Open Access

Strand-specific transcriptomes of Enterohemorrhagic Escherichia coli in response to interactions with ground beef microbiota: interactions between microorganisms in raw meat

BMC Genomics201718:574

https://doi.org/10.1186/s12864-017-3957-2

Received: 13 March 2017

Accepted: 24 July 2017

Published: 3 August 2017

Abstract

Background

Enterohemorrhagic Escherichia coli (EHEC) are zoonotic agents associated with outbreaks worldwide. Growth of EHEC strains in ground beef could be inhibited by background microbiota that is present initially at levels greater than that of the pathogen E. coli. However, how the microbiota outcompetes the pathogenic bacteria is unknown. Our objective was to identify metabolic pathways of EHEC that were altered by natural microbiota in order to improve our understanding of the mechanisms controlling the growth and survival of EHECs in ground beef.

Results

Based on 16S metagenomics analysis, we identified the microbial community structure in our beef samples which was an essential preliminary for subtractively analyzing the gene expression of the EHEC strains. Then, we applied strand-specific RNA-seq to investigate the effects of this microbiota on the global gene expression of EHEC O2621765 and O157EDL933 strains by comparison with their behavior in beef meat without microbiota. In strain O2621765, the expression of genes connected with nitrate metabolism and nitrite detoxification, DNA repair, iron and nickel acquisition and carbohydrate metabolism, and numerous genes involved in amino acid metabolism were down-regulated. Further, the observed repression of ftsL and murF, involved respectively in building the cytokinetic ring apparatus and in synthesizing the cytoplasmic precursor of cell wall peptidoglycan, might help to explain the microbiota’s inhibitory effect on EHECs. For strain O157EDL933, the induced expression of the genes implicated in detoxification and the general stress response and the repressed expression of the peR gene, a gene negatively associated with the virulence phenotype, might be linked to the survival and virulence of O157:H7 in ground beef with microbiota.

Conclusion

In the present study, we show how RNA-Seq coupled with a 16S metagenomics analysis can be used to identify the effects of a complex microbial community on relevant functions of an individual microbe within it. These findings add to our understanding of the behavior of EHECs in ground beef. By measuring transcriptional responses of EHEC, we could identify putative targets which may be useful to develop new strategies to limit their shedding in ground meat thus reducing the risk of human illnesses.

Keywords

RNA-Seq EHEC Ground beef Natural microbiota 16S metagenomics analysis

Background

Bacteria rapidly program their gene expression to survive changing environments and resist challenging conditions. Such adaptation through gene expression can radically change bacterial physiology and pathogenicity. RNA-sequencing has revolutionized the study of gene expression because, in addition to quantifying transcriptional output, it allows the detection and characterization of all transcripts in a genome. This innovative, non a priori tool is rapidly becoming the method of choice for revealing new functional genes and pathways in individual microbes [1, 2] as well as in complex environmental communities, e.g. from the sea [3, 4] and the human gut [5, 6]. However, deciphering the behavior of one population within a complex microbiota comprising related taxonomic species is a real challenge.

EHECs are food-borne zoonotic agents associated with disease outbreaks worldwide and represent a serious public health concern. They are strongly associated with severe forms of infection such as hemorrhagic colitis and, in some extreme cases, hemolytic-uremic syndrome (HUS) [7, 8]. Human infection is typically acquired through ingestion of contaminated food (undercooked ground beef, dairy products, vegetables, etc.) and water. In 2014, 5955 confirmed cases of EHEC infection were reported in the EU. That year as in previous years, the most commonly reported EHEC serogroup was O157 (46.3% of cases with known serogroup), followed by serogroups O26, O103, O145, O91, O146 and O111 (EFSA, Zoonosis 2014). Beef cattle are the primary reservoir of EHECs [9] which they carry in and excrete from their gastrointestinal tract. These animals stay without any symptoms of disease [10]. Cattle faeces is considered the main source of EHEC contamination of carcasses during slaughter [11].

Data from EFSA (Zoonosis, 2014) [12] show that the highest proportion of STEC-positive samples were reported from ruminant meat (goat, sheep, cattle and deer). Over the same period, STEC were reported in about 1% of cheese samples, particularly sheep and goat milk cheeses, while contamination was rare in RTE (Ready-To-Eat) food of plant origin. In France, the majority of STEC infections point to the consumption of undercooked minced meat [13]. We hypothesized that biotic and abiotic factors in meat could have an impact on the growth, physiology and virulence of EHEC strains. The meat microbiota might antagonize human pathogens by many different mechanisms, including production of compounds with antimicrobial activity as well as competition for attachment sites or nutrients [14, 15]. To date, few data about the interaction between bovine meat microflora and EHECs are available.

The aim of this study was to develop a new RNA-seq tool for use with polymicrobial samples, among others. We identified metabolic pathways of EHEC that were altered by natural microbiota, to improve our understanding of the mechanisms controlling the growth and survival of EHECs in ground beef. We first conducted a 16S metagenomics analysis to identify the natural microbiota present in some ground meat. We then used strand-specific RNA-seq to investigate the effects of the ground beef background microbiota on the overall gene expression of EHEC strains O2621765 and O157EDL933 by comparison with their behavior in beef meat without microbiota.

Methods

Bacterial strains

The EHECs used in this study were 0157:H7 strain EDL933 (O157EDL933) and O26:H11 strain 21,765 (O2621765), isolated from Michigan ground beef and human fecal samples respectively and linked to human EHEC infections [16, 17].

Sample preparation

A single beef piece of about 11 kg taken from a local slaughterhouse was kept 72 h at 4 °C before processing in our laboratory. The outer part of the muscle was aseptically separated from the inner part, considering the first to be highly contaminated and the second to be devoid of bacteria. Each part was aseptically minced and divided into 4 portions of 200 g before homogenization in a sterile stomacher bag (Spiral Biotech, Bethesda, Md.). For each trial, the ground beef was checked for the absence of E. coli O26:- and O157:- [18].

Frozen cultures of EHEC were revived on plate-count agar (PCA, BioMérieux, Marcy l’Etoile, France) by incubation at 37 °C for 24 h. One colony of each strain was selected and cultured in BHI (Brain Heart Infusion) overnight at 37 °C without agitation. Cells were harvested by centrifugation at 4000×g for 15 min at 12 °C. Pellets were suspended in 0.1% peptone (Oxoid) water (12 °C).

The ground beef was then inoculated as follows: each 200 g portion was inoculated with 20 ml (108 CFU/ml) of EHEC strains O157EDL933 or O2621765 to obtain a final concentration of about 107 CFU/g. For the control experiment, a similar amount of sterile peptone water was added. Each ground beef sample was hand mixed to distribute the inoculum evenly. Samples were then incubated in a sterile stomacher bag without agitation at 12 °C for 7 days. This temperature (12 °C) was chosen as being sufficient for identifying inter-organism interactions without halting EHEC growth [19]. Three biological replicates were performed for each condition.

Ground beef microbiota and EHEC enumeration

After serial dilutions of the samples stomached 1 min in BPW (Buffered Peptone Water, bioMerieux, Marcy l’Etoile, France), decimal dilutions of the samples were plated and incubated in appropriate conditions: for aerobic bacteria, soy agar plates were incubated at 37 °C for 24 h; for LAB, MRS (deMan Rogosa Sharpe; Difco) agar plates were incubated in a GasPak jar (BBL, Becton Dickinson Microbiology Systems) at 30 °C for 48 h, complemented with a gas generating sachet (BD GasPak EZ Gas Generating Sachet). For enumeration of O157EDL933 and O2621765 strains, dilutions were spread respectively onto ChromID O157:H7 agar (ChromID O157:H7 + 0.05 mg/l cefixime +2.5 mg/l tellurite, bioMérieux) and chromogen agar E. coli Brillance. The plates were incubated at 37 °C for 24 h. All counts were performed in duplicate.

Collection of EHEC cells and ground beef microbiota for metagenomic and transcriptomic analysis

Beef samples were diluted (1:4 w/v) in cold BPW containing 0.1% (v/v) Tween X-20 and pummeled for 1 min using a stomacher lab blender (Seward Medical, London, UK). Following pummeling, the filtered samples were clarified by centrifugation at 200 g for 5 min at 4 °C. The supernatants were removed to fresh tubes, and cells were collected by centrifugation at 3500 g for 10 min at 4 °C.

Metagenomic analysis

Genomic DNA extraction, PCR amplification and amplicon sequencing

DNA was extracted from samples using a Genomic DNA kit (NucleoSpin Tissue, Germany). To minimize PCR bias, DNA extracts from the triplicate samples were pooled and used as the template DNA. DNA purity (A260/A280 and A260/A230) and quantity were measured with a NanoDrop ND-1000 spectrophotometer (Thermo Scientific). The V1-V3 region of the 16S rRNA gene was amplified by PCR using the primers 27F (5′-AGAGTTTGATYMTGGCTCAG-3′) and 519R (5′-GWATTACCGCGGCKGCTG-3′) [20]. The libraries were prepared with Nextera XT version 3 chemistry (Illumina) and the Illumina MiSeq platform sequencer produced 300 bp paired-end reads.

Quality trimming, taxonomic assignments, and operational taxonomic unit clustering

The paired-end sequences were assembled into a single sequence, ca. 510-bp. The sequences were trimmed using MOTHUR software [21] to remove primers, long homo-polymers and sequences shorter than 100 bases (Fig. 1). Reads with ambiguous bases were removed from the data set. Sequence alignment, chimeras checking, distance calculation, OTU (Operational Taxonomic Unit) clustering and selection of non-redundant sequences were performed using MOTHUR version 1.34.4 [21]. The SILVA reference file for bacteria [22] was used to align the sequences. Indeed, to assign our sequence reads to the prokaryotic taxonomy, the Naïve Bayesian Classifier tool hosted by Ribosomal Database Project (RDP Classifier) [23] and described by Wang et al. (2007) was used [24]. This method runs a bootstrapping algorithm to find the confidence limit of the assignment of each query. A bootstrap cutoff value of 0.6 was used in our study. The percentage of each bacterial genus or phylum was calculated individually for each sample, providing information on relative abundance within the individual samples based on the relative numbers of reads in each one. Sequencing reads are available at the European Nucleotide Archive under study number PRJEB13580 (ERS1119543-ERS1119545).
Fig. 1

Workflow for studying the overall gene expression of Enterohemorrhagic Escherichia coli (EHEC) strains O157:H7 EDL933, O26:H11 21,765 in ground meat with or without background microbiota

RNA sequencing analysis

RNA extraction and DNase treatment

Briefly, cell pellets were re-suspended in 10 mL of TRIzol Reagent (Ambion) and incubated for 5 min at 20 °C. One volume of chloroform was added to 5 volumes of homogenate; nucleic acids were then separated from proteins by centrifugation at 12,000 g for 15 min at 4 °C. One volume of the aqueous supernatant was mixed with 1 volume of ethanol 70%. RNA extractions were then performed according to the Qiagen recommendations, including on-column RNA purification (RNeasy Midi Kit, QIAGEN).

