Metabolic adaptation of adherent-invasive Escherichia coli to exposure to bile salts

The adherent-invasive Escherichia coli (AIEC), which colonize the ileal mucosa of Crohn’s disease patients, adhere to intestinal epithelial cells, invade them and exacerbate intestinal inflammation. The high nutrient competition between the commensal microbiota and AIEC pathobiont requires the latter to occupy their own metabolic niches to survive and proliferate within the gut. In this study, a global RNA sequencing of AIEC strain LF82 has been used to observe the impact of bile salts on the expression of metabolic genes. The results showed a global up-regulation of genes involved in degradation and a down-regulation of those implicated in biosynthesis. The main up-regulated degradation pathways were ethanolamine, 1,2-propanediol and citrate utilization, as well as the methyl-citrate pathway. Our study reveals that ethanolamine utilization bestows a competitive advantage of AIEC strains that are metabolically capable of its degradation in the presence of bile salts. We observed that bile salts activated secondary metabolism pathways that communicate to provide an energy benefit to AIEC. Bile salts may be used by AIEC as an environmental signal to promote their colonization.

Utilization of EA as a nitrogen source was investigated by addition of EA hydrochloride (5 mM) and glucose (0.1 %) to the mM9b medium (mM9b-EA). To test the capacity of the LF82 strain to use EA as the sole carbon source, the mM9b medium was supplemented with EA hydrochloride (5 mM) and NH4Cl (20 mM). For each condition, three lysogeny broth (LB) cultures were each started from a single colony and grown overnight at 37°C with aeration.
Cells were pelleted by centrifugation, resuspended in medium with EA as the sole nitrogen or carbon source and diluted 50-fold in the corresponding medium. Cultures were then incubated at 37°C, and growth was monitored in three parallel cultures by following the optical density (λ=600 nm).

RNA extraction
Total RNAs were extracted from bacteria using a Direct-zol RNA MiniPrep kit (Zymo research) and treated with a Turbo DNA-free kit (Ambion) to remove any contaminating genomic DNA.
DNase-treated RNA samples were purified with RNA clean and concentrator-25 (Zymo research). For RNA-seq analysis, purified RNAs were then quantified using a NanoDrop 1000 spectrophotometer (Thermo Scientific). The integrity of the results (RNA integrity number [RIN]) was assessed using an Agilent 2100 bioanalyzer. A Ribo-zero Magnetic kit for Gramnegative bacteria (Epicentre) was used according to the manufacturer's recommendations to remove the 23S and 16S rRNA from the total RNA samples. The samples were then purified using RNA clean and concentrator-5 (Zymoresearch). To evaluate the degree of rRNA depletion, the samples were analyzed using an Agilent 2100 bioanalyzer. The remaining RNA was sequenced using Illumina HiSeq 2500 technology with the genomic ProfileXpert platform (Claude Bernard University, Lyon, France). Three replicates from each experimental condition (with or without bile salts) were employed.

RNA-seq data analysis
We used RNA-seq to compare transcriptomes to understand the regulatory networks that control gene expression in AIEC strains during ileal colonization. Briefly, reads were mapped against the genomic sequence of E. coli LF82 (GenBank accession n° NC_011993; 3 ).
Reads mapped to several positions and reads mapped to rRNA were removed from further analysis. The number of reads overlapping each gene based on GenBank annotation was recorded. Reads from replicate samples were pooled, and the number of reads per gene was normalized according to the total number of reads in each library and the gene size. To normalize the expression of genes in different RNA-seq samples, values corresponding to the number of reads per kilobase per million mapped (RPKM) were calculated as follows: (number of reads for the gene x 10 9 ) / (total number of reads x size of the gene). A p-value adjustment for differentially expressed genes (DEGs) was performed to take into account multiple testing and control the false positive rate to a chosen level α < 0.001. Products of DEGs were classified by functional category according to metabolism pathways of the Ecocyc database (http://ecocyc.org) 4 . The RKPM values for each gene were plotted and visualized as a circle using the Circos program 5 . Metabolic pathways were analyzed using Ecocyc 6 .
A total of 53,900,605 reads from minimal medium and 56,327,631 reads from medium supplemented with bile salts were obtained for each cDNA library. Among them, 50,353,549 (93.4%) and 53,246,854 (94.5%) reads mapped to the genome of Escherichia coli LF82, which has a total size of 4,773,108 bp. The average numbers of reads per region were 462 (without bile salts) and 747 (with bile salts), with a coverage of 98.4% and 98.6% of the 4376 encoding DNA-encoding sequences (CDS) of E. coli LF82 represented by at least one single read. The absolute and relative distributions of reads in the two media for the annotated genes of the LF82 strain are shown in Fig. S1. A dendrogram used to represent all the samples shows the grouping of the replicates and a difference between the biological conditions. Box plots and the volcano plot are also represented to show the quality of the normalization and the differential expression of genes (Fig. S1). To assess the reliability of RNA-seq for determining the relative abundances of individual transcripts in the absence and presence of bile salts, we used absolute quantification of mRNAs for three up-regulated genes (prpB, eutB and LF82_715), two downregulated genes (pfkA and cfa) and one unchanged gene (folX) by qRT-PCR. The data for these genes provided an r2 value of 0.9708, confirming the data obtained by RNA-seq ( Fig. S2; Table S1).

