Proteomic View of Interactions of Shiga Toxin-Producing Escherichia coli with the Intestinal Environment in Gnotobiotic Piglets

Background Shiga toxin (Stx)-producing Escherichia coli cause severe intestinal infections involving colonization of epithelial Peyer’s patches and formation of attachment/effacement (A/E) lesions. These lesions trigger leukocyte infiltration followed by inflammation and intestinal hemorrhage. Systems biology, which explores the crosstalk of Stx-producing Escherichia coli with the in vivo host environment, may elucidate novel molecular pathogenesis aspects. Methodology/Principal Findings Enterohemorrhagic E. coli strain 86–24 produces Shiga toxin-2 and belongs to the serotype O157:H7. Bacterial cells were scrapped from stationary phase cultures (the in vitro condition) and used to infect gnotobiotic piglets via intestinal lavage. Bacterial cells isolated from the piglets’ guts constituted the in vivo condition. Cell lysates were subjected to quantitative 2D gel and shotgun proteomic analyses, revealing metabolic shifts towards anaerobic energy generation, changes in carbon utilization, phosphate and ammonia starvation, and high activity of a glutamate decarboxylase acid resistance system in vivo. Increased abundance of pyridine nucleotide transhydrogenase (PntA and PntB) suggested in vivo shortage of intracellular NADPH. Abundance changes of proteins implicated in lipopolysaccharide biosynthesis (LpxC, ArnA, the predicted acyltransferase L7029) and outer membrane (OM) assembly (LptD, MlaA, MlaC) suggested bacterial cell surface modulation in response to activated host defenses. Indeed, there was evidence for interactions of innate immunity-associated proteins secreted into the intestines (GP340, REG3-γ, resistin, lithostathine, and trefoil factor 3) with the bacterial cell envelope. Significance Proteomic analysis afforded insights into system-wide adaptations of strain 86–24 to a hostile intestinal milieu, including responses to limited nutrients and cofactor supplies, intracellular acidification, and reactive nitrogen and oxygen species-mediated stress. Protein and lipopolysaccharide compositions of the OM were altered. Enhanced expression of type III secretion system effectors correlated with a metabolic shift back to a more aerobic milieu in vivo. Apparent pathogen pattern recognition molecules from piglet intestinal secretions adhered strongly to the bacterial cell surface.


Introduction
Enterohemorrhagic E. coli (EHEC) are a group of bacteria containing many serotypes responsible for outbreaks of bloody diarrhea occasionally leading to hemolytic uremic syndrome (HUS) and neurological abnormalities which in severe cases can be fatal [1].These complications are attributed to Shiga toxins, one of which -Shiga toxin-1 (Stx-1) -is shared with Shigella dysenteriae type 1, acquired via horizontal gene transfer through phages [2].The most common cause of outbreaks and sporadic cases of bloody diarrhea in the U.S. are strains of serotype O157:H7.EHEC strain 86-24 caused an outbreak of hemorrhagic colitis with a high mortality rate in 1986 [3] and produces only Shiga toxin-2 (Stx-2) [4].Pig is the only species besides human naturally susceptible to the toxins' systemic effects, and Stx-2 producers in particular have caused severe neurologic symptoms in the piglet model [5].Although conventional piglets are naturally not susceptible to EHEC, we consider the gnotobiotic piglet model of infection with EHEC strains particularly useful to obtain systems-level insights into the molecular pathogenesis associated with Stx-production and with the characteristic bacterial attaching-effacing colonic lesions.which resembles that observed in humans [6].
The first genome of an E. coli serotype O157:H7 strain (EDL933) was published by Perna et al. [2].A pathogenicity island in the genome was termed ''locus of enterocyte effacement'' (LEE).LEE-expressed proteins include the type III secretion system (T3SS) needle and translocator pore proteins (i.e., EspA, EspB and EspD), which fuse with the mammalian host cell membrane to allow the injection of effectors [7].One LEEencoded virulence factor is the transmembrane intimin receptor (Tir), which integrates into the host cell membrane and binds to the bacterial outer membrane (OM) protein intimin (Eae).This interaction triggers actin polymerization and pedestal formation, a pathological process characteristic of EHEC invasion of the colonic epithelium [7,8].The genes encoding the T3SS structural subunits, additional effectors manipulating host cell functions (e.g., EspF, EspG, and Map) and chaperones for the T3SS effectors also reside in the LEE gene cluster [7,8].The A/E lesions, which are formed following the epithelial cytoskeletal rearrangements, disrupt the boundary function of the mucosa, allow systemic release of Shiga toxins, and cause infiltration of inflammatory host cells.Nearly 40 potentially functional effectors not encoded by LEE were also identified [9] of which EspF U [10] and NleH [11] were further characterized.StcE is a secreted metalloproteinase cleaving and inactivating host glycoproteins such as mucin-7 and glycoprotein 340 [12].This enzyme is exported by a type II secretion system (T2SS) encoded by the virulence plasmid pO157.The transcription factor Ler regulates stcE and many of the LEE genes transcriptionally [13].Protein abundance profiles of secreted effectors confirmed the important regulatory functions of Ler and the integration host factor Ihf [14].Most of the effectors were absent in the secreted bacterial fraction of Dler and Dihf mutants of EHEC strain EDL933.
While the functional roles of many effectors and their transcriptional regulation in vitro and in vivo have been studied, far less information is available on system-wide adaptations of EHEC strains to their in vivo host environments.We reported the first comprehensive proteomic analysis of an EHEC strain  and identified more than 2500 proteins, but that study was not focused on environmental adaptation [15].In vivo host adaptation surveys were previously performed for Shigella dysenteriae serotype 1 (SD1) using a similar gnotobiotic piglet model; acid resistance systems, anaerobic energy metabolism pathways and the expression of a T3SS system that mediates epithelial cell invasion were strongly induced in SD1 cells isolated from the piglet intestinal environment [16,17].Transcriptome analysis was applied to examine the adaptation of E. coli O157 strains to environments simulating in vivo conditions, such as the presence of red blood cell membranes [18] and the uptake by human macrophages [19].Red blood cell membrane adherence resulted in higher gene expression for stress response regulators (rpoE, rseB, phoQ) and T3SS effectors (eae, tir, cesT, map, and espA), whereas the NtrC regulon responsible for nitrogen metabolism was strongly downregulated in the examined E. coli O157 strains [18].EHEC strain EDL933 revealed expression changes for 22% of all arrayed genes when residing in human macrophages compared to growth in vitro [19].In addition, genes encoding iron transporters (fit and chu), acid resistance systems, and the Shiga toxin subunits stx1B/2B were increased in expression in the macrophage milieu, while the SOS response repressor lexA and several SOS response genes were down-regulated.Oxidative burst in macrophages is associated with increased production of reactive oxygen and nitrogen species.Gene expression levels for the peroxide stress-associated regulator oxyR and the oxidative stress response enzymes katP, katE and katG were surprisingly unchanged [19].In a different study, decreased production of the signaling molecule indole and increased Curli biosynthesis were linked to enhanced biofilm formation in strain EDL933 compared to strain 86-24 [20].
The purpose of this study was to identify changes in the proteome of E. coli O157:H7, strain 86-24, resulting from physiological adaptations to piglet conditions that mimic hemorrhagic colitis in humans.Unlike macrophage and epithelial cell models, the gnotobiotic piglet model does not disregard the complexity and stages of infection as well as defensive responses by the host.We hypothesized that the analysis of proteome-wide changes allowed molecular insights into the pathogen's survival and at least partial reconstruction of metabolic networks important in vivo.Data on proteomic host adaptations were previously reported for Salmonella enterica Serovar Typhimurium (S. typhimurium) [21,22], and were extended towards the building of models of the pathogen's in vivo metabolic capacity [23].Here, we are not only taking steps into that direction, but also discovered innate immune response-associated proteins binding to the EHEC cell surface.

Ethics Statement
Experiments in piglets were performed to allow analysis of the E. coli O157:H7 proteome from in vivo environments.Piglets were delivered by caesarian section and housed at the Division of Infectious Disease of Tufts University School of Veterinary Medicine in accordance with approved procedures of the Institutional Animal Care and Use Committee at Tufts University.The animal studies were specifically approved by this committee.The respective protocols were No. G770-6, titled gnotobiotic piglet model of E. coli infection, date 4-24-2006, and No. G2009-52, titled gnotobiotic piglet model of E. coli infection, date 5-4-2009.