DNase treatment

The TURBO DNA-free kit (Ambion) was used for the DNase treatment. Total RNA was treated using a rigorous protocol that includes a double quantity of DNase (6 units). Reactions were terminated with the addition of the DNase inactivation reagent (0.2 × the reaction volume) according to the kit instructions. The presence of DNA contamination was assessed by real time PCR targeting of the eae gene [25]. RNA integrity (RIN) values [26] were determined by running 1 μl aliquots on a Bioanalyzer 2100 (Agilent, Technologies, Inc.) resulting in clear patterns with prominent 16S and 23S ribosomal bands and RIN values of ≥8.

RNA-Seq library preparation and sequencing

Library construction and sequencing were performed at the MGX-Montpellier GenomiX platform. Ribo-Zero rRNA Removal Kit Bacteria (Illumina, San Diego, CA) was used to remove ribosomal RNA from 2 μg of total RNA. For each sample, 100 ng mRNA-enriched fraction was used to construct sequencing libraries using Illumina’s TruSeq Stranded mRNA Sample Prep Kit (Low throughput). The mRNAs were fragmented into small pieces using divalent cations under elevated temperature. The cleaved RNA fragments were copied into first strand cDNA using SuperScript II reverse transcriptase, Actinomycin D and random hexamer primers. The addition of Actinomycin D prevents spurious DNA-dependent synthesis, while allowing RNA-dependent synthesis, improving strand specificity. The second strand cDNA was synthesized by replacing dTTP with dUTP. The incorporation of dUTP quenches the second strand during amplification, ensuring that only the first cDNA strand is efficiently amplified [27, 28]. These cDNA fragments then had the addition of a single ‘A’ base and subsequent ligation of the adapter. The products were then purified and enriched with 15 cycles of PCR. The final cDNA libraries were validated with a DNA 1000 Labchip on a Bioanalyzer (Agilent) and quantified with a KAPA qPCR kit. For each sequencing lane of a flowcell V4, three to six libraries were pooled in equal proportions, denatured with NaOH and diluted to 7.5 pM or 8 pM before clustering. Cluster formation, primer hybridization and pair end-read 125 cycles sequencing were performed on cBot and HiSeq2500 (Illumina, San Diego, CA) respectively. Sequencing reads are available at the European Nucleotide Archive under study number PRJEB13600 (ERS1127228-ERS1127233; ERS1127235-ERS1127239; ERS1138027).

RNA-Seq data processing

The raw reads of RNA-seq data were processed using the galaxy.prabi.fr web service and the computing facilities of the LBBE/PRABI (Fig. 1) [2931]. To clean the sequences, the adapters that were ligated to the 5′ or 3′ end with a maximum allowed error rate of 0.1 and a minimum overlap length of 3 bp were removed using Cutadapt (version 1.6) [32]. Low quality sequences were then removed using Trimmomatic (version 0.32.1) [33] with the following explicit parameters: paired-end data: the minimum quality required to keep a base was 20 and the reads were trimmed when the average quality over a 4 bp window dropped below 22, starting from the 5′ end of the read. Moreover, trimmed/clipped sequences were kept only if they were at least 30 bp long. The reads were then aligned and assigned to the reference genome using the Bowtie 2 package (version 0.2) [34], with the following parameters: paired-end alignment modes with a maximum fragment length for valid paired-end alignments of 500 bp, and end-to-end read alignment. We then used the mapped files to run HTSeq-count (version 0.4.1) [35]. This script takes an alignment file in BAM format and a feature file in GFF format and calculates the number of reads mapping to each feature. The paired-end strand-specific RNA-Seq reads were assigned to these features based on their overlapping genomic coordinates and strand orientation. Reads with more than one reported alignment were not counted for any feature. Where the reads overlapped more than one feature, the intersection-nonempty mode was used. Counts of RNA-Seq fragments were computed for each coding DNA sequence (CDS) according to the genome annotations provided by the MicroScope platform (https://www.genoscope.cns.fr/agc/microscope/) (E. coli O26:H11 21,765: chromosome ECO26H.gff, plasmid ECO26H_p.gff; E. coli O157:H7 EDL933: chromosome NC_002655.gff, plasmid AF074613.gff) [16, 36]. Then the count tables generated by the HTSeq count were used as input to run DESeq2 (version 2.1.6.0) [37]. This tests for differential expression based on a model using the negative binomial distribution. The DESeq2 package provides its own normalization approach. The read count Kij for gene i in sample j is described with a generalized linear model (GLM). DESeq2 models read counts Kij with mean μij and dispersion αi. The mean is taken as a quantity qij, proportional to the concentration of cDNA fragments from the gene in the sample, scaled by a correction factor sij. To estimate these factors, the DESeq2 package uses the median-of-ratios method [38]. The principal idea is that non differential expression (DE) genes should have similar read counts across samples, leading to a ratio of 1. Supposing most genes are not DE, the median of this ratio for the sample provides an estimate of the normalization factor that should be accounted for differences in sequencing depth between samples. This method is advantageous to calculate gene-specific normalization factors sij to account for further sources of technical biases such as differing dependence on GC content or gene length. The False Discovery Rate (FDR) was controlled by adjusting p-values with the Benjamini Hochberg method [39]. In this report, only genes showing a ≥ 2 fold up-regulation or down-regulation with a minimum normalized read count = 10 and a Benjamini-Hochberg FDR adjusted p-value of ≤0.005 were considered to be differentially regulated. The transcript lists resulting from DESeq2 were then analyzed using the MicroScope platform, a web-based framework for functional analysis of large numbers of genes [40].

Results and discussion

Microbial community structures in raw ground beef inoculated or not with strains O2621765 or O157EDL933

Our experiment was designed so that EHEC strains would have sufficient time to adapt to the beef environment and that we could retrieve sufficient RNA(s) for analysis. A large initial EHEC inoculum (3 log greater than the natural microbiota counts) was chosen to ensure that the meat microbiota, present at an initial level of about 104 CFU/g, would not impair the growth of either O157EDL933 or O2621765 [19]. As expected, after 7 days of incubation, no flora were detected in the inner part of the muscle, used as control, whereas in the outer part about 8 log CFU/g of each type of bacteria (aerobic and LAB) were counted (Additional file 1: Table S1). In the non-inoculated ground meat samples (outer part of the muscle) commensal E. coli counts were about 3 logs below the level of the other natural microbiota species. In artificially contaminated minced meat obtained from the outer part, after 7 days’ incubation at 12 °C the commensal E. coli were no longer detectable. They had probably lost out in competition with the two EHEC strains (O157EDL933 and O2621765) inoculated at an initial level of about 107 CFU/g. The counts of both these EHEC strains were 1 log higher than the natural microbiota counts. The presence of the meat microbiota at an initial level of about 104 CFU/g did not impair the growth of either O157EDL933 or O2621765 (See Additional file 2: Table S2).

Using 16S metagenomics analysis enabled us to characterize the bacterial populations of our samples and evaluate their relative abundances after 7 days of incubation at 12 °C. A total of 2,852,952 bacterial 16S ribosomal RNA sequences, obtained from 3 samples (ground beef with natural microbiota and ground beef inoculated with strain O2621765 or strain O157EDL933) were analyzed. At the genus level, 2,827,664 sequences were classified (bootstrap value cutoff = 0.6). However, 0.9% of total sequences were either unclassified or classified at a higher taxonomic level. In the natural microbiota of the ground beef samples, our data clearly showed that Serratia dominated (79.4% incidence). Sequences belonging to the genus Carnobacterium and the enteric bacteria cluster amounted to 8.8% and 6.8% respectively. Carnobacterium is a psychrotrophic bacterium frequently isolated from foods, including meat [41]. Sequences from Kurthia and other bacteria were represented in smaller proportions in the control samples. In experimentally contaminated samples, Escherichia dominated the microbiota (74.9% for O2621765 and 76.4% for O157EDL933) and outnumbered Serratia, whose sequences were detected in less than 5% (Fig. 2; Additional file 3: Table S3). The number of sequences belonging to the genus Carnobacterium was not affected by the EHEC strains.
Fig. 2

Genus level distribution of sequences based on the 16S rRNA gene libraries constructed from a ground beef sample with natural microflora (a) and ground beef samples inoculated with E. coli O26:H11 21,765 (b) or E. coli O157:H7 EDL933 (c) strains. Only taxa with a percentages above 1% are displayed in the histograms

The analysis of the microbial composition of our samples is not discussed any further in this study. For our purpose, identification of the microbial community structure at the genus level was an essential preliminary for subtractively analyzing the gene expression of the EHEC strains. It was required to identify the genes of EHEC strains expected to be differentially expressed due to the possibility of spurious alignment and crosstalk between the EHEC genes and any other similar gene identified as being part of genome sequencing data provided by GenBank and corresponding to the most abundant genus identified by 16S metagenomics analysis (Fig. 1).

Global gene expression of EHEC strains O2621765 and O157EDL933 in ground meat with or without background microbiota

We set out to investigate the effects of ground beef background microbiota on the overall gene expression of EHEC strains O2621765 and O157EDL933, by comparison with their behavior in beef without microbiota. Even if eukaryotic rRNA acquired from bovine tissue cells could not be depleted owing to partial degradation, we were able to detect differential expression of numerous genes with high statistical significance thanks to the biological replicates and robust bacterial rRNA depletion. Bacterial rRNA depletion was nearly complete in all samples, since fewer than 2.3% of all high quality reads were aligned with bacterial rRNA-encoding genes (See Additional file 4: Table S4). Haas et al. (2012) has also claimed that changes in gene expression could be analyzed by RNA-Seq data analysis when the number of fragments per sample is reduced to 2–3 million [42].

