The recent outbreak of severe infections with Shiga toxin (Stx) producing Escherichia coli (STEC) serotype O104:H4 highlights the need to understand horizontal gene transfer among E. coli strains, identify novel virulence factors and elucidate their pathogenesis. Quantitative shotgun proteomics can contribute to such objectives, allowing insights into the part of the genome translated into proteins and the connectivity of biochemical pathways and higher order assemblies of proteins at the subcellular level.
We examined protein profiles in cell lysate fractions of STEC strain 86-24 (serotype O157:H7), following growth in cell culture or bacterial isolation from intestines of infected piglets, in the context of functionally and structurally characterized biochemical pathways of E. coli. Protein solubilization in the presence of Triton X-100, EDTA and high salt was followed by size exclusion chromatography into the approximate Mr ranges greater than 280 kDa, 280-80 kDa and 80-10 kDa. Peptide mixtures resulting from these and the insoluble fraction were analyzed by quantitative 2D-LC-nESI-MS/MS. Of the 2521 proteins identified at a 1% false discovery rate, representing 47% of all predicted E. coli O157:H7 gene products, the majority of integral membrane proteins were enriched in the high Mr fraction. Hundreds of proteins were enriched in a Mr range higher than that predicted for a monomer supporting their participation in protein complexes. The insoluble STEC fraction revealed enrichment of aggregation-prone proteins, including many that are part of large structure/function entities such as the ribosome, cytoskeleton and O-antigen biosynthesis cluster.
Nearly all E. coli O157:H7 proteins encoded by prophage regions were expressed at low abundance levels or not detected. Comparative quantitative analyses of proteins from distinct cell lysate fractions allowed us to associate uncharacterized proteins with membrane attachment, potential participation in stable protein complexes, and susceptibility to aggregation as part of larger structural assemblies.
Citation: Pieper R, Zhang Q, Clark DJ, Huang S-T, Suh M-J, Braisted JC, et al. (2011) Characterizing the Escherichia coli O157:H7 Proteome Including Protein Associations with Higher Order Assemblies. PLoS ONE 6(11): e26554. https://doi.org/10.1371/journal.pone.0026554
Editor: Gunnar F. Kaufmann, The Scripps Research Institute, United States of America
Received: July 29, 2011; Accepted: September 28, 2011; Published: November 7, 2011
This is an open-access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.
Funding: The funder was the National Institute of Allergy and Infectious Diseases (NIAID), National Institutes of Health (NIH). The funder was involved in the selection process of this White Paper research project. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. At Tufts University, the work was funded by the National Institute of Allergy and Infectious Diseases (NIAID), National Institutes of Health (NIH) under contract number N01-AI-30050. At the J Craig Venter Institute, the work was funded by the NIAID, NIH, under contract number N01-AI15447. The funders had no role in study design, data collection and analysis or preparation of the manuscript. Because the work was funded via two NIH contracts, the funder reserved the right to approve submission of the manuscript for publication, but the research teams made the decision to publish.
Competing interests: The authors have declared that no competing interests exist.
Enterohemorrhagic E. coli (EHEC) are a group of bacteria containing many serotypes that are responsible for outbreaks of bloody diarrhea often leading to serious systemic complications such as the hemolytic uremic syndrome (HUS) and neurological abnormalities which in severe cases can be fatal . These complications are attributed to Shiga toxins (Stx), shared with Shigella dysenteriae and acquired via horizontal gene transfer through phages . The most common cause of outbreaks and sporadic cases of bloody diarrhea and HUS in the US are strains of the serotype O157:H7. The recent outbreak in Europe highlights the fact that Stx producing E. coli (STEC) serogroups, other than E. coli O157:H7, unable to form attachment and effacement lesions in the colon which are characteristic for EHEC , are also capable of causing severe systemic illness in humans . STEC is defined by the CDC as a pathogenic E. coli isolate/strain expressing an active Shiga toxin and causing diarrhea or more severe disease symptoms. STEC strain 86-24 caused an outbreak of hemorrhagic colitis with a high mortality rate in 1986  and was shown to produce only Shiga toxin 2 (Stx2) . While epidemiological data suggest that Stx2 producers are more frequently associated with HUS than Stx1 producers, it is less clear whether Stx2 or Stx1 and Stx2 producers are more prevalent among strains causing severe disease complications , . Pig is the only species besides human naturally susceptible to the systemic effects of Stx, and Stx2 producers in particular have caused severe neurologic symptoms, beyond bloody diarrhea, in the piglet model .
Shotgun proteomics is a term used for experiments in which complex protein samples are digested into peptide fragments followed by one or several dimensions of liquid chromatography (LC) to separate the peptides and tandem mass spectrometry (MS) to identify these peptides via mapping of peptide-spectral matches (PSMs) back to the proteins of origin , . Mapping the PSMs is facilitated by powerful computational algorithms such as Mascot , Protein/Peptide Prophet  and InsPecT  and availability of accurate protein sequence databases derived from annotated genomes. 2D-LC-MS/MS refers to two orthogonal dimensions of peptide separations prior to tandem MS (also termed MS2), the latter one typically performed with a nano-flow C18-reversed phase capillary column. Peptide separation methods continue to evolve and include specialized applications using affinity and size exclusion chromatography (SEC) and biphasic columns . The diversity of methods developed for protein quantification from shotgun proteomics data is considerable, as recently reviewed . Among label-free quantification techniques, the major ones are based on MS1 normalized peak intensity measurements, such as the accurate mass and time tag approach , and MS2 based spectral counting techniques, such as emPAI  and APEX . Due to experimental variability, e.g. in the peptide LC elution times in highly complex mixtures and because of stochastic sampling and ion suppression effects during MS1 and MS2 scans, quality control metrics are important to maintain reproducible conditions across all analyses compared in a given experiment , . To compute the probability of detection for each peptide theoretically present in a shotgun proteomics dataset, the APEX method employs machine learning techniques . The computations result in peptide/protein Oi values (correction factors that define detection probabilities) that are integrated into an equation to normalize spectral count values. This method provides added flexibility in that it allows quantitative comparisons of the same protein across datasets (samples) and of different proteins in the same dataset. To validate the latter analysis feature, we recently assessed the stoichiometry of several Shigella dysenteriae protein complexes with known subunit ratios experimentally and obtained CVs of less than 20% for several soluble intracellular complexes .
Shotgun proteomics experiments without fractionation on the subcellular or protein level do inherently not produce information on protein assembly in larger entities (such as complexes and organelles), co-fractionation, solubility, membrane attachment and subcellular localization. Non-denaturing protein fractionation performed prior to 2D-LC-MS/MS affords the opportunity to evaluate patterns of protein distributions in a distinct fraction and among several fractions. To our knowledge, only two publications have described the use of SEC to separate proteins via Mr prior to shotgun 2D-LC-MS/MS analysis. The human mammary epithelial cell proteome  and the human liver tissue  were profiled, in each case resulting in expanded proteome coverage with ∼1,600 protein identifications. These data were not examined with the purpose to establish patterns of protein fractionation. Although SEC needs to be used in combination with light scattering and refractive index measurements to determine reliable Mr values for protein complexes , the technique is quite useful to estimate whether the native Mr of a protein corresponds to a monomer, a homo-oligomer or a larger structural entity involving multiple proteins, phospholipids, glycans or ribonucleoproteins. Proteomic profiling of soluble versus insoluble cell lysate fractions was applied to Escherichia coli strain MG1655, resulting in 1,147 protein identifications with enrichment of some in the insoluble fraction . Insoluble E. coli lysate fractions have also been studied to assess the propensity of proteins to form aggregates , . The aforementioned studies were elegant but did not focus on robust global quantification of proteins comparing lysate fractions or correlation of data with protein participation in higher order complexes.
The proteomics literature on E. coli is extensive . In addition to protein profiling by two-dimensional gel electrophoresis (2D-GE) and LC-MS/MS, studies employing large-scale affinity purification followed by mass spectrometry (AP-MS) have contributed to determine novel functional roles, interactions and networks of proteins , , . The combined number of physical E. coli protein-protein interactions from these studies was ∼17,500. Single-cell E. coli proteomic and transcriptomic data revealed surprisingly low correlations of protein and mRNA copy numbers . At least a third of the in silico predicted E. coli proteome remains to be functionally characterized. Even for proteins attributed to have a specific function, novel functions continue to emerge, e.g. the roles of elongation factor Tu in cell shape maintenance  and enolase as a component of the mRNA degradosome . The first genome of an E. coli serotype O157:H7 strain was published by Perna et al. . Remarkably, 1,387 new genes encoded in strain-specific clusters were discovered in E. coli O157:H7 compared to strain MG1655. Most of the lateral gene transfer appeared to have occurred by phage integration. A key pathogenicity island in EHEC is the locus of enterocyte effacement (LEE). Some LEE-expressed proteins contribute to intimate bacterial adherence to intestinal epithelial cells, and the resident type 3 secretion system (T3SS) injects effectors into the host cell cytoplasm . E. coli O157:H7 strains also feature a repertoire of effectors whose genes are spread across the genome . The literature pertaining to STEC and EHEC proteomes and their adaptations to the host environment is sparse. In addition to a study discovering novel T3SS effectors , protein-based responses to oxidative stress in a stress-sensitive E. coli O157:H7 strain were examined .
Two major objectives were pursued in the research described here. First, using fractionation on the protein level and shotgun proteomics, the intention was to profile the E. coli serotype O157:H7 (strain 86-24) proteome. Secondly, we systematically examined whether biochemically relevant information can be derived from cell lysate fractionations, taking advantage of existing information in databases for E. coli protein functions, localizations, large subcellular assemblies and biosynthetic and metabolic pathways. Instead of using only cultured bacteria to accomplish these objectives, we also processed bacteria directly from the large bowel of gnotobiotic piglets infected with E. coli O157:H7.
Experiments in piglets were performed to allow analysis of the E. coli O157:H7 proteome from an in vivo environment. Piglets were delivered by cesarian 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 growth and recovery from suspension culture 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 (OD600∼2.0) in a shaking incubator at 37°C, pelleted by centrifugation at 7,000×g for 10 min at 4°C, washed with a 20-fold volume of PBS and re-concentrated at 7,000×g for 15 min at 4°C. 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 1×109 and 2×1010 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 14,500×g for 30 min at 4°C, as previously described . Up to 2×1010 bacterial cells were re-suspended in 1 ml of a hypotonic lysis buffer (25 mM Tris-OAc, pH 7.8, 0.05% Triton X-100, 5 mM Na-EDTA, and protease inhibitors benzamidine and AEBSF in 1 mM concentrations). Chicken lysozyme (150 µg/ml) was added to initiate cell lysis, agitating the suspension at 20°C for 1 h followed by quick-freezing and storage at −80°C until further processing. We did not observe any contamination by pig proteins in bacterial lysates derived from piglet intestines, judged by analysis of the 500 most dominant spots in 2D gels, all of which were identified as STEC proteins.
Fractionation of cell lysates
To complete cell lysis, disintegrate post-lysis aggregates and degrade nucleic acids, each thawed suspension was sonicated (ten 30-sec on/60-sec off cycles at amplitude 5; Misonex 3000 sonicator) on ice. After addition of DNAse I and RNAse at 5 µg/ml and 15 mM MgCl2, the suspension was gently agitated for 1 h at 20°C and centrifuged at 16,100×g for 30 min at 4°C. Supernatant and insoluble pellet fractions were retained. The pellet was re-suspended in 300 µl of a high salt solution (50 mM Tris-OAc, pH 7.8, 2.5 M NaBr and 5 mM Na-EDTA), agitated for 1 h at 20°C, and spun again at 16,100×g for 30 min at 4°C. This supernatant was added to the 1st lysate supernatant. The pellet was briefly washed with a 10-fold excess of PBS, and the wash solution was discarded. Small aliquots of the insoluble pellet and the lysate supernatant were analyzed in 4–12%T Coomassie Brilliant Blue G-250 (CBB)-stained SDS-PAGE gels to assess protein recovery. Following concentration to ca. 2 mg/ml protein, 250 µl aliquots of the supernatant were applied to the analytical SEC column G3000-SWXL (7.8 mm×30 cm, 5 µm particle size, TOSOH Bioscience, USA) and separated at a flow rate of 0.8 ml/min using PBS supplemented with 0.01% Triton X-100 at pH 7.5. Calibration runs on the SEC column were performed using the following protein standards: thyreoglobulin (670 kDa), bovine IgG (158 kDa), ovalbumin (45 kDa), myoglobin (18 kDa) and cobalamin (1.7 kDa). Elution peaks detected at A280 were plotted against native Mr values of proteins (log Mr versus elution volume VE), permitting a rough calculation of Mr ranges for the VE segments derived from the SEC fractionation of samples. Fractions were collected in 2 min increments and analyzed in CBB-stained 4–12%T SDS-PAGE gels; those corresponding to 6.8–8.8 min, 8.8–10.8 min and 10.8–14.8 min elution times contained nearly all protein solubilized from lysates and were recovered for further analyses. The fractions termed F1-sSEC, F2-sSEC and F3-sSEC, respectively, from here on were devoid of lysozyme, which eluted in VE segments beyond the 15 min mark due to its low Mr. The insoluble lysate pellet termed F4-p from here on was not subjected to SEC, but protein aliquots containing ca. 60–100 µg protein were also prepared for LC-MS/MS.
Iodixanol density gradient centrifugation of insoluble E. coli fractions
Bacterial lysates derived from ca. 4×109 cells (500 µl) grown in vitro to stationary phase were subjected to hydrostatic pressure cycling in the Barocycler NEP3229 (PBI Inc., South Easton, MA), using 15 cycles of exposure to 35,000 psi over 55 sec and 15 psi over 5 sec. This lysis method is gentler than sonication and minimizes membrane denaturation and aggregate formation. The lysis buffer and nucleic acid digestions were identical to conditions described above, except the reduction of post-lysis salt concentrations with only 150 mM NaCl. Iodixanol step gradient layers were prepared with an Optiprep™ stock solution (Axis Shield, Oslo, Norway) containing 50% iodixanol in a 50 mM HEPES buffer, pH 7.8, supplemented with 125 mM sucrose, 0.5 M NaOAc and 5 mM MgCl2. An unfractionated lysate in an aliquot of ca. 500 µl was mixed with 800 µl Optiprep™ stock solution (40% w/v iodixanol concentration) followed by overlays with 33.3%, 25%, 16.7% and 8.3% iodixanol (v/v), each in a 1 ml volume. The sample was spun in a SW60 rotor at 50,000 rpm (257,000×g) for 3 h at 18°C. Turbid bands of non-solubilized material detected in the gradients were carefully aspirated followed by 10-fold dilution with PBS to recover the solid pellet fractions by centrifugation at 16,000×g for 45 min.