Construction and transcomplementation of isogenic mutants
Isogenic mutants of E. coli LF82 was generated by using the lambda red recombination system. E. coli LF82 was transformed with pKOBEG, a plasmid encoding the Red proteins that protect linear DNA from degradation in bacteria. The plasmid was maintained in bacteria at 30°C with 25 mg/l of chloramphenicol and 1 mM of L-arabinose. The Flp recognition target-flanked cassette harboring the kanamycin resistance cassette was generated by PCR from E. coli BW25141 with d-eutB-F/d-eutB-R, d-eutE-F/d-eutE-R and d-citF-F/citF-R primers (Table S2) and High Fidelity Platinum Taq polymerase (Invitrogen) according to the manufacturer's instructions. The PCR products were electroporated into previously glycerol-washed E. coli LF82. The resulting LF82ΔeutB, LF82ΔeutE and LF82 ΔcitF isogenic mutants (Km R ) were selected on LB agar containing 50 mg/L kanamycin. Replacement of the eut and citF genes by the kanamycin resistance cassette was confirmed by PCR. The kanamycin resistance cassette was then removed from LF82ΔeutB bacteria by the transient expression of the Flp recombinase from the pCP20 plasmid, creating the LF82ΔeutB (Km S ) strain.
The eutB gene was amplified by PCR from E. coli LF82 genomic DNA using eutBEcoR1-F and eutBBamH1-R primers ( Table S2). The amplified DNA was purified with a NucleoSpin extract kit (Macherey-Nagel), digested with EcoRI and BamHI (New England Biolabs), and ligated to the EcoRI-BamHI-digested expression vector pBK-CMV (Agilent Technologies).
This construct was electroporated into LF82ΔeutB (Km S ) electrocompetent strains and selected on Mueller Hinton agar containing 50 mg/L kanamycin. The presence of the eutB gene was confirmed by PCR. The construction was checked by double-stranded DNA sequencing (GATC biotech, Germany).

Murine model of gut colonization
For the in vivo experiments, we used C57BL/6 mice, which were housed in specific pathogenfree conditions in the animal care facility at the Université Clermont Auvergne, Clermont-Ferrand, France. For in vivo competition assays, ten twelve-week-old mice (body weight ≈26-28 g) were pretreated by administering oral amoxicillin (1 g/L), vancomycin (500 mg/L), metronidazole (500 mg/L) and neomycin (1 g/L) for four days, and 3% dextran sulfate sodium salt (Sigma) for the last day. At 24 h after stopping the antibiotic treatment, the animals were orally challenged with 10 9 bacteria (50% LF82 -50% LFΔeutB). Three days after bacterial infection, fresh fecal pellets (100-200 mg) were collected from individual mice and resuspended in PBS. After serial dilutions, the bacteria were enumerated by plating on TS agar medium containing amoxicillin to isolate the two bacteria and amoxicillin + kanamycin to isolate LFΔeutB. The plates were then incubated overnight at 37°C before counting the CFU.
The CFU count of the LF82 strain for each mouse was calculated by subtracting the number of CFU that were resistant to kanamycin from the number of CFU counted on an agar plate containing only amoxicillin. Three days after infection, the mice were anesthetized with isoflurane and then euthanized by cervical dislocation. Colonization of the two strains was studied by enumerating the mucosa-associated AIEC bacteria by homogenizing 0.5 cm of ileum and 1 cm of colon, beginning at 0.5 cm from the cecal junction, in sterile PBS solution. Samples were plated on TS agar containing amoxicillin or amoxicillin + kanamycin and incubated overnight at 37°C.

Biofilm formation assay
Biofilm assays were used as previously described with some modifications 7,8 . Briefly, 6 µL of a 3-hour culture in mM9-EA medium supplemented or without 1% bile salts was inoculated into 144 µL of the same medium in a 96-well culture-treated polystyrene microtiter plate (Nunc). Wells filled with growth medium alone were included as negative controls. After 4 h and 30 min of incubation at 37°C, surface-adherent biofilm formation was measured by staining bound cells for 15 min with a 0.5% (w/v) aqueous solution of crystal violet. After rinsing with distilled water, the bound dye was released from the stained cells using 95% ethanol, and the OD at 540 nm was determined.

Autoaggregation assay
After an overnight culture at 37°C in M9 minimal medium supplemented with glucose (0.1%) and bile salts (1%) when needed, cells were diluted in the same medium and grown at 37°C to the exponential phase. Next, 20 µL of each culture was Gram-stained and visualized by microscopy.   Table S1). (C) Global analysis of increased transcript levels of genes involved in metabolism in E. coli LF82 with connections with each metabolic pathway. The gray circular lines represent a log2 = 0.5 (i.e., RPKM bile salts containing medium/minimal medium ratio values) of each gene among the upregulated genes. (D) Global analysis of decreased transcript levels of genes involved in metabolism in E. coli LF82 with associations with each metabolic pathway. The gray circular lines represent a log2 = 0.5 (i.e., RPKM bile salts containing medium/minimal medium ratio values) of each gene among the downregulated genes.    Ethanolamine is metabolized by AIEC to produce nitrogen and acetyl-CoA. Citrate is converted to acetate and oxaloacetate by citrate lyase. Oxaloacetate, which is necessary for the methylcitrate pathway, is also supplied for L-tartrate fermentation and amino acid degradation.

Supplemental data
Propionyl-coA from various degradation products is metabolized to produce pyruvate, which can be converted to acetyl-coA. The genes that exhibited an increase in mRNA concentrations in bacteria grown with bile salts in comparison to without bile salts encoded proteins involved in ethanolamine degradation (green boxes), 2-methylcitrate pathway (blueberry boxes), citrate and L-tartrate fermentation (cyan boxes) and 1,2-propandediol degradation (pink boxes).