Sequencing depths of 3.8 ± 1.5 and 4.3 ± 2.1 million gene mapped fragments were obtained in the O2621765 and O157EDL933 datasets respectively (Additional file 4: Table S4). Based on mean normalized counts of gene-mapped fragments, the numbers of ORFs with at least 10 fragments that DESeq2 detected were 5035 for O157EDL933 and 4865 for O2621765, amounting to 88.1% and 86.7% of their annotated genes respectively.

Analysis of gene expression with significant expression levels 2-fold and adjusted P-value 0.005 revealed that the majority of genes did not differ significantly between conditions. Strain O2621765 had a large number of significantly changed genes than O157EDL933. Overall, of 5611 coding sequences of strain O2621765, 95 were expressed at significantly lower levels in samples with microbiota compared to those without microbiota, and 28 were significantly up-regulated (Fig. 3; Tables 1 and 3; Additional file 5: Table S5 and Additional file 6: Table S6). The corresponding numbers for down- and up-regulated genes in strain O157EDL933 were 21 and 25 respectively (Fig. 4; Tables 2 and 4; Additional file 7: Table S7 and Additional file 8: Table S8).
Fig. 3

Summary of the main pathways and regulations affected by background microbiota in Enterohemorrhagic Escherichia coli O26:H11 strain 21,765 in ground beef. Green arrows with head indicate up-regulation and red arrows without head down-regulation. * Genes expected to be up-regulated owing to the possibility of spurious alignment and crosstalk between the reference genome and other similar genomes identified at genus level by 16S meta-genomic analysis, were discarded from the study (For discarded genes see Additional file 6: Table S6)

Table 1

Genes significantly down-regulated in Escherichia coli O26:H11 21,765 in samples with microbiota compared to those without microbiota

Locus (ECO26H_v1_ #)

Gene name

Mean of normalized counts

FCa

adj. p b

Function or product

Nitrate metabolism and nitrite detoxification

 180,052

hcr

90

−2.3

0.003

HCP oxidoreductase, NADH-dependent

 110,034

ybdL

1542

−2.1

3.3E-05

Methionine aminotransferase, PLP-dependent

 180,029

grxA

180

−2.1

0.0006

Glutaredoxin 1, redox coenzyme for ribonucleotide reductase (RNR1a)

 520,050

norR

126

−2.0

0.0002

NO-responsive regulator

790008

ytfE

47

−2.3

0.005

Regulator of cell morphogenesis and NO sensing

DNA repair

 740,026

nfi

590

−2.0

1.6E-10

Endonuclease V

Amino acid transport and metabolism

 10,028

carB

490

−2.3

3.8E-08

Carbamoyl-phosphate synthase, large subunit

 180,040

artJ

878

−4.6

6.2E-25

Arginine transporter subunit; periplasmic-binding component of ABC superfamily

 20,008

leuC

96

−2.1

0.0001

3-isopropylmalate dehydratase (isomerase), subunit with LeuD

 270,021

trpB

302

−2.0

0.0003

Tryptophan synthase, beta subunit

 270,022

trpC

244

−2.0

0.002

Fused indole-3-glycerolphosphate synthetase; N-(5-phosphoribosyl) anthranilate isomerase

 270,023

trpD

222

−2.1

0.0003

Fused glutamine amidotransferase (component II) of anthranilate synthase; anthranilate phosphoribosyl transferase

 270,024

trpE

173

−2.8

1.0E-06

Component I of anthranilate synthase

 40,007

metQ

954

−2.1

1.9E-05

DL-methionine transporter subunit; periplasmic-binding component of ABC superfamily

 40,009

metN

322

−2.8

8.1E-11

DL-methionine transporter subunit; ATP-binding component of ABC superfamily

 430,005

hisD

257

−2.3

1.0E-09

Bifunctional histidinal dehydrogenase and histidinol dehydrogenase

 430,006

hisC

59

−2.5

0.0004

Histidinol-phosphate aminotransferase

 530,047

argA

172

−2.6

6.7E-07

Fused acetylglutamate kinase homolog (inactive); amino acid N-acetyltransferase

 530,066

lysA

568

−2.1

4.9E-05

Diaminopimelate decarboxylase, PLP-binding

 610,023

argG

754

−2.5

6.0E-09

Argininosuccinate synthetase

 610,063

gltD

168

−2.0

6.2E-05

Glutamate synthase small chain

 660,006

metL

2959

−3.0

0.0003

Fused asparto kinase II; homoserine dehydrogenase II

 660,007

metB

251

−2.1

0.0003

Cystathionine gamma-synthase, PLP-dependent

 700,021

metE

3087

−2.5

2.1E-06

5-methyl tetra hydro pteroyl tri glutamate-homocysteine S-methyltransferase

 700,022

metR

275

−2.1

3.4E-05

DNA-binding transcriptional activator, homocysteine-binding

 700,077

ilvC

611

−3.2

4.8E-11

Ketol-acid reductoisomerase, NAD(P)-binding

 700,079

ilvA

444

−2.5

5.5E-09

Threonine deaminase

 700,081

ilvE

402

−2.0

1.1E-05

Branched-chain amino-acid aminotransferase

 700,083

ilvB(G)

273

−2.0

0.0002

Acetolactate synthase I, large subunit

 700,084

ilvL

5169

−2.6

4.9E-17

ilvG operon leader peptide

 710,012

asnA

3488

−2.1

1.1E-12

Asparagine synthetase A

 730,015

argC

450

--3.2

1.9E-15

N-acetyl-gamma-glutamylphosphate reductase, NAD(P)-binding

 730,016

argB

154

−2.6

2.6E-07

Acetylglutamate kinase

 730,017

argH

1040

−3.5

2.1E-13

Argininosuccinate lyase

 760,070

adiC

582

−3.7

2.1E-13

Arginine/agmatine antiporter

 760,072

adiA

224

−2.6

3.1E-08

Biodegradative arginine decarboxylase

 800,012

argI

112

−6.5

1.9E-20

Ornithine carbamoyltransferase 1

Carbohydrate metabolism and energy production

 470,020

glpA

363

−2.0

0.001

sn-glycerol-3-phosphate dehydrogenase (anaerobic), large subunit, FAD/NAD(P)-binding

 710,165

lldD

973

−2.1

1.1E-12

L-lactate dehydrogenase, FMN linked

 710,166

lldR

123

−2.5

6.5E-06

DNA-binding transcriptional repressor

 710,167

lldP

430

−2.1

8.1E-06

L-lactate permease

bacterial pathogenicity

 230,055

yccC

1840

−2.1

9.6E-07

Cryptic autophosphorylating protein tyrosine kinase Etk

 230,056

etp

214

−2.1

0.002

Phosphotyrosine-protein phosphatase

Inorganic ion transport and metabolism

 430,088

rncA

68

−2.0

0.003

Nickel and cobalt resistance

 560,020

fecR

116

−3.0

1.7E-08

Transmembrane signal transducer for ferric citrate transport; KpLE2 phage-like element

 620,109

feoB

997

−2.0

1.1E-05

Fused ferrous iron transporter, protein B: GTP-binding protein; membrane protein

 620,179

nikB

148

−2.1

0.0004

Nickel transporter subunit; membrane component of ABC superfamily

 620,180

nikC

113

−2.3

1.7E-05

Nickel transporter subunit; membrane component of ABC superfamily

 620,181

nikD

180

−2.0

0.0006

Nickel transporter subunit; ATP-binding component of ABC superfamily

Cell wall/membrane biogenesis

 20,017

ftsL

485

−2.3

1.3E-07

Membrane bound cell division protein at septum containing leucine zipper motif

 20,020

murF

436

−2.3

5.5E-08

UDP-N-acetylmuramoyl-tripeptide:D-alanyl-D-alanine ligase

Transcription

 180,070

lrp

2442

−3.0

1.7E-15

DNA-binding transcriptional dual regulator, leucine-binding

aFC is the fold change of the genes that exhibit significant (FC ≤ −2, false discovery rate (FDR) ≤ 0.005, minimum normalized read count = 10) differential expression. Only discussed genes in our study are shown on this table

bAdjusted p-value for multiple testing with the Benjamini-Hochberg procedure which controls FDR

Fig. 4

Summary of the main pathways and regulations affected by background microbiota in Enterohemorrhagic Escherichia coli O157:H7 strain EDL933 in ground beef. Green arrows with head indicate up-regulation and red arrows without head down-regulation. * Genes expected to be up-regulated owing to the possibility of spurious alignment and crosstalk between the reference genome and other similar genomes identified at genus level by 16S meta-genomic analysis, were discarded from the study (For discarded genes see Additional file 8: Table S8)

Table 2

Genes significantly down-regulated in Escherichia coli O157:H7 EDL933 in samples with microbiota compared to those without microbiota

Locus (Z #)

Gene name

Mean of normalized counts

FCa

adj. p b

Function or product

Nitrate metabolism and nitrite detoxification

 1107

hcp

270

−2.5

3.2E-08

Hybrid-cluster [4Fe-2S-2O] protein in anaerobic terminal reductases

Amino acid transport and metabolism

 0037

carA

369

−2.3

1.8E-07

Carbamoyl phosphate synthetase small subunit, glutamine amidotransferase

 1090

artJ

1305

−4.0

8.4E-15

Arginine transporter subunit; periplasmic-binding component of ABC superfamily

 1093

artI

407

−2.0

0.0002

Arginine transporter subunit; periplasmic-binding component of ABC superfamily

 3181

hisG

221

−2.0

0.001

ATP phosphoribosyltransferase

 3569

hisM

208

−2.0

0.0008

Histidine/lysine/arginine/ornithine transporter subunit; membrane component of ABC superfamily

 4135

argA

737

−2.5

2.7E-07

Fused acetylglutamate kinase homolog (inactive); amino acid N-acetyltransferase

 4534

argG

972

−3.0

9.9E-12

Argininosuccinate synthetase

 5285

ilvC

773

−2.0

2.1E-05

Ketol-acid reductoisomerase. NAD(P)-binding

 5515

argE

448

−2.5

2.8E-09

Acetylornithine deacetylase

 5516

argC

1480

−3.7

1.4E-16

N-acetyl-gamma-glutamylphosphate reductase, NAD(P)-binding

 5517

argB

533

−2.8

1.1E-09

Acetylglutamate kinase

 5518

argH

1496

−2.5

7.2E-06

Argininosuccinate lyase

 5719

adiA

539

−2.1

0.001

Biodegradative arginine decarboxylase

 5866

argI

404

−2.5

0.0006

Ornithine carbamoyltransferase 1

bacterial pathogenicity

 3200

peR

739

−2.0

6.9E-05

Perosamine synthetase Per

aFC is the fold change of the genes that exhibit significant (FC ≤ −2, false discovery rate (FDR) ≤ 0.005, minimum normalized read count = 10) differential expression. Only genes discussed in our study are shown on this table

bAdjusted p-value for multiple testing with the Benjamini-Hochberg procedure which controls FDR