Digestion of protein fractions with trypsin
Soluble and insoluble protein fractions were digested using a method termed filter-aided sample preparation (FASP) . Briefly, pellets were re-suspended in 200 µl of 25 mM Tris-OAc (pH 7.8), 0.05% Triton X-100 and 5 mM EDTA. Soluble SEC fractions were transferred stepwise to and concentrated in a Microcon YM-10 membrane filter unit (10 kDa Mr cut-off; Millipore, Billerica, MA) and re-equilibrated in 200 µl of 25 mM Tris-OAc (pH 7.8), 0.05% Triton X-100 and 5 mM EDTA. To each sample, 20 µl of a 1 M DTT stock solution and 12 µl of a 10% SDS stock solution were added, and proteins were denatured by incubation for 3 minutes at 95°C. Room temperature-adjusted samples were transferred back into Microcon YM-10 units and subjected to proteolytic digestion with trypsin (trypsin/bacterial protein ratio of 1∶30 to 1∶50) at 20°C overnight  following alkylation. Each filtrate contained a complex peptide mixture, collected by centrifugation at 14,000×g for 40 minutes in the flow-through. Each Microcon YM-10 filter unit was rinsed twice with 50 µl of 500 mM NH4HCOO and once with 100 µl of 50 mM NH4HCO3. The rinses were added back to the respective protein digestion filtrate, which was then reduced to dryness.
Strong cation exchange chromatography (SCX)
The lyophilized fractions (F1-sSEC, F2-sSEC, F3-sSEC and F4-p) were subjected to peptide separation via SCX on a Polysulfoethyl-Aspartamide column (4.6 mm×50 mm, 5 µm particle size; Nest Group, USA). Each protein digestion mix was reconstituted in SCX buffer A (5 mM KH2PO4, 25% CH3CN, pH 3.0). Linear gradients with SCX buffer B (5 mM KH2PO4, 0.7 M KCl, 25% CH3CN, pH 3.0) were run at a flow rate of 0.5 ml/min: 10 min at 0% buffer B; increase of buffer B concentration to 20% over 30 min, 50% over 20 min and 100% over 5 min; maintain flow at 100% B for 5 min. Fractions were collected in 3 min increments. Fractions 1+2, 3−5, 6+7, 8, 9, 10, 11, 12, 13, 14, 15, 16+17 and 18+19 were collected, individually or in pools as indicated, and lyophilized overnight. Following reconstitution in 75 µl of 1% formic acid in 2% CH3CN, fractions with a high salt content (15, 16+17 and 18+19) were desalted using ZipTip C18 cartridges using a vendor-provided protocol (Millipore, Billerica, MA).
Reversed phase C18 LC-MS/MS and database searches
Peptide mixtures contained in the 52 SCX fractions (13 fractions for the samples F1-sSEC, F2-sSEC, F3-sSEC and F4-p each) were subjected to 2nd dimension separation on a BioBasic C18 column (75 µm×10 cm; New Objective, Woburn, MA). Protein digestions derived from iodixanol gradient insoluble fractions contained ca. 25–50 µg protein and were directly analyzed by C18 LC-MS/MS. The instrument set-up of the nano-ESI LTQ linear ion trap mass spectrometer (LTQ; Thermo, San Jose, CA) coupled to an upfront Agilent 1100 solvent delivery system was described previously . It was calibrated prior to each LC-MS/MS run with 200 nmol human [Glu1]-fibrinopeptide B (M.W. 1570.57), verifying that elution times with a CH3CN gradient varied less than 10% and that peaks representing ion counts had widths at half-height of <0.25 min, signal/noise ratios >200 and heights >107. Briefly, loading a 25 µl sample volume was followed by peptide binding and wash steps on a C18 trapping cartridge at a flow rate of 0.01 ml/min for 13 min. Peptides were eluted from the C18 cartridge and separated on the C18 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) at a flow rate of 350 nl/min. Eluates entered the nano-ESI source, and mass spectra were acquired starting at 10% solvent B. 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 with a dynamic exclusion setting of 90 sec, deselecting repeatedly observed ions for MS/MS. All peptide fractions from a given sample were run consecutively on the LC-MS/MS system. The LTQ search parameters (+1 to +3 ions) included mass error tolerances of ±1.4 Da for peptide precursor ions and ±0.5 Da for peptide fragment ions. The search engine used for peptide identifications was Mascot v.2.3 (Matrix Science). Search parameters allowed one missed tryptic cleavage, and were set for carbamidomethyl-modifications of cysteine and oxidation of methionine residues as fixed and variable modifications, respectively. The protein sequence database for the enterohemorrhagic E. coli O157:H7 strain EDL933 genome was downloaded from the RefSeq dataset in NCBI to enable Mascot searches, because the genome sequence of strain 86-24 has not been published.
Quantitative protein analysis using spectral counting (APEX)
Protein identifications derived from Mascot v.2.3 searches required at least one unique peptide with an e-value <0.1. Mascot search peptide false discovery rates (FDRs) were determined searching an in silico randomized E. coli O157:H7 EDL933 protein sequence database. The average FDR over 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™ and ProteinProphet™ , . Prot.xml files were analysed using an open-source software termed the APEX quantitative proteomics tool v1.1 that we developed in 2008 . Using a pre-defined set of MS analysis parameters and 30 physicochemical properties provided by the APEX tool (default settings), ARFF files for Oi computations were generated for a set of 100 highly abundant bacterial proteins with Mr values >30 kDa (according to 2D gel data) to obtain 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 Oi 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 APEXi values from 2D-LC-MS/MS data includes the Oi correction factors, probability scores for protein identification (pi) and spectral counts (ni) as variables: Setting the protein FDR at 1%, only protein identified at a 99% confidence level were used for spectra counting. A factor of 2.5×106, roughly the estimated number of protein molecules per E. coli cell, was used to normalize APEX scores as protein molecules per cell .
Bioinformatics and statistics methods
Proteins were analyzed using various software tools to predict subcellular localizations, membrane attachment, molecular functions and pathway associations. Functions were derived from information in CMR (http://cmr.jcvi.org/), Swiss-Prot (www.uniprot.org/) and Ecocyc (http://ecocyc.org) databases, pathways examined in Ecocyc. For the in silico prediction of subcellular protein localizations, the software tools Cell PLoc (http://chou.med.harvard.edu/bioinf/Cell-PLoc) and PSORTb v.2.0 (www.psort.org/psortb/) were used. The algorithms of SignalP (export signal peptides), TMHMM (transmembrane domains, TMDs) and LipoP (lipid anchor motifs), all available at www.cbs.dtu.dk/, were used to detect motifs for export and membrane integration. BOMP was used for outer membrane (OM) β-barrel protein prediction (www.bioinfo.no/tools/bomp).
Data from eleven experiments pertaining to biological and technical replication with samples derived from cell growth in vitro were included in statistical analyses to assess quantitative patterns of protein fractionation. There were three biological cell culture replicates. Technical replication was performed prior to SEC and at the final step (C18 LC-MS/MS). Experimental reproducibility with respect to protein quantification was assessed calculating squared Pearson's Product Moment correlation coefficients (R2). Proteins not observed in both of the compared experimental pairs were not considered. Protein abundance data and associated annotations were imported into the MultiExperiment Viewer (MeV, http://www.tm4.org/mev/)  to perform statistical analysis and to cluster and visualize patterns of protein abundances across SEC fractions. Non-parametric statistical methods were applied to identify proteins that showed significant and consistent abundance changes between the SEC fractions under study. The Kruskal-Wallis test was performed to identify proteins with differential abundance among the three SEC fractions (F1-s versus F2-s versus F3-s). The two-sample Wilcoxon Rank Sum test was performed to find proteins differentially abundant comparing the collection of soluble protein fractions (F1-s+F2-s+F3-s) with the insoluble protein fraction (F4-p). Significant proteins from the Kruskal-Wallis test were re-imported into the MeV software to correlate the data with sequence-based, calculated Mrs and Mrs derived from protein complexes of a homo- or heteromultimeric nature that a given protein subunit is known to participate in in E. coli. Annotation data from UniProt and EcoCyc were considered, and the highest Mr value for a complex was chosen when variability in subunit composition was reported in the literature. Furthermore, hierarchical clustering (Euclidian distance and Pearson correlation) was performed using MeV to cluster ‘significant proteins’ based on similarity of their abundance patterns across all samples (n = 11) for each comparison. Clusters enriched in proteins that are part of multi-subunit complexes or larger subcellular assemblies were examined in depth.
Results and Discussion
Expanding a commonly used 2D-LC-MS/MS shotgun proteomics strategy for the analysis of bacterial lysates, we added a less frequently used soluble protein fractionation step, SEC (Figure 1). To establish a framework for comparisons of protein abundances, the computationally adjusted spectral counting method APEX that leverages quantitative comparisons of different proteins in a given dataset, in addition to relative comparisons of the same protein present in different fractions, was employed. Quantitative proteomic analysis was followed by data mining, including the correlation of enrichment of proteins in a distinct cell lysate fraction (soluble vs. insoluble) or a distinct SEC fraction (Mr ranges of >280 kDa vs. 280-80 kDa vs. 80-10 kDa) with its biochemical properties such as sequence-based and native Mr value, participation as a subunit in higher order protein assemblies, subcellular localization and attachment to or integration in phospholipid membranes. Since E. coli is one of the most extensively characterized prokaryotic organisms, plenty of information stored in various databases was available to assess the value of our approach and to build hypotheses. Based on the observed fractionation patterns of characterized E. coli proteins, we were able to infer similar properties for many uncharacterized E. coli O157:H7 proteins.
Abbreviations: EHEC, enterohemorrhagic E. coli; SEC, size exclusion chromatography; FASP, filter-aided sample preparation for digestion of complex protein mixtures; SCX, strong cation exchange chromatography; BR, biological replicate; T1R and T4R, technical replicate stages. The total number of SCX fractions subjected to LC-MS/MS sequentially was 52 in a given experiment.
A schematic of the experimental strategy, which we term SEC-2D-LC-MS/MS here, is shown in Figure 1. E. coli O157:H7 strain 86-24 cells were lysed using a lysozyme/EDTA spheroplasting method in the presence of 0.05% Triton X-100, followed by subsequent protein extraction of the pellet with a high salt solution (2.5 M NaBr). The combined solubilized fraction was separated by SEC yielding fractions in the Mr ranges >280 kDa, 280-80 kDa and 80-10 kDa, with estimated Mrs based on retention times for five globular protein standards. As depicted in Figure 2, the separation of proteins by SEC was also evident after denaturation and analysis in SDS-PAGE gels. These fractions, termed F1-sSEC, F2-sSEC and F3-sSEC (highest to lowest Mr), and the insoluble lysate pellet (F4-p) were subjected to tryptic digestion. Four fractionations of tryptic peptide mixtures via SCX resulted in 13 samples each, which were sequentially analyzed by nano-flow C18-LC-MS/MS. This workflow was applied to 19 experiments yielding on average 270,000 mass spectra, with 35.2% of the mass spectra assigned to peptides predicted from the E. coli O157:H7 strain EDL933 protein sequence database. Using the Mascot algorithm, the average peptide FDR was 1.3%. For APEX-based protein quantification, Mascot-analyzed datasets or subsets of data derived from a distinct lysate fraction were rescored with Peptide/Protein Prophet, and quantities of proteins computed with a normalization factor based on ∼2.5×106 proteins per E. coli cell .
The fractions (F1-s, F2-s and F3-s) were eluted from the G3000-SWXL column in PBS and 0.01% Triton X-100 and pooled as shown in the chromatogram. The A280 profiles are shown for an EHEC lysate sample and a mixture of five Mr protein standards, the latter of which pertains to the line with a low void volume peak at 667 kDa. Protein separations occurred in the 500-10 kDa range. On the right, protein bands visualized by Coomassie Blue staining in a 4–12%T SDS-PAGE gel are shown in the same order as the SEC fractions (with F3-s before pooling in two consecutive SEC fractions).
Protein quantification reproducibility
Eleven of the 19 SEC-2D-LC-MS/MS datasets were derived from extracts of bacterial cells grown in LB media to an OD600 of ∼2.0 and used for quantitative reproducibility assessments. Three stages of replication were considered, including the cell lysis stage (biological replicates, BR), the SEC separation stage (technical replicates, T1R) and the LC-MS/MS stage (technical replicates, T4R), as depicted in Figure 1. Applying the Pearson correlation metric to APEXi protein quantities of two datasets, the average R2 values were 0.71 (STDEV 0.05), 0.73 (STDEV 0.06) and 0.78 (STDEV 0.05) for the BR, T1R and T4R stages, respectively (n≥5 in each case). That the R2 values increased in this order was not surprising, since biological variation (at the BR stage) and more sample handling (at the T1R vs. the T4R stage) are expected to negatively impact reproducibility. The total number of proteins identified ranged from 1115 to 1861, averaging 1634 (STDEV 0.13). Limiting the analysis to proteins assigned to a specific subcellular or membrane localization using the PSORTb tool, the average R2 values were 0.79 (β-barrel outer membrane group, 47 proteins; STDEV 0.05), 0.79 (lipid-anchored group, 72 proteins; STDEV 0.02), 0.82 (soluble periplasmic group, 106 proteins; STDEV 0.05), 0.57 (group with two or more cytoplasmic TMDs, 236 proteins; STDEV 0.07) and 0.75 (soluble cytoplasmic group, 1308 proteins; STDEV 0.04). These calculations pertain to the T1R stage of replication (n = 5). Quantitative reproducibility levels for β-barrel OM and lipid-anchored proteins were relatively high, suggesting that these proteins were as effectively solubilized and digested as soluble periplasmic and cytoplasmic proteins. Proteins with two or more TMDs were less reproducibly quantified, clearly due to lower sequence coverage that results from lower abundance, lower solubility and less effective digestion in highly hydrophobic regions. The digestion method (FASP) was previously described as introducing little quantitative bias against hydrophobic membrane proteins  and was largely confirmed by our data. Non-ionic detergents (Triton X-100) and reagents with chaotropic properties (EDTA, 2.5 M NaBr) enhanced membrane protein solubilization , , . Details of the R2 analyses are presented in Table S1. R2 values for our entire dataset were lower compared to 2D-LC-MS/MS without the upfront SEC separation step , where the corresponding digestion, SCX and C18LC steps in addition to similar LTQ analysis parameters and sampling depth were applied. Those 2D-LC-MS/MS data resulted in R2 values of 0.84–0.88 at the T4R stage. Inserting more fractionation steps apparently decreased the repeatability of protein quantification. Repeatability and reproducibility of label-free shotgun proteomics are recognized as important criteria to assess data quality, particularly in biomarker discovery studies when quantitative accuracy is critical . Further method improvements of the SEC-2D-LC-MS/MS workflow are desirable to reduce variability in protein quantification.
