(i) Animals.

Protocols used were reviewed and approved by the USDA-ARS- Institutional Animal Care and Use Committee, National Animal Disease Center, Ames, IA, United States. A total of four rumen-fistulated Holstein heifers/cows (#A - #D), 3–4 years of age and routinely used as rumen fluid donors at the NADC, were utilized for this project. All experiments were conducted in duplicate. For the in vitro experiments, rumen fluid was collected from animals #A and #D, on the lactation and maintenance diets respectively, to evaluate the four different E. coli strains. For the in vivo experiments, animals #A and #B were used and bacteria exposed to the rumens of both animals, sequentially, to factor in effects of any in vivo host-related variations on bacteria being tested. Both animals were on the lactation diet for the first half of the in vivo experiments, then transitioned and acclimated to the maintenance diet over 2 weeks to keep them on this maintenance diet for the second half. Animal #C, used as a control for the in vivo experiments, remained unexposed to the test bacteria but experienced all the dietary changes along with the test animals to comparatively monitor diet-related changes in rumen fluid. Feed intake and body temperatures were monitored daily for all animals. Animals were housed in barns at the NADC; in a field barn initially and subsequently in BSL-2 containment barns when ready to introduce the E. coli strains into the rumen. At the end of the study, following confirmation of their negative-shedding status for all E. coli strains tested, all animals were returned to the field barn.

(ii) Diet.

As in the previous study [28] the animals were fed one of two diets typically provided to dairy cattle: the lactation diet (L diet) usually provided at the time of lactation to meet the high energy needs of the cow, or the total mixed ration maintenance diet (M diet) usually formulated to maintain the overall weight of the animal when dry [30,31]. All dietary compositions were monitored by an animal nutritionist on site and prepared as reported previously [28]. The M diet comprising 25% grass hay, 65% corn silage, and 10% Steakmaker 40–20 (S1 Fig; Land O’Lakes, Inc., Arden Hills, MN) was limit fed at ~19 lbs/head with access to pasture or at ~30 lbs/head without pasture. The L diet with 2.5% grounded corn, 62% corn silage, 2.5% soybean meal, 12% legume hay and 21% lactation premix (S1 Fig) was limit fed at ~99 lbs/head. The cattle were initially fed the L diet for 2 weeks following which, when needed, the diet was gradually switched to the M diet over 2 weeks to allow for acclimation. The M diet was then continued for another 2 weeks. Water was provided ad libitum throughout.

(i) Inoculum preparation.

Log-phase cultures of each bacterial strain were prepared in Luria-Bertani (LB) broth, with or without nalidixic acid (100 μg/ml; LB-Nal), at 39^oC with aeration, as described previously [28]. Bacteria harvested from the log-phase cultures at an OD₆₀₀ 0.5–0.6, washed and re-suspended in sterile phosphate-buffered saline (PBS), were used to charge dialysis cartridges as described below under “Evaluation of STEC strains”. All STEC cultures were confirmed by culture on Rainbow Agar O157 (Biolog, Hayward, CA) as described below, and serologically by using latex agglutination tests (E. coli O157 latex, Oxoid Diagnostic Reagents, Oxoid Ltd., Hampshire, UK and E. coli non-O157 identification kit, Pro-Lab Diagnostics, Ontario, Canada).

(ii) Isolation of bacteria from dialysis cartridges.

Post-incubation, the retrieved dialysis cartridges were wiped down with 70% ethanol before puncturing the dialysis membrane with a sterile 18-gauge needle attached to a syringe for aspiration. Bacteria adhering to the dialysis cartridge walls were scrapped off into the suspension within before aspirating the same. Serial dilutions of the aspirated bacterial suspensions were prepared using sterile 0.9% saline and plated on LB-Nal and/or sorbitol MacConkey (BD Biosciences) with 4-methylumbelliferyl-β-d-glucuronide (MUG, 100 mg/liter; Sigma; SMAC-MUG) agar to determine viable colony counts in CFU/ml. Well-isolated sorbitol-fermenting, and MUG-utilizing (fluorescent under UV light; O26 and O145) or non-utilizing (non-fluorescent; O111) colonies on SMAC-MUG were individually plated on Rainbow Agar O157 (Biolog). The Rainbow Agar O157 plates were incubated overnight at 37^oC and colonies were selected based on color for further serological verification of the serogroup as follows: O157- black; O26 – purple/magenta; O111 – bluish-gray; O145 – purplish-gray and E. coli Nal^R-purplish-pink colonies (Biolog; [32,33]).

