Enterohemorrhagic Escherichia coli responds to gut microbiota metabolites by altering metabolism and activating stress responses

ABSTRACT

Enterohemorrhagic Escherichia coli (EHEC) is a major cause of severe bloody diarrhea, with potentially lethal complications, such as hemolytic uremic syndrome. In humans, EHEC colonizes the colon, which is also home to a diverse community of trillions of microbes known as the gut microbiota. Although these microbes and the metabolites that they produce represent an important component of EHEC’s ecological niche, little is known about how EHEC senses and responds to the presence of gut microbiota metabolites. In this study, we used a combined RNA-Seq and Tn-Seq approach to characterize EHEC’s response to metabolites from an in vitro culture of 33 human gut microbiota isolates (MET-1), previously demonstrated to effectively resolve recurrent Clostridioides difficile infection in human patients. Collectively, the results revealed that EHEC adjusts to growth in the presence of microbiota metabolites in two major ways: by altering its metabolism and by activating stress responses. Metabolic adaptations to the presence of microbiota metabolites included increased expression of systems for maintaining redox balance and decreased expression of biotin biosynthesis genes, reflecting the high levels of biotin released by the microbiota into the culture medium. In addition, numerous genes related to envelope and oxidative stress responses (including cpxP, spy, soxS, yhcN, and bhsA) were upregulated during EHEC growth in a medium containing microbiota metabolites. Together, these results provide insight into the molecular mechanisms by which pathogens adapt to the presence of competing microbes in the host environment, which ultimately may enable the development of therapies to enhance colonization resistance and prevent infection.

KEYWORDS:

• Enterohemorrhagic Escherichia coli
• EHEC
• gut microbiota
• metabolites
• RNA-seq
• Tn-seq
• bacterial metabolism
• stress response

Introduction

Enterohemorrhagic Escherichia coli (EHEC) is a major cause of foodborne illness worldwide, with serotype O157:H7 causing over 63,000 illnesses per year in the United States alone ¹. Symptoms of EHEC infection include abdominal pain and bloody diarrhea; in addition, up to 15% of EHEC-infected individuals develop hemolytic uremic syndrome, a severe complication caused by EHEC’s Shiga toxin that can result in acute renal failure². Along with enteropathogenic E. coli (EPEC) and the mouse pathogen Citrobacter rodentium, EHEC is a member of the attaching and effacing (A/E) group of pathogens³. These pathogens are characterized by their ability to tightly adhere to intestinal epithelial cells and modify the host cytoskeleton, resulting in localized destruction (effacement) of brush border microvilli and the formation of a pedestal structure beneath the adherent bacteria⁴. The ability to produce this A/E lesion is conferred by a pathogenicity island known as the Locus of Enterocyte Effacement (LEE)^5,6, which encodes a type III secretion system (T3SS) that is essential for host infection^7–9. EHEC delivers at least 39 different effector proteins into host cells via the T3SS¹⁰, which collectively modulate numerous host processes including cytoskeleton dynamics, phagocytosis, inflammatory signaling cascades, and apoptosis¹¹, thereby producing a more favorable environment for pathogen colonization.

Within the human gut, EHEC’s major site of colonization is the colon^12,13, which is inhabited by a community of trillions of microbes collectively known as the gut microbiota. These resident gut microbes often interfere with the ability of pathogens to infect their host, a phenomenon termed colonization resistance¹⁴. Numerous in vivo studies using the related mouse pathogen C. rodentium have demonstrated that the gut microbiota provides colonization resistance against A/E pathogens. For example, disruption of the gut microbiota by treating mice with the antibiotic metronidazole increases the severity of subsequent C. rodentium infection;¹⁵ moreover, germ-free mice colonized with C. rodentium are unable to clear the infection, unlike mice possessing a normal gut microbiota¹⁶. Colonization resistance results from many direct and indirect interactions between the gut microbiota and invading pathogens, including competition for nutrients, production of bacteriocins and other antimicrobials by the gut microbiota, and promotion of host immune defenses¹⁷. Importantly, the gut microbiota shape the ecological niche available for pathogen colonization through their use of dietary and host resources, as well as through production and secretion of a wide variety of metabolites^18–20.