Comparing in silico predicted and experimental E. coli O157:H7 protein profiles
Nine distinct pathovars have been recognized in E. coli, EHEC and EPEC (enteropathogenic E. coli) being the main pathovars that contain LEE in the genome and cause attachment and effacement (A/E) lesions in the colonic epithelium . O157:H7 is the most prominent serotype associated with outbreaks of bloody diarrhea and severe complications including HUS, two of which were linked to strains 86-24 and EDL933, respectively , . The strain 86-24, a producer of Stx2, was shown to be more virulent than EDL933, a producer of Stx1 and Stx2, in the gnotobiotic piglet model where oral infection was followed by severe neurologic injuries . We used the strain 86-24 whose genome has not been sequenced for proteomic analyses of in vitro and piglet intestine-derived bacterial samples. The obvious choice of a strain whose genome has been sequenced for MS database searches was EDL933. Its genome is well characterized , the 86-24 and EDL933 strain differences in virulence in the piglet model seemed to be primarily linked to the Shiga toxin expression profile , both strains caused significant disease outbreaks in humans and are associated with the same MLST type st23, assessing the conservation of 15 distinct housekeeping gene sequences . Finally, gene expression profiles of the two strains were recently compared, and substantial sequence similarities in several variable prophage regions can be inferred from successful DNA microarray analyses .
The subunit Stx2A of Shiga toxin 2 was identified in only one SEC-2D-LC-MS/MS experiment. Temporally repressed toxin expression, as reported for STEC-associated edema disease , and rapid secretion of the toxin, which limits intracellular Stx2 concentrations, are reasons that could account for low Stx2 abundance observed in the 86-24 proteome. The pathogenicity locus LEE encodes virulence factors contributing to A/E lesions upon contact with intestinal epithelial cells . Twenty-one LEE-encoded proteins were identified in this survey, some with relatively high abundances including the outer membrane adhesion protein intimin (Eae), T3SS subunits (Esc proteins), translocator pore-forming Esp proteins, effector proteins (Tir and EspF), as well as proteins with chaperone (CesT) and regulatory functions (Ler, SepL). EHEC strains have an interesting repertoire of putative T3SS effectors encoded by genes outside of LEE . Using the Sakai strain in the latter study, 39 proteins were experimentally confirmed to be secreted effectors, and the largest group was termed the NleG family. Fifteen of these effectors, half of them encoded by prophage regions in the EDL933 genome, were identified at low abundance levels in strain 86-24 lysates. The most reproducibly profiled of these effectors were NleG5-1 and NleG2-3. Interestingly, these and two other NleG family members were recently characterized as U-box E3 ubiquitin ligases with strong auto-ubiquitination activity in vitro . This finding suggests that, upon injection into the epithelial host cell cytoplasm, these enzymes target the ubiquitin proteasome system, supporting the notion of molecular mimicry in order to perturb host cell functions. Generally low T3SS effector abundances surveyed here were not unexpected since secretion rapidly follows effector gene expression in the ribosome, and secreted fractions were not examined here. In a comparative analysis of wild-type strains of EHEC and EPEC, secretome profiles yielded higher abundance levels of several Esp effectors, whereas the respective Δler mutants, a key transcription factor inducing T3SS expression , did not. Indeed, two effectors that remain associated with the cell surface (Tir and Eae) were detected with higher abundance levels in the E. coli strain 86-24 proteome.
FliC, the antigenic determinant of the H antigen (H7 for strain 86-24) was reproducibly identified. Half of the 31 identified proteins encoded by genes of the virulence-associated plasmid pO157 were profiled with moderate to high abundances, including catalase/peroxidase (KatP), a putative α/β fold hydrolase (Ytp2) whose biological role is unknown and StcE. StcE is a metalloprotease that cleaves human C1 esterase inhibitor, mucin 7 and glycoprotein 340. The enzyme compromises human T-cell function and contributes to intimate adherence to the gut epithelium , . It is a substrate of the pO157-encoded type 2 secretion system of which several subunits (EtpO, EtpG, EtpM and EtpE) were also identified in this survey. There are 16 and 19 prophage regions in the EDL933 and Sakai E. coli O157 genomes, respectively . Searching the EDL933 database, we identified proteins in all but one of the 15 prophage regions and the infective phage BP-933W segment. However, only 16% of the phage-encoded proteins were identified from strain 86-24 lysates, and most of these low in abundance compared to 46.7% coverage for the entire EDL933 proteome. Low abundance and detection frequency are not necessarily signs of mutational decay, as a report on inducibility and release of E. coli O157 prophage DNA evidenced , . In the LEE and CP-933T prophage segments, the proportion of identified proteins was relatively high (37% and 24%, respectively). It was lowest in the CP-933N prophage region (6.6%). Shotgun proteomic analysis of the E. coli K12 revealed that, while horizontally transferred genes mapping to the (phage) K-loops were as effectively transcribed as ‘core’ genes, proteins were expressed at a low percentage of ∼10% , which corresponds well with our data.
With a coverage of 46.7% (2521 of the 5397 predicted E. coli EDL933 proteins identified at a 1% FDR), this is the most comprehensive E. coli O157:H7 proteome analysis to date. Despite impressive technology advances, reports in which bacterial proteomes are profiled at a percentage above 50% of the genome-derived in silico proteome are rare. Among the reasons are: (1) silent genes, annotated as ORFs but not translated into proteins; (2) non-constitutively expressed genes repressed under most physiological conditions; (3) low abundance of proteins missed due to the stochastic peptide sampling mode in shotgun proteomics; (4) short polypeptides that are expressed but yield few or no proteotypic peptides. Table S2 provides the entire list of proteins identified here with functional, structural, localization, physicochemical and APEXi quantification data. A 2nd worksheet (Table S2) pertains to use of a combination of two algorithms, InsPecT and MS-GF, which resulted in 1966 protein identifications at a 0.2% peptide FDR. Table S3 contains the peptide dataset, also published in its entirety in the data repository PRIDE (http://www.ebi.ac.uk/pride/, Acc. No. 17108-17114). We are not reporting on differentially abundant proteins comparing the two physiological environments from which the STEC cells were isolated. These differences were extensive, influencing the activity of numerous biochemical pathways. Focusing here on the overall proteome and methodological improvements described in the following sections, we will cover proteomic adaptations of STEC to the pig intestinal environment in a separate publication.
Fate of E. coli O157:H7 membrane proteins during cell lysate fractionation
Proteins localized in the inner membrane (IM) and OM were profiled comprehensively, with 458 of 1163 predicted integral IM proteins (39.4%) and 54 of 128 predicted OM proteins (42.2%). In silico predictions were based on subcellular assignments by the PSORTb tool. Lysis and solubilization reagents used here (Triton X-100, EDTA, lysozyme, 2.5 M NaBr) and tryptic digestions under denaturing conditions were sufficient to generate soluble peptides from many hydrophobic membrane proteins. Previously, expression of only 680 bitopic and polytopic IM proteins has been experimentally confirmed in 14 E. coli proteomic studies . Low abundance membrane proteins were quantified less reproducibly which is a general concern of shotgun proteomics . Membrane proteins were markedly enriched in fraction F1-sSEC compared to the other three fractions, likely caused by high Mr mixed micelle formation with the detergent Triton X-100 and elution near the SEC void volume (∼500 kDa). Nearly 65% and 90% of all integral IM and OM proteins, respectively, were identified in fraction F1-sSEC. Given this observation, our data support the localization of orphan proteins such as YhcB, YjdB, YijP, YagU and YdgA, previously assigned to cell envelope protein complexes , in the IM. Examining APEXi quantities of membrane proteins thought to be present in distinct stoichiometric ratios, we note that such data seem to be accurate in some, but less accurate in other cases. Succinate dehydrogenase (SdhABCD) is a tetrameric complex with a 1∶1∶1∶1 stoichiometry. SdhA and SdhB are peripherally bound to the membrane, while SdhC and SdhD have three TMDs in the IM. APEXi quantities were 10156, 9057, 421 and 9943, respectively. SdhC was quantitatively underrepresented. The integral IM subunits of the cytochrome D terminal oxidase, CydA and CydB, were profiled in the expected 1∶1 ratio (APEXi values of 3680 and 3220, respectively). The tetrameric ATP-binding cassette transporter Opp has two peripheral (OppD and OppF) and two integral (OppB and OppC) IM protein subunits, which are expected to have a 1∶1∶1∶1 stoichiometric ratio. The APEXi quantities were 484, 343, 102 and 114, respectively, with both integral membrane subunits 3- to 4-fold less abundant. Interestingly, the soluble subunit OppA had a much higher APEXi score (7116). This protein docks onto permease subunits (OppB/C) at the periplasmic surface to deliver oligopeptides for cytoplasmic import. While the observed quantitative ratios for subunits of membrane protein complexes need to be interpreted cautiously, the approach offers an opportunity to gain preliminary insights into protein complex composition. Data in Tables S2 and S4 can be parsed to assess stoichiometric ratios of known protein complexes and, sorting by gene loci, potentially novel complexes formed by protein neighbors.
Table S4 lists APEXi abundance values for proteins separately in the four surveyed fractions. Abundant β-barrel OM proteins such as OmpA, OmpC, OmpX, YciD and OmpT were retained to a higher extent in fraction F4-p than integral IM and lipoproteins, which is in agreement with studies on E. coli and Campylobacter membrane extracts where treatment of cells with Triton X-100 and EDTA resulted only in partial OM protein solubilization , . Less hydrophobic β-barrel OM proteins appear to form mixed micelles with Triton X-100 less effectively. Peripheral membrane proteins were often profiled across the spectrum of SEC fractions, indicative of solubilization in the absence of Triton X-100: for example, subunits SdhA and SdhB, also detected as an E. coli succinate dehydrogenase sub-complex by blue native PAGE , subunits of ATP synthase (AtpH, AtpA and AtpD) and subunits of the main protein secretion system (SecA and SecB) were quite abundant in fractions F2-sSEC and F3-sSEC, as opposed to the corresponding integral IM subunits of the complexes.
More than 50 uncharacterized low Mr proteins devoid of any predicted TMD or β-barrel motifs were enriched in fraction F1-sSEC. Due to the membrane protein enrichment in this fraction, we searched for the presence of N-terminal lipid anchor motifs associated with conserved signal peptidase II (Sp II) sequences. This analysis gave supporting evidence for lipid attachment of 38 proteins with the N-terminal Sp II motifs and 17 proteins with either less conserved motifs or unusual distances of the motif form the N-terminus not identified as lipoproteins by the algorithm LipoP. The proteins with an APEXi score greater than 50 are listed in Table 1. In some cases, we suspect incorrect prediction of the translational start site in the protein annotation of strain EDL933 as the reason for the failure to recognize the motif. Examples of proteins with atypical lipid anchor motifs are Z1474 (motif ITGC17A), YijI (motif CASC49S), Z0955 (motif FTTC18S), Z3360 (motif MKAC44S) and Z0984 (motif LCGC27G). Further support of membrane attachment was obtained for twelve of the 55 proteins via LC-MS/MS analysis of an iodixanol density gradient fraction highly enriched in bacterial membranes (see Table S5). Data for all the 82 identified and assigned lipoproteins, including those previously characterized, are provided in Table S6. Our approach did not allow specifying either IM or OM attachment of the proteins. Relying on transcript detection and lipoprotein motif searches, a study on E. coli strain MG1655 reported a roughly equal number of lipoproteins with unknown functions . To our knowledge, lipoproteins of pathogenic E. coli strains have not been globally surveyed to date. Some of the lipoproteins, such as those encoded by prophage regions and absent in E. coli MG1655, may have unique functional roles in E. coli O157:H7. Unlike these lipid-anchored periplasmic proteins, soluble periplasmic proteins were almost invariably enriched in fraction F3-sSEC. The exceptions were proteins reported to interact with the OM: TolB and YbgF, which interact with the cell envelope complex Tol-Pal , and AsmA, which has been associated with the OM protein assembly complex .
Fractionating soluble proteins associated with higher order complexes
Abundance comparisons of proteins predicted to be soluble, either cytoplasmic or periplasmic, across the SEC fractions F1-sSEC, F2-sSEC and F3-sSEC were used to assess their enrichment in Mr ranges discordant with that of monomeric forms. Incorporating information on homooligomeric and heterooligomeric assemblies for many E. coli proteins, we were able to compare the experimentally determined Mr intervals for protein elution via SEC with protein sequenced-derived monomeric and native Mr values described for protein complexes in E. coli databases. EcoCyc and the E. coli data subset in Uniprot were queried to extract annotated native Mr values for more than 500 proteins part of complexes. As described in detail in Table S4, the data revealed excellent agreements, but also differences between our experimental Mr estimates and those reported in the databases: a rank matrix for APEXi quantities of each protein observed in a given fraction was established to facilitate comparing the SEC-derived ‘protein peaks’ with reported oligomeric Mrs. It is noteworthy mentioning a few confounding factors: (1) protein complexes fall apart upon cell lysis and solubilization in a largely unpredictable manner; (2) SEC permits only estimates of true Mr values of protein complexes and tends to be less accurate if non-globular shaped proteins are involved ; (3) our methodology was unable to reveal specific protein interaction partners when a Mr value for a protein suggested a complex.