Bacterial Cell Recovery from Cell Cultures and Animal Experiments
Cells of the E. coli O157:H7 strain 86-24 were isolated from in vitro cell cultures and in vivo animal experiments.Cells were grown in vitro in Luria-Bertani (LB) broth to stationary phase (OD 600 ,2.0) in a shaking incubator at 37uC, pelleted by centrifugation at 7,0006g for 10 min at 4uC, washed with a 20fold volume of PBS and re-concentrated at 7,0006g for 15 min at 4uC.As for the animal experiments, several piglets developed diarrhea 18-30 h post-inoculation and were euthanized 3 to 5 days later.Bacterial cells (between 1610 9 and 2610 10 cells) were recovered from the piglets' gut contents, repeatedly washed with PBS and purified via density gradient centrifugation with an isotonic 65% Percoll solution at 14,5006g for 30 min at 4uC, as previously described [17].Up to 2610 10 bacterial cells were resuspended either in 1 ml of TTE lysis buffer (25 mM Tris-OAc, pH 7.8, 0.05% Triton X-100, 5 mM Na-EDTA, and benzamidine and AEBSF in 1 mM concentrations) to process samples via shotgun proteomics or in 0.5 ml GR lysis buffer (8 M urea, 2 M thiourea, 4% (w/v) CHAPS, 18 mM DTT and 0.5% (v/v) Bio-Lyte pH 3-10 carrier ampholytes) for 2D gel analysis.2D gel samples were frozen at -80uC until further use.

Protein Extraction from Cell Lysates and Digestion
Samples to be subjected to shotgun proteomics were supplemented with chicken lysozyme (150 mg/ml) and agitated at 20uC for 1 h followed by quick-freezing and storage at 280uC until further use.To disintegrate cells further and solubilize proteins, sonication of the samples was followed by nucleic acid degradation (DNAse I and RNAse, at 5 mg/ml) and lysate agitation for 1 h at 20uC.The supernatant and insoluble pellet for each sample were separated by centrifugation at 16,1006g, and the pellet fraction re-extracted with a solution including 2.5 M NaBr.The supernatant was concentrated to 2 mg/ml protein and applied to an analytical SEC column (G3000-SWXL; 7.8 mm630 cm; TO-SOH Bioscience, USA).Proteins were chromatographically separated in PBS supplemented with 0.01% Triton X-100 into fractions representing the M r segments .280kDa, 280-80 kDa and 80-10 kDa (fractions termed F1-s SEC , F2-s SEC and F3-s SEC , respectively).The insoluble pellet was termed F4-p.These four fractions containing roughly 60-100 mg protein were each subjected to digestion with trypsin using a method termed filteraided sample preparation (FASP) [24].Briefly, all four fractions for a given sample were concentrated in a Microcon YM-10 membrane filter unit (10 kDa M r cut-off; Millipore, Billerica, MA).Twenty ml of a 1 M DTT stock solution and 12 ml of a 10% SDS stock solution were added to denature the proteins for 3 minutes at 95uC.Following alkylation, proteolytic digestions (trypsin/bacterial protein ratio of 1:30 to 1:50) were performed at 20uC overnight.Filtrates contained the peptide mixture collected by centrifugation at 14,0006g and rinsed three times with 500 mM and 50 mM NH 4 HCO 3 solutions to recover the entire protein digestion mixture, and samples were lyophilized.Experimental details for all of the procedures presented here were published previously [15].

Two-dimensional Electrophoresis and Differential Gel Display
EHEC lysates in GR buffer were thawed, incubated at 20uC for 30 min and vortexed intermittently to complete protein solubilization.Whole cell lysates were centrifuged at 16,1006g for 30 min, and supernatants were subjected to protein quantification using the 2D Quant Kit (GE Healthcare, Piscataway, NJ).Supernatant samples were subjected to 2D gel electrophoresis in batches of 12 gels using published procedures [25,26].Briefly, 1 st dimension protein analysis in 24 cm immobilized linear pH gradient strips (pH range 4-7; GE Healthcare) included gel rehydration loading of samples with ,150 mg protein and electrophoresis for ,60,000 Vh.Following reduction and alkylation steps, re-equilibrated strips were applied to 2 nd dimension SDS-PAGE slab gel electrophoresis (25619.560.15cm; 8-18%T) for ,1,800 Vh.Gels were fixed, stained with Coomassie Brilliant Blue G250 (CBB), de-stained, subjected to gel image analysis (data acquisition as 16 bit TIFF images) and imported into the software tool Proteomweaver v4 (Bio-Rad, Hercules, CA).As described previously [25,26], the gel image analysis proceeded with spot detection, matching, normalization, intensity averaging and spot annotation steps.The in vitro group consisted of eight gels, derived from three biological culture replicates.The in vivo group consisted of 15 gels, derived from four biological replicates (bacterial isolates were from the colon and, in one case, from the ileum of the infected piglets).The Mann-Whitney Test was used for statistical significance analysis of protein spot differences.It is a nonparametric two sample distribution-free t-test and assesses whether two independent samples of observations come from the same distribution.The p-values determined by this test were based on 8 (in vitro spot) and 15 (in vivo spot) intensity observations.Spot abundance ratios $1.5 with a p-value ,0.05 were imported into the Multiple Experiment Viewer (MeV) software suite.Significant protein changes with an adjusted p-value ,0.05 were recorded.The positions of 25 cytoplasmic protein spots were used as landmarks for M r and pI calibrations in 2D gels.Analysis of the 500 most abundant spots resulted in high Mascot scores for E. coli EDL933 proteins suggesting lack of contamination with Sus scrofa (piglet) proteins.

Two-dimensional LC-MS/MS Analysis and Database Searches
Protein digestion mixtures (four fractions for each bacterial lysate) were subjected to 1 st dimension peptide separation ''offline'' on a polysulfoethyl-aspartamide column (4.6 mm650 mm, Nest Group, USA).Strong cation exchange chromatography (SCX) was performed at a flow rate of 0.5 ml/min as follows: 100% buffer A (5 mM KH 2 PO 4 , 25% CH 3 CN, pH 3.0) for 10 min; linear gradient to 20% buffer B (5 mM KH 2 PO 4 , 0.7 M KCl, 25% CH 3 CN, pH 3.0) over 30 min; to 50% buffer B over 20 min; to 100% buffer B over 5 min; maintain flow at 100% B for 5 min.Nineteen fractions were collected, and some of the early-and late-eluting fractions were pooled due to low peptide content based on the UV trace at 214 nm.The final number of SCX fractions was 13 for each of the FASP digests.A total of 52 SCX fractions for one bacterial lysate were lyophilized and stored at 280uC.On the day of 2 nd dimension peptide separation (LC-MS/MS) using a nano-ESI LTQ linear ion trap mass spectrometer (Thermo-Fisher, San Jose, CA), the set of 52 fractions was resuspended in 0.1% formic acid and transferred into a 96-well plate.In sequential, automated mode, a BioBasic C 18 column (75 mm610 cm; New Objective, Woburn, MA) was used for LC separations at a 350 nl/min flow rate [27].The calibration was performed with 200 nmol human [Glu 1 ]-fibrinopeptide B (M.W. 1570.57),verifying that elution times with a CH 3 CN gradient varied less than 10% and peaks representing ion counts had widths at half-height of ,0.25 min, signal/noise ratios .200and peak heights .10 7.The loaded sample volume was 25 ml, and peptides were eluted from the C 18 trapping cartridge into the C 18 column with 53 min binary gradient runs from 97% solvent A (0.1% formic acid) to 80% solvent B (0.1% formic acid, 90% AcCN).Spectra were acquired in automated MS/MS mode, with the top five parent ions selected for fragmentation in scans of the m/z range 300-1,500 and a dynamic exclusion setting of 90 sec.LTQ search parameters (+1 to +3 ions) included mass error tolerances of 61.4 Da for peptide precursor ions and 60.5 Da for fragment ions (monoisotopic mass values), allowed one missed tryptic cleavage and were set for Cys carbamidomethyl modification and Met oxidation as fixed and variable modifications, respectively.The search engine used for peptide identifications was Mascot v2.3 (Matrix Science).The protein sequence databases derived from the genomes of E. coli O157:H7 strain EDL933 and Sus scrofa were downloaded from RefSeq in NCBI.

