozoan cytoplasmic fermentive metabolism, releasing the reduced NADP+ molecules and maintaining redox balance. When compared to the virulent H. meleagridis culture, there seems to be no elevation in ethanol concentration in the media. Based on the serum-free growth and previous reports [9,16], it is likely that the virulent parasite, due its' overall in vitro maladaptive state, dramatically slowed down its' metabolism just for mere survival purpose. Overall, this underlines the complexity of the bacteria-parasite interaction.

Conclusions
The in vitro secretion kinetics assay of proteases demonstrated the significance of these lytic molecules in H. meleagridis extracellular milieu, showing a time-dependent secretion from as early as 2 hours post-incubation. Comparative 1D-IGZ analysis between the virulent and the attenuated H. meleagridis, displayed the robustness of proteases secreted by the virulent strain. The protease inhibitor susceptibility assay revealed cysteine proteases to be the predominant lytic molecules in the H. meleagridis exoproteome, with the higher potency in cultures of virulent parasites. The comprehensive analysis of the H. meleagridis exoproteome using high-resolution proteomics identified a pool of 176 proteins, providing an extensive molecular overview of the H. meleagridis extracellular milieu and indicating its' dynamic micro-environment. The quantitative exoproteome analysis of the virulent and the attenuated H. meleagridis cultures displayed differences in proteins of co-cultivating E. coli, which in turn reflected the protozoabacteria interaction for the first time at the molecular level. Differences between H. meleagridis-derived exoproteins were almost non-existent, suggesting that in in vitro conditions the majority of proteomic alterations are endogenous, as previously observed in our comparative proteome studies. Overall, the present study investigating the exoproteome of H. meleagridis provides an attractive avenue for further research, hypothesis testing and comparative analysis of strains with different biological properties.
Supporting information S1 Table. List of proteins identified in the H. meleagridis exoproteome. The list contains H. meleagridis specific proteins identified in the exoproteome with �2 confident peptides in at least 3 replicates. Information on SignalP and SecretomeP data, number of confident peptides, the % of protein sequence coverage (>95% confidence) and the number of ambiguous accessions is given. The total ProtScore means the score with all peptides (unique and shared). The Unused ProtScore is the score using only the unique peptides. Term "ALL" represents combined data for all 6 replicates. The LP1, LP2, and LP3 are 3 replicates of the H. meleagridis virulent strain (low passage, 28x). The HP1, HP2 and HP3 are 3 replicates of the H. meleagridis attenuated strain (high passage, 319x). (XLSX) S2 Table. Extended information on protein identifications. The file contains 15 folders. Folders named "Prot_HMx_all" and "Ambiguous Accessions_HMx_all" contain lists of all identified proteins and their ambiguous accessions. The total ProtScore means the score with all peptides (unique and shared). The Unused ProtScore is the score using only the unique peptides. Folders named "Prot_HMx_28_A", "Prot_HMx_28_B", "Prot_HMx_28_C" contain lists of proteins identified in the exoproteome of each biological replicate (A, B, C) of the low passsage H. meleagridis strain. Folders named "Prot_HMx_319_A", "Prot_HMx_319_B", "Prot_HMx_319_C" contain lists of proteins identified in the exoproteome of each biological replicate (A, B, C) of the high passsage H. meleagridis strain. Folders named "Ambiguous Accessions_HMx_28 or 319_A or B or C" contain lists of all ambiguous accessions for the given protein identified in the exoproteome of each biological replicate (A, B, C) of either LP (28)