The comparisons revealed stronger experimental evidence for intact protein homooligomers than for large heterooligomeric protein complexes. For example, the pyruvate dehydrogenase multienzyme complex consists of 60 subunits: AceE24AceF24Lpd12 (the subscript number indicates the number of subunits in a given complex). All three subunits were profiled over the entire Mr range (SEC void volume to 10 kDa). Proteins enriched in a SEC fraction concordant with the Mr estimate of known complexes were: Ftn24, Kbl12, KatG4, GlpK4, GltA4, FbaB10, MaeB6, RpoABCDE and GlnA12 (fraction F1-sSEC); Udp6, IcdA2, TnaA4, GapA4, GlpQ2, Fba2, AcnB2, TktA2, AspC2, Gnd2 and Ppa6 (fraction F2-sSEC). Complexes that were largely disintegrated were RibH60, ClpB4, ClpA12ClpP14, Lpd2GcvP2GcvH1GcvT1, HtrA6 and Dps12. Having verified intact complexes, all profiled proteins not characterized as subunits of higher order assemblies to date were examined, resulting in ∼190 proteins for which participation in a complex is now predicted. We required the monomeric Mr value to be least 25% below that of the lower assigned Mr limit in SEC experiments (80 and 280 kDa for F1-sSEC and F2-sSEC, respectively) to assign a new complex-associated protein. The two categories of proteins, those reported vs. not reported to participate in protein complexes, are distinguished by their color codes in Table S4.
We took this analysis a step further, classifying soluble proteins that were enriched with statistical significance in a distinct SEC fraction. Proteins with predicted/experimental evidence for membrane integration or attachment were excluded from the analysis. A Kruskal-Wallis test returned 760 proteins as significant with a p-value set at <0.02 in a three-group analysis (F1-sSEC vs. F2-sSEC vs. F3-sSEC). To visualize how the proteins were quantitatively distributed among the fractions (with analytical replicates) the data were displayed in a heat map, ordered either based on monomeric or native (protein complex) Mr values (Figure 3). While the patterns were noisy, in part due to numerous proteins for which native Mr data are unavailable, the expected Mr distribution was more evident in the heat map ordered based on native protein Mr values. The exception was proteins at the very bottom of the heat maps, many of which were assigned a maximum Mr of 999 kDa in the pattern analysis. A lower level of correlation with such high native Mr values was likely caused by disintegration of large complexes, e.g. the ribosome whose subunits were the majority of proteins in this part of the graphic. Ribosomal proteins were most abundant in fraction F3-sSEC. The reagent EDTA, added to the lysis buffer, lowers free Mg2+ which probably destabilized the ribosome and led to the release of disassembled low Mr subunits . The dataset pertaining to these heat maps is provided in Table S7. Then we performed a Pearson correlation-based hierarchical clustering analysis to assess whether many of the 190 predicted, complex-associated proteins clustered with proteins part of established complexes in E. coli. This was indeed observed, as shown for a selected list of proteins in Table 2 and in a graphic revealing protein clusters in Figure S1. The clusters C1 and C2 featured protein enrichment in fraction F1-sSEC (like Ftn24); the cluster C3 represented proteins most abundant in fraction F2-sSEC (like Udp6). For several proteins associated here with higher order assemblies (including some in Table 2), protein-protein interactions have been suggested. The E. coli interactome published by Hu et al.  reported interactions of the proteins PrsA, YbeZ, YeeX, YggE and YfbU with ribosome subunits, of YbeZ with RraA and of YfbU with YeeX. PrsA localized in a small cluster with a majority of proteins being ribosomal. YbiB and YbiC, each a protein of unknown function, and YlbA and YlbC, each believed to contribute to allantoin assimilation, are encoded by adjacent genes whose products may be subunits of a complex. Only a few proteins encoded by prophage loci or plasmid pO157 were identified as protein candidates for oligomeric assemblies: catalase KatP and the uncharacterized proteins L7060 (plasmid pO157) and Z2386 (CP933-R region). In summary, this data lends further support to the notion that our methodology to associate uncharacterized proteins with complex formation is biologically meaningful.
Each heat map column under the broad blue arrow indicating a Mr range represents 11 individual, replicate APEXi datasets. Each row corresponds to one of 760 EHEC proteins, quantified with the APEXi tool, that were altered with statistical significance using the Kruskal-Wallis test (p-value<0.02) comparing protein abundances in the three SEC fractions (F1-sSEC, F2-sSEC and F3-sSEC). The range of colors in the heat map corresponds to APEXi values from 0 to 5000. Left heap map: proteins are ordered based on calculated, amino acid sequence-based Mr values; right heat map: proteins are ordered based on native Mr values including homooligomers and multi-protein complexes (for calculations, see Table S7). The maximum Mr value was 999 kDa, including larger complexes such as the ribosome. Protein positions are indicated for the monomer/complex as follows: Ftn/Ftn24, Udp/Udp6 and Hfq/Hfq6. The correlation of protein elution patterns by SEC with calculated native Mrs is improved.
Proteins enriched in insoluble lysate fractions and prone to aggregation
More than 1,000 proteins were routinely identified in LC-MS/MS analyses of the insoluble fraction (F4-p). We have already referred to the retention of membrane proteins - particularly those of the OM - in the insoluble E. coli O157:H7 pellet, likely caused by incomplete protein solubilization. F4-p protein profiles also revealed high prevalence of numerous proteins not associated with membranes, which was supported by statistical evidence using a Wilcoxon rank sum test that compared protein abundances in the three SEC fractions vs. the F4-p fraction (Table S7). At a p-value of <0.02, 800 proteins displayed a statistically significant pattern of enrichment in either F4-p or the combined SEC fractions. Given the considerable knowledge on insoluble protein aggregates formed in E. coli cells, our proteomic and literature data were mined to determine whether a pattern of protein characteristics evolved that linked specific types of proteins with aggregation/insolubility behavior. Data related to protein propensities to aggregate are included in Table S4 (worksheet F4-p).
To determine that membrane proteins prevalent in fraction F4-p were indeed a fraction discernible from aggregates of intracellular proteins, the bacterial cell lysate was subjected to iodixanol density gradient centrifugation. Two opaque layers were isolated at a density of ρ = 1.21 g/cm2 (HD) and at a lower density of ρ = 1.13 g/cm2 (LD). LC-MS/MS analysis (Table S5) revealed that the LD layer was enriched in membrane proteins including highly prevalent OM proteins, whereas the HD layer was enriched in cytoplasmic proteins which apparently aggregated and were rendered insoluble after cell lysis. The overlap of proteins identified in the HD gradient fraction with those of high abundance in F4-p fractions from the main analysis was extensive, further evidence in support of the assumption that F4-p was enriched in proteins forming insoluble aggregates. Soluble periplasmic proteins were underrepresented in F4-p and absent in the HD fraction, supporting the notion that such proteins, which are subject to folding pathways in more oxidative environments , have a lower propensity to aggregate. Similar conclusions were drawn in a study on protein aggregation using healthy E. coli cells . This is not to say that protein aggregation does not occur in the periplasm of E. coli and other enteric pathogens. Under acid stress conditions, two small acid stress response chaperones (HdeA and HdeB) act as a pH-optimized periplasmic protein disaggregation system  that is induced in Shigella dysenteriae in the mammalian gut environment . Ninety% of the ∼120 E. coli proteins described as aggregate-forming thermolabile substrates of the chaperones DnaK, GroEL, ClpB and ClpX/P ,  were also detected, and typically enriched, in the F4-p fraction in our survey.
Ribosomal protein subunits were markedly enriched in F4-p, tempting us to explore whether proteins participating in large organelle-like structures are generally prone to form aggregates, as observed for non-pathogenic E. coli aggregates . A third of the 100 proteins of highest abundant in fraction F4-p were ribosomal or ribosome-associated elongation factors (Table S4). GlpD and GlpA/GlpB, two glycerol-3-phosphate dehydrogenases , , and glycerol kinase (GlpK) were very abundant in fraction F4-p. Interaction of GlpK, but not of the dehydrogenases, with the inner membrane glycerol facilitator protein was reported . We speculate that a large transport-and-degradation complex mediating glycerol import and oxidation, involving these proteins, exists at the IM cytoplasmic surface.
We also noticed significant enrichment in F4-p and the HD fractions of several enzymes contributing to O-antigen building block biosynthesis and assembly. Most of the enzymes are part of a large gene cluster encompassing the gene loci Z3189-Z3206. Table 3 lists the 16 O-antigen biosynthesis proteins detected here and three additional enzymes associated with mobilization of galactose, an O-antigen sugar precursor. We hypothesize that the proteins co-localize at the cytoplasmic surface of the IM since some components required for biosynthesis of LPS, the outermost OM layer, are integrated in the IM, thus forming an ‘O-antigen biosynthesome’. Enzymes responsible for lipid A-core biosynthesis and encoded by the waa and lpx genes displayed less consistent enrichment in fraction F4-p. These enzymes represent the biosynthetic branch for synthesis of the LPS core molecule of E. coli O157 and have been linked to an integrated structure in the IM periphery . To our knowledge, there is no other evidence for an integrated O-antigen biosynthesis apparatus. The precise pathway of the serotype O157 antigen biosynthesis remains to be elucidated, since several of the glycosyltransferase (WbdR, WbdQ, WbdP, WbdO and GalF) and the epimerase (Gne) activities have not been verified . Purifying the predicted ‘O-antigen biosynthesome’ and protein-protein interaction screens will be useful to reveal structure-function relationships, with the potential of small molecule inhibitor design targeting one or several of the biosynthetic steps. Abrogation of O-antigen biosynthesis is an attractive pathway of interference with the pathogen's recognition by or evasion from the human immune system.
We also observed enrichment in F4-p of proteins with major roles in the biogenesis of the cytoskeleton of rod-shaped bacteria, including the bacterial actins MreB and FtsA , , the MreB-interacting elongation factor Tu , the bacterial tubulin FtsZ , , the cell division-regulating system MinCDE  and the polymer-forming SopA/SopB system that controls plasmid segregation during cell division . With the exception of SopA/SopB, these structures have also been associated with the IM periphery. In addition to FtsA/FtsZ, the core structure of cell division septa, two other components of the cell division-initiating Z-ring (FtsE and YhdE) were enriched in F4-p. Most proteins associated with O-antigen biosynthesis or cytoskeletal elements were enriched in fraction F4-p with statistical significance (Table 3). In further support of the aggregation behavior of such large subcellular protein assemblies, many ribosomal, O-antigen biosynthesis and cytoskeletal proteins were identified in the density gradient fraction HD (see Notes, Table S5). Finally, most subunits part of these higher order protein assemblies also clustered in a Euclidian distance clustering analysis, which is depicted in Figure S2.
Many DNA-binding subunits of two-component transcriptional regulators were also enriched in fraction F4-p (e.g. ArcA, PhoP, OmpR and RstA), as were enzyme complexes harboring iron-sulfur cofactors (e.g. AcnA, AcnB, FumA, SdaA and PflA/PflB). We examined whether other organelle-like assemblies in bacteria had enrichment patterns similar to those mentioned previously. Although individual proteins were enriched in fraction F4-p, this was not generally observed for the tRNA synthetase complex , the mRNA degadosome  and the purinosome. Purinosome (purine biosynthesis) subunits enriched in F4-p were Add, PurC, PurM and PurT. Unlike O-antigen biosynthesis and cell division, these structures are not known to be IM-associated. Among the uncharacterized proteins whose genes clustered at a specific locus, we noticed that YeaD, YeaG and YeaK (loci Z2820-27) were enriched in fraction F4-p, and speculate that these proteins form a functional unit in the E. coli O157:H7 cell.
We developed and assessed the reproducibility of a shotgun proteomics approach incorporating a SEC step to separate proteins according to their Mr values prior to LC-MS/MS. The strategy was applied to a virulent E. coli O157:H7 strain, isolated from the large bowel of infected piglets and from cell cultures, and included quantification of more than 2500 proteins via computationally modified spectral counting. Only 13% of the ORFs encoded by phage/prophage regions were expressed at the protein level, mostly at low abundance. Defective prophages are not simply genetic remnants in the course of EHEC evolution but are inducible and released from cells as particulate DNA, disseminating virulence-related genes . Coverage of proteins encoded by plasmid pO157 and the LEE pathogenicity island, each producing functional virulence factors, was markedly higher (31 and 36%, respectively). To examine whether the fractionation strategy allowed us to profile intact or partially intact protein complexes, we correlated protein abundances with reported E. coli monomeric and native (protein complex) Mr values. Supported by clustering data, abundance patterns for more than 120 uncharacterized proteins suggested participation in complexes not characterized to date. In insoluble E. coli O157:H7 lysate fractions, high abundances of subunits of large organelle-like assemblies and oligomer-forming proteins previously shown to be prone to aggregation were detected.
Hierarchical clustering with the Pearson correlation metric of STEC proteins with different abundances comparing three size exclusion chromatography fractions - F1-sSEC (+280), F2-sSEC (280-80) and F3-sSEC (80-10 kDa) denoting novel proteins predicted to be part of oligomeric protein complexes. The software tool MeV was used for the analysis using the Hierarchical Pearson Correlation Metric (average linkage clustering) visualized at the intensity scale of protein abundance APEXi scores from 1.0 to 5000). The soluble cell lysate fractions were F1-sSEC (+280 kDa), F2-sSEC (280-80 kDa) and F3-sSEC (80-10 kDa). Prior to the HCL analysis, 760 proteins of the total EHEC dataset (2521 proteins) observed to be differentially abundant with statistical significance (F1-sSEC vs. F2-sSEC vs. F3-sSEC) at a p-value of <0.02 using the Kruskal Wallis test were selected. In the three clusters, more than 70% of the proteins were part of characterized protein complexes in the approximated Mr range.