Although the gut microbiota clearly has a major impact on A/E pathogens, little is known about how the pathogens sense and respond to these competing microbes in order to mount a successful infection. There is some evidence that EHEC overcomes competition for nutrients through its ability to metabolize certain mucin-derived sugars (such as galactose, mannose, and ribose), which are poorly catabolized by competing microbes such as commensal E. coli ²¹. In addition, several studies have demonstrated that EHEC virulence gene expression is affected by contact with microbiota or their secreted metabolites. For example, expression of the LEE is increased by butyrate²², succinate²³, and contact with Enterococcus faecalis cells²⁴, while LEE expression is repressed by indole²⁵ and fucose²⁶. Furthermore, several microbiota-produced organic acids have been shown to increase EHEC motility and chemotaxis²⁷, while Bacteroides thetaiotaomicron-mediated depletion of vitamin B₁₂ reduces expression of Shiga toxin 2²⁸. Although the ability to use microbiota-derived compounds as cues for virulence gene regulation is undoubtedly beneficial to EHEC, none of these studies have systematically explored whether EHEC has evolved additional strategies for growth in the presence of the competing microbiota. We hypothesized that EHEC possesses specific genes or gene regulation patterns that promote growth or survival in the presence of the gut microbiota.

In this study, we used a simple in vitro model to identify and characterize EHEC’s response to the intestinal microbiota by using a defined Microbial Ecosystem Therapeutic (MET-1) previously used to cure recurrent Clostridioides difficile disease in humans²⁹. Our major objectives were (1) to examine how EHEC gene expression changes in response to gut microbiota metabolites and (2) to determine which EHEC genes are needed for growth in these conditions, with the overarching goal of identifying genetic pathways that may contribute to A/E pathogen colonization or survival in the gut environment.

Materials and methods

Bacterial strains and growth conditions

Bacterial strains and plasmids used in this study are listed in Table S1. Unless otherwise stated, all strains were grown in lysogeny broth (LB; 10 g/L tryptone, 5 g/L yeast extract, 10 g/L NaCl) at 37ºC with aeration at 225 rpm or on LB agar at 37ºC. Antibiotics and supplements were used when necessary at the following concentrations: ampicillin, 100 µg/mL; kanamycin, 50 µg/mL; chloramphenicol, 30 µg/mL; diaminopimelic acid (DAP), 0.3 mM.

Preparation of MET-1 and Donor 5 microbiota community supernatants

The Research Ethics Board of the University of Guelph approved this study (REB no. 09AP011). Microbial strains were cultured from the stool of two healthy donors at the University of Guelph; the MET-1 strains originated from a 41-year-old female²⁹, and the Donor 5 community originated from a 44-year-old Caucasian male³⁰. MET-1 and Donor 5 microbiota communities were cultured in continuous flow bioreactors under conditions emulating the human distal gut³¹. Briefly, MET-1 strains²⁹ augmented with additional microbial species from the same donor to increase representation of the fecal ecosystem (strain 32–6-I 28 D6 FAA 94% ID to Anaerotignum lactatifermentans; strain 32–6-I 30 D6 FAA 96% ID to Mediterraneibacter glycyrrhizinilyticus; strain 32–6-I 16 TSA 98% ID to Dorea formicigenerans; strain 32–6-I 16 NA 99% ID to Coprococcus comes; strain 32–6-I 11 D6 FAA 99% ID to Bifidobacterium pseudocatenulatum; strain 16–6-S BF 5 99% ID to Faecalicatena fissicatena; and strain 16–6-I 43 FAA 99%ID to Roseburia inulinivorans, where each strain %ID was informed through BLAST identity of the V3-V6 16S rRNA gene region) were individually cultured on fastidious anaerobe agar (LabM) with 10% defibrinated sheep’s blood (Hemostat Laboratories) in a Ruskinn Concept Plus anaerobe chamber at 37°C supporting an atmosphere of CO₂:H₂:N₂ 10:10:80 until biomass was readily visible. Biomass was aseptically scraped into 10 mL sterile, degassed phosphate buffered saline to make a homogenous slurry, and this was introduced into a pre-prepared 500 mL Multifors vessel (Infors, Switzerland), with a working volume of 400 mL. Conditions were set to mimic the distal human gut (37°C, pH 7, gentle agitation, oxygen-free, and fed with a constant supply of mucin (4 g/L) and insoluble starch substrates (12 g/L) at a flow rate of ∼400 mL/day). Vessels were maintained anaerobically by bubbling nitrogen through them. Following inoculation, cultures were gently agitated and the pH adjusted to 6.9 − 7.0. Media pumps were switched on 24 h following inoculation, and the culture was allowed to reach compositional equilibrium over 28 days. Human subject recruitment for this study was approved by The Research Ethics Board of the University of Guelph (REB no. 10JL002). The donor 5 fecal ecosystem community was prepared as described by Yen et al.³⁰. Samples from both microbial communities were withdrawn from the vessel at equilibrium, ultracentrifuged at 31,400 × g at 4ºC for 2 h, and then the supernatants were filter-sterilized by passing through 0.22 µm syringe filters.