Genes significantly down-regulated in ground beef with microbiota

Genes involved in amino acid transport and biosynthesis

For O2621765, 31 of the 97 genes known, in E. coli, to encode the enzymes needed for amino acid biosynthesis [43] were down-regulated in ground meat with microbiota (Fig. 3; Table 1). For example, genes implicated in tryptophan biosynthesis (trpE, trpD, trpC, trpB and trpA), methionine biosynthesis, transport or regulation (metBELQNR), biosynthesis of glutamate (gltD), lysine (lysA), and arginine (argABCIHG and carB) and genes which encode the enzymes of the isoleucine and valine synthesis pathway (ilvGEALG) were down-regulated. For O157EDL933, eight genes implicated in the arginine biosynthesis pathway were down-regulated (argABCEIHG and carA) (Table 2). The artI gene encoding arginine transporter subunit and the hisM gene encoding histidine/lysine/arginine/ornithine transporter subunit were also down-regulated. The hisCD and hisG genes of the histidine biosynthetic pathway were down-regulated in strains O2621765 and O157EDL933 respectively. The ilvC and artJ genes, which encode Ketol-acid reductoisomerase and arginine transporter subunit respectively, were down-regulated in both strains. The adiC and adiA genes were also both down-regulated in O2621765. Arginine produced or imported into the cell by AdiC is decarboxylated inside the cell by AdiA to form agmatine, releasing CO2 and replacing it with a proton. The leuC and asnA genes, which encode 3-isopropylmalate dehydratase and asparagine synthetase A respectively, were also down-regulated in O2621765. The lrp gene encoding leucine-responsive regulatory protein (Lrp) which is a global regulator of E. coli metabolism was down-regulated in O2621765. The repression of this regulatory gene correlated well with the reduced transcript levels of its target genes, notably the decreased expression of several genes of the amino acid biosynthesis pathways (Table 1). Expression of the lrp gene is known to be regulated in part by the nutrients available to the cell, and is decreased in glucose minimal media enriched with amino acids [44]. The repression of genes involved in amino acid biosynthesis can be explained by the repression of lrp in ground meat with microbiota. The hypothesis proposed is that the hydrolyzed proteins produced by the activity of proteolytic enzymes and peptidases of the background microbiota may enrich the meat with peptides and amino acids and so inhibit the expression of lrp. These data are in accordance with a study led by Tao et al. (1999) which indicated that the expression of many genes involved in the biosynthesis of building blocks, most notably the amino acid biosynthesis pathways, is repressed in high-nutrient environments, as appears to be the case in the present study.

Genes involved in carbohydrate metabolism and energy production

Genes implicated in L-lactate metabolism were down-regulated in O2621765 (lldR, lldP and lldD). L-lactate dehydrogenase (encoded by lldD) is a peripheral membrane protein that catalyzes the oxidation of L-lactate to pyruvate. This protein and L-lactate permease (encoded by the lldP gene) allow E. coli to grow in a medium containing L-lactate as the sole carbon source [45]. Transcription of glpA, which encodes a sn-glycerol-3-phosphate (G3P) dehydrogenase, GlpA, was also repressed in O2621765 (Table 1). Under anaerobic conditions, G3P dehydrogenase GlpA converts G3P to dihydroxyacetone phosphate [46]. This latter compound is converted into glyceraldehyde-3P entering the gluconeogenesis pathway [47]. Thus, in ground meat, the presence of microbiota seems to repress the transcription of several genes involved in the assimilation of gluconeogenic substrates such as glycerol-3P and L-lactate. These metabolites could be consumed by the background microbiota.

Genes implicated in cell wall/membrane biogenesis

The murF gene encoding the MurF amide ligase enzyme was repressed in O2621765 (Table 1). MurF has been shown to be required for catalyzing the final step in the synthesis of bacterial cell wall peptidoglycan and to be essential for bacterial survival [4851]. The ftsL gene encoding the conserved division protein FtsL was also down-regulated in O2621765 (Table 1). FtsL, rather than serving simply as a protein scaffold within the divisome along with its partners FtsB and FtsQ, functions as part of a sensing mechanism that promotes the onset of cell wall remodeling processes needed for the initiation of cell constriction once assembly of the divisome complex is deemed complete [5254].

In the present study we demonstrated that murF and ftsL genes were down-regulated in strain O2621765 in meat with microbiota. The microbiota may exert an inhibitory action on EHEC strains by interfering with the cytokinetic ring apparatus and the synthesis of bacterial cell wall peptidoglycan. Many studies have shown that the growth of EHEC strains in ground beef may be inhibited by background microbiota initially present at levels greater than that of the pathogen E. coli [55, 56]. In our study, the low level of natural microbiota compared to strain O2621765 might not sufficiently affect the expression of murF and ftsL genes to halt growth. Both strains (O157EDL933 and O2621765) were added at an initial level of about 3 logs above that of the natural microbiota. After 7 days of incubation at 12 °C, about 8 log CFU/g of each type of natural microbiota (aerobic and LAB) were counted whereas 9 log CFU/g of each EHEC strains were enumerated. In the tested conditions, the presence of the meat microbiota at an initial level of about 104 CFU/g did not impair the growth of either O157EDL933 or O2621765 (See Additional file 1: Table S1; Additional file 2: Table S2).

Genes involved in nitrate metabolism and nitrite detoxification

The mean raw nitrate content of fresh meat ranged from 18.7 to 38.5 mg/kg [57]. Microorganisms can reduce nitrate to nitrite, which is partially oxidized to nitrate by sequestering oxygen [58]. The nitrite undergoes chemical reactions that lead to reactive nitrogen species (RNS), including NO [58, 59]. These RNSs can interact with and damage numerous targets, including tyrosine residues, thiols and nucleotide bases [60]. Two down-regulated genes, hcr and hcp, were identified in strains O2621765 and O157EDL933 respectively (Tables 1 and 2). They encode the hybrid cluster protein (HCP) and its NADH oxidoreductase (HCR), respectively, which are supposed to be involved in nitrite or nitrate reduction by an as yet unidentified reaction [61]. In E. coli, hcp has been found to be induced in presence of nitric oxide (NO) under both aerobic and anaerobic conditions [62]. Because of the cytotoxic effects of NO and its derivatives, E. coli has two prominent ways to detoxify NO: a nitric-oxide reductase (NorVW) and the flavohemoglobin Hmp. The norVW genes encode flavorubredoxin and flavorubredoxin reductase, respectively, which convert NO to nitrous oxide. In O157EDL933, the norV gene has a 154 bp deletion, resulting in a frameshift mutation. In O2621765, the norR gene, encoding an NO-responsive regulator (NorR) [63, 64], was down-regulated (Table 1).

In our study the ytfE gene was down-regulated in O2621765. The ytfE gene is predicted to encode for a cytoplasmic protein, and is annotated as a regulator of cell morphology and NO sensing [16, 65]. Those authors also showed that YtfE is of major importance in E. coli’s response to NO [65].

Other genes that have shown up-regulation in response to nitrosative stress [62, 66] were down-regulated in O2621765. These were the four metBELR genes, involved in methionine biosynthesis or regulation; the gltD gene, which encodes glutamate synthase (NADPH) small chain precursor; grxA, encoding glutaredoxin 1 redox coenzyme ribonucleotide reductase; and ybdL, encoding a PLP-dependent methionine aminotransferase (Table 1).

Expression of the nfi gene, encoding endonuclease V (EndoV), was also repressed in O2621765 grown in ground beef with microbiota. EndoV is a ubiquitous protein responsible for the specific cleavage at the second phosphodiester bond 3′ to inosine [67, 68]. In E. coli, many studies have shown that EndoV prevents mutations from nitrosative deamination during nitrate/nitrite respiration [69].

In ground beef with microbiota, raw nitrate content could be metabolized by background microbiota in the meat samples prior to their contamination with EHECs. The up-regulated NO response mechanisms in ground beef without microbiota may allow the pathogen to adapt to sublethal environmental conditions and enhance its resistance to chemicals typically used as preservatives.

Genes implicated in the bacterial pathogenicity of EHECs

The peR gene was down-regulated in strain O157EDL933 (Table 2). This gene encodes a perosamine synthetase essential for the production of O antigen. As described by Bilge et al., (1996), an O157:H7∆peR mutant is significantly more adherent to HeLa cells than is its E. coli O157:H7 parent [70], the mutant strain being deficient in the O antigen. The O side chains of bacterial lipopolysaccharide (LPS) may physically hinder contact between the outer membrane and eukaryotic cells. Because of the conserved nature of most EHEC O157 strains, such factors might explain the higher prevalence of O157 EHECs in bovine meat (Zoonose EFSA, 2014). Nevertheless, more data are needed on other O157 EHEC strains.

With O2621765, the etk and etp genes, which encode a protein-tyrosine kinase, Etk, and a phosphotyrosine-protein phosphatase, Etp, respectively, were down-regulated (Table 1). Various studies have shown that pathogenicity seems to be positively correlated with the level of tyrosine phosphorylation [71, 72]. Increasing evidence supports the idea that tyrosine phosphorylation contributes to several key steps in the infection process, such as adhesion to the host and regulation of pathogenic functions [7375].

In our study we demonstrated that the expression of the O side chains of bacterial LPS in O157EDL933, and the level of tyrosine phosphorylation in O2621765 were down-regulated in meat with microbiota. Various studies have shown that, the O side chains of bacterial LPS and the level of tyrosine phosphorylation are correlated with the capacity of bacteria to adhere to the eukaryotic cells. Based on these studies, our hypothesis supports the idea that strain O2621765 might adhere less well to beef tissue with microbiota than O157EDL933. These data are in accordance with the higher prevalence of O157 compared to O26 EHECs in ground beef.