Euclidian distance clustering of EHEC proteins with different abundances in size exclusion chromatography fractions (F1-sSEC to F3-sSEC) compared to the insoluble lysate fraction (F4-p) denoting proteins part of the ribosome, the O-antigen biosynthesis apparatus and cell division/cytoskeletal elements. The software tool MeV was used for the analysis using the Hierarchical Euclidian Distance Clustering metric (average linkage clustering) visualized at the intensity scale of protein abundance APEXi scores from 1.0 to 5000. The fractions were F1-sSEC (+280 kDa), F2-sSEC (280-80 kDa), F3-sSEC (80-10 kDa) - all from the soluble fraction - and F4-p, the insoluble cell lysate fraction). Prior to the HCL analysis, 802 proteins of the total EHEC dataset (2521 proteins) observed to be differentially abundant with statistical significance comparing F1–F3 with F4-p using the Wilcoxon Rank Sum test (p-values<0.02) were selected. Proteins that are part of the three clusters are also listed in Table 3. The markers (stripes) to the right of the heat maps each denote the position of a protein in the cluster, in the same order as the listed protein names to the right: - blue, left column: ribosome; - brown, middle column: cell division system and cytoskeleton; - green, right column: O-antigen biosynthesis apparatus. The highlighted cluster at the bottom of the image on the right was enriched in particular in proteins part of the ribosome and O-antigen biosynthesis system, the proteins are listed in the order of their appearance in the HCL tree.
Quantitative reproducibility of shotgun proteomics analyses of STEC strain 86-24. BR: biological replicate; T1R: technical replicate at size exclusion chromatography stage; T4R: technical replicate at LC-MS/MS stage; STDEV: standard deviation; R2: Pearson correlation coefficient; FDR: protein false discovery rate determined with Mascot algorithm.
Shotgun proteomics analysis of STEC strain 86-24 cell lysates using the modified spectral counting method APEX. Worksheet A - 52 Fractions_Mascot (1% FDR). Theme colors used in worksheet: pO157 plasmid-encoded proteins (orange); proteins encoded by prophage genome regions (green and, if non-LEE effector proteins, turquoise); non-LEE effectors outside of prophage regions (blue). Column A: GenBank accession number in NCBI - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column B: Swiss-Prot accession number - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column C: Ref Seq. accession number - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column D: Locus tag - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column E: Gene name - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column F: Replicon - Escherichia coli O157:H7 EDL933 genome. Column G: Protein description derived from Escherichia coli O157:H7 EDL933 protein sequence database. Column H: APEXi protein quantities (mean averages) from 19 individual 2D-LC-MS/MS experiments using the APEX Quantitative Proteomics Tool (http://www.biomedcentral.com/1471-2105/9/529). The method calculates protein quantities computationally, essentially based on spectral counts, normalizes based on the total number of spectral counts in a given dataset followed by multiplication with an adjustment factor (2.5×106, for the estimated number of protein molecules per cell). Column I-AA: APEXi protein quantities from 19 individual 2D-LC-MS/MS experiments using the APEX method. Column AB: Subcellular localization of proteins derived from assignments with the algorithms in PSORTb (http://www.psort.org/psortb/). Column AC: Subcellular localization of proteins derived from assignments with the algorithms in Cell PLoc (Chou & Shen, Nature Protocols 3, p. 153–162; 2008. Column AD: protein sequence length (in amino acids). Column AE-AI: Subcellular localization of proteins derived from individual bioinformatics searches. Each search parameter is incorporated in the Cell-Ploc tool. Column AJ: Subcellular localization of proteins derived from assignments with the algorithms in Sosui from Gram-negative bacteria (http://bp.nuap.nagoya-u.ac.jp/sosui/). Column AK: Prediction of a type III secretion system component or effector using the Database for type III secretion systems (DTTSS, sdbi.sdut.edu.cn/ttss/). Column AL: Main functional role category based in data in the CMR database (JCVI); selection for E. coli EDL933: http://cmr.jcvi.org/tigr-scripts/CMR/GenomePage.cgi?org=ntec02. Column AM: Sub functional role category based in data in the CMR database (JCVI); selection for E. coli EDL933: http://cmr.jcvi.org/tigr-scripts/CMR/GenomePage.cgi?org=ntec02. Color codes: pO157 plasmid-encoded proteins (orange); proteins encoded by prophage genome regions (green); if these were predicted non-LEE effectors (turquoise); non-LEE effectors not in prophage regions (blue). Worksheet B - 52 Fractions_InsPecT (0.2% FDR). Column A: GenBank accession number in NCBI - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column B: Swiss-Prot accession number - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column C: Ref Seq. accession number - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column D: Locus tag - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column E: Gene name - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column F: Replicon - Escherichia coli O157:H7 EDL933 genome. Column G: Protein description derived from Escherichia coli O157:H7 EDL933 protein sequence database. Column H: Protein scoring was performed with the algorithms InsPecT and MS-GF (http://proteomics.ucsd.edu/Software/MSGeneratingFunction.html). APEXi protein quantities (mean averages) from 19 individual 2D-LC-MS/MS experiments were determined uploading the MS-GF rescored datasets (identified at the 0.2% FDR) into the APEX Quantitative Proteomics Tool (http://www.biomedcentral.com/1471-2105/9/529) and using the APEX method. The method calculates protein quantities computationally essentially based on spectral counts, normalizes based on the total number of spectral counts in a given experiment followed by multiplication with an adjustment factor (2.5×106, for the estimated number of protein molecules per cell). Column I-AA: APEXi protein quantities from 19 individual 2D-LC-MS/MS experiments using the APEX method. Column AB: Subcellular localization of proteins derived from assignments with the algorithms in PSORTb (http://www.psort.org/psortb/). Column AC: Subcellular localization of proteins derived from assignments with the algorithms in Cell PLoc (Chou & Shen, Nature Protocols 3, p. 153–162; 2008. Column AD: protein sequence length (in amino acids). Column AE-AI: Subcellular localization of proteins derived from individual bioinformatics searches. Each search parameter is incorporated in the Cell-Ploc tool. Column AJ: Subcellular localization of proteins derived from assignments with the algorithms in Sosui from Gram-negative bacteria (http://bp.nuap.nagoya-u.ac.jp/sosui/). Column AK: Prediction of a type III secretion system component or effector using the database for type III secretion systems (DTTSS, sdbi.sdut.edu.cn/ttss/). Column AL: Main functional role category based in data in the CMR database (JCVI); selection for E. coli EDL933: http://cmr.jcvi.org/tigr-scripts/CMR/GenomePage.cgi?org=ntec02. Column AM: Sub functional role category based in data in the CMR database (JCVI); selection for E. coli EDL933: http://cmr.jcvi.org/tigr-scripts/CMR/GenomePage.cgi?org=ntec02.
STEC O157:H7 proteins/peptides identified by LC-nESI-MS/MS on a Linear Ion Trap instrument at a 1% protein false discovery rate. A: gi| accession numbers are equivalent to those listed in Table S2, Supplemental Information (genome E. coli O157:H7, strain EDL933); B: highest Mascot score for a given protein in the dataset at a 1% FDR; C: .dat file from Mascot search result; D: amino acid sequence for peptide-spectral match; E: start and end amino acid positions within protein sequence as annotated for the respective ORF in the E. coli EDL933 protein sequence database; F: peptide ion score in tandem MS/MS experiment, peptide fragment ions with scores <20 are not listed; G: peptide mass value; H: peptide identification output file.
Shotgun proteomics analysis of STEC strain 86-24 using the modified spectral counting method APEX comparing protein profiles in four lysate fractions. Worksheet A: F1-sSEC samples, proteins identified in the 13 SCX fractions (F1–13) derived from the first size exclusion chromatography fraction of the soluble cell lysate. Worksheet B: F2-sSEC samples, proteins identified in the 13 SCX fractions (F14–26) derived from the second size exclusion chromatography fraction of the soluble cell lysate. Worksheet C: F3-sSEC samples, proteins identified in the 13 SCX fractions (F27–39) derived from the third size exclusion chromatography fraction of the soluble cell lysate. Worksheet D: F4-p samples, proteins identified in the 13 SCX fractions (F40–52) derived from the insoluble cell lysate fraction. General descriptions for all worksheets: Column A: GenBank accession number in NCBI - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column B: Swiss-Prot accession number - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column C: Ref Seq. accession number - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column D: Locus tag - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column E: Gene name - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column F: Subcellular localization of proteins derived from assignments of a combination of algorithms to predict subcellular localizations - PSORTb (http://www.psort.org/psortb/), Cell PLoc (Chou & Shen, Nature Protocols 3, p. 153–162; 2008), the Ecocyc database (http://ecocyc.org/) and SwissProt; Protein complex assignments in homooligomeric complexes (e.g. ‘Ho-2’) or heterooligomeric complexes (e.g. ‘Ht-7’, if subunit score unknown ‘Ht-Mu’) were made from data in Ecocyc and Swiss-Prot, if available. Abbr.: CY, cytoplasmic; IM, inner membrane; OM, outer membrane; L, lipoprotein; PP, periplasmic; Ex, extracellular; PP or CY/IM, IM/OM periphery; 50S and 30S Rib, ribosome or ribosome subunit; TMD: transmembrane domain; larger cellular assemblies (ribosome, degradosome, cell divisome) may be experimentally unproven. In worksheet D, the symbol for aggregate formation is AGG. Column G: protein description from EDL933 genome database or, if more information, from Ecocyc and Swiss-Prot databases. Column H: APEXi protein abundance ranking: this data is based on the mean averages of spectral counts derived from 12 individual 2D-LC-MS/MS experimental replicates using the APEX quantitative proteomics tool. The method calculates protein quantities via computationally adjusted spectral counts, normalizes based on the total number of spectral counts in a given experiment followed by multiplication with an adjustment factor (2.5×106, for the estimated number of protein molecules per cell); the protein abundances are ranked using the total number of proteins identified at a 5% protein FDR confidence level. Column I: molecular weight in kDa based on protein sequence annotated in the EDL933 genome. Column J-U: APEXi protein quantities from 12 individual 2D-LC-MS/MS experiments using the APEX method (11 experiments in the case of fraction F4-p). Column V: absolute APEXi protein quantities averaged from 12 replicate 2D-LC-MS/MS experiments using the APEX method (11 experiments in the case of fraction F4-p). Column W: Subcellular localization of proteins derived from assignments with the algorithms in PSORTb (http://www.psort.org/psortb/). Column X: Subcellular localization of proteins derived from assignments with the Cell PLoc tool (Chou & Shen, Nature Protocols 3, p. 153–162; 2008). Column Z-AE: Individual assignments using individual algorithms components of Cell PLoc. Column AF: Prediction of a type III secretion system component or effector using the Database for type III secretion systems (DTTSS, sdbi.sdut.edu.cn/ttss/). Column AG: Main functional role category based in data in the CMR database (JCVI); selection for E. coli EDL933: http://cmr.jcvi.org/tigr-scripts/CMR/GenomePage.cgi?org=ntec02. Column AH: Sub functional role category based in data in the CMR database (JCVI); selection for E. coli EDL933: http://cmr.jcvi.org/tigr-scripts/CMR/GenomePage.cgi?org=ntec02. Specific descriptions for worksheets: Worksheet A: color codes. Orange, proteins experimentally shown to be localized in E. coli protein complexes and with the appropriate Mr range of >280 kDa in fraction F1-sSEC; blue, peripheral membrane proteins experimentally shown to be localized in membrane-associated protein complexes (inner or outer membrane); yellow, proteins not experimentally shown to be localized in E. coli complexes, but with protein sequence-based Mr values well below the Mr range of 280 kDa assigned to proteins in fraction F1-sSEC. Worksheet B: color codes. Orange, proteins experimentally shown to be localized in E. coli protein complexes and with an appropriate Mr range of 280-80 kDa in fraction F2-sSEC; yellow, proteins not experimentally shown to be localized in E. coli complexes, but with protein sequence-based Mr values well below the Mr range from 280 to 80 kDa assigned to proteins in fraction F2-sSEC. Worksheet C: color codes. Yellow, periplasmic and extracellular proteins in the 80-10 kDa in fraction (F3-sSEC); many periplasmic proteins appear not to be part of oligomeric complexes according to this data; gray, proteins experimentally shown to be localized in E. coli complexes in Mr ranges above 80 kDa and relatively abundant in fraction F3-sSEC; this data indicates that the protein complexes these protein subunits belong to are not stable under the experimental lysis and fractionation conditions used here. Worksheet D: color codes. Color codes are used to assess the enrichment of certain proteins and protein families in the insoluble cell lysate fraction F4-p: these proteins are part of known or predicted large subcellular protein or protein-ribonucleic acid assemblies; orange, E. coli ribosome: ribosomal protein subunits and ribosome-associated proteins; green, other large subcellular assemblies not as well characterized as the ribosome, but with some evidence reported: mRNA degradosome, cell division apparatus, cytoskeletal elements, O-antigen biosynthesis system; glycerol utilization system; blue, subunits of membrane-associated protein complexes; red text in gene name column: cytoplasmic proteins reported to form aggregates or are substrates of chaperones DnaK and GroEL.
Shotgun proteomics analysis of STEC strain 86-24 comparing protein profiles in two insoluble, iodixanol density gradient fractions. Legends for worksheets A and B. Column A: MS sample ID - mass spectrometry database search file name. Column B: Accession No - gi| protein accession number in NCBI - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column C: Protein description - according to protein annotations in NCBI - Escherichia coli O157:H7 EDL933. Column D: Mascot protein score - with at least one peptide e-value<0.1. Column E: Total peptides Rank = 1−Rank 1 peptides according to Mascot search results (peptide e-values<0.1). Column F: Protein sequence coverage - according to the Mascot search results (peptide e-values<0.1). Column G: pI - Isoelectric point based on protein sequence annotated in the EDL933 database. Column H: M.W. - Molecular weight in kDa based on protein sequence annotated in the EDL933 database. Column I: TMHMM - Bioinformatic prediction of transmembrane domains with algorithm TMHMM (www.cbs.dtu.dk) to predict integral inner membrane proteins. Column J: LipoP - Bioinformatic prediction of lipoprotein anchor motifs and SpII signal sequences with algorithm LipoP (www.cbs.dtu.dk) to predict lipid-anchored proteins. Column K: Sp I - Bioinformatic prediction of signal cleavage site motifs with algorithm SignalP (www.cbs.dtu.dk) to predict exported (periplasmic/outer membrane) proteins. Column L: PSORTb - Bioinformatic prediction of subcellular localization with the set of algorithms in PSORTb (www.psort.org/psortb/). Column M: Notes about membrane association, aggregation, larger assemblies with protein subunits - to establish protein features causing enrichment in one of the two iodixanol density gradient fractions derived from insoluble EHEC cell lysate matter, we assessed protein biophysical and biochemical traits based on data in the APEX abundance analysis (Table S3) and descriptions for proteins in Ecocyc and SwissProt databases. Abbreviations: iodx_low density worksheet: CY, cytoplasm; IM, cytoplasmic membrane; CY/IM, peripheral membrane protein; Abbreviations: iodx_high density worksheet: categories - ‘aggregation-prone cytoplasmic protein’, ‘membrane protein complex’ (mostly peripheral or monotopic membrane proteins without transmembrane regions), ‘larger assembly’ (larger protein assembly e.g. ribosome, tRNA synthase complex, ‘oligomeric in cytoplasm’ (soluble protein complex in cytoplasm), ‘LP’ (lipoprotein), ‘TMD’ (transmembrane domain proteins), and ‘OM’ (β-barrel outer membrane protein).