Semi-quantitative Proteomic Analysis using Computationally Modified Spectral Counting
Protein identifications derived from Mascot v2.3 searches required at least one unique peptide with an e-value ,0.1.Mascot search peptide false discovery rates (FDRs) were determined searching a randomized E. coli O157:H7 EDL933 decoy protein sequence database.The average FDR for 19 2D-LC-MS/ MS experiments was 1.3%.Following file conversion into the mzXML format, MS data were re-scored using the algorithms PeptideProphet TM and ProteinProphet TM [28,29].Prot.xml files were analysed using an in house-developed software tool termed APEX quantitative proteomics tool v1.1 [30].Using a pre-defined set of MS analysis parameters and 30 physicochemical properties provided by the APEX tool (default settings), ARFF files for O i computations were generated for a set of 100 highly abundant bacterial proteins with M r values .30kDa (according to 2D gel data) to achieve a good representation of tryptic peptides.Observation of tryptic peptides in the training dataset were correlated with the 30 physicochemical peptide properties, resulting in the computation of O i values (the expected number of unique proteotypic peptides for a given protein i) for all protein sequences, in this case 5397 annotated proteins in the E. coli O157:H7 EDL933 sequence database.The equation to calculate APEX i values from 2D-LC-MS/MS data includes the O i correction factors, probability scores for protein identification (p i ) and spectral counts (n i ) as variables: APEXi~pi(ni=Oi)= X N k~1 pk(nk=Ok)|C Setting the protein FDR at 1%, only proteins identified at a 99% confidence level were used for spectra counting.A factor of 2.5610 6 , the estimated number of protein molecules per E. coli cell, was used to normalize APEX scores as protein molecules per cell [31].

Bioinformatics and Statistical Methods
Spectral counting datasets analyzed by the APEX tool were imported into the software suite MeV (http://www.tm4.org/mev/ ) [32].In MeV, 11 in vitro datasets (three biological replicates) and 7 in vivo datasets (from three piglets) were subjected to the nonparametric two sample Wilcoxon Rank Sum test to determine statistically significant abundance changes of proteins (p-value ,0.05) comparing the two groups.The Benjamini-Hochberg procedure was applied for multiple testing corrections, calculating a FDR for each p-value [33].The acceptable FDR for statistical significance of protein abundance changes was generally set at an adjusted p-value threshold of 0.05.Some protein changes were accepted to enhance biological data interpretations despite an adjusted p-value .0.05.A Wilcoxon Rank Sum was used to determine significant protein abundance ratios with a p-value ,0.05 for the comparison of two distinct in vivo sample groups (the HB-and the HB+ groups, see Results; n = 4 and n = 3, respectively).Computational tools and literature evidence contributed to assignments of subcellular localizations, membrane attachment, molecular functions and pathway associations, as described previously [20].Information on E. coli gene regulation, protein functional data and biochemical pathways was taken from EcoCyc and Uniprot (www.ecocyc.organd www.uniprot.org,respectively) to allow protein network analyses.

Results and Discussion
Our study had three objectives.First, proteomic data from two sample groups were compared: EHEC cells cultured in vitro to mid-stationary phase versus those isolated post-mortem from the colon of gnotobiotic piglets inoculated with 5610 9 bacterial cells.This dose usually results in severe diarrhea (as compared to mild or moderate diarrhea at lower infectious doses), which invariably lead to systemic disease.Shotgun proteomic data consisted of 1,500 bacterial proteins on average, identified at a 1.3% peptide FDR using Mascot v2.3 [15].Since APEX data estimate protein abundances as copies/cell, semi-quantitative comparisons of different proteins, e.g.those associated with a specific pathway, complex, or cellular localization, within a given sample group were possible.The differential 2D gel display of ,500 identified protein spots provided relative abundances of proteins comparing the two sample groups.2D gel profiles were less comprehensive, but very often confirmed the direction of a protein abundance change computed from the APEX datasets, as shown in Dataset S1 and the 2D gel maps in Figure S1.Proteins in the Dataset S1 are ordered according to functional categories, with annotation data for proteins and statistically significant protein abundance differences, if observed, for both APEX and 2D gel data.Second, we compared proteomic profiles of two piglet sample subgroups based on MS identification of Sus scrofa hemoglobin.The rationale was that hemoglobin was indicative of a more invasive stage of infection (Dataset S2).Third, predicting host protein-bacterial surface interactions, the in vivo MS data were searched with a Sus scrofa protein sequence database (Dataset S3).Data from cell cultures of strain 86-24, never in contact with host proteins, served to eliminate false positive protein hits such as trypsin, cytokeratin, and peptide sequences identical among EHEC and Sus scrofa protein orthologs.For the purpose of brevity, the terms ''increased, decreased or altered in vivo'' in the following paragraphs denote a protein abundance increase, decrease or change comparing in vivo and in vitro conditions, respectively.Quantitative data discussed in the following paragraphs pertain to APEX data, often included in Tables 1, 2 and 3, unless otherwise indicated.

EHEC Metabolism, Homeostasis and Nutrient Acquisition
Figure 1 depicts fifteen broad biological role categories which, based on summed abundances of proteins assigned to the categories, revealed quantitative differences comparing the in vitro and in vivo conditions.Marked changes were related to ribosome structural integrity; phosphate and ammonium mobilization and transport; de novo nucleotide biosynthesis; galactose, fructose, and mannose transport and metabolism; fatty acid lipid biosynthesis and metabolism; small organic acid and glycerol fermentation; aerobic carbon metabolism and TCA cycle; aerobic electron transport; and anaerobic electron transport.The schematic in Figure 2 serves to establish pathway links among the proteins assigned to these categories, with a focus on molecular processes highly active in EHEC cells isolated from the piglets' guts.The import and metabolism of several sugars were activated in vivo, consistent with carbon resources included in the piglets' diet.Lactose, fructose, and mannose were taken up, metabolized and fed into the glycolytic pathway (GP) at higher rates in vivo.Transporters for 2-deoxy-D-galactose (GalP) and lactose (LacY) were expressed only in vivo; b-galactosidase (LacZ) was increased 322-fold in vivo (Table 1).Enzymes converting b-D-galactose to a-D-glucose-1-phosphate (GalM, GalK, GalE, GalT) and MelA whose product is a-D-galactose were also increased in vivo.FruA and FruB, responsible for fructose transport, were 24-fold more abundant in vivo.FruK, the enzyme phosphorylating b-D-fructose-1-phosphate prior to GP entry [34], and phosphofructokinase (PfkA), the main regulatory GP enzyme, were strongly increased in vivo.The class I D-fructose-6-phosphate aldolase FsaA revealed a 163-fold abundance increase in vivo (Table 1).Its biological pathway associations are not entirely understood [35].We assume that FsaA has a metabolic bypass function when D-fructose-6phosphate accumulates as a result of insufficient acetyl-CoA turnover in the citric acid cycle.This notion is supported by a concurrent 3.5-fold in vivo abundance increase of GldA (Table 1), the enzyme that reduces one of the two end-products of FsaA, dihydroxyacetone, to glycerol [36] (Figure 2).Also consistent with this conclusion were decreased in vivo quantities of all citric acid cycle and glyoxalate bypass enzymes.FsaA was previously reported to be active at low phosphate concentrations in E. coli [35].As described later, there was strong evidence of a limited phosphate supply in intestinal EHEC cells.
Catabolic enzymes and transporters for C 2 -to C 6 -organic acids and glycerol derivatives were more abundant in vivo, suggesting bacterial responses to high quantities of the acids in the piglet intestine.The periplasmic subunit of the sn-glycerol-3-phosphate ABC transporter (UgpB) was increased in vivo.This C 3 -carbon intermediate may result from the turnover of bacterial or host phospholipids in vivo.Via oxidation to dihydroxyacetone-phosphate (DHAP), the molecule enters the GP as an energy source.At Table 1.Selection of differentially abundant EHEC proteins involved in metabolic and transport activities.