Genes implicated in inorganic ion transport and metabolism

Iron, nickel and other inorganic ions have various roles in cellular biochemistry. In particular, they are well known as co-factors with a key role in bacterial virulence. For example, it has been established that iron acquisition genes are important contributors to the virulence of uropathogenic E. coli (UPECs) [76]. In our study, several relevant genes were down-regulated in O2621765: the nikBCD genes, encoding one part of the E. coli NikABCDE permease. This permease is known to be maximally expressed under anaerobic condition when intracellular nickel is scarce [77, 78]; the rcnA gene, encoding a membrane bound polypeptide, RcnA, described as a cobalt and nickel efflux system in E. coli, [79]; the feoB gene, encoding the polytopic membrane protein, FeoB, which is essential for Fe(II) uptake [80] (note that feoB has been described as one of the most prevalent virulence genes in UPECs isolated from patients with community-acquired urinary tract infection [81]); and the fecR gene, encoding the FecR protein, the sensor that recognizes iron (III) dicitrate in the periplasm (Table 1) [82].

The results of the transcriptomic analysis of O2621765 suggest that iron and nickel could be more bioavailable in ground beef with microbiota and their transporters might be repressed in response to these external stimuli.

Other down-regulated genes

Three (ECO26H_v1_590016; 690,023 and 710,001) and two (Z4363 and 5408) genes annotated as putative transcriptional regulators were down-regulated respectively in O2621765 and O157EDL933 strains (Fig. 3; Fig. 4; Additional file 5: Table S5 and Additional file 7: Table S7). Among them, one (yihL) is shared by the two strains. Several other genes (ECO26H_v1_110033; 200,020; 370,007; 430,155; 710,101 and 750,064 in O2621765; Z1966 and 3199 in O157EDL933) were also down-regulated (Additional file 5: Table S5 and Additional file 7: Table S7). Their probable effects as a response to an environmental signal were not discussed in the present study because their functional role has not yet been fully characterized. Moreover, their annotation is based on the presence of conserved amino acid motifs and structural features or limited homology.

Eleven other genes (ECO26H_v1_120023; 210,002; 220,017; 340,120; 430,059; 500,297; 580,076; 60,012; 670,016; 80,078 and p30071), annotated as hypothetical or conserved proteins of unknown function, were down-regulated in the O2621765 strain (Additional file 5: Table S5). Only one (ECO26H_v1_ p30071) did not show any significant sequence identity with any of the known genes in the O157EDL933 strain.

Genes significantly up-regulated in ground meat with microbiota

Preamble. This is the first study to use strand-specific RNA-seq to investigate the effects of ground beef background microbiota on the overall gene expression of EHEC strains O2621765 and O157EDL933. To reduce the effect of spurious alignment and crosstalk between the reference genome of EHEC and other similar genomes, the genes that were transcriptionally up-regulated in the presence of microbiota were filtered for E. coli specific genes (Fig. 1). The down-regulated genes we identified could not be affected by spurious alignment or crosstalk between the reference genomes and other similar genomes, so it was not necessary to apply this filter. A spurious alignment could not generate a false down-regulated gene.

We therefore discarded from our study genes that we expected to be up-regulated in EHEC strains owing to the possibility of spurious alignment and crosstalk between the reference genomes and other similar genomes identified at the genus level by 16S metagenomics analysis. The criterion whether or not to discard an up-regulated gene was a score of nucleotide identity (between the altered gene and any other gene provided by GenBank as being part of the genome sequencing data of the identified genus by 16S meta-genomic analysis) of more than 50% across 80% or more of their length.

This constraint also influenced our sequencing strategy, sequencing mode (paired-end sequencing), library-type (strand-specific cDNA library) and choice of the read length (125 bp). The paired-end strand-specific RNA-Seq with a read length of 125 bp allows for higher levels of specific alignments, thus reducing the possibility of spurious alignment between the reference genome and other similar genomes. Reads with more than one reported alignment were not counted for any feature. We also discarded up-regulated genes with identity to any other gene identified as being part of genome sequencing data provided by GenBank and corresponding to the most abundant bacteria identified by 16S metagenomics analysis (Figs. 3 and 4). We discarded genes with more than 50% nucleotide identity across 80% or more of their length (For discarded genes see Additional file 6: Table S6 and Additional file 8: Table S8). Of the 28 and 35 genes induced in strains O2621765 and O157EDL933 respectively, 14 and 12 passed the filter and are discussed in this study (Tables 3 and 4).
Table 3

Genes significantly up-regulated in Escherichia coli O26:H11 21,765 in samples with microbiota compared to those without microbiota

Locus (ECO26H_v1_ #)

Gene name

Mean of normalized counts

FCa

adj. p b

Function or product

Unknown or hypothetical proteins

260022

ymgB

101

2.1

0.0001

Hypothetical protein

310027

ydfP

34

2.5

0.003

Conserved hypothetical protein; Qin prophage

340082

ydhS

506

2.0

0.0002

Conserved hypothetical protein; putative FAD/NAD(P)-binding domain

400008

yecT

105

2.1

0.0002

Hypothetical protein

420013

yeeE

55

2.3

0.001

Conserved hypothetical protein; putative inner membrane protein

50014

_

102

2.1

0.002

Conserved protein of unknown function

510037

_

162

2.5

3.6E-06

Protein of unknown function

590009

_

61

2.1

0.004

Conserved protein of unknown function

p40014

_

591

2.0

0.0002

Conserved protein of unknown function

Other up-regulated genes

250005

_

48

2.3

0.001

T3SS secreted effector NleG-like protein

270179

ydcX

105

2.3

8.8E-06

Putative inner membrane protein

500243

yfhR

250

2.0

7.1E-05

Putative peptidase

50040

yhhI

106

2.1

0.0007

Putative transposase

p40035

_

99

2.3

0.0002

Pilus assembly protein

aFC is the fold change of the genes that exhibit significant (FC ≥ 2, false discovery rate (FDR) ≤ 0.005, minimum normalized read count = 10) differential expression. Only genes passed the filter are shown on this table. Expected genes to be differentially expressed due to the possibility of spurious alignment, and crosstalk between the reference genome and other similar genomes identified at genus level by 16S meta-genomic analysis, were discarded from this study (For discarded genes see Additional file 6: Table S6)

bAdjusted p-value for multiple testing with the Benjamini-Hochberg procedure which controls FDR

Table 4

Genes significantly up-regulated in E. coli O157:H7 EDL933 in samples with microbiota compared to those without microbiota

Locus (Z #)

Gene name

Mean of normalized counts

FCa

adj. p b

Function or product

Antibiotic resistance

2169

marB

78

2.6

8.2E-06

Conserved hypothetical protein associated with multiple antibiotic resistance operon

2170

marA

133

2.1

2.1E-05

DNA-binding transcriptional dual activator of multiple antibiotic resistance

Response to stress

2478

pspD

74

2.6

2.2E-05

Peripheral inner membrane phage-shock protein

Hypothetical proteins

1402

ymcB

102

2.1

0.0004

Conserved hypothetical protein

2695

ydhS

1115

2.0

0.0001

Conserved hypothetical protein; putative FAD/NAD(P)-binding domain

2774

ydjR

206

2.3

0.0004

Conserved hypothetical protein

3175

yeeE

143

2.8

1.7E-07

Conserved hypothetical protein; putative inner membrane protein

5712

yjcZ

173

2.1

0.0002

Conserved hypothetical protein

Other up-regulated genes

0043

caiC

204

2.0

0.0005

Putative crotonobetaine CoA ligase:carnitine CoA ligase

1987

ldrB

176

2.0

0.001

Fragment of small toxic polypeptide (partial)

3266

yegX

100

2.3

0.0002

Putative membrane-bound hydrolase

3660

yfeA

3551

2.0

2.9E-05

Putative diguanylate cyclase

aFC is the fold change of the genes that exhibit significant (FC ≥ 2, false discovery rate (FDR) ≤ 0.005, minimum normalized read count = 10) differential expression. Only genes passed the filter are shown on this table. Expected genes to be differentially expressed due to the possibility of spurious alignment, and crosstalk between the reference genome and other similar genomes identified at genus level by 16S meta-genomic analysis, were discarded from this study (For discarded genes see Additional file 8: Table S8)

bAdjusted p-value for multiple testing with the Benjamini-Hochberg procedure which controls FDR

Genes implicated in a detoxification role in EHECs

For O157EDL933, the marA and marB genes, which encode a DNA-binding transcriptional dual activator of multiple antibiotic resistance and a conserved hypothetical protein associated with a multiple antibiotic resistance operon, respectively, were up-regulated (Table 4). The expression of these genes as part of the antibiotic resistance machinery could be activated in response to external stimuli [83], presence of an antibiotic at a sub-lethal concentration being the most common stimulus. In our 16S metagenomics study, the genus Serratia represented about 1.5% of the bacterial population in the ground beef sample inoculated with O157EDL933 (Fig. 2; Additional file 3: Table S3). Most Serratia spp. have been shown to produce antibiotics. For example, production of carbapenem antibiotics has been demonstrated in Serratia sp. ATCC 39006 [84]. Production of antibiotics has also been reported in Serratia plymuthica RVH1 [85] and Serratia marcescens strain 12 [86]. These observations suggest that the presence of ground beef background microbiota might induce multidrug resistance in EHECs by stimulating transcription of E. coli drug resistance genes such as marA, which has been described as sufficient to give E. coli multiple antibiotic resistance [87].

Genes implicated in stress response

The pspD gene encoding the peripheral inner membrane phage-shock protein pspD was up-regulated in O157EDL933 (Table 4). Expression of the pspD gene is known to be associated with membrane stress and in response to conditions such as phage attack, heat shock and hyperosmotic stress, and exposure to certain metabolites [88, 89]. Thus, the presence of adventitious microbial activity or phage could induce specific stresses in E. coli. However, this needs further investigation.

Other up-regulated genes

Several other genes were up-regulated (ECO26H_v1_250,005; p40035; 500,243; 50,040 and Z0043; 3266 respectively, in the O2621765 and O157EDL933 strains), but their functional roles, based on the presence of conserved amino acid motifs, structural features or limited identity, have not yet been fully characterized (Tables 3 and 4). The effect of their over-expression in response to an environmental signal needs to be identified and they are now new candidates for detailed functional description.