Lipoprotein assignments of STEC 86-24 proteins based on enrichment in a size exclusion chromatography fraction of high Mr and presence of putative lipid anchor motifs. Columns A–E, G: accession numbers, locus tags and gene names derived from the annotated genome database of enterohemorrhagic E. coli strain EDL933 (NCBI); gene name in parentheses are from Swiss-Prot or EcoCyc annotations. Column F: data on lipoprotein participation in multi-protein complexes; L: lipoprotein; IM/OM, inner and outer membrane; L?: no predictive information on IM or OM attachment and unusual lipid anchor motifs; Ht-Mu: hetero-multimeric, ‘2’, ‘3’ etc. indicating a defined number of protein subunits in a membrane-localized complex. Columns I/J: APEXi abundance of protein; column I: the proteins identified in each of the three size exclusion chromatography fractions (F1-sSEC, F2-sSEC, F3-sSEC) and the insoluble pellet (Fp-4) were ranked based on their abundance rank within a given fraction; column J: absolute APEXi protein abundance normalized by 2.5×106 protein molecules per cell in fraction F1-sSEC. Column K: The PSORTb prediction for subcellular localization (http://www.psort.org/psortb/). Column L: The LipoP algorithm predicts lipoprotein anchor motifs based on a hydrophobic N-terminal sequence followed by a moderately conserved five-amino acid motif with a cysteine in core position 4. This cysteine is covalently coupled to a phospholipid in the mature lipoprotein (http://www.cbs.dtu.dk/services/LipoP/). Column M: Δseq, apparently incorrect prediction of translational start site in ORF, as annotated in EDL933 genome; the amino acid (aa) number indicated the predicted aa residue shift. Color codes: experimentally verified lipoproteins (yellow); strongly conserved Sp II motif with hydrophobic leader sequence in N-terminal Sp II signal sequence (no color); putative Sp II motif for a protein enriched in F1-sSEC fraction (gray). Color themes, worksheet A. Yellow, outer membrane proteins including outer membrane β-barrel and lipoproteins; blue, inner membrane proteins, including transmembrane domain and lipoproteins; gray, peripheral inner membrane proteins and non-membrane contaminant proteins. Color themes, worksheet B. Yellow, proteins part of membrane protein complexes but without transmembrane domains; orange, proteins part of well or partially characterized larger subcellular assemblies; gray, soluble cytoplasmic proteins denoted in the literature or shown in this study to be prone to aggregation or subunits of cytoplasmic protein complexes. The three categories may overlap to some extent.
Statistical evaluation of STEC strain 86-24 protein enrichment patterns in different size exclusion chromatography and cell lysate fractions. Legend for worksheet A (KW_F123_p02): Correlation of protein molecular weights and their relative abundances in size exclusion chromatography fractions F1-sSEC, F2-sSEC, F3-sSEC. Column A: protein ID in Multiple Experiment Viewer (MeV) experiment. Column B: descriptive protein input in Multiple Experiment Viewer (MeV) experiment. Column C: GenBank accession number in NCBI - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column D: Locus tag - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column E: Gene name - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column F: Subcellular localization of proteins derived from assignments via a combination of algorithms to predict subcellular localizations - PSORTb (http://www.psort.org/psortb/), Cell PLoc (Chou & Shen, Nature Protocols 3, p. 153–162; 2008), the Ecocyc database (http://ecocyc.org/) and Swiss-Prot; Protein complex assignments in homooligomeric complexes (e.g. ‘Ho-2’) or heterooligomeric complexes (e.g. ‘Ht-7’, if subunit score unknown ‘Ht-Mu’) were made from data in Ecocyc and Swiss-Prot, if available. Abbr.: CY, cytoplasmic; IM, inner membrane; OM, outer membrane; L, lipoprotein; PP, periplasmic; Ex, extracellular; PP or CY/IM, IM/OM periphery; 50S and 30S Rib, ribosome or ribosome subunit. Column G: molecular weight in Da calculated from protein sequence length (in amino acids). Column H: molecular weight in Da calculated from a native protein complex (if subunit composition known); if homooligomeric structure was not known (‘Ho-Mu’), we calculated M.W. for a hexamer; if heterooligomeric structure was not known (‘Ht-‘Mu’), we calculated M.W. from the sum the proteins/subunits; for large assemblies (cytoskeletal elements, cell division septa and ribosomes, we used a large M.W. 999,999 Da by default). Column I: Proteins were ordered based on their average abundance ranking according to APEXi values in each of the fractions F1-sSEC, F2-sSEC, and F3-sSEC. Column J: Proteins isoelectric points. Column K: Protein descriptions in the EDL933 database. Column L-O: Kruskal-Wallis statistical test (three groups, group 1: fractions F1-sSEC; group 2: F2-sSEC; group 3: F3-sSEC); proteins with p-values<0.02 were accepted as statistically significant. Column P-Z, AB-AL, AN-AX: APEXi protein quantities from 11 individual 2D-LC-MS/MS replicates using the APEX method, for each of the fractions F1-sSEC, F2-sSEC, and F3-sSEC. Column AA, AM, AY: Protein quantities (mean averages, fractions F1-sSEC, F2-sSEC and F3-sSEC) from 11 individual 2D-LC-MS/MS experiments using the APEX method. The method calculates protein quantities computationally essentially based on spectral counts, normalizes based on the total number of spectral counts in a given experiment followed by multiplication with an adjustment factor (2.5×106, for the estimated number of protein molecules per cell). Color code (yellow): proteins part of higher order protein complexes judged form the Mr of the analyzed SEC fraction and the monomeric Mr. Legend for worksheet B (KW_F123vsF4_p02): Correlation of protein properties and APEXi abundances in fractions F1-sSEC, F2-sSEC, and F3-sSEC versus insoluble lysate fraction F4-p Column A: protein ID in Multiple Experiment Viewer (MeV) experiment. Column B: descriptive protein input in Multiple Experiment Viewer (MeV) experiment. Column C: GenBank accession number in NCBI - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column D: Locus tag - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column E: Gene name - Escherichia coli O157:H7 EDL933 protein sequence database (chromosome and pO157 plasmid). Column F: Subcellular localization of proteins derived from assignments via a combination of algorithms to predict subcellular localizations - PSORTb (http://www.psort.org/psortb/), Cell PLoc (Chou & Shen, Nature Protocols 3, p. 153–162; 2008), the Ecocyc database (http://ecocyc.org/) and Swiss-Prot; Protein complex assignments in homooligomeric complexes (e.g. ‘Ho-2’) or heterooligomeric complexes (e.g. ‘Ht-7’, if subunit score unknown ‘Ht-Mu’) were made from data in Ecocyc and Swiss-Prot, if available. Abbr.: CY, cytoplasmic; IM, inner membrane; OM, outer membrane; L, lipoprotein; PP, periplasmic; Ex, extracellular; PP or CY/IM, IM/OM periphery; 50S and 30S Rib, ribosome or ribosome subunit. Column G: molecular weight in Da calculated from protein sequence length (in amino acids). Column H: molecular weight in Da calculated from a native protein complex (if subunit composition known); if homooligomeric structure was not known (‘Ho-Mu’), we calculated M.W. for a hexamer; if heterooligomeric structure was not known (‘Ht-‘Mu’), we calculated M.W. from the sum of the proteins/subunits; for large assemblies (cytoskeletal elements, cell division septa and ribosomes, we used a large M.W. 999,999 Da by default). Column I: Proteins were ordered based on their average abundance ranking according to APEXi values in the sum of the fractions F1-sSEC, F2-sSEC, and F3-sSEC versus F4-p. Column J: Proteins isoelectric points. Column K: Protein descriptions in the EDL933 database. Column L-P: Wilcoxon rank sum statistical test (two groups, group 1: fractions F1-sSEC, F2-sSEC, and F3-sSEC; group 2: F4-p); the groups essentially represent the soluble fraction and the insoluble fraction of EHEC cell lysates; proteins with p-values<0.02 were accepted as statistically significant. Column R-AX, AZ-BJ: APEXi protein quantities from 33 and 11 2D-LC-MS/MS experimental replicates for all combined fractions (F1-sSEC, F2-sSEC, F3-sSEC) and the F4-p fraction, respectively. Column AY and BK: Protein quantities (mean averages) from the fraction combinations described above, using the APEX method. The method calculates protein quantities computationally essentially based on spectral counts, normalizes based on the total number of spectral counts in a given experiment followed by multiplication with an adjustment factor (2.5×106, for the estimated number of protein molecules per cell).
At Tufts University, we thank D. Girouard for performing the animal C-sections. At the JCVI, we thank T. Dracheva for suggestions regarding statistical tools for proteomic analysis and L. Losada for assessments of taxonomic relationships of E. coli strains. The MS data have been published in NCBI Peptidome (Study PSE142) and PRIDE (http://www.ebi.ac.uk/pride/, Acc. No. 17108-17114).
Conceived and designed the experiments: RP QZ RDF SNP ST. Performed the experiments: RP QZ DJC MJS. Analyzed the data: RP STH JCB SHP. Contributed reagents/materials/analysis tools: JCB SHP. Wrote the paper: RP STH ST.