RNA-Seq

An RNA-Seq approach was used to compare EHEC gene expression in the presence and absence of gut microbiota-produced metabolites to gain insight into EHEC’s response to the gut microbiota. Biological triplicate cultures of EHEC strain EDL933 were grown overnight in LB as described above. Each replicate culture was inoculated at a starting OD_600 nm of 0.02 into three different media: (1) M9 minimal medium [10× M9 concentrate (0.48 M Na₂HPO₄∙7 H₂O; 0.22 M KH₂PO₄; 85.6 mM NaCl; 0.187 M NH₄Cl; 20 mM MgSO₄; 1 mM CaCl₂; 4% glucose) diluted with dH₂O to a final concentration of 1×]; (2) rich medium (10× M9 concentrate diluted with sterile chemostat medium to a final concentration of 1×); and (3) microbiota metabolite medium (10× M9 concentrate diluted with filter-sterilized MET-1 culture supernatant to a final concentration of 1×). Cultures were incubated at 37ºC with aeration until reaching an OD_600 nm of 0.3–0.5. Cultures were treated with RNAProtect Bacteria Reagent (Qiagen) to stabilize RNA, then RNA was extracted using the RNeasy Mini Kit (Qiagen). 5 μg of total RNA per sample was submitted for rRNA depletion, library preparation, and sequencing by GENEWIZ (Plainfield, NJ). Sequencing was performed using the Illumina HiSeq 2500 platform with 50-bp single-end reads.

RNA-Seq data were analyzed using Rockhopper version 2.0.3 with default settings³². Transcripts were considered to be differentially regulated in the presence of microbiota metabolites if: (i) there was a two-fold or greater difference in expression in microbiota metabolite medium relative to both controls; (ii) expression was either higher or lower in microbiota metabolite medium than in both of the controls; and (iii) the changes in expression in microbiota metabolite medium vs. both controls was significant at the level of q < 0.05.

Pathway enrichment analysis of Gene Ontology (GO)³³ terms in the GO Biological Processes database was performed using Cytoscape 3.9.1³⁴ by using a separate pre-ranked list of up- and down-regulated genes against the background universe of all detected genes (after pre-filtering) arranged according to log2FC. Redundant GO terms were removed by using a redundancy cutoff of 0.5. Only pathways with an FDR <0.05 were considered to be significantly enriched. GeneRatio was calculated as the proportion of enriched genes per pathway.

Reverse transcriptase quantitative PCR (RT-qPCR)

For RT-qPCR validation of RNA-Seq results, cultures were grown under the same conditions as those used for RNA-Seq (described above). For the analysis of T3SS and pilus gene expression (Figure S2), EHEC strains were cultured overnight in LB at 37ºC with aeration, then subcultured 1:50 into DMEM and incubated statically at 37ºC with 5% CO₂ for 4 h. In all cases, RNA was extracted from 500 µL of culture using RNAprotect Bacteria Reagent (Qiagen) followed by the GeneJET RNA Purification Kit (Thermo Fisher Scientific). Contaminating genomic DNA was removed from 2-µg aliquots of purified RNA using the TURBO DNA-free Kit (Thermo Fisher Scientific), followed by reverse transcription with the QuantiTect Reverse Transcription Kit (Qiagen). RT-qPCR was performed using the QuantiTect SYBR Green PCR Kit (Qiagen) on an Applied Biosystems 7500 Fast Real-Time PCR System, using the ΔΔC_T relative quantitation method with truB (which was experimentally verified to be expressed at equal levels under all growth conditions used in this study; data not shown) as the endogenous control. The primers used for RT-qPCR are listed in Table S2; the efficiency of all primer pairs was verified to be within the range of 90–110% (data not shown). No template and no reverse transcriptase controls were included in each RT-qPCR plate to confirm the absence of primer dimer and contaminating genomic DNA, respectively.