(iii) Rumen fluid and fecal sample testing for O157 and non-O157 STEC.

As done previously, rumen fluid and fecal samples were cultured over two weeks, to determine the presence of any STEC in the animal, both at the beginning and end of the study [28]. Standardized non-enrichment and selective-enrichment culture protocols were used with modifications [34–36]. Briefly,10 ml rumen fluid or 10-g fecal sample was added to 50 ml Trypticase soy broth (BD Bioscience, San Jose, CA) supplemented with cefixime (50 μg/liter; U.S. Pharmacopeia, Washington D.C), potassium tellurite (2.5 mg/liter; Sigma-Aldrich Corp., St. Louis, MO), and vancomycin (40 mg/liter; Alfa Aesar, Haverhill, MA) (TSB-CTV) or LB-Nal broth and mixed well. Serial dilutions of each sample were prepared with sterile 0.9% saline both before and after overnight incubation of the TSB-CTV or LB-Nal suspension at 37°C with aeration. The dilutions prepared before incubation (non-enrichment cultures) and after overnight incubation (selective-enrichment cultures) were spread plated onto SMAC-MUG and LB-Nal plates. SMAC-MUG plates were read after overnight incubation at 37^oC and colonies that did not ferment sorbitol or utilize MUG (non-fluorescent under UV light) were further evaluated to be O157 serologically. Fluorescent (MUG-utilizing) and non-fluorescent colonies that fermented sorbitol were cultured on the Rainbow Agar O157 plates and serologically tested for O26, O111 and O145 serogroups as described above.

(i) In vitro evaluation in rumen fluid.

Rumen fluid was collected from two different rumen-fistulated animals, one on the L diet (animal #A) and the other on the M diet (animal #D), for the respective in vitro experiments, and distributed separately in 300 ml aliquots per sterile flasks with rubber stoppers. To create anaerobic culture conditions, as described previously, the flasks were transferred into the anaerobic Coy Chamber for 72 hr, charged with bacteria containing dialysis cartridges within the chamber, sealed and then removed for incubation at 39^oC for 48 hr [28]. Sterile dialysis cartridges (Ultra-test Dialyzer Chamber, Harvard Apparatus, Hollister, MA), charged with the 4 ml of the bacterial suspension in PBS and sealed at both ends with polycarbonate membranes (0.01um pore size, Harvard Apparatus), were used to contain the test bacteria within allowing for easy recovery while permitting diffusion of rumen fluid through the cartridge [28]. A total of 4 dialysis cartridges per bacterial strain were used.

(ii) In vivo evaluation in the rumen.

Two rumen-fistulated animals, #A and #B, were used to set up the in vivo studies; both animals were fed the lactation diet in the first half of the in vivo study and the maintenance diet in the second half, post-acclimation. In each half of the study, the test bacteria were exposed to both rumens sequentially, by alternating the animal used for the experiment, to account for possible host-related influences on the bacteria besides the diet itself. A total of 4 dialysis cartridges per bacterial strain were used. As done previously, the dialysis cartridges charged with bacteria were placed in mesh laundry bags tethered to a 3 ft. fishing line attached to the rumen cannula cap to ensure retrievability from the rumen [28]. All bags with the charged dialysis cartridges were introduced into the rumen and retrieved only after 48 hr, then transported to the lab on ice where the integrity of the dialysis cartridges was verified and contents processed [28].

(i) Protein sample preparation and labelling for isobaric tags for relative and absolute quantification (iTRAQ) analysis.

Following each in vitro and in vivo experiment, harvested bacteria of the same serogroup and growth condition were pelleted together. The pellets were washed three times with ice-cold sterile PBS (pH 7.4), frozen and subsequently processed to obtain bacterial cellular and membrane proteins for proteomic analysis as described previously [28,40]. For iTRAQ, the protein fractions were quantified [41] and dissolved in denaturant buffer (0.1% SDS (w/v) and dissolution buffer (0.5 M triethylammonium bicarbonate, pH 8.5) in the iTRAQ 8-plex kit (AB SCIEX Inc., Foster City, CA). A total of 100 μg of protein per sample were reduced, alkylated, trypsin-digested, and resulting peptides labeled with either of the iTRAQ tags 114, 115, 116,117, 118, 119 or 121, according to the manufacturer’s instructions (AB SCIEX Inc.). Peptides from bacteria grown in MRF-in vitro and MRF-in vivo were labeled separately based on strain-growth condition and then mixed in one tube; the LRF-in vitro and LRF-in vivo peptides were processed likewise. The mixed labeled peptides were dried and held at -80°C until ready for processing as follows. The labeled peptides were desalted with C18-solid phase extraction and dissolved in strong cation exchange (SCX) solvent A (25% (v/v) acetonitrile, 10 mM ammonium formate, and 0.1% (v/v) formic acid, pH 2.8).