- 1. Paton JC, Paton AW (1998) Pathogenesis and diagnosis of Shiga toxin-producing Escherichia coli infections. Clin Microbiol Rev 11: 450–479.JC PatonAW Paton1998Pathogenesis and diagnosis of Shiga toxin-producing Escherichia coli infections.Clin Microbiol Rev11450479
- 2. Perna NT, Plunkett G 3rd, Burland V, Mau B, Glasner JD, et al. (2001) Genome sequence of enterohaemorrhagic Escherichia coli O157:H7. Nature 409: 529–533.NT PernaG. Plunkett 3rdV. BurlandB. MauJD Glasner2001Genome sequence of enterohaemorrhagic Escherichia coli O157:H7.Nature409529533
- 3. Denamur E (2011) The 2011 Shiga toxin-producing Escherichia coli O104:H4 German outbreak: a lesson in genomic plasticity. Clin Microbiol Infect 17: 1124–1125.E. Denamur2011The 2011 Shiga toxin-producing Escherichia coli O104:H4 German outbreak: a lesson in genomic plasticity.Clin Microbiol Infect1711241125
- 4. Hyde R (2011) Germany reels in the wake of E coli outbreak. Lancet 377: 1991.R. Hyde2011Germany reels in the wake of E coli outbreak.Lancet3771991
- 5. Griffin PM, Ostroff SM, Tauxe RV, Greene KD, Wells JG, et al. (1988) Illnesses associated with Escherichia coli O157:H7 infections. A broad clinical spectrum. Ann Intern Med 109: 705–712.PM GriffinSM OstroffRV TauxeKD GreeneJG Wells1988Illnesses associated with Escherichia coli O157:H7 infections. A broad clinical spectrum.Ann Intern Med109705712
- 6. Jarvis KG, Kaper JB (1996) Secretion of extracellular proteins by enterohemorrhagic Escherichia coli via a putative type III secretion system. Infect Immun 64: 4826–4829.KG JarvisJB Kaper1996Secretion of extracellular proteins by enterohemorrhagic Escherichia coli via a putative type III secretion system.Infect Immun6448264829
- 7. Elliott EJ, Robins-Browne RM, O'Loughlin EV, Bennett-Wood V, Bourke J, et al. (2001) Nationwide study of haemolytic uraemic syndrome: clinical, microbiological, and epidemiological features. Arch Dis Child 85: 125–131.EJ ElliottRM Robins-BrowneEV O'LoughlinV. Bennett-WoodJ. Bourke2001Nationwide study of haemolytic uraemic syndrome: clinical, microbiological, and epidemiological features.Arch Dis Child85125131
- 8. Pickering LK, Obrig TG, Stapleton FB (1994) Hemolytic-uremic syndrome and enterohemorrhagic Escherichia coli. Pediatr Infect Dis J 13: 459–475; quiz 476.LK PickeringTG ObrigFB Stapleton1994Hemolytic-uremic syndrome and enterohemorrhagic Escherichia coli.Pediatr Infect Dis J13459475; quiz 476
- 9. Donohue-Rolfe A, Kondova I, Oswald S, Hutto D, Tzipori S (2000) Escherichia coli O157:H7 strains that express Shiga toxin (Stx) 2 alone are more neurotropic for gnotobiotic piglets than are isotypes producing only Stx1 or both Stx1 and Stx2. J Infect Dis 181: 1825–1829.A. Donohue-RolfeI. KondovaS. OswaldD. HuttoS. Tzipori2000Escherichia coli O157:H7 strains that express Shiga toxin (Stx) 2 alone are more neurotropic for gnotobiotic piglets than are isotypes producing only Stx1 or both Stx1 and Stx2.J Infect Dis18118251829
- 10. Link AJ (2002) Multidimensional peptide separations in proteomics. Trends Biotechnol 20: S8–13.AJ Link2002Multidimensional peptide separations in proteomics.Trends Biotechnol20S813
- 11. Wolters DA, Washburn MP, Yates JR 3rd (2001) An automated multidimensional protein identification technology for shotgun proteomics. Anal Chem 73: 5683–5690.DA WoltersMP WashburnJR Yates 3rd2001An automated multidimensional protein identification technology for shotgun proteomics.Anal Chem7356835690
- 12. Perkins DN, Pappin DJ, Creasy DM, Cottrell JS (1999) Probability-based protein identification by searching sequence databases using mass spectrometry data. Electrophoresis 20: 3551–3567.DN PerkinsDJ PappinDM CreasyJS Cottrell1999Probability-based protein identification by searching sequence databases using mass spectrometry data.Electrophoresis2035513567
- 13. Keller A, Eng J, Zhang N, Li XJ, Aebersold R (2005) A uniform proteomics MS/MS analysis platform utilizing open XML file formats. Mol Syst Biol 1: 2005 0017.A. KellerJ. EngN. ZhangXJ LiR. Aebersold2005A uniform proteomics MS/MS analysis platform utilizing open XML file formats.Mol Syst Biol12005 0017
- 14. Tanner S, Shu H, Frank A, Wang LC, Zandi E, et al. (2005) InsPecT: identification of posttranslationally modified peptides from tandem mass spectra. Anal Chem 77: 4626–4639.S. TannerH. ShuA. FrankLC WangE. Zandi2005InsPecT: identification of posttranslationally modified peptides from tandem mass spectra.Anal Chem7746264639
- 15. Swanson SK, Washburn MP (2005) The continuing evolution of shotgun proteomics. Drug Discov Today 10: 719–725.SK SwansonMP Washburn2005The continuing evolution of shotgun proteomics.Drug Discov Today10719725
- 16. Elliott MH, Smith DS, Parker CE, Borchers C (2009) Current trends in quantitative proteomics. J Mass Spectrom 44: 1637–1660.MH ElliottDS SmithCE ParkerC. Borchers2009Current trends in quantitative proteomics.J Mass Spectrom4416371660
- 17. Zimmer JS, Monroe ME, Qian WJ, Smith RD (2006) Advances in proteomics data analysis and display using an accurate mass and time tag approach. Mass Spectrom Rev 25: 450–482.JS ZimmerME MonroeWJ QianRD Smith2006Advances in proteomics data analysis and display using an accurate mass and time tag approach.Mass Spectrom Rev25450482
- 18. Ishihama Y, Oda Y, Tabata T, Sato T, Nagasu T, et al. (2005) Exponentially modified protein abundance index (emPAI) for estimation of absolute protein amount in proteomics by the number of sequenced peptides per protein. Mol Cell Proteomics 4: 1265–1272.Y. IshihamaY. OdaT. TabataT. SatoT. Nagasu2005Exponentially modified protein abundance index (emPAI) for estimation of absolute protein amount in proteomics by the number of sequenced peptides per protein.Mol Cell Proteomics412651272
- 19. Lu P, Vogel C, Wang R, Yao X, Marcotte EM (2007) Absolute protein expression profiling estimates the relative contributions of transcriptional and translational regulation. Nat Biotechnol 25: 117–124.P. LuC. VogelR. WangX. YaoEM Marcotte2007Absolute protein expression profiling estimates the relative contributions of transcriptional and translational regulation.Nat Biotechnol25117124
- 20. Rudnick PA, Clauser KR, Kilpatrick LE, Tchekhovskoi DV, Neta P, et al. (2010) Performance metrics for liquid chromatography-tandem mass spectrometry systems in proteomics analyses. Mol Cell Proteomics 9: 225–241.PA RudnickKR ClauserLE KilpatrickDV TchekhovskoiP. Neta2010Performance metrics for liquid chromatography-tandem mass spectrometry systems in proteomics analyses.Mol Cell Proteomics9225241
- 21. Tabb DL, Vega-Montoto L, Rudnick PA, Variyath AM, Ham AJ, et al. (2010) Repeatability and reproducibility in proteomic identifications by liquid chromatography-tandem mass spectrometry. J Proteome Res 9: 761–776.DL TabbL. Vega-MontotoPA RudnickAM VariyathAJ Ham2010Repeatability and reproducibility in proteomic identifications by liquid chromatography-tandem mass spectrometry.J Proteome Res9761776
- 22. Kuntumalla S, Braisted JC, Huang ST, Parmar PP, Clark DJ, et al. (2009) Comparison of two label-free global quantitation methods, APEX and 2D gel electrophoresis, applied to the Shigella dysenteriae proteome. Proteome Sci 7: 22.S. KuntumallaJC BraistedST HuangPP ParmarDJ Clark2009Comparison of two label-free global quantitation methods, APEX and 2D gel electrophoresis, applied to the Shigella dysenteriae proteome.Proteome Sci722
- 23. Jacobs JM, Mottaz HM, Yu LR, Anderson DJ, Moore RJ, et al. (2004) Multidimensional proteome analysis of human mammary epithelial cells. J Proteome Res 3: 68–75.JM JacobsHM MottazLR YuDJ AndersonRJ Moore2004Multidimensional proteome analysis of human mammary epithelial cells.J Proteome Res36875
- 24. Zhang J, Xu X, Gao M, Yang P, Zhang X (2007) Comparison of 2-D LC and 3-D LC with post- and pre-tryptic-digestion SEC fractionation for proteome analysis of normal human liver tissue. Proteomics 7: 500–512.J. ZhangX. XuM. GaoP. YangX. Zhang2007Comparison of 2-D LC and 3-D LC with post- and pre-tryptic-digestion SEC fractionation for proteome analysis of normal human liver tissue.Proteomics7500512
- 25. Arakawa T, Wen J (2001) Size-exclusion chromatography with on-line light scattering. Curr Protoc Protein Sci Chapter 20: Unit 20 26.T. ArakawaJ. Wen2001Size-exclusion chromatography with on-line light scattering.Curr Protoc Protein Sci Chapter20Unit 20 26
- 26. Corbin RW, Paliy O, Yang F, Shabanowitz J, Platt M, et al. (2003) Toward a protein profile of Escherichia coli: comparison to its transcription profile. Proc Natl Acad Sci U S A 100: 9232–9237.RW CorbinO. PaliyF. YangJ. ShabanowitzM. Platt2003Toward a protein profile of Escherichia coli: comparison to its transcription profile.Proc Natl Acad Sci U S A10092329237
- 27. Maisonneuve E, Fraysse L, Moinier D, Dukan S (2008) Existence of abnormal protein aggregates in healthy Escherichia coli cells. J Bacteriol 190: 887–893.E. MaisonneuveL. FraysseD. MoinierS. Dukan2008Existence of abnormal protein aggregates in healthy Escherichia coli cells.J Bacteriol190887893
- 28. Mogk A, Tomoyasu T, Goloubinoff P, Rudiger S, Roder D, et al. (1999) Identification of thermolabile Escherichia coli proteins: prevention and reversion of aggregation by DnaK and ClpB. EMBO J 18: 6934–6949.A. MogkT. TomoyasuP. GoloubinoffS. RudigerD. Roder1999Identification of thermolabile Escherichia coli proteins: prevention and reversion of aggregation by DnaK and ClpB.EMBO J1869346949
- 29. Han MJ, Lee SY (2006) The Escherichia coli proteome: past, present, and future prospects. Microbiol Mol Biol Rev 70: 362–439.MJ HanSY Lee2006The Escherichia coli proteome: past, present, and future prospects.Microbiol Mol Biol Rev70362439
- 30. Arifuzzaman M, Maeda M, Itoh A, Nishikata K, Takita C, et al. (2006) Large-scale identification of protein-protein interaction of Escherichia coli K-12. Genome Res 16: 686–691.M. ArifuzzamanM. MaedaA. ItohK. NishikataC. Takita2006Large-scale identification of protein-protein interaction of Escherichia coli K-12.Genome Res16686691
- 31. Butland G, Peregrin-Alvarez JM, Li J, Yang W, Yang X, et al. (2005) Interaction network containing conserved and essential protein complexes in Escherichia coli. Nature 433: 531–537.G. ButlandJM Peregrin-AlvarezJ. LiW. YangX. Yang2005Interaction network containing conserved and essential protein complexes in Escherichia coli.Nature433531537
- 32. Hu P, Janga SC, Babu M, Diaz-Mejia JJ, Butland G, et al. (2009) Global functional atlas of Escherichia coli encompassing previously uncharacterized proteins. PLoS Biol 7: e96.P. HuSC JangaM. BabuJJ Diaz-MejiaG. Butland2009Global functional atlas of Escherichia coli encompassing previously uncharacterized proteins.PLoS Biol7e96
- 33. Taniguchi Y, Choi PJ, Li GW, Chen H, Babu M, et al. (2010) Quantifying E. coli proteome and transcriptome with single-molecule sensitivity in single cells. Science 329: 533–538.Y. TaniguchiPJ ChoiGW LiH. ChenM. Babu2010Quantifying E. coli proteome and transcriptome with single-molecule sensitivity in single cells.Science329533538
- 34. Defeu Soufo HJ, Reimold C, Linne U, Knust T, Gescher J, et al. (2010) Bacterial translation elongation factor EF-Tu interacts and colocalizes with actin-like MreB protein. Proc Natl Acad Sci U S A 107: 3163–3168.HJ Defeu SoufoC. ReimoldU. LinneT. KnustJ. Gescher2010Bacterial translation elongation factor EF-Tu interacts and colocalizes with actin-like MreB protein.Proc Natl Acad Sci U S A10731633168
- 35. Kaberdin VR, Lin-Chao S (2009) Unraveling new roles for minor components of the E. coli RNA degradosome. RNA Biol 6: 402–405.VR KaberdinS. Lin-Chao2009Unraveling new roles for minor components of the E. coli RNA degradosome.RNA Biol6402405
- 36. Schmidt MA (2010) LEEways: tales of EPEC, ATEC and EHEC. Cell Microbiol 12: 1544–1552.MA Schmidt2010LEEways: tales of EPEC, ATEC and EHEC.Cell Microbiol1215441552
- 37. Tobe T, Beatson SA, Taniguchi H, Abe H, Bailey CM, et al. (2006) An extensive repertoire of type III secretion effectors in Escherichia coli O157 and the role of lambdoid phages in their dissemination. Proc Natl Acad Sci U S A 103: 14941–14946.T. TobeSA BeatsonH. TaniguchiH. AbeCM Bailey2006An extensive repertoire of type III secretion effectors in Escherichia coli O157 and the role of lambdoid phages in their dissemination.Proc Natl Acad Sci U S A1031494114946
- 38. Asakura H, Panutdaporn N, Kawamoto K, Igimi S, Yamamoto S, et al. (2007) Proteomic characterization of enterohemorrhagic escherichia coli O157:H7 in the oxidation-induced viable but non-culturable state. Microbiol Immunol 51: 875–881.H. AsakuraN. PanutdapornK. KawamotoS. IgimiS. Yamamoto2007Proteomic characterization of enterohemorrhagic escherichia coli O157:H7 in the oxidation-induced viable but non-culturable state.Microbiol Immunol51875881
- 39. Pieper R, Zhang Q, Parmar PP, Huang ST, Clark DJ, et al. (2009) The Shigella dysenteriae serotype 1 proteome, profiled in the host intestinal environment, reveals major metabolic modifications and increased expression of invasive proteins. Proteomics 9: 5029–5045.R. PieperQ. ZhangPP ParmarST HuangDJ Clark2009The Shigella dysenteriae serotype 1 proteome, profiled in the host intestinal environment, reveals major metabolic modifications and increased expression of invasive proteins.Proteomics950295045
- 40. Wisniewski JR, Zougman A, Nagaraj N, Mann M (2009) Universal sample preparation method for proteome analysis. Nat Methods 6: 359–362.JR WisniewskiA. ZougmanN. NagarajM. Mann2009Universal sample preparation method for proteome analysis.Nat Methods6359362
- 41. Pieper R, Gatlin-Bunai CL, Mongodin EF, Parmar PP, Huang ST, et al. (2006) Comparative proteomic analysis of Staphylococcus aureus strains with differences in resistance to the cell wall-targeting antibiotic vancomycin. Proteomics 6: 4246–4258.R. PieperCL Gatlin-BunaiEF MongodinPP ParmarST Huang2006Comparative proteomic analysis of Staphylococcus aureus strains with differences in resistance to the cell wall-targeting antibiotic vancomycin.Proteomics642464258