Strain and plasmid construction

EHEC ΔyigM, ΔbhsA, and ΔyhcN and C. rodentium ΔbhsA and ΔyhcN deletion mutants were generated by allelic exchange. Briefly, in-frame deletion constructs for each gene were generated by overlap-extension PCR³⁵ using the UpF-UpR and DnF-DnR primers listed in Table S2. Overlap PCR products were restriction digested and ligated into pUC18. All inserts were confirmed by Sanger sequencing, then subcloned into suicide vector pRE112³⁶. Suicide plasmids were transferred into EHEC or C. rodentium by biparental mating using MFDpir as the donor³⁷ with transconjugants selected on LB chloramphenicol plates. Loss of the pRE112 plasmid from the EHEC or C. rodentium chromosome was subsequently selected for by growth on LB agar (without NaCl) with 5% sucrose. Sucrose-resistant and chloramphenicol-sensitive colonies were screened for the presence of the intended deletion by PCR.

To generate lux reporters pJW15P_bioB and pJW15P_glnK, promoter regions were amplified from EHEC genomic DNA using primers PbioB_F/PbioB_R and PglnK_F/PglnK_R (Table S2); PCR products were then restriction digested and ligated into pJW15 by standard techniques.

EHEC ΔudhA::kan^R and EDL933 ΔglnLG::kan^R mutants were constructed by λ Red recombination³⁸. The PCR template for λ Red recombination was generated by amplifying the kanamycin resistance cassette from pKD13 using primers whose 5’ end contained 50 nt of sequence complementary to the flanking regions of the genes of interest. The purified PCR products were electroporated into EDL933 harboring λ Red recombinase plasmid pKD46. Transformants were selected on LB agar containing 40 µg/mL kanamycin; the correct chromosomal location of insertions was verified by PCR using both gene-specific and kanamycin cassette-specific primers. Loss of pKD46 was verified by plating transformants on LB plates containing ampicillin to ensure a loss of pKD46-encoded ampicillin resistance.

Biotin assay

Biotin concentrations in culture media and supernatants were measured using the Abcam Biotin Assay Kit – Colorimetric (catalog number ab185441) with three technical replicates per sample.

Luminescence assay

Strains harboring lux reporters were cultured in triplicate overnight in LB with kanamycin. For the experiments shown in Figure 2c, strains were subcultured 1:100 into M9-glucose. After 4 h incubation, 100 µL of each culture was transferred to a black/clear bottom 96-well plate, biotin was added at indicated concentrations, and luminescence and OD_600 nm were measured in a Tecan Infinite 200 plate reader after a further 4 h incubation. For the experiments shown in Figure 2d, strains were subcultured into the indicated media, and luminescence and OD_600 nm were measured after 4 h incubation. For the experiments shown in Figure S5, EHEC harboring pJW15P_glnK was subcultured 1:100 into M9-glucose for 4 h; SCFAs were then added at the indicated concentrations and luminescence and OD_600 nm were measured every 5 min for 2 h. Normalized luminescence was calculated by dividing raw luminescence by the OD_600 nm of the same well.

HeLa infections, microscopy, cell death assays

HeLa human epithelial cells were seeded into eight-well LabTek chamber slide or 12-well tissue culture plates using DMEM supplemented with 10% fetal bovine serum and incubated in an atmosphere of 95% air–5% CO₂ and allowed to grow to 80% confluence. EHEC strains were grown in LB broth overnight, and preinduced by subculturing 1:20 in prewarmed DMEM without phenol red (G.E. Healthcare) at 37°C in 5% CO₂ for 3.5 h without shaking. HeLa cells in serum-free DMEM were infected at a multiplicity of infection (MOI) of 100 for 1 h. After incubation, the cells were washed three times and incubated for additional 2 or 4 hours. Following the infection, cells were washed with PBS, fixed with 4% paraformaldehyde for 15 min, permeabilized with 0.1% of Triton X-100 for 10 min, and blocked with 3% BSA – PBS for 30 min. HeLa cells were immunostained using a polyclonal rabbit anti-E. coli (Cat. No. ab137967; Abcam) antibody for 1 h at room temperature, followed by a goat anti-rabbit IgG