Gene name
Protein description

In vivo average
In low oxygen concentrations, GlpA/GlpB/GlpC, a membranebound enzyme complex also increased in abundance in vivo, catalyzes sn-glycerol-3-phosphate oxidation [37] (Figure 2).The C 4 -dicarboxylate transporters DcuA and DcuB were increased in abundance more than 10-fold in vivo; their substrates include tartrate and fumarate [38].Consistent with increased import activities for these acids, tartrate dehydrogenase (TtdA/TtdB) and fumarase B (FumB) (Table 1) were detected only in vivo.High activity of the enzymes have been linked to anaerobic bacterial biofilm formation [39].Oxygen concentrations are generally low in the intestinal lumen.The ability to form a strong biofilm was considered a key virulence attribute of the EAEC O104:H4 strain that caused an epidemic systemic disease outbreak three years ago [40].Proteins facilitating ascorbate uptake and catabolism also revealed 4 to 6-fold abundance increases in vivo (Figure 2).The product of the Ula pathway, D-xylulose-5-phosphate, is fed into the GP via the pentose phosphate pathway [41].Transcriptional regulators involved in activating short organic acid metabolism were also profiled.FruR, the catabolite repressor activator, was 2.2-fold decreased in vivo.FruR represses many of the genes/ operons encoding aforementioned transporters and enzymes [42] (Figure 2), and its de-repression appears to be a critical factor in redirecting bacterial carbon flow in the intestinal milieu.There was strong evidence for phosphate and ammonia starvation in EHEC cells in vivo.In addition to UgpB which contributes to phosphate mobilization, the anion-specific outer membrane porin PhoE, subunits of the high affinity phosphate transporter Pst, periplasmic alkaline phosphatase (PhoA) and the auxiliary factor PhoU were strongly induced in vivo (Table 1/ Figure 2).These proteins are part of the Pho regulon controlled by the two-component signal transduction system (2-CST) PhoB/ PhoR [43].Both the response regulator (PhoB) and the sensor histidine kinase (PhoR) were only detected in bacterial cells from the intestine.Several observations were indicative of strong ammonia mobilization in EHEC cells in vivo, supporting the biosynthesis of compounds essential for bacterial growth such as peptides, proteins and nucleotides.Aspartate-ammonia lyase was highly abundant in vivo.Subunits of the ethanolamine ammonialyase complex (EutL/EutC/EutB/EutE/EutM) were detected only in vivo.The inner membrane NH 4 + transporter AmtB, highly increased in bacterial cells from the intestine (Table 1), facilitates ammonia import followed by cytoplasmic NH 4 + assimilation via Lglutamine synthetase (GlnA) as shown in Figure 2 [44]; GlnA was also increased nearly 3-fold in vivo (Table 1).In contrast, enzymes with catalytic roles in amino acid metabolism (DadA, DadX, GcvP/GcvT, SdaA, Tdh, and Kbl) were strongly decreased in vivo.
The nitrogen-regulatory protein PII-2 (GlnK) interacts with AmtB after ammonia shock [45].GlnK (Table 2) and a 2-CST implicated in the response to nitrogen starvation (GlnG/GlnL) [46] were identified only in the intestinal EHEC proteome.In summary, the bacterial cells metabolically adjusted to increased availability of C 2 -to C 6 -organic acids and decreased availability of important nutrients such as PO 4 32 and NH 4 + in vivo.There was evidence for a re-directed EHEC energy metabolism as a consequence of microaerophilic conditions in the pig intestinal lumen.Citric acid cycle enzymes (Table 1) and NADH dehydrogenase (Nuo), the main E. coli membrane complex for aerobic respiration, were decreased strongly and moderately, respectively.Two global regulators sensing O 2 concentrations in bacterial cells (ArcA/ArcB and Fnr) [47] were identified (Figure 2).Unexpectedly, ArcA was decreased 2-fold in abundance in vivo.Post-translational events modulate activities of 2-CST response regulators, and their abundances do not necessarily correlate with their activities.Enzymes positively regulated transcriptionally by ArcA, and particularly Fnr, were increased in vivo (Table 1/ Figure 2).Some of the enzymes catalyze the catabolism of two glycolytic products, pyruvate and phosphoenolpyruvate.Branches supporting the break-down of these metabolites are designated mixed acid fermentation (MAF) [48].The fermentation products are lactate (activity of LdhA), acetate (activities of Pta and AckA), formate (activity of formate dehydrogenase-O; FdoG/FdoH/ FdoI), fumarate (activity of fumarate reductase; FrdA/FrdB/ FrdC/FrdD) and ethanol (activity of AdhE) (Figure 2).Additional enzymes metabolizing MAF intermediates (Ppc, PykF, PflB, TdcE, AdhE, FumB, and YfiD) were also invariably increased in vivo.AdhE, PflB and YfiD were among the in vivo most abundant cytoplasmic proteins of strain 86-24 (Table 1).High abundance of cytochrome bd terminal oxidase (CydAB), an oxidoreductive complex that depends on O 2 as electron acceptor, suggested that the EHEC-colonized environment was microaerophilic rather than anaerobic [49].Major electron donors for O 2 -independent electron transport in vivo were formate dehydrogenase-O and hydrogenase-2 (HybA/HybC/HybD/HybO) whose subunits (e.g., FdoG and HybC, Table 1) were increased in vivo.Electron acceptors contributing to O 2 -independent electron transport were fumarate reductase (Frd) and formate-dependent nitrite reductase (NrfA/NrfB/NrfC/NrfD).Both Hyb and Nrf are oxidoreductases localized at the E. coli inner membrane-periplasmic interface [50,51].Activity of the Fdo complex is coupled to Nrf and dimethylsulfoxide reductase (DmsA/DmsB), another membraneassociated electron carrier elevated in abundance in vivo.As depicted in the schematic of Figure 2, these redox complexes  branches and electron transporters indirectly via NarL, a transcription factor active at elevated nitrate or nitrite concentrations.NarL induces expression of nuo and nitrate reductase genes, whereas it represses pflB, adhE, hybACDO, fumB, and dmsA/dmsB expression [53].A three-fold decrease of abundance of NarL was detected in vivo (Table 2), and de-repression of NarL-controlled genes in vivo can be inferred from numerous abundance changes, including five such proteins listed in Table 1.Marked abundance differences were detected for two E. coli pyridine nucleotide transhydrogenases.Cytoplasmic transhydrogenase SthA specialized on re-oxidation of NADPH [54] was 10fold decreased; membrane-bound PntA/PntB increased 3.4-fold in vivo (Table 1).PntA/PntB activities are associated with a balanced intracellular supply of NADPH for various biosynthetic processes [54].Biosynthesis of amino acids (Table 1) and nucleotides de novo, each requiring NADPH in several catalytic steps, was indeed activated in vivo.In agreement with this data, the NADPH-consuming menaquinone-menaquinol redox cycle [55] also seemed to be mobilized in EHEC cells in vivo, judged from 1.7-and 2.5-fold increases of the enzymes MdaB and YgiN, respectively (Figure 2).

EHEC Responding to Stress Derived from the Intestinal Environment
During the distinct infectious stages, the piglet's defensive responses likely challenge EHEC stress response systems in different ways.Bacterial cells are exposed to various local secretion products in the gastro-intestinal tract.There is ex vivo and in vivo evidence for interactions of EHEC cells with specialized microfold (M) cells in Peyer's Patches of the colon, contributing to localized innate immune defenses [56].Penetration of the mucosa would expose the bacteria to erythrocytes, direct attacks by phagocytes, release of oxidative stressors and toxins, and the adaptive immune system.Here, the stress responses are examined as ''averages'' over distinct infectious stages (we isolated EHEC cells from the colon and ileum, with few proteomic differences); in vitro vs. in vivo comparisons lose some spatial and temporal resolution.

Osmotic Stress
Ten osmotic stress response proteins were quantitatively changed in abundance nine of which were decreased considerably in intestinal vs. in vitro (stationary phase) bacterial cells.This included OsmC (Table 2), OsmY, YjbJ, YbaY and entericidin B (EcnB).EcnB, 3-fold decreased in vivo, is a bacteriolysin and causes cell death at high osmolarity, a process inhibited by the lipopeptide EcnA and a toxin-antitoxin binding mechanism [57].EcnA was neither detected in vivo nor in vitro.EcnB expression is repressed by OmpR, a global cell envelope stress regulator.However, OmpR was 3-fold decreased in vivo.Without data on its phosphorylation, fewer insights in transcriptional control activities are obtained.OmpR also controls the expression of OmpC and OmpF, trimeric porins facilitating small hydrophilic solute diffusion across the OM [58].OmpC was highly abundant in the EHEC proteome in vivo (Table 2), whereas OmpF was not detected.In the intestinal milieu, a high OmpC/OmpF ratio may limit diffusion of toxic molecules including colicins, bile acids and antimicrobial peptides into bacterial cells because OmpF has a larger pore size.Indeed, phosphorylated OmpR induces ompC expression, lending support to the notion of high in vivo OmpR activity [59].MscS, a mechanosensitive channel involved in water homeostasis and part of the RpoS regulon, was increased 3-fold in vivo (Table 3).MscS has been linked to osmotic down-shock responses [60], supporting osmotic down-shock in EHEC cells exposed to the intestinal environment.

Protease-induced Stress
EHEC cells encounter secreted proteases (pancreatic and intestinal) passing through the mammalian gut.Since the OM is a porous barrier towards the cells' exterior, these insults implicate proteolytic damage in the bacterial cell envelope.There was little evidence for elevated abundances of chaperone-protease systems (e.g.PpiD, DegQ, HtrA, and SurA), which remove misfolded and aggregated proteins in the periplasmic space in vivo.Ecotin and the chaperone HchA were 14-fold and 5-fold decreased, respectively, in vivo (Table 2).E. coli HchA is part of the RpoS regulon and responds to osmotic up-shock [61].The function of ecotin as a trypsin inhibitor [62] in the bacterial periplasm must be questioned considering the exposure of EHEC in vivo to secreted trypsin.A predicted peptidase (YgeY) and a transcarbamylase (YgeW), possibly co-expressed since the gene loci are adjacent to each other, were detected only in vivo (Table 2).Four small cold shock proteins, CspA in particular (Table 2) attributed to have cold shock and RNA chaperone functions [63] were strongly increased in abundance in vivo.