- 42. Keller A, Nesvizhskii AI, Kolker E, Aebersold R (2002) Empirical statistical model to estimate the accuracy of peptide identifications made by MS/MS and database search. Anal Chem 74: 5383–5392.A. KellerAI NesvizhskiiE. KolkerR. Aebersold2002Empirical statistical model to estimate the accuracy of peptide identifications made by MS/MS and database search.Anal Chem7453835392
- 43. Nesvizhskii AI, Keller A, Kolker E, Aebersold R (2003) A statistical model for identifying proteins by tandem mass spectrometry. Anal Chem 75: 4646–4658.AI NesvizhskiiA. KellerE. KolkerR. Aebersold2003A statistical model for identifying proteins by tandem mass spectrometry.Anal Chem7546464658
- 44. Braisted JC, Kuntumalla S, Vogel C, Marcotte EM, Rodrigues AR, et al. (2008) The APEX Quantitative Proteomics Tool: generating protein quantitation estimates from LC-MS/MS proteomics results. BMC Bioinformatics 9: 529.JC BraistedS. KuntumallaC. VogelEM MarcotteAR Rodrigues2008The APEX Quantitative Proteomics Tool: generating protein quantitation estimates from LC-MS/MS proteomics results.BMC Bioinformatics9529
- 45. Neidhardt FCU (1996) Escherichia coli and Salmonella Typhimurium: Cellular and Molecular Biology. 2nd ed. Neidhardt FCea, editor. Washington D.C.: American Association of Microbiology. pp. 13–16.FCU Neidhardt1996FCea NeidhardtEscherichia coli and Salmonella Typhimurium: Cellular and Molecular Biology. 2nd edWashington D.C.American Association of Microbiology1316
- 46. Saeed AI, Bhagabati NK, Braisted JC, Liang W, Sharov V, et al. (2006) TM4 microarray software suite. Methods Enzymol 411: 134–193.AI SaeedNK BhagabatiJC BraistedW. LiangV. Sharov2006TM4 microarray software suite.Methods Enzymol411134193
- 47. Page WJ, Taylor DE (1988) Comparison of methods used to separate the inner and outer membranes of cell envelopes of Campylobacter spp. J Gen Microbiol 134: 2925–2932.WJ PageDE Taylor1988Comparison of methods used to separate the inner and outer membranes of cell envelopes of Campylobacter spp.J Gen Microbiol13429252932
- 48. Schnaitman CA (1971) Solubilization of the cytoplasmic membrane of Escherichia coli by Triton X-100. J Bacteriol 108: 545–552.CA Schnaitman1971Solubilization of the cytoplasmic membrane of Escherichia coli by Triton X-100.J Bacteriol108545552
- 49. Schnaitman CA (1971) Effect of ethylenediaminetetraacetic acid, Triton X-100, and lysozyme on the morphology and chemical composition of isolate cell walls of Escherichia coli. J Bacteriol 108: 553–563.CA Schnaitman1971Effect of ethylenediaminetetraacetic acid, Triton X-100, and lysozyme on the morphology and chemical composition of isolate cell walls of Escherichia coli.J Bacteriol108553563
- 50. Kuntumalla S, Zhang Q, Braisted JC, Fleischmann RD, Peterson SN, et al. (2011) In vivo versus in vitro protein abundance analysis of Shigella dysenteriae type 1 reveals changes in the expression of proteins involved in virulence, stress and energy metabolism. BMC Microbiol 11: 147.S. KuntumallaQ. ZhangJC BraistedRD FleischmannSN Peterson2011In vivo versus in vitro protein abundance analysis of Shigella dysenteriae type 1 reveals changes in the expression of proteins involved in virulence, stress and energy metabolism.BMC Microbiol11147
- 51. Riley LW, Remis RS, Helgerson SD, McGee HB, Wells JG, et al. (1983) Hemorrhagic colitis associated with a rare Escherichia coli serotype. N Engl J Med 308: 681–685.LW RileyRS RemisSD HelgersonHB McGeeJG Wells1983Hemorrhagic colitis associated with a rare Escherichia coli serotype.N Engl J Med308681685
- 52. Whittam (2006) EcMLST a multi-locus sequence typing database system for pathogenic E. coli. East Lansing, MI: Michigan State University. Whittam2006EcMLST a multi-locus sequence typing database system for pathogenic E. coliEast Lansing, MIMichigan State University
- 53. Lee JH, Kim YG, Cho MH, Wood TK, Lee J (2011) Transcriptomic analysis for genetic mechanisms of the factors related to biofilm formation in Escherichia coli O157:H7. Curr Microbiol 62: 1321–1330.JH LeeYG KimMH ChoTK WoodJ. Lee2011Transcriptomic analysis for genetic mechanisms of the factors related to biofilm formation in Escherichia coli O157:H7.Curr Microbiol6213211330
- 54. Cornick NA, Matise I, Samuel JE, Bosworth BT, Moon HW (2000) Shiga toxin-producing Escherichia coli infection: temporal and quantitative relationships among colonization, toxin production, and systemic disease. J Infect Dis 181: 242–251.NA CornickI. MatiseJE SamuelBT BosworthHW Moon2000Shiga toxin-producing Escherichia coli infection: temporal and quantitative relationships among colonization, toxin production, and systemic disease.J Infect Dis181242251
- 55. Wu B, Skarina T, Yee A, Jobin MC, Dileo R, et al. (2010) NleG Type 3 effectors from enterohaemorrhagic Escherichia coli are U-Box E3 ubiquitin ligases. PLoS Pathog 6: e1000960.B. WuT. SkarinaA. YeeMC JobinR. Dileo2010NleG Type 3 effectors from enterohaemorrhagic Escherichia coli are U-Box E3 ubiquitin ligases.PLoS Pathog6e1000960
- 56. Li M, Rosenshine I, Tung SL, Wang XH, Friedberg D, et al. (2004) Comparative proteomic analysis of extracellular proteins of enterohemorrhagic and enteropathogenic Escherichia coli strains and their ihf and ler mutants. Appl Environ Microbiol 70: 5274–5282.M. LiI. RosenshineSL TungXH WangD. Friedberg2004Comparative proteomic analysis of extracellular proteins of enterohemorrhagic and enteropathogenic Escherichia coli strains and their ihf and ler mutants.Appl Environ Microbiol7052745282
- 57. Grys TE, Siegel MB, Lathem WW, Welch RA (2005) The StcE protease contributes to intimate adherence of enterohemorrhagic Escherichia coli O157:H7 to host cells. Infect Immun 73: 1295–1303.TE GrysMB SiegelWW LathemRA Welch2005The StcE protease contributes to intimate adherence of enterohemorrhagic Escherichia coli O157:H7 to host cells.Infect Immun7312951303
- 58. Lathem WW, Grys TE, Witowski SE, Torres AG, Kaper JB, et al. (2002) StcE, a metalloprotease secreted by Escherichia coli O157:H7, specifically cleaves C1 esterase inhibitor. Mol Microbiol 45: 277–288.WW LathemTE GrysSE WitowskiAG TorresJB Kaper2002StcE, a metalloprotease secreted by Escherichia coli O157:H7, specifically cleaves C1 esterase inhibitor.Mol Microbiol45277288
- 59. Casjens S (2003) Prophages and bacterial genomics: what have we learned so far? Mol Microbiol 49: 277–300.S. Casjens2003Prophages and bacterial genomics: what have we learned so far?Mol Microbiol49277300
- 60. Asadulghani M, Ogura Y, Ooka T, Itoh T, Sawaguchi A, et al. (2009) The defective prophage pool of Escherichia coli O157: prophage-prophage interactions potentiate horizontal transfer of virulence determinants. PLoS Pathog 5: e1000408.M. AsadulghaniY. OguraT. OokaT. ItohA. Sawaguchi2009The defective prophage pool of Escherichia coli O157: prophage-prophage interactions potentiate horizontal transfer of virulence determinants.PLoS Pathog5e1000408
- 61. Taoka M, Yamauchi Y, Shinkawa T, Kaji H, Motohashi W, et al. (2004) Only a small subset of the horizontally transferred chromosomal genes in Escherichia coli are translated into proteins. Mol Cell Proteomics 3: 780–787.M. TaokaY. YamauchiT. ShinkawaH. KajiW. Motohashi2004Only a small subset of the horizontally transferred chromosomal genes in Escherichia coli are translated into proteins.Mol Cell Proteomics3780787
- 62. Bernsel A, Daley DO (2009) Exploring the inner membrane proteome of Escherichia coli: which proteins are eluding detection and why? Trends Microbiol 17: 444–449.A. BernselDO Daley2009Exploring the inner membrane proteome of Escherichia coli: which proteins are eluding detection and why?Trends Microbiol17444449
- 63. Stenberg F, Chovanec P, Maslen SL, Robinson CV, Ilag LL, et al. (2005) Protein complexes of the Escherichia coli cell envelope. J Biol Chem 280: 34409–34419.F. StenbergP. ChovanecSL MaslenCV RobinsonLL Ilag2005Protein complexes of the Escherichia coli cell envelope.J Biol Chem2803440934419
- 64. Hobb RI, Fields JA, Burns CM, Thompson SA (2009) Evaluation of procedures for outer membrane isolation from Campylobacter jejuni. Microbiology 155: 979–988.RI HobbJA FieldsCM BurnsSA Thompson2009Evaluation of procedures for outer membrane isolation from Campylobacter jejuni.Microbiology155979988
- 65. Pan JY, Li H, Ma Y, Chen P, Zhao P, et al. (2010) Complexome of Escherichia coli envelope proteins under normal physiological conditions. J Proteome Res 9: 3730–3740.JY PanH. LiY. MaP. ChenP. Zhao2010Complexome of Escherichia coli envelope proteins under normal physiological conditions.J Proteome Res937303740
- 66. Brokx SJ, Ellison M, Locke T, Bottorff D, Frost L, et al. (2004) Genome-wide analysis of lipoprotein expression in Escherichia coli MG1655. J Bacteriol 186: 3254–3258.SJ BrokxM. EllisonT. LockeD. BottorffL. Frost2004Genome-wide analysis of lipoprotein expression in Escherichia coli MG1655.J Bacteriol18632543258
- 67. Walburger A, Lazdunski C, Corda Y (2002) The Tol/Pal system function requires an interaction between the C-terminal domain of TolA and the N-terminal domain of TolB. Mol Microbiol 44: 695–708.A. WalburgerC. LazdunskiY. Corda2002The Tol/Pal system function requires an interaction between the C-terminal domain of TolA and the N-terminal domain of TolB.Mol Microbiol44695708
- 68. Xiong X, Deeter JN, Misra R (1996) Assembly-defective OmpC mutants of Escherichia coli K-12. J Bacteriol 178: 1213–1215.X. XiongJN DeeterR. Misra1996Assembly-defective OmpC mutants of Escherichia coli K-12.J Bacteriol17812131215
- 69. Folta-Stogniew E, Williams KR (1999) Determination of molecular masses of proteins in solution: Implementation of an HPLC size exclusion chromatography and laser light scattering service in a core laboratory. J Biomol Tech 10: 51–63.E. Folta-StogniewKR Williams1999Determination of molecular masses of proteins in solution: Implementation of an HPLC size exclusion chromatography and laser light scattering service in a core laboratory.J Biomol Tech105163
- 70. Zundel MA, Basturea GN, Deutscher MP (2009) Initiation of ribosome degradation during starvation in Escherichia coli. RNA 15: 977–983.MA ZundelGN BastureaMP Deutscher2009Initiation of ribosome degradation during starvation in Escherichia coli.RNA15977983
- 71. Missiakas D, Raina S (1997) Protein folding in the bacterial periplasm. J Bacteriol 179: 2465–2471.D. MissiakasS. Raina1997Protein folding in the bacterial periplasm.J Bacteriol17924652471
- 72. Malki A, Le HT, Milles S, Kern R, Caldas T, et al. (2008) Solubilization of protein aggregates by the acid stress chaperones HdeA and HdeB. J Biol Chem 283: 13679–13687.A. MalkiHT LeS. MillesR. KernT. Caldas2008Solubilization of protein aggregates by the acid stress chaperones HdeA and HdeB.J Biol Chem2831367913687
- 73. Tomoyasu T, Mogk A, Langen H, Goloubinoff P, Bukau B (2001) Genetic dissection of the roles of chaperones and proteases in protein folding and degradation in the Escherichia coli cytosol. Mol Microbiol 40: 397–413.T. TomoyasuA. MogkH. LangenP. GoloubinoffB. Bukau2001Genetic dissection of the roles of chaperones and proteases in protein folding and degradation in the Escherichia coli cytosol.Mol Microbiol40397413
- 74. Cole ST, Eiglmeier K, Ahmed S, Honore N, Elmes L, et al. (1988) Nucleotide sequence and gene-polypeptide relationships of the glpABC operon encoding the anaerobic sn-glycerol-3-phosphate dehydrogenase of Escherichia coli K-12. J Bacteriol 170: 2448–2456.ST ColeK. EiglmeierS. AhmedN. HonoreL. Elmes1988Nucleotide sequence and gene-polypeptide relationships of the glpABC operon encoding the anaerobic sn-glycerol-3-phosphate dehydrogenase of Escherichia coli K-12.J Bacteriol17024482456
- 75. Yeh JI, Chinte U, Du S (2008) Structure of glycerol-3-phosphate dehydrogenase, an essential monotopic membrane enzyme involved in respiration and metabolism. Proc Natl Acad Sci U S A 105: 3280–3285.JI YehU. ChinteS. Du2008Structure of glycerol-3-phosphate dehydrogenase, an essential monotopic membrane enzyme involved in respiration and metabolism.Proc Natl Acad Sci U S A10532803285
- 76. Voegele RT, Sweet GD, Boos W (1993) Glycerol kinase of Escherichia coli is activated by interaction with the glycerol facilitator. J Bacteriol 175: 1087–1094.RT VoegeleGD SweetW. Boos1993Glycerol kinase of Escherichia coli is activated by interaction with the glycerol facilitator.J Bacteriol17510871094
- 77. Raetz CR, Reynolds CM, Trent MS, Bishop RE (2007) Lipid A modification systems in gram-negative bacteria. Annu Rev Biochem 76: 295–329.CR RaetzCM ReynoldsMS TrentRE Bishop2007Lipid A modification systems in gram-negative bacteria.Annu Rev Biochem76295329
- 78. Samuel G, Hogbin JP, Wang L, Reeves PR (2004) Relationships of the Escherichia coli O157, O111, and O55 O-antigen gene clusters with those of Salmonella enterica and Citrobacter freundii, which express identical O antigens. J Bacteriol 186: 6536–6543.G. SamuelJP HogbinL. WangPR Reeves2004Relationships of the Escherichia coli O157, O111, and O55 O-antigen gene clusters with those of Salmonella enterica and Citrobacter freundii, which express identical O antigens.J Bacteriol18665366543
- 79. van den Ent F, Amos LA, Lowe J (2001) Prokaryotic origin of the actin cytoskeleton. Nature 413: 39–44.F. van den EntLA AmosJ. Lowe2001Prokaryotic origin of the actin cytoskeleton.Nature4133944
- 80. Vats P, Yu J, Rothfield L (2009) The dynamic nature of the bacterial cytoskeleton. Cell Mol Life Sci 66: 3353–3362.P. VatsJ. YuL. Rothfield2009The dynamic nature of the bacterial cytoskeleton.Cell Mol Life Sci6633533362
- 81. Adams DW, Errington J (2009) Bacterial cell division: assembly, maintenance and disassembly of the Z ring. Nat Rev Microbiol 7: 642–653.DW AdamsJ. Errington2009Bacterial cell division: assembly, maintenance and disassembly of the Z ring.Nat Rev Microbiol7642653
- 82. Lim GE, Derman AI, Pogliano J (2005) Bacterial DNA segregation by dynamic SopA polymers. Proc Natl Acad Sci U S A 102: 17658–17663.GE LimAI DermanJ. Pogliano2005Bacterial DNA segregation by dynamic SopA polymers.Proc Natl Acad Sci U S A1021765817663
- 83. Kuhner S, van Noort V, Betts MJ, Leo-Macias A, Batisse C, et al. (2009) Proteome organization in a genome-reduced bacterium. Science 326: 1235–1240.S. KuhnerV. van NoortMJ BettsA. Leo-MaciasC. Batisse2009Proteome organization in a genome-reduced bacterium.Science32612351240