(ii) Strong cation exchange fractionation and reverse-phase liquid chromatography with tandem mass spectrometry (LC-MS/MS).

The peptides were fractionated using an Agilent HPLC 1260 with a polysulfoethyl A column (2.1 × 100 mm, 5 µm, 300 Å; PolyLC, Columbia, MD). Peptides were eluted with a linear gradient of 0–20% solvent B (25% (v/v) acetonitrile and 500 mM ammonium formate, pH 6.8) over 50 min followed by ramping up to 100% solvent B for 5 min. The absorbance at 280 nm was monitored and a total of 10 fractions per sample were collected. The fractions were lyophilized and resuspended in LC solvent A (0.1% formic acid in 97% water (v/v), 3% acetonitrile (v/v)). A hybrid quadrupole Orbitrap (Q Exactive Plus) MS system (Thermo Fisher Scientific, Bremen, Germany) was used with high energy collision dissociation (HCD) in each MS and MS/MS cycle. The MS system was interfaced with an automated Easy-nLC 1000 system (Thermo Fisher Scientific). Each sample fraction was loaded onto an Acclaim Pepmap 100 pre-column (20 mm × 75 μm; 3 μm-C18) and separated on a PepMap RSLC analytical column (250 mm × 75 μm; 2 μm-C18) at a flow rate of 350 nl/min during a linear gradient from solvent A (0.1% formic acid (v/v)) to 30% solvent B (0.1% formic acid (v/v) and 80.0% acetonitrile (v/v)) for 95 min, to 98% solvent B for 15 min, and hold 98% solvent B for an additional 30 min. Full MS scans were acquired in the Orbitrap mass analyzer over m/z 400–2000 range with resolution 70,000 at 200 m/z. The top ten most intense peaks with charge state ≥ 2 were isolated (with 2 m/z isolation window) and fragmented in the high energy collision cell using a normalized collision energy of 28% selected. The maximum ion injection time for the survey scan and the MS/MS scans were 250 ms, and the ion target values were set to 3e6 and 1e6, respectively. The selected sequenced ions were dynamically excluded for 60 sec.

(i) Database searching.

Tandem mass spectra were extracted by Proteome Discoverer (Thermo-Fisher) version 2.2.0.388. Charge state deconvolution and deisotoping were not performed. All MS/MS samples were analyzed using Mascot (Matrix Science, London, UK; version 2.4.1). Mascot was set up to search the E. coli O26:H11, E. coli O111:H8, and E. coli O145:NM databases, that are also part of the E. coli pan proteome, individually, assuming the digestion enzyme trypsin. Mascot was searched with a fragment ion mass tolerance of 0.60 Da and a parent ion tolerance of 20 PPM. Methylthio of cysteine and iTRAQ8plex of lysine and the n-terminus were specified in Mascot as fixed modifications. Gln->pyro-Glu of the n-terminus, deamidated of asparagine and glutamine, methyl of lysine, oxidation of methionine, phospho of serine, threonine and tyrosine, nmethylmaleimide of cysteine and iTRAQ8plex of tyrosine were specified in Mascot as variable modifications.

(ii) Criteria for protein identification.

Scaffold (version Scaffold_4.11.1, Proteome Software Inc., Portland, OR) was used to validate MS/MS based peptide and protein identifications. Peptide identifications were accepted if they could be established at greater than 80–95% probability by the Peptide Prophet algorithm [42] with Scaffold delta-mass correction. Protein identifications were accepted if they could be established at greater than 80–95% probability and contained at least 2 identified peptides. False discovery rate (FDR) was at medium confidence; 0.2–0.5% for proteins post-MR exposure and 2.5–3% for proteins post-LRF exposure. Protein probabilities were assigned by the Protein Prophet algorithm [43]. Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony. Proteins sharing significant peptide evidence were grouped into clusters.