Acid Stress
Several acid resistance stress pathways are implicated in the survival of E. coli strains when challenged by intracellular acidification.Our data suggest that EHEC strain 86-24 utilizes two glutamate decarboxylases to neutralize intracellular H + ions and GadC for aminobutyrate export in the intestinal environment [64] (Figure 2).Due to 99% sequence identity, the decarboxylases GadA and GadB could not be discerned based on APEX i data.However, their respective pI values are 5.23 and 5.29, and 2D gel profiles suggested that GadA was increased 2-fold in vivo vs. in vitro, whereas GadB was highly abundant in both experimental groups.GadC was increased 3-fold in vivo (Table 2).Arginine decarboxylase (AdiA) was an order of magnitude less abundant than GadA and GadB in vivo, implying that this enzyme was not essential to counteract an acidifying intracellular pH in bacterial cells.While the ambient pH is neutral to basic in the colon, MAF pathway activities and lactate, acetate, and formate production likely cause a pH drop in the bacterial cytoplasm supporting the notion of enhanced acid stress.

Stress Associated with Reactive Oxygen and Nitrogen Species
Enzymes removing peroxides and oxygen radicals were abundant under in vitro and in vivo conditions, including alkyl hydroperoxide reductase (AhpC), superoxide dismutase (SodB) and two catalases, KatP and KatG (Figure 3).A 2.3-fold in vivo increase of the oxidoreductase AhpF, which can form a complex with oxidized AhpC in the presence of an excess of NADH, was observed.AhpF binding may recycle the oxidized form of AhpC more effectively and increase AhpC activity in vivo [65].SodB is an iron-dependent dismutase.Since iron starvation induces decreased sodB expression in many c-proteobacteria [26], it is reasonable to conclude that EHEC cells are generally not iron-starved in vivo.Periplasmic SodC and cytoplasmic SodA were strongly decreased in vivo (Table 2), which was expected considering their transcriptional repression by Fnr and ArcA in microaerophilic environments.Higher abundances of nitric oxide dioxygenase (HmpA) and NADPH-dependent nitrite reductase (NirB/NirD) in vivo supported the notion that nitrite and nitric oxide (NO) stress rose considerably in vivo (Table 2/Figure 3).These enzymes detoxify NO [66,67], a cytotoxic product produced by activated macrophages and neutrophils to kill invasive bacteria.In addition, NO is presumably a catalytic by-product of the in vivo-activated electron transporter nitrite reductase, an enzyme complex more abundant in vivo (e.g.NrfA, Table 1).

Bacterial Cell Envelope Changes in the Intestinal Milieu
E. coli strains have complex regulatory systems modifying the cell envelope to withstand external stresses in hostile environments.In addition to OmpR, the transcription factor network coordinating the cellular responses to extracellular signals involves PhoQ/PhoP, RcsB/RcsD, RstA/RstB, OxyR, CpxA/CpxR, CheY/CheZ, LexA, BaeR/BaeS, BolA, BasR/BasS and EvgA/ EvgS [59].Nine of these regulators (PhoP, RstA, and CheY are listed in Table 2) were surprisingly decreased in abundance in intestinal EHEC cells.Complex interactions with proteins, RNAs and post-translational modifications influence the activities of these transcription factors such that expression data derived from host environmental stress does not offer conclusive insights.
For example, CpxA and CpxR form a 2-CST.CpxR was one of the more abundant extracytoplasmic stress response regulators in vivo.This transcription factor is involved in functions as diverse as pilus assembly, secretion, motility, chemotaxis, adherence, biofilm formation, multidrug resistance and toxin efflux [68,69].Two periplasmic chaperones whose respective genes are under control of CpxR were less abundant in vivo (PpiA and Spy).LpxD and LpxB (Table 2), two enzymes involved in lipid A core biosynthesis, a protein of unknown function (YebE) and the regulator BaeR were also decreased in abundance in vivo.Autokinase activity of CpxA is inhibited by CpxP, a periplasmic protein that was profiled only in intestinal EHEC cells [70].CpxP might deactivate CpxR in bacterial cells in vivo, by reducing the level of phosphorylation of CpxR by CpxA.Proteomic data generally supported the notion of reduced chemotaxis and quorum sensing in vivo; e.g., the methyl-accepting chemotaxis protein Tar, the flagellar protein FliC, and the autoinducer (AI)-2 uptake system Lsr were lower in abundance in vivo.Tryptophanase (Table 1) and LuxS, enzymes involved in biosynthesis of indole and autoinducer-2 (4,5-dihydroxy-2,3-pentanedione), respectively, were strongly decreased in vivo vs. in vitro.The factor for inversion stimulation (Fis), a small DNA-binding protein which affects processes such as cell mobility, aggregation and biofilm formation via modulation of nucleoid structures [71], was quite abundant in vivo, but not detected in vitro.Regarding its regulatory capabilities, Fis overlaps with Fnr and NarL (two anaerobic respiration regulators), CpxA/CpxR, IhfB, and DksA.This is evident surveying transcription factor specificities listed in Table 2.The role of the signaling molecule acetyl phosphate (Ac-P) in EHEC cells is also interesting, as the schematic in Figure 3 reveals.Ac-P release affects the stress response transcriptional regulator network [72].Accumulation of intracellular pyruvate resulting from lower citrate cycle metabolism and lower activities of acetyl-CoA synthetase (Acs) (Table 1) and pyruvate oxidase (PoxB) in vivo, combined with highly active Pta/AckA, support the intracellular accumulation of Ac-P.We hypothesize that Ac-P is an important signaling molecule in EHEC cells in the infectious process.
Five transcriptional regulators, Fis, BolA, CpxR, CpxP and RcsB, influence E. coli biofilm formation.Three fermentationrelated enzymes (TtdA, TtdB and FumB), more abundant in EHEC in vivo, have also been associated with biofilms in E. coli [39].Biofilm formation requires capsular polysaccharides which embed bacterial cells in an extracellular matrix structure.YgiB, YmcA, YccZ and L7026, four proteins tentatively linked to exopolysaccharide biosynthesis, were increased in EHEC cells in vivo and may therefore also be implicated in biofilms (Figure 3).PgaA, an export protein for poly-beta-1,6-N-acetyl-D-glucosamine identified only in vivo (with low abundance), is required for exopolysaccharide biosynthesis [73].Several enzymes involved in lipopolysaccharide (LPS) biosynthesis, including the lipid A core, the O-antigen and OM phospholipid asymmetry, were also altered in abundance: ArnA and a putative lipid IVA N-acyltransferase (L7029); the O-antigen biosynthetic enzymes (Ugd, ManB, WbdP, and WecE); three proteins with functions in OM asymmetry maintenance (MlaA, MlaC, and LptD).Targeted biochemical studies may reveal whether these proteins (see also Table 2/ Figure 3) have functional roles in biofilm formation and/or bacterial cell surface modulation in the host environment.