Genes encoding hypothetical proteins

Of the 14 and 12 genes studied, 9 and 5 were annotated as unknown or hypothetical proteins in O2621765 and O157EDL933 strains, respectively (Tables 3 and 4). Two genes (yeeE and ydhS) were shared between the two strains. Six other (ECO26H_v1_260,022; 310,027; 50,014; 510,037; 590,009 and p40014) identified in O2621765 strain didn’t show any significant sequence identity with any gene of the other one. In EHEC, about one third of the genes are still annotated as hypothetical [89]. Genes annotated as hypothetical and up-regulated in presence of naturel microbiota, are now new candidates for a detailed functional description.

Conclusions

This article demonstrates how RNA-Seq coupled with 16S metagenomics analysis can be used to discover new and relevant impacts on the functions of an individual microbe within a complex environmental community. The methodology developed was designed for describing the effects of ground beef background microbiota on the overall gene expression of EHEC strains O2621765 and O157EDL933 by comparison with their behavior in beef with no microbiota. For our experimental design, the choice of two strains, one isolated from Michigan ground beef (O157EDL933) and one from human fecal samples (O2621765), enabled us to gain information of continuing importance for understanding the biology of EHECs in general and the strain-specific adaptations of strains O157EDL933 and O2621765 in particular.

In strain O2621765, 95 genes were expressed at significantly lower levels in ground beef samples with microbiota, and 28 were significantly up-regulated. The corresponding numbers for down- and up-regulated genes in O157EDL933 were 21 and 25. Together, all the genes identified in the present study serve as a starting point for developing testable hypotheses about the mechanisms underlying the organism’s adaptations to different habitats, including the evolution of its virulence. Moreover, genes that are altered in presence of natural microbiota, notably genes for the category of the hypothetical proteins, are now new targets for studying interactions between microorganisms. Finally, murF and ftsL exemplarily show that transcriptome profiling will be a powerful technique for designing novel therapeutic approaches, since treating an EHEC infection with antibiotics can potentiate the disease.

Tremendous diversity exists within the meat microbiota and changing experimental conditions such as storage conditions will alter the microbial population [90, 91] and so change its interaction mechanisms. As this new methodology is not tied to particular conditions, it can be used with any other conditions, where it would screen for new interaction mechanisms.

Abbreviations

BHI: 

Brain heart infusion

bp: 

Base pairs

BPW: 

Buffered peptone

cDNA: 

Complementary DNA

CDS: 

Coding DNA sequence

CFU: 

Colony-forming units

DE: 

Differential expression

DNA: 

Deoxyribonucleic acid

E.

coli

EFSA: 

European food safety authority

EHEC: 

Enterohemorrhagic Escherichia coli

EU: 

European Union

FDR: 

False discovery rate

GLM: 

Generalized linear model

HCP: 

Hybrid cluster protein

HUS: 

Hemolytic-uremic syndrome

LAB: 

Lactic acid bacteria

LPS: 

LipoPolySaccharide

NO: 

Nitric oxide

ORF: 

Open reading frame

OUT: 

Operational taxonomic unit

PCA: 

Plate-count agar

PCR: 

Polymerase chain reaction

RNS: 

Reactive nitrogen species

rRNA: 

Ribosomal ribonucleic acid

RTE: 

Ready-To-Eat

STEC: 

Shiga toxine E. coli

UPEC: 

UroPathogenic

Declarations

Acknowledgments

The authors are grateful to the LABGeM team and the National Infrastructure “France Genomique” for the functional annotation. This work was performed using the galaxy.prabi.fr web service and the computing facilities of the LBBE/PRABI. The authors thank Harriet Coleman for English proofreading. The authors also thank Patricia Mariani (Hospital Robert Debré, France) who kindly provided the strain O26:H11 21765.

Funding

This study was supported by “FEDER (Fonds Européen de Développement Régional) and CPER (Contrat de Plan État-Région) ASTERisk (Auvergne STEC Risque)”, Auvergne region. This work was also supported by the institute of VetAgro Sup. The funding bodies had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Delphine Thevenot-Sergentet obtained additive funding for performing experiments. Francoise Leriche, Stéphanie Blanquet-Diot and Delphine Thevenot-Sergentet obtained funding for Postdoc salary.

Availability of data and materials

Sequencing reads from RNA-seq analysis are available at the European Nucleotide Archive under study number PRJEB13600 (ERS1127228-ERS1127233; ERS1127235-ERS1127239; ERS1138027). Sequencing reads from 16S metagenomics analysis are available at the European Nucleotide Archive under study number PRJEB13580 (ERS1119543-ERS1119545).

Authors’ contributions

DS, FL, SB and WG involved in steering committee of the project. DS and WG conceived and designed the experiments. WG performed the experiments. WG and CG prepared the samples. SC, VN and AD provided tools for bioinformatics analysis. WG performed bioinformatics analyses, interpreted the data and wrote the manuscript. FL, DS and SB significantly contributed to the revision of the manuscript. All authors participated in the acquisition of data and the revision of the manuscript. All authors read and approved the final version to be published. All authors agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Ethics approval and consent to participate

All data were collected in accordance with the European Parliament and Council decision for the epidemiological surveillance and control of communicable disease in the European community. Ethical approval and informed consent were not required.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Authors’ Affiliations

(1)
UMR 5557 Ecologie Microbienne, Research Group on Bacterial Opportunistic Pathogens and Environment, CNRS, VetAgro Sup and Université de Lyon
(2)
Université Clermont Auvergne, INRA, UMRF
(3)
UMR UCA INRA 454 MEDIS Microbiota Digestive environment and Health, Université Clermont Auvergne
(4)
VetAgro Sup, Campus Agronomique de Lempdes
(5)
Alternative Energies and Atomic Energy Commission (CEA), Genomic Institute Genoscope & CNRS-UMR8030 & Evry University, Laboratory of Bioinformatics Analysis in Genomics and Metabolism
(6)
PRABI, Rhône Alpes Bioinformatics Center, UCBL, Lyon1, Université de Lyon
(7)
UMR 5557 Ecologie Microbienne, CNRS, Université de Lyon
(8)
Reference Laboratory for Escherichia coli including Shiga Toxin-Producing E. coli, VetAgro Sup, Campus Vétérinaire de Lyon, Université de Lyon