(iii) Intensity Based Absolute Quantification (iBAQ) analysis of the LC-MS/MS data.

MaxQuant [44] was used for label-free peptide and protein identification, protein iBAQ quantification, and reporter ion intensity quantification. The Uniprot E. coli pan proteome UP000000625 [45] was used as the search database. Only proteins identified by more than one peptide with Q values less than 0.05 were considered for these analyses. iBAQ values comparing the abundance of proteins between the two diets were transformed into relative iBAQ (riBAQ) values (S2 Fig). Each protein was represented by a single riBAQ value within each diet. To detect proteins which were differentially abundant within each diet the log2FoldChange (L2FC) was calculated by: log2(lact_riBAQ/ maint_riBAQ). Proteins with an absolute L2FC of greater than 1 were considered differentially abundant.

(iv) iTRAQ quantitation without reference comparison.

Two different methods were tested to normalize reporter ion intensities. NOMAD, normalization of mass spectrometry data, was used to normalize peptide reporter ion intensities and aggregate peptide reporter ion intensities to protein level intensities. This is an ANOVA type of normalization and is similar to what can be performed in Scaffold. In addition, centered log ratios (clr) were used to normalize aggregated protein intensities. This is a more general normalization that can be performed on many different ‘-omics’ data [46]. Differentially abundant proteins between in vivo and in vitro conditions were determined using T-tests and P values were corrected by the FDR method.

(i) In LRF.

In vitro, the pH of the rumen fluid collected from animal #A on the L diet was at 5.6 prior to addition of dialysis cartridges with STEC. The rumen fluid pH ranged between 5.6–6.0, post-incubation, in the different flasks with STEC containing dialysis cartridges suggesting absence of any bacterial effects (Table 1). Likewise, no bacterial influence on ruminal pH was observed in the in vivo experiments with animals #A and #B on the L diet; pH ranged from 5.2–6.0 prior to introduction of dialysis cartridges into the rumen and remained between 5.5–6.0 after 48 h (Table 2). The ruminal pH of the control animal #C ranged between 5.7 and 5.5 at timepoints matching pre- and post- challenge of the test animals (Table 2).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. pH and volatile fatty acid compositions of LRF used in in vitro experiments.

https://doi.org/10.1371/journal.pone.0313978.t001

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. pH and volatile fatty acid compositions of LRF in in vivo experiments.

https://doi.org/10.1371/journal.pone.0313978.t002

The total VFA concentrations ranged from 141–230 μM/ml in the in vitro samples with higher values occurring post-exposure to test bacteria (Table 1). In vivo, the ruminal total VFA concentrations were between 144–193 μM/ml for the test animals and between 121–168 μM/ml for the control animal (Table 2). In contrast to the observed increase in VFA post-bacterial exposure in vitro, VFA values were lower in vivo post-exposure to the test bacteria. Since this decrease was also observed with the control animal it is unlikely this was a bacterial effect, and more host related.

(ii) In MRF.

In the in vitro experiments, the pH of the rumen fluid collected from animal #D on the maintenance diet ranged from 6.0 to 6.2 irrespective of the presence of the test bacteria (Table 3). The total VFA concentrations ranged between 141–190 μM/ml (Table 3), with increased concentrations observed post-exposure to test bacteria as with LRF-in vitro. In the in vivo experiments, the ruminal pH ranged between 6.0–7.0 and total VFA between 87–174 μM/ml (Table 4). The test bacteria did not impact the pH or VFA concentrations in the in vivo experiments with similar changes observed in the control animal (Table 4).

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. pH and Volatile fatty acid compositions of MRF used in in vitro experiments.

https://doi.org/10.1371/journal.pone.0313978.t003

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. pH and Volatile fatty acid compositions of MRF in in vivo experiments.

https://doi.org/10.1371/journal.pone.0313978.t004

Overall, butyrate concentrations were higher in LRF than MRF while the opposite was true for acetate concentrations, as expected [28,30,31]. The differences in the total VFA concentrations, between the pre- and post-exposure samples, were significant (p<0.05) in all instances except for the LRF-in vivo samples.

(i) In vitro survival in rumen fluid.