EHEC Exposed to Infiltrating Immune Cells Following Mucosal Injury
Anticipating the ability to identify host proteins, shotgun proteomic data were queried with a Sus scrofa protein sequence database.We noticed that, in a subset of three in vivo samples, hemoglobin subunits were identified whereas, in a subset of four samples, this was not the case (here termed HB+ and HB-, respectively).Hemoglobin is a biomarker for vascular damage following EHEC infection since the protein is highly abundant only in red blood cells.Comparing the proteomic profiles of the HB+ and HB-subgroups, 138 proteins were quantitatively changed (Wilcoxon rank sum test; P,0.05).Four Ler-regulated virulence factors (StcE, Tir, EspA and Eae) and two T3SS effector chaperones (CetT and CesF) were increased in abundance at least 6-fold in the HB+ group (Table 3).The metalloprotease StcE interacts with and cleaves mucosal glycoproteins (Figure 3).EspA is a T3SS needle filament protein and essential for Tir translocation into the host cell membrane [7].It is plausible that increased abundances of T3SS effectors and StcE facilitate bacterial translocation through M cell layers, and exposure to phagocytic cells prompts the release of Stx-2 [56].The coincidental detection of hemoglobin supports this notion, but the lack of detection of Shiga toxin-2, supposed to be initiated during macrophage apoptosis, does not.In our experiments, EHEC cells were mostly derived from the colon where A/E lesions are reported to be less prevalent than in the ileum [56]; we can only speculate that the observed T3SS induction coincides with M cell translocation and exposure to phagocytes.Other proteins altered in abundance comparing the HB+ and HB-groups were the IhfA subunit of integration host factor which is implicated in induction of the same set of effectors controlled by Ler [14].IhfA was increased 2.2-fold in the HB+ group.The carbon storage regulator CsrA (Table 3) showed a 3-fold increase in the HB+ group.Interestingly, a DcsrA mutant strain of S. typhimurium was attenuated in its ability to invade HEp-2 epithelial cells [74].Three proteins linked to intracellular survival of S. typhimurium in macrophages [75,76] were highly expressed in the HB+ group, but not in the HB-group: the heat shock proteins IbpA and IbpB and the biotin synthase BioB (Table 3).These proteins may promote the intracellular survival of EHEC cells in macrophages.An intracellular life stage in those macrophages that communicate with M cells and are associated with submucosal invasion is being debated.E. coli O157:H7 were described to translocate into the murine ileal submucosa via extensive interactions with M cells and macrophages [56].However, a transcriptome survey of E. coli O157:H7 cells invading THP-1 monocytes revealed few agreements with our proteomic data [19].Invasion by an EPEC strain caused extensive changes in the proteome of Caco-2 intestinal cells, and this was reported to be mediated by pedestal formation and translocation of T3SS effectors [77].
Increased access to nutrients and oxygen may be attributable to the different metabolic states of HB+ compared to HB-bacterial cells, thus supporting higher cell growth rates: two carbon starvation proteins (Slp and YjiY) were decreased, whereas 21 proteins associated with aerobic energy metabolism and 11 proteins implicated in amino acid biosynthesis were increased in the HB+ group.These adaptations are not unexpected considering more nutrient-and oxygen-rich environments available to the bacteria when erythrocytes and leukocytes infiltrate the site of infection.Activation of the innate immune system in the piglets could result in the release of toxins such as reactive oxygen and nitrogen species (ROS, RNS) and antimicrobial peptides.There was little proteomic evidence for rising ROS and RNS detoxification activities mediated by SodA, SodC, AhpC, KapP, KatG and HmpA, or for the increased expression of antimicrobial efflux pumps (e.g.MdtE/MdtF) in the HB+ group.HmpA is a key enzyme neutralizing RNS.While hmpA induction was reported to be absent in intracellular S. flexneri bacteria, S. typhimurium revealed increased hmpA transcription in parallel with the onset of NO production following the invasion of macrophages [75,78].

Glycoprotein Adhesion to the EHEC Cell Surface
Identification of hemoglobin subunits suggested intestinal mucosal damage and enteric hemorrhage inflicted by EHEC cells.Unlike hemoglobin, most of the other identified Sus scrofa proteins (Table 4) were present in both in vivo sample subgroups.The functional roles of a few of these proteins and their tissuespecific expression were reassuring aspects as it regards the concern of unspecific binding to bacterial cells.Seven of the proteins were reported to be expressed in the intestinal or colonic epithelium of mammals.Two proteins are eosinophil products (eosinophil peroxidase, proteoglycan 3-like protein) secreted in response to microbial insults [79,80].Three proteins have C-type lectin domains known to have intercellular adhesive properties, including the regenerating islet-derived proteins REG3-c and REG1-a [81,82].In murine models, REG3-c is released from Paneth secretory cell granules into the intestinal lumen, unfolding antimicrobial activities against Gram-positive bacteria [81].REG1-a (synonymous to lithostathine) is primarily secreted from acinar pancreatic cells, inhibits CaCO 3 precipitation and also has tissue-regenerative activities.REG1-a induces bacterial aggregation in vitro [82].To our knowledge for the first time, we provide evidence for adherence of these C-type lectins to Stx-producing E. coli in vivo.Resistin is a short cysteine-rich protein with proinflammatory activities that competes with LPS for binding to the Toll-like receptor 4 (TLR-4) [83].Finally, proteoglycan 3-like protein also features a C-type lectin domain and may be engaged in pathogen pattern recognition as a co-receptor to Toll-like receptors [84] (Figure 3).
The strongest evidence for the specificity of bacterial surface binding pertains to the glycoprotein 340 (also termed GP340, deleted in malignant brain tumors 1-protein, and DMBT1).
Human saliva GP340, a highly O-glycosylated and sulfated mucin contributing to the mucosal defense against pathogens, was identified as an in vitro substrate for StcE [12].GP340 is also expressed in intestinal epithelial cells and responds to proinflammatory stimuli such as TNF-a and bacterial lipopolysaccharides (LPS).The data here provide the first in vivo evidence that S. scrofa GP340 is able to adhere to the surface of an enterohemorrhagic E. coli strain.The host-pathogen cross-talk results in binding of the StcE protease to GP340, likely followed by glycoprotein cleavage.These events may alter the extracellular matrix density and diminish protection of the mucosa from the formation of A/E lesions.Interestingly, Caco-2 intestinal cells infected by an EPEC strain in a T3SS-dependent manner revealed a significant GP340 abundance decrease [77].The hypothesis that EPEC and EHEC cells send out signals manipulating intestinal secretion of critical host defense proteins, such as GP340, needs to be verified.Eosinophil peroxidase is a pro-inflammatory enzyme released during eosinophil degranulation in the gut mucosa [85].The trefoil factor 3, a protein known to be secreted by colonic Goblet cells and involved in mucosal repair, the tyrosine-protein kinase JAK1, a signal transduction enzyme implicated in interleukin-2 and interferon-c production, and two other proteins with unknown functions were identified (Table 4).

Systems-level Interpretation of Proteomic Data
Investigating the EHEC proteome under human-like disease conditions is of interest to unravel molecular mechanisms of pathogenesis and identify vaccine and therapeutic drug targets.The only non-primate model with human-like pathology is the gnotobiotic piglet.What are the three main limitations of this model and our experimental approach?First, a normal intestinal microbiota is beneficial to the maturation of immune system and provides protection from pathogens [86].The microbiota are absent in gnotobiotic piglets, and conventional piglets are not susceptible to EHEC disease, that seen in gnotobiotic piglets.The Stx receptor differences in pig vs. human may also contribute: Stx released during edema disease (in pigs) binds to receptor Gb4, Stx released during hemorrhagic colitis (in humans) binds to receptor Gb3.Second, bacterial cells are exposed to different host environment-induced stresses.The preferred sites for colonization, adhesion and invasion, in humans and animal models, are still under debate.The milieu acidifies in the stomach and neutralizes during passage in the small intestine, while oxygen access for respiration is gradually reduced.EHEC cells reveal tropism for the colon and ileum in humans and piglets, but the sites of adhesion and A/E lesion formation vary [87,88].M cells located in ileal and colonic Peyer's Patches allow bacterial translocation in murine and ex vivo human models.Macrophage apoptosis triggers release of Stx in the lamina propria [19,56].These stages of infection cannot be separately sampled in an in vivo animal model.Analysis of postmortem intestinal tissue yields averaged proteomic profiles.Third, host proteome profiling was limited to adherent proteins co-fractionating with in vivo-isolated bacterial cells.Immune responses of piglets to EHEC cell invasion were not profiled globally.We assessed proteomic evidence in support of the more invasive pathogenesis model proposed by Livrelli et al. [56], as opposed to the paradigm that EHEC cells largely remain in the intestinal lumen when A/E lesions are formed.It was useful to compare this data with SD1 proteome data also examined using a gnotobiotic piglet model [16,17].S. dysenteriae adapts to intracellular survival in and the spread from epithelial cells and macrophages.Unlike S. typhimurium [89], SD1 does not infect the host systemically with an outcome of septicemia [90].

EHEC and SD1 Metabolize Intestinal Carbon Sources in Similar Ways
Figure 1 depicts energy metabolism pathways apparently activated in EHEC cells in vivo in a microaerophilic milieu.Carbon sources were discernible from those utilized during in vitro growth.Transporters and metabolism for lactose, galactose, mannose, N-acetylglucosamine, fructose, ascorbate, sn-glycerol-3-phosphate and C 4 dicarboxylates were increased in vivo, consistent with the transcriptional modulation by 2-CSTs responsive to low oxygen concentrations [38,41,91,92].For example, DcuA and DcuB were strongly induced in vivo.DcuA was also increased in the SD1 proteome in vivo [16].This data was not in agreement with the constitutive expression reported for E. coli DcuA [93].E. coli dcuA/dcuB double knock-out mutants were growth-impaired when exposed to anaerobic conditions [49], suggesting this transport pathway to be a target for antibiotic inhibitor design.Short chain fatty acids also contribute to lowering the pH to 5-6 in the proximal colon [94].EHEC is well-adapted to acid stress, and the high abundance of GadA and GadB/GadC in vivo revealed the pathogen's reliance on this acid resistance system.SD1 cells also activated in vivo an arginine decarboxylase acid resistance system (AdiA/AdiC) and periplasmic disaggregation proteins (HdeA and HdeB) linked to acid stress [16,17].Stronger acid stress responses of SD1 may result from exposure to phagolysosomal conditions inside macrophages.The ability of E. coli O157:H7 to survive in macrophages was also examined, and transcriptome data revealed intracellular up-regulation of hdeA, hdeD, and adiC [19].Increased transport and metabolism of fructose, lactose and galactose by EHEC and SD1 cells in the intestine may result from high lactose and fructose content of Similac Sensitive Formula, a diet the piglets were fed.The nutrient adaptations were stronger in EHEC compared to SD1 cells (a 9fold vs. 4-fold change for DcuA; a 322-fold vs. 21-fold change for LacZ; a 24-fold vs. 1.7-fold for FruA/FruB).Several mixed acid fermentation enzymes and revealed stronger in vivo increases, citrate cycle enzymes stronger decreases in EHEC vs. SD1.The electron transport chains Fdo, Hyb, and Dms active at low oxygen levels were also induced more strongly in vivo in EHEC vs. SD1 cells.While strain 86-24 expressed a periplasmic nitrite reductase (deleted from the SD1 genome) in vivo, SD1 up-regulated a periplasmic nitrate reductase (NapA/NapB) in vivo [16].Both enzymes operate at low intracellular nitrate concentrations.Proteome data for SD1 and EHEC revealed a strongly induced PHO regulon suggesting phosphate starvation in vivo.Evidence for both Mg and phosphate starvation in S. typhimurium-containing (intracellular) vacuoles [75] and in S. flexneri following invasion of macrophages and epithelial cells [78] was reported.The PHO regulon, apparently important in all four pathogens, may have promising targets for drug design.There was evidence for ammonia starvation in EHEC in vivo, considering increased expression of the AmtB-GlnK system, which controls the internal NH 4 + /glutamine pool and the response to ammonia shock [45].Interestingly, these proteins were not identified in SD1 in vivo.To our knowledge, the GlnK-AmtB system has not been associated with increased fitness of the two pathogens during mucosal invasion processes.
Acetyl phosphate (Ac-P) may be an interesting intracellular signaling molecule in EHEC cells.Ac-P has been implicated in the functional modulation of cell envelope stress regulators which, in turn, triggers changes in cell envelope structure, motility and chemotaxis.E. coli Ac-P was reported to phosphorylate a few 2-CST response regulators by direct PO 4  32 transfer [95].Vilu et al. characterized acetate overflow metabolism in E. coli, when citrate cycle flux is decreased, acetyl-CoA synthetase (Acs) activity is repressed, and the enzymes Pta and AckA remain active.The disruption of the Pta-Acs node results in accumulation of Ac-P under these conditions [72].Whereas the intestinal milieu is microaerophilic (E. coli growth conditions used in the aforementioned study were aerobic), proteomic data also suggested Pta-Acs node disruption in EHEC cells in vivo (Figure 3).Citric cycle flux and, apparently, the activities of Acs, the enzymes catalyzing subsequent fatty acid biosynthesis steps and PoxB, which converts Ac-CoA directly to acetate, were reduced in vivo.Increased Ac-P concentration levels may then trigger the phosphorylation of 2-CST response regulators, e.g.OmpR, PhoBR, RcsB, CheY, GlnG and CpxR [72] (Figure 3).Modulation of their DNA binding activities affects cell envelope functions [96].RcsB, for example, is part of an E. coli phosphorelay system that influences biofilm formation, osmotic stress and capsular polysaccharide biosynthesis.RcsB activity is influenced by Ac-P [72].Five proteins implicated in capsular polysaccharide synthesis were decreased in abundance in EHEC in vivo (Figure 3).Periplasmic OsmC, active under osmotic stress conditions and repressed transcriptionally by RcsB in E. coli [97], was also decreased.CpxR is another 2-CST response regulator phosphorylated by Ac-P.This post-translational modification results in the activation of expression of an adaptor protein, CpxP [98].CpxP was identified in EHEC cells in vivo, but at low abundance.CpxR phosphorylation and CpxP expression in vivo did not result in the induction of genes implicated in cell envelope functions and believed to be under positive transcriptional control of CpxR (LpxB, LpxD, Spy, PpiA, and In fact, the proteins were decreased in abundance in vivo supporting the notion that 2-CSTs are part of an intricate network rather than acting alone.Finally, enzymes part of the tryptophan biosynthesis pathway including its regulator TrpR and the indole-producing tryptophanase were strongly decreased in vivo.Indole-based signaling was linked to biofilm formation in EHEC strain 86-24 [20], and the molecule may also modulate EHEC cell communication in the intestinal lumen.Genetic and biochemical tools are required to understand the pathogen cells' communication patterns with each other and host defenses.

Direct Host-pathogen Interactions
We demonstrate that host-pathogen interactions can be examined without exhaustively profiling the host cell proteome.As the example of GP340 shows, the method has biological significance.Saliva GP340 was previously identified as a substrate of metalloprotease StcE.Grys et al. [12] proved that StcE adheres to and cleaves this glycoprotein.The process appears to facilitate epithelial infection by EHEC [56].The structure of the mucinase and its O-glycan chain binding site were recently characterized [99].GP340 is a human (and apparently piglet) defense factor binding to the bacterial cell surface in a calcium-dependent manner to inhibit invasion.Other piglet proteins identified here as adhesion proteins may have similar roles in the innate immune defense: the trefoil factor 3 (TFF3) is secreted by mucin-producing goblet cells and involved in mucosal integrity [100].Interestingly, Citrobacter rodentium, a model organism producing A/E lesions in mice, triggered goblet cell depletion in the colon and decreased secretion of cell-protective TFF-3 peptides [101].Our data are the first, tentative evidence that piglet (and possibly human) TFF-3 peptides have a role in protecting Goblet cells from EHEC invasion by directly binding to the bacterial surface.Our data also provide evidence of resistin binding to the pathogen's surface.Resistin was determined to compete with LPS for TLR-4 binding in THP-1 monocyte cell cultures [83].We hypothesize that secreted pig resistin has similar danger signaling and immunomodulatory roles when bacterial cells come into contact with the luminal surface of ileum and colon.Three C-type lectins were identified.The bactericidal lectin REG3-c is released from gut mucosal cells, reported to bind to Gram (+) bacteria and recognize peptidoglycan structures [81].Our study suggests that pig REG3-c effectively binds to EHEC cell surfaces, but its bactericidal activity needs to be verified.The proteoglycan 3-like protein PRG3 and lithostathine, an abundant pancreatic juice protein, have not been associated with innate immunity to date.PRG3 may have a functional role similar to that of REG3-c.An intriguing role as a protein causing aggregation of E. coli cells was attributed to lithostathine, but not verified in vivo [82].Lithostathine adheres to the EHEC surface, but a bacterial aggregation function and interference with mucosal wall damage needs to be examined.Ctype lectins are a diverse family of proteins implicated in pattern recognition, TLR surface receptor cross-linking, and bacterial opsonization.These proteins modulate innate and adaptive immune responses [84,102].Interactions at the cell surface (Figure 3) may occur to a lesser degree when commensal gut microbes are present in conventional piglets.The competing bacteria provide a crucial line of defense against infection with pathogens and inflammation [94].
Our data provide tentative support for the murine EHEC pathogenesis model proposed by Livrelli et al., where invasion of M-cells is followed by bacterial uptake by and temporary survival in macrophages that underlie M-cells [56].Our proteomic data support two bacterial sample subgroups, one with a more aerobic metabolism hemoglobin binding (HB+), the other with a less aerobic metabolism and no hemoglobin (HB-).Identification of hemoglobin suggested vascular damage and a more invasive stage of the bacteria where exposure to more oxygen re-activates aerobic respiration and NADH oxidase activity of infiltrating inflammatory cells mediates the respiratory burst [103].In the HB+ group, we noticed increased expression of three proteins considered particularly important for intracellular survival in macrophages.Biotin synthase B expression was proposed to be essential for S. typhimurium replication inside biotin-starved macrophages [76].The chaperones IbpA and IbpB were abundance-increased in S. typhimurium upon macrophage entry and functionally associated with stress responses to oxidative burst.We also observed high abundance of SodB and HmpA in the HB+ group, but not the expression increases reported for the two protein orthologs of S. typhimurium residing in macrophages [75].The catalytic activities of SodB and HmpA, which detoxify ROS and RNS, respectively, are important for intracellular survival of S. flexneri, S. typhimurium and likely also E. coli O157:H7 strains [19,75,104].An E. coli O157:H7 transcriptome analysis revealed low THP-1 macrophage invasion rates 24h post-infection, but in macrophages wherein bacteria were detected clusters of E. coli cells had formed suggesting bacterial replication.This study by Poirier et al. [19] did not support strong T3SS effector abundance increases during the time course of infection; our in vivo data were indicative of increased effector expression (Tir, Eae, EspA, and StcE) and their regulators (Ler and IhfA) in the more invasive stage (HB+ group).In contrast, expression of Shiga toxin (Stx2b) and iron transporters (Fit, Chu) were found to be induced in the course of infection [19], but not observed in the comparison of the HB+ vs. HB-proteome profiles.Finally, the pathogen patterns recognition molecules were not identified specifically in one of the two EHEC sample subgroups.Their binding to bacterial cells is expected to occur in the environment of the intestinal lumen.

Concluding Remarks
A comprehensive proteomic analysis comparing EHEC cells derived from cell cultures and following bacterial collection from the gut lumen of infected gnotobiotic piglets unraveled multiple biochemical pathways including microaerophilic respiration, carbon utilization, acid resistance, and responses to several stresses (RNS, phosphate, ammonium and NADPH starvation) that were activated in vivo.There was evidence for extensive adaptations of the bacterial envelope to the intestinal milieu, influenced by a dedicated transcription factor network and acetyl phosphate signaling.For the first time, a set of piglet adhesion factors with likely functions in pathogen recognition and innate immune system activation was identified.The study resulted in the identification of proteins potentially useful for targeted drug design and set the stage for a significant systems biology role to understand interactions of pathogens with their mammalian hosts in an environment reflecting the normal infectious process.

Supporting Information
Figure S1 Differential 2D gel display of proteins from lysates of EHEC cells isolated from cell culture versus piglet intestines.Gel image B368.This 2D gel image visualizes the proteome of EHEC strain 86-24 cells grown to the stationary phase in LB media (in vitro) with no exposure to the piglet host organism.Gel image B276.This gel visualizes the proteome of bacterial cells isolated from piglet intestinal lavage after infection with EHEC bacteria, observation of symptoms such as bloody diarrhea and fever, and post-mortem recovery of colon and caecum sections of the intestines (in vivo).The gel images range from the Mr 200 kDa (top) to 10 kDa (bottom) and the pI of 4.5 (left to the pI of 7 (right).The protein spots marked in RED are those increased in abundance in the denoted gel (sample) group, those in BLUE were decreased in the denoted gel (sample group).The spot numbers are equivalent to those in the column ''AO'' of Supplementary Dataset S1.EHEC cells were suspended in a 2D gel rehydration buffer (GR lysis buffer).This solution contained 8 M urea, 2 M thiourea, 4% (w/v) CHAPS, 18 mM DTT and 0.5% (v/v) Bio-Lyte pH 3-10 carrier ampholytes).Samples were frozen at 280uC until further use.On the day before 2D gel separation, bacterial lysates were thawed, incubated at 20uC for 30 min and vortexed intermittently to complete protein solubilization.Whole cell lysates were centrifuged at 16,1006g for 30 min, and supernatants were subjected to protein quantification.Supernatant samples were subjected to 2D gel electrophoresis in batches of 12 gels using procedures described in the main text.Briefly, 1 st dimension protein analysis in 24 cm immobilized linear pH gradient strips (pH range 4-7; GE Healthcare) included gel rehydration loading of samples with ,150 mg protein and electrophoresis for ,60,000 Vh.Following reduction and alkylation steps, re-equilibrated strips were applied to 2 nd dimension SDS-PAGE slab gel electrophoresis (25619.560.15cm; 8-18%T) for ,1,800 Vh.Gels were fixed, stained with Coomassie Brilliant Blue G250 (CBB), de-stained, subjected to gel image analysis (data acquisition as 16 bit TIFF images) and imported into the software tool Proteomweaver v4 (Bio-Rad, Hercules, CA).The differential display analysis was performed comparing at least six gel replicate experiments (images) for the in vitro and in vivo EHEC groups.The softwarebased quantitative and statistical analysis method is described in the main text.(PDF) Dataset S1 Protein abundance differences comparing EHEC strain 86-24 cells cultured in vitro versus EHEC strain 86-24 cells isolated from the intestine of infected piglets.This Table includes data from a shotgun proteomics analysis computing protein values using the APEX Quantitative Proteomics Tool and from a differential 2D gel display analysis computing protein values in the gel image analysis software Proteomweaver v4.Detailed legends are provided in the 1 st worksheet of the dataset.(XLSX) Dataset S2 Protein abundance differences comparing EHEC strain 86-24 cells isolated from a piglet subgroup positive for hemoglobin versus a piglet subgroup negative for hemoglobin.This Table includes data from a shotgun proteomics analysis computing protein values using the APEX Quantitative Proteomics Tool.Detailed legends are provided in the 1 st worksheet of the dataset.

Figure 1 .
Figure 1.Global adaptation of EHEC cells to the intestinal milieu.Fifteen biological role categories, as defined in Dataset S1, are displayed in the graph.The bar length represents the sum of APEX i quantities of all proteins with a statistically significant abundance change (in vitro versus in vivo) assigned to a given biological role category.Blue bars represent the in vitro (cell culture) growth, red bars the in vivo (intestinal) environments.doi:10.1371/journal.pone.0066462.g001

Figure 2 .
Figure 2. Metabolic pathways active in intestinal EHEC cells.From the left to the right: import and metabolism of glycerol derivatives and dicarboxylic acids; import of mono-and disaccharides and their entry into the glycolytic pathway (GP); metabolism of ascorbic acid; mixed acid fermentation utilizing GP products; high affinity phosphate and ammonia import and metabolism; NADPH generation and menaquinone/ menaquinol cycle; glutamate decarboxylase acid resistance system (Gad); oxygen-dependent and -independent electron transport chains; *periplasmic electron transporters.Symbols/color codes: substrates and products of pathways are denoted in black script, proteins in red script (if increased in abundance in vivo), in blue script (if there was no abundance change but high expression in vitro and in vivo), in green script (if decreased in abundance in vivo); full arrows indicate a catalytic step, dotted arrows a transport step.In the top right corner, seven global transcriptional regulators are depicted; their color-coded symbols are included in the graphic to illustrate where they influence gene expression.The area between the green line (outer membrane) and inner line (inner membrane) represents the periplasmic space.doi:10.1371/journal.pone.0066462.g002

Figure 3 .
Figure 3. Changes in the EHEC cell envelope during effacement of the host environment and molecular responses initiated by the piglet host.At the top of the schematic, generally from left to right: Sus scrofa proteins identified as adhesion factors to the bacterial surface; proteins implicated in capsular poly-N-acetylglucosamine (PNAG) and lipopolysaccharide (LPS) biosynthesis; proteins implicated in b-barrel OM protein synthesis (YaeT and NlpB), lipoprotein export (LolA, LolB and LolD), OM asymmetry (Imp, MlaA and MlaC); T2SS and T3SS effectors and some of the established interactions with host cells and the extracellular matrix (StcE, EspA/B, Eae/Tir and OmpA); putative peptidoglycan-modulating proteins (YdhA and YkfE).Below from left to right: various extracytoplasmic stress responses; regulators of cell envelope stress and the connection to acetyl phosphate (Ac-P) signaling; tryptophan/indole biosynthesis; ROS and NOS mediated stress responses by peroxidases, catalases, dismutases and NOdetoxifying enzymes.doi:10.1371/journal.pone.0066462.g003 (XLSX) Dataset S3 Protein identifications searching the LC-MS/MS data from in vivo and in vitro EHEC samples with the Sus scrofa protein sequence database (RefSeq in NCBI).Detailed legends are provided in the 1 st worksheet of the dataset.(XLSX)

Table 1 .
Cont.Differential abundance refers to in vivo versus in vitro protein abundance changes of the EHEC proteins.Transcription factors involved in expression control of the respective genes are listed.+/2, positive/negative regulation. *'

Table 2 .
[52]ction of differentially abundant EHEC proteins involved in stress response and cell envelope structure.generateanelectrochemicalgradientacrossthe inner membranes and a proton motive force, resulting in ATP production.A ninefold increase of the periplasmic subunit of molybdate ABC transporter ModA was measured in vivo (Table1).Molybdate is a DmsA cofactor[52].Its import may be needed for Dms function in EHEC cells.Fnr regulates gene expression for some of the MAF

Table 2 .
Cont.Transcription factors involved in expression control of the respective genes are listed.+/2, positive/negative regulation. '

Table 3 .
Selection of EHEC proteins differentially abundant in the hemoglobin-positive (HB+) group versus the hemoglobinnegative (HB-) in vivo group.
'Average (avg) abundance changes derived from HB+ and HB-in vivo sample groups (p,0.05) were also compared with the in vitro protein abundance data (*).The statistical significance of the in vivo vs. the in vitro change and an observed increase qor decrease Q are also provided.Details are provided in Dataset S2. doi:10.1371/journal.pone.0066462.t003

Table 4 .
Sus scrofa proteins identified from purified intestinal EHEC cells.