References

  1. Van Vliet AHM. Next generation sequencing of microbial transcriptomes: challenges and opportunities. FEMS Microbiol Lett. 2010;302:1–7.View ArticlePubMedGoogle Scholar
  2. Creecy JP, Conway T. Quantitative bacterial transcriptomics with RNA-seq. Curr Opin Microbiol 2015;0:133–140.Google Scholar
  3. Frias-Lopez J, Shi Y, Tyson GW, Coleman ML, Schuster SC, Chisholm SW, et al. Microbial community gene expression in ocean surface waters. Proc Natl Acad Sci U S A. 2008;105:3805–10.View ArticlePubMedPubMed CentralGoogle Scholar
  4. Gilbert JA, Field D, Huang Y, Edwards R, Li W, Gilna P, et al. Detection of large numbers of novel sequences in the Metatranscriptomes of complex marine microbial communities. PLoS One. 2008;3:e3042.View ArticlePubMedPubMed CentralGoogle Scholar
  5. Rey FE, Faith JJ, Bain J, Muehlbauer MJ, Stevens RD, Newgard CB, et al. Dissecting the in vivo metabolic potential of two human gut Acetogens. J Biol Chem. 2010;285:22082–90.View ArticlePubMedPubMed CentralGoogle Scholar
  6. Turnbaugh PJ, Quince C, Faith JJ, McHardy AC, Yatsunenko T, Niazi F, et al. Organismal, genetic, and transcriptional variation in the deeply sequenced gut microbiomes of identical twins. Proc Natl Acad Sci U S A. 2010;107:7503–8.View ArticlePubMedPubMed CentralGoogle Scholar
  7. Rowe PC, Orrbine E, Lior H, Wells GA, Yetisir E, Clulow M, et al. Risk of hemolytic uremic syndrome after sporadic Escherichia Coli O157:H7 infection: results of a Canadian collaborative study. J Pediatr. 1998;132:777–82.View ArticlePubMedGoogle Scholar
  8. Tserenpuntsag B, Chang H-G, Smith PF, Morse DL. Hemolytic uremic syndrome risk and Escherichia Coli O157:H7. Emerg Infect Dis. 2005;11:1955–7.View ArticlePubMedPubMed CentralGoogle Scholar
  9. Sarimehmetoglu B, Aksoy MH, Ayaz ND, Ayaz Y, Kuplulu O, Kaplan YZ. Detection of Escherichia Coli O157:H7 in ground beef using immunomagnetic separation and multiplex PCR. Food Control. 2009;20:357–61.View ArticleGoogle Scholar
  10. Renter DG, Sargeant JM, Oberst RD, Samadpour M. Diversity, frequency, and persistence of Escherichia Coli O157 strains from range cattle environments. Appl Environ Microbiol. 2003;69:542–7.View ArticlePubMedPubMed CentralGoogle Scholar
  11. Barkocy-Gallagher GA, Arthur TM, Rivera-Betancourt M, Nou X, Shackelford SD, Wheeler TL, et al. Seasonal prevalence of Shiga toxin-producing Escherichia Coli, including O157:H7 and non-O157 serotypes, and salmonella in commercial beef processing plants. J Food Prot. 2003;66:1978–86.View ArticlePubMedGoogle Scholar
  12. European Food Safety Authority, European Centre for Disease Prevention and Control (ECDC). The European Union summary report on trends and sources of zoonoses, zoonotic agents and food‐borne outbreaks in 2014. EFSA J. [Internet]. 2015 [cited 2017 Jul 28];13. Available from: http://doi.wiley.com/10.2903/j.efsa.2015.4329.Google Scholar
  13. King LA, Loukiadis E, Mariani-Kurkdjian P, Haeghebaert S, Weill F-X, Baliere C, et al. Foodborne transmission of sorbitol-fermenting Escherichia Coli O157:[H7] via ground beef: an outbreak in northern France, 2011. Clin. Microbiol. Infect. Off. Publ. Eur. Soc. Clin. Microbiol. Infect. Dis. 2014;20:O1136–44.Google Scholar
  14. Kostrzynska M, Bachand A. Use of microbial antagonism to reduce pathogen levels on produce and meat products: a review. Can J Microbiol. 2006;52:1017–26.View ArticlePubMedGoogle Scholar
  15. Hartmann HA, Wilke T, Erdmann R. Efficacy of bacteriocin-containing cell-free culture supernatants from lactic acid bacteria to control listeria monocytogenes in food. Int J Food Microbiol. 2011;146:192–9.View ArticlePubMedGoogle Scholar
  16. Galia W, Mariani-Kurkdjian P, Loukiadis E, Blanquet-Diot S, Leriche F, Brugère H, et al. Genome sequence and annotation of a human infection isolate of Escherichia Coli O26:H11 involved in a raw milk cheese outbreak. Genome Announc. 2015;3Google Scholar
  17. Riley LW, Remis RS, Helgerson SD, McGee HB, Wells JG, Davis BR, et al. Hemorrhagic colitis associated with a rare Escherichia Coli serotype. N Engl J Med. 1983;308:681–5.View ArticlePubMedGoogle Scholar
  18. Perelle S, Dilasser F, Grout J, Fach P. Detection by 5′-nuclease PCR of Shiga-toxin producing Escherichia Coli O26, O55, O91, O103, O111, O113, O145 and O157:H7, associated with the world’s most frequent clinical cases. Mol Cell Probes. 2004;18:185–92.View ArticlePubMedGoogle Scholar
  19. Berry ED, Koohmaraie M. Effect of different levels of beef bacterial microflora on the growth and survival of Escherichia Coli O157:H7 on beef carcass tissue. J Food Prot. 2001;64:1138–44.View ArticlePubMedGoogle Scholar
  20. Andreotti R, Pérez de León AA, Dowd SE, Guerrero FD, Bendele KG, Scoles GA. Assessment of bacterial diversity in the cattle tick Rhipicephalus (Boophilus) microplusthrough tag-encoded pyrosequencing. BMC Microbiol. 2011;11:6.View ArticlePubMedPubMed CentralGoogle Scholar
  21. Schloss PD, Westcott SL, Ryabin T, Hall JR, Hartmann M, Hollister EB, et al. Introducing mothur: open-source, platform-independent, community-supported software for describing and comparing microbial communities. Appl Environ Microbiol. 2009;75:7537–41.View ArticlePubMedPubMed CentralGoogle Scholar
  22. Pruesse E, Quast C, Knittel K, Fuchs BM, Ludwig W, Peplies J, et al. SILVA: a comprehensive online resource for quality checked and aligned ribosomal RNA sequence data compatible with ARB. Nucleic Acids Res. 2007;35:7188–96.View ArticlePubMedPubMed CentralGoogle Scholar
  23. Bacci G, Bani A, Bazzicalupo M, Ceccherini MT, Galardini M, Nannipieri P, et al. Evaluation of the performances of ribosomal database project (RDP) classifier for taxonomic assignment of 16S rRNA Metabarcoding sequences generated from Illumina-Solexa NGS. J Genomics. 2015;3:36–9.View ArticlePubMedPubMed CentralGoogle Scholar
  24. Wang Q, Garrity GM, Tiedje JM, Cole JR. Naive Bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy. Appl Environ Microbiol. 2007;73:5261–7.View ArticlePubMedPubMed CentralGoogle Scholar
  25. Nielsen EM, Andersen MT. Detection and characterization of verocytotoxin-producing Escherichia Coli by automated 5′ nuclease PCR assay. J Clin Microbiol. 2003;41:2884–93.View ArticlePubMedPubMed CentralGoogle Scholar
  26. Schroeder A, Mueller O, Stocker S, Salowsky R, Leiber M, Gassmann M, et al. The RIN: an RNA integrity number for assigning integrity values to RNA measurements. BMC Mol Biol. 2006;7:3.View ArticlePubMedPubMed CentralGoogle Scholar
  27. Levin JZ, Yassour M, Adiconis X, Nusbaum C, Thompson DA, Friedman N, et al. Comprehensive comparative analysis of strand-specific RNA sequencing methods. Nat Methods. 2010;7:709–15.View ArticlePubMedPubMed CentralGoogle Scholar
  28. Parkhomchuk D, Borodina T, Amstislavskiy V, Banaru M, Hallen L, Krobitsch S, et al. Transcriptome analysis by strand-specific sequencing of complementary DNA. Nucleic Acids Res. 2009;37:e123.View ArticlePubMedPubMed CentralGoogle Scholar
  29. Blankenberg D, Von Kuster G, Coraor N, Ananda G, Lazarus R, Mangan M, et al. Galaxy: a web-based genome analysis tool for experimentalists. In: Ausubel Al FM, editor. Curr. Protoc. Mol. Biol. Ed; 2010. Chapter 19:Unit 19.10.1–21.Google Scholar
  30. Giardine B, Riemer C, Hardison RC, Burhans R, Elnitski L, Shah P, et al. Galaxy: a platform for interactive large-scale genome analysis. Genome Res. 2005;15:1451–5.View ArticlePubMedPubMed CentralGoogle Scholar
  31. Goecks J, Nekrutenko A, Taylor J. Galaxy team. Galaxy: a comprehensive approach for supporting accessible, reproducible, and transparent computational research in the life sciences. Genome Biol. 2010;11:R86.View ArticlePubMedPubMed CentralGoogle Scholar
  32. Martin M. Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnetjournal. 2011;17:10–2.Google Scholar
  33. Bolger AM, Lohse M, Usadel B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinforma Oxf Engl. 2014;30:2114–20.View ArticleGoogle Scholar
  34. Langmead B, Trapnell C, Pop M, Salzberg SL. Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Biol. 2009;10:R25.View ArticlePubMedPubMed CentralGoogle Scholar
  35. Anders S, Pyl PT, Huber W. HTSeq—a python framework to work with high-throughput sequencing data. Bioinformatics. 2015;31:166–9.View ArticlePubMedGoogle Scholar
  36. Perna NT, Plunkett G, Burland V, Mau B, Glasner JD, Rose DJ, et al. Genome sequence of enterohaemorrhagic Escherichia Coli O157:H7. Nature. 2001;409:529–33.View ArticlePubMedGoogle Scholar
  37. Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. bioRxiv. 2014;002832.Google Scholar
  38. Anders S, Huber W. Differential expression analysis for sequence count data. Genome Biol. 2010;11:R106.View ArticlePubMedPubMed CentralGoogle Scholar
  39. YH Benjamini YH. Controlling the false discovery rate - a practical and powerful approach to multiple testing. J R Stat Soc Ser B. 1995;57:289–300.Google Scholar
  40. Vallenet D, Belda E, Calteau A, Cruveiller S, Engelen S, Lajus A, et al. MicroScope--an integrated microbial resource for the curation and comparative analysis of genomic and metabolic data. Nucleic Acids Res. 2013;41:D636–47.View ArticlePubMedGoogle Scholar
  41. Leisner JJ, Laursen BG, Prévost H, Drider D, Dalgaard P. Carnobacterium: positive and negative effects in the environment and in foods. FEMS Microbiol Rev. 2007;31:592–613.View ArticlePubMedPubMed CentralGoogle Scholar
  42. Haas BJ, Chin M, Nusbaum C, Birren BW, Livny J. How deep is deep enough for RNA-Seq profiling of bacterial transcriptomes? BMC Genomics. 2012;13:734.View ArticlePubMedPubMed CentralGoogle Scholar
  43. Tao H, Bausch C, Richmond C, Blattner FR, Conway T. Functional genomics: expression analysis of Escherichia Coli growing on minimal and Rich Media. J Bacteriol. 1999;181:6425–40.PubMedPubMed CentralGoogle Scholar
  44. Chen CF, Lan J, Korovine M, Shao ZQ, Tao L, Zhang J, et al. Metabolic regulation of Irp gene expression in Escherichia Coli K-12. Microbiology. 1997;143:2079–84.View ArticlePubMedGoogle Scholar
  45. Dong JM, Taylor JS, Latour DJ, Iuchi S, Lin EC. Three overlapping lct genes involved in L-lactate utilization by Escherichia Coli. J Bacteriol. 1993;175:6671–8.View ArticlePubMedPubMed CentralGoogle Scholar
  46. Cole ST, Eiglmeier K, Ahmed S, Honore N, Elmes L, Anderson WF, et al. 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. 1988;170:2448–56.View ArticlePubMedPubMed CentralGoogle Scholar