Post-incubation in LRF, the STEC serogroups demonstrated greater reduction (5–7-log) in average viable counts compared to E. coli Nal^R (3-log), with STEC O26:H11 having the highest reduction by 7-log (Fig 1; S1 Table). In MRF, the STEC serogroups had 2-log reduction in average viable counts post-incubation compared to the 3-log reduction in E. coli Nal^R viable counts (Fig 2A; S2 Table). Based on these results, it appeared that the non-O157 STEC serogroups have different survival dynamics than E. coli Nal^R, in vitro, in both rumen fluids (Figs 1A and 2A) and were more sensitive to the low pH, higher VFA conditions in LRF-in vitro. The test bacteria were not recovered from either rumen fluids, pre- or post- exposure to the dialysis cartridges, thereby verifying the containment of the test bacteria within.

(ii) In vivo survival in the rumen.

In contrast to the in vitro results, all three STEC serogroups had a survival pattern similar to that of E. coli Nal^R. Post-48 hr incubation within the rumen of fistulated cows fed the L diet, an average of 1–3-log reduction in the viable counts was observed among the tested bacteria (Fig 1B; S1 Table). Likewise, a 1–2-log reduction in the averaged viable counts was recorded post-48 hr exposure to the rumen of cows on the M diet (Fig 2B, S2 Table). These results further confirm the importance of conducting in vivo experiments as previously reported [28]. The test bacteria were not recovered from the rumen fluid and fecal samples of the animals, including the control animal, prior to and post-exposure to the dialysis cartridges containing the test bacteria.

(i) iBAQ analysis highlighted overall differences in proteins expressed in LRF and MRF, irrespective of growth condition.

Using iBAQ analysis proteins expressed collectively by STEC O26:H11, O111:H8 and O145:NM, uniquely or differentially, in MRF and LRF were identified (S3–S6 Tables). Since all samples pooled in each MS run belonged to the same rumen fluid type, in both the in vitro and in vivo culture conditions, iBAQ data for a particular rumen fluid type could be used to estimate the overall differences in protein expression between MRF and LRF. As done previously, to avoid any false interpretations due to differences in intensities, iBAQ values derived from the LC-MS/MS data were converted to riBAQ values to look at relative protein abundance for each rumen fluid type [28] (S2 Fig) without focusing on individual strains or the in vitro/in vivo growth conditions. A total of 446 bacterial proteins were identified by riBAQ of which 371 proteins were detected in both rumen fluids, 34 were unique to LRF and 41 were unique to MRF (S3–S6 Tables). In addition, any protein with an absolute L2FC value of 1 or greater was considered to be enriched in the respective rumen fluid (S3 Fig); 91 proteins were found to be enriched in LRF (S3 and S5 Tables) and 185 enriched in MRF (S3 and S4 Tables). These differentially expressed proteins were used to identify GO terms that were significantly enriched in each rumen fluid (S5 and S6 Tables).

(ii) Reference-free iTRAQ analysis identified proteins enriched by growth conditions in each rumen fluid.

Using reference-free iTRAQ data analysis, along with centered log ratios (clr) normalization, a total of 405 and 412 proteins were identified as being differentially expressed by the three serotypes in LRF and MRF, respectively (Fig 3). In LRF, of the 405 proteins, 25 were more highly expressed in the in vivo condition and 37 were more highly expressed in the in vitro condition (Fig 3, S7 and S8 Tables). In MRF, 26 of the 412 proteins were more highly expressed in the in vivo condition and 41 were more highly expressed in vitro (Fig 3; S9 and S10 Tables). NOMAD normalization yielded minimal data except with proteins expressed in vitro in LRF and hence, not used in the final analysis (S11–S14 Tables). Multivariate similarity analysis aligned the serotype-proteomes more with growth conditions and showed minimal strain-based variations supporting the observed in vivo and in vitro survival patterns (S4 Fig, 1 and 2). All the same analysis of the top fifty proteins expressed with most variation between strains, in LRF or MRF, did reveal subtle serotype-related differences (S5 Fig; S15–S16 Tables).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Venn diagrams of total differentially expressed proteins, up regulated in vivo and in vitro, in LRF and MRF.

https://doi.org/10.1371/journal.pone.0313978.g003

(iii) iBAQ and iTRAQ analyses provided a broader and in-depth perspective on proteins expressed in LRF and MRF, respectively.

Based on the iBAQ analysis, overall, a total of 91 STEC proteins were identified to be enriched in LRF and 185 proteins in MRF (S3–S6 Tables), irrespective of the growth conditions. Analysis of GO-terms for these proteins indicated an enrichment of transport, ribosome assembly, and membrane assembly pathways in LRF and MRF (examples: AcrB, BamD, BtuB, DegP, FadL, FepA, FhuA, HisJ, LamB, OmpF, OmpT, OmpW, SlyB, TolC, WzzB) while biosynthetic pathways were primarily dominant in MRF (S5–S6 Tables). Interestingly, more membrane associated proteins were enriched (L2FC >1) in LRF than in MRF including NmpC, GfcE, LamB, OmpA, CirA, Ag43, Eae, FhuA, FepA, GfcD, BtuB, LptD, OmpX, TolC, BamD, OmpT, BamB, OmpF, SlyB, OmpW, FadL, NuoG, WzzB, PtsG, LpoA, AcrB, PspA, and NarG (S3–S6 Tables). On the other hand, 11 membrane-associated proteins were enriched in MRF, (GdhA, ZipA, SecD, MetQ, FabF2, Alkyl hydroperoxide reductase subunit F, YhcB_1, AtpB, SdhB, FrdA, XdhD) (S3–S6 Tables).

Centered log ratios (clr) normalization helped sort the reference-free-iTRAQ data better than NOMAD (S7–S14 Tables). Following clr normalization, the iTRAQ data could be used to reliably distinguish proteins expressed exclusively in vivo or in vitro, in both LRF and MRF (Fig 3, Table 5, S7–S10). While bacterial proteins identified by iBAQ overlapped with those identified by reference- free-iTRAQ, the data analysis using the latter was nuanced in the context of growth conditions. For instance, membrane-associated proteins expressed in LRF, determined as in vivo-expressed included, AtpE, TolC, YdgA, DmsB, BamA, Lpp and those determined as in vitro-expressed were FabI, HtpG, AtpF (S7 and S8 Tables). Likewise, proteins expressed in MRF were identified as in vivo-expressed (FadL, FhuA, LptD, TolC, SlyB) or in vitro-expressed (PspA) (Table 5, S9 and S10). All other membrane proteins identified by iBAQ were determined to be expressed under both growth conditions by iTRAQ analysis in either one of the strains tested. Additionally, as observed with iBAQ analysis, proteins associated with metabolic/biosynthetic pathways were enriched in vitro and in vivo in MRF (Table 5). While virulence-associated proteins were not highlighted by iTRAQ analysis, iBAQ identified proteins that support pathogenicity, namely, intimin (LRF) involved in STEC adherence [53] and glutamine/arginine transport proteins (LRF/MRF) associated with acid resistance [54] (S3–S6 Tables).

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. E. coli proteins differentially enriched in rumen fluid influenced by the maintenance (MRF) and lactation (LRF) diets, in vitro and in vivo, based on reference-free iTRAQ data analysis.

https://doi.org/10.1371/journal.pone.0313978.t005

Few of the top 50 proteins with differential expression may have contributed to the outlier serotype-proteomes observed in the multivariate analysis (S4–S5 Figs; S15–S16 Tables). For instance, in LRF, proteins that were expressed by STEC O26:H11 and STEC O111:H8 in vitro (OmpA, OmpT, OmpX, NmpC, OmpC, FecA) were expressed in vivo by serotype O145:NM (S5 Fig). Similarly, some proteins were expressed in vitro by serotype O145:NM alone (UbiD_1, DcyD, HupB, RplL, NapA, NanA, RplC); these differences may have resulted in the serotype O145:NM being an outlier in vitro in LRF (S4 and S5 Figs). On the other hand, STEC O111:H8 was an outlier in LRF-in vivo, as it did not express RpS6, YjbH proteins under this growth condition (S4 and S5 Figs). In MRF the only outlier was the serotype O145:NM, under the in vitro condition (S4 Fig), expressing proteins otherwise expressed by the other two serotypes in vitro under in vivo growth conditions (ArgT, DcyD, NanA, SucC, YbiB, FumA, GldA_1, XdhD, AspA, EspC, UbiD; S5 Fig). While serotype O145:NM being an outlier under in vitro conditions in LRF and MRF may be associated to its porcine origin or lack of toxin genes, no such differences were observed in vivo ruling out this possibility.