  47. Bertin Y, Deval C, de la Foye A, Masson L, Gannon V, Harel J, et al. The gluconeogenesis pathway is involved in maintenance of Enterohaemorrhagic Escherichia Coli O157:H7 in bovine intestinal content. PLoS One. 2014;9:e98367.View ArticlePubMedPubMed CentralGoogle Scholar
  48. Baum EZ, Crespo-Carbone SM, Abbanat D, Foleno B, Maden A, Goldschmidt R, et al. Utility of muropeptide ligase for identification of inhibitors of the cell wall biosynthesis enzyme MurF. Antimicrob Agents Chemother. 2006;50:230–6.View ArticlePubMedPubMed CentralGoogle Scholar
  49. Baum EZ, Crespo-Carbone SM, Klinger A, Foleno BD, Turchi I, Macielag M, et al. A MurF inhibitor that disrupts cell wall biosynthesis in Escherichia Coli. Antimicrob Agents Chemother. 2007;51:4420–6.View ArticlePubMedPubMed CentralGoogle Scholar
  50. Baum EZ, Crespo-Carbone SM, Foleno BD, Simon LD, Guillemont J, Macielag M, et al. MurF inhibitors with antibacterial activity: effect on muropeptide levels. Antimicrob Agents Chemother. 2009;53:3240–7.View ArticlePubMedPubMed CentralGoogle Scholar
  51. Paradis-Bleau C, Lloyd A, Sanschagrin F, Clarke T, Blewett A, Bugg TD, et al. Phage display-derived inhibitor of the essential cell wall biosynthesis enzyme MurF. BMC Biochem. 2008;9:33.View ArticlePubMedPubMed CentralGoogle Scholar
  52. Gonzalez MD, Akbay EA, Boyd D, Beckwith J. Multiple interaction domains in FtsL, a protein component of the widely conserved bacterial FtsLBQ cell division complex. J Bacteriol. 2010;192:2757–68.View ArticlePubMedPubMed CentralGoogle Scholar
  53. Liu B, Persons L, Lee L, de Boer PAJ. Roles for both FtsA and the FtsBLQ subcomplex in FtsN-stimulated cell constriction in E scherichia coli: cell fission regulation by FtsN. FtsA and FtsBLQ Mol Microbiol. 2015;95:945–70.View ArticlePubMedGoogle Scholar
  54. Tsang M-J, Bernhardt TG. A role for the FtsQLB complex in cytokinetic ring activation revealed by an ftsL allele that accelerates division. Mol Microbiol. 2015;95:925–44.View ArticlePubMedPubMed CentralGoogle Scholar
  55. Palumbo SA, Pickard A, Call JE. Population changes and Verotoxin production of Enterohemorrhagic Escherichia Coli strains inoculated in milk and ground beef held at low temperatures. J Food Prot. 1997;60:746–50.View ArticleGoogle Scholar
  56. Vold L, Holck A, Wasteson Y, Nissen H. High levels of background flora inhibits growth of Escherichia Coli O157:H7 in ground beef. Int J Food Microbiol. 2000;56:219–25.View ArticlePubMedGoogle Scholar
  57. Hsu J, Arcot J, Alice LN. Nitrate and nitrite quantification from cured meat and vegetables and their estimated dietary intake in Australians. Food Chem. 2009;115:334–9.View ArticleGoogle Scholar
  58. Honikel K-O. The use and control of nitrate and nitrite for the processing of meat products. Meat Sci. 2008;78:68–76.View ArticlePubMedGoogle Scholar
  59. Hammes WP. Metabolism of nitrate in fermented meats: the characteristic feature of a specific group of fermented foods. Food Microbiol. 2012;29:151–6.View ArticlePubMedGoogle Scholar
  60. Gaupp R, Ledala N, Somerville GA. Staphylococcal response to oxidative stress. Front Cell Infect Microbiol. 2012;2:33.View ArticlePubMedPubMed CentralGoogle Scholar
  61. Filenko NA, Browning DF, Cole JA. Transcriptional regulation of a hybrid cluster (prismane) protein. Biochem Soc Trans. 2005;33:195–7.View ArticlePubMedGoogle Scholar
  62. Flatley J, Barrett J, Pullan ST, Hughes MN, Green J, Poole RK. Transcriptional responses of Escherichia Coli to S-nitrosoglutathione under defined chemostat conditions reveal major changes in methionine biosynthesis. J Biol Chem. 2005;280:10065–72.View ArticlePubMedGoogle Scholar
  63. Gardner AM, Helmick RA, Gardner PR. Flavorubredoxin, an inducible catalyst for nitric oxide reduction and detoxification in Escherichia Coli. J Biol Chem. 2002;277:8172–7.View ArticlePubMedGoogle Scholar
  64. Hutchings MI, Mandhana N, Spiro S. The NorR protein of Escherichia Coli activates expression of the Flavorubredoxin gene norV in response to reactive nitrogen species. J Bacteriol. 2002;184:4640–3.View ArticlePubMedPubMed CentralGoogle Scholar
  65. Justino MC, Vicente JB, Teixeira M, Saraiva LM. New genes implicated in the protection of anaerobically grown Escherichia Coli against nitric oxide. J Biol Chem. 2005;280:2636–43.View ArticlePubMedGoogle Scholar
  66. Roos V, Klemm P. Global gene expression profiling of the asymptomatic bacteriuria Escherichia Coli strain 83972 in the human urinary tract. Infect Immun. 2006;74:3565–75.View ArticlePubMedPubMed CentralGoogle Scholar
  67. Gates FT, Linn S. Endonuclease V of Escherichia Coli. J Biol Chem. 1977;252:1647–53.PubMedGoogle Scholar
  68. Guo G, Ding Y, Weiss B. Nfi, the gene for endonuclease V in Escherichia Coli K-12. J. Bacteriol. 1997;179:310–6.View ArticleGoogle Scholar
  69. Weiss B. Endonuclease V of Escherichia coli prevents mutations from nitrosative deamination during nitrate/nitrite respiration. Mutat Res Repair. 2001;461:301–9.View ArticleGoogle Scholar
  70. Bilge SS, Vary JC, Dowell SF, Tarr PI. Role of the Escherichia Coli O157:H7 O side chain in adherence and analysis of an rfb locus. Infect Immun. 1996;64:4795–801.PubMedPubMed CentralGoogle Scholar
  71. Ravichandran A, Sugiyama N, Tomita M, Swarup S, Ishihama Y. Ser/Thr/Tyr phosphoproteome analysis of pathogenic and non-pathogenic pseudomonas species. Proteomics. 2009;9:2764–75.View ArticlePubMedGoogle Scholar
  72. Ruiguang Ge WS. Bacterial Phosphoproteomic analysis reveals the correlation between protein phosphorylation and bacterial pathogenicity. Genomics Proteomics Amp Bioinforma. 2011;9:119–27.View ArticleGoogle Scholar
  73. Jers C, Soufi B, Grangeasse C, Deutscher J, Mijakovic I. Phosphoproteomics in bacteria: towards a systemic understanding of bacterial phosphorylation networks. Expert Rev Proteomics. 2008;5:619–27.View ArticlePubMedGoogle Scholar
  74. Kenny B, DeVinney R, Stein M, Reinscheid DJ, Frey EA, Finlay BB. Enteropathogenic E. coli (EPEC) transfers its receptor for intimate adherence into mammalian cells. Cell 1997;91:511–520.Google Scholar
  75. Soufi B, Jers C, Hansen ME, Petranovic D, Mijakovic I. Insights from site-specific phosphoproteomics in bacteria. Biochim Biophys Acta. 1784;2008:186–92.Google Scholar
  76. Snyder JA, Haugen BJ, Buckles EL, Lockatell CV, Johnson DE, Donnenberg MS, et al. Transcriptome of Uropathogenic Escherichia Coli during urinary tract infection. Infect Immun. 2004;72:6373–81.View ArticlePubMedPubMed CentralGoogle Scholar
  77. Chivers PT, Sauer RT. Regulation of high affinity nickel uptake in bacteria Ni2+−DEPENDENT INTERACTION OF NikR WITH WILD-TYPE AND MUTANT OPERATOR SITES. J Biol Chem. 2000;275:19735–41.View ArticlePubMedGoogle Scholar
  78. Pina KD, Desjardin V, Mandrand-Berthelot M-A, Giordano G, Wu L-F. Isolation and characterization of thenikR gene encoding a nickel-responsive regulator inEscherichia coli. J Bacteriol. 1999;181:670–4.PubMedPubMed CentralGoogle Scholar
  79. Rodrigue A, Effantin G, Mandrand-Berthelot M-A. Identification of rcnA (yohM), a nickel and cobalt resistance gene in Escherichia Coli. J Bacteriol. 2005;187:2912–6.View ArticlePubMedPubMed CentralGoogle Scholar
  80. Marlovits TC, Haase W, Herrmann C, Aller SG, Unger VM. The membrane protein FeoB contains an intramolecular G protein essential for Fe(II) uptake in bacteria. Proc Natl Acad Sci U S A. 2002;99:16243–8.View ArticlePubMedPubMed CentralGoogle Scholar
  81. Zhao L, Chen X, Zhu X, Yang W, Dong L, Xu X, et al. Prevalence of virulence factors and antimicrobial resistance of uropathogenic Escherichia Coli in Jiangsu province (China). Urology. 2009;74:702–7.View ArticlePubMedGoogle Scholar
  82. Van Hove B, Staudenmaier H, Braun V. Novel two-component transmembrane transcription control: regulation of iron dicitrate transport in Escherichia Coli K-12. J Bacteriol. 1990;172:6749–58.View ArticlePubMedPubMed CentralGoogle Scholar
  83. Levy SB. Active efflux, a common mechanism for biocide and antibiotic resistance. Symp Ser Soc Appl Microbiol. 2002:65S–71S.Google Scholar
  84. Parker WL, Rathnum ML, Wells JS, Trejo WH, Principe PA, Sykes RB. SQ 27,860, a simple carbapenem produced by species of Serratia and Erwinia. J. Antibiot. (Tokyo). 1982;35:653–60.View ArticleGoogle Scholar
  85. Van Houdt R, Moons P, Aertsen A, Jansen A, Vanoirbeek K, Daykin M, et al. Characterization of a luxI/luxR-type quorum sensing system and N-acyl-homoserine lactone-dependent regulation of exo-enzyme and antibacterial component production in Serratia plymuthica RVH1. Res Microbiol. 2007;158:150–8.View ArticlePubMedGoogle Scholar
  86. Coulthurst SJ, Williamson NR, Harris AKP, Spring DR, Salmond GPC. Metabolic and regulatory engineering of Serratia Marcescens: mimicking phage-mediated horizontal acquisition of antibiotic biosynthesis and quorum-sensing capacities. Microbiol Read Engl. 2006;152:1899–911.View ArticleGoogle Scholar
  87. Gambino L, Gracheck SJ, Miller PF. Overexpression of the MarA positive regulator is sufficient to confer multiple antibiotic resistance in Escherichia Coli. J Bacteriol. 1993;175:2888–94.View ArticlePubMedPubMed CentralGoogle Scholar
  88. Jovanovic G, Lloyd LJ, Stumpf MPH, Mayhew AJ, Buck M. Induction and function of the phage shock protein extracytoplasmic stress response in Escherichia Coli. J Biol Chem. 2006;281:21147–61.View ArticlePubMedGoogle Scholar
  89. Landstorfer R, Simon S, Schober S, Keim D, Scherer S, Neuhaus K. Comparison of strand-specific transcriptomes of enterohemorrhagic Escherichia Coli O157:H7 EDL933 (EHEC) under eleven different environmental conditions including radish sprouts and cattle feces. BMC Genomics. 2014;15:353.View ArticlePubMedPubMed CentralGoogle Scholar
  90. Ercolini D, Ferrocino I, Nasi A, Ndagijimana M, Vernocchi P, La Storia A, et al. Monitoring of microbial metabolites and bacterial diversity in beef stored under different packaging conditions. Appl Environ Microbiol. 2011;77:7372–81.View ArticlePubMedPubMed CentralGoogle Scholar
  91. Yang X. Microbial ecology of beef carcasses and beef products. In: de Souza Sant’ Ana A, editor. Quant. Microbiol. Food Process. [Internet]. Chichester, UK: John Wiley & Sons, Ltd; 2016. [cited 2017 Jun 9]. p. 442–62. Available from: http://doi.wiley.com/10.1002/9781118823071.ch22.Google Scholar

Copyright

© The Author(s). 2017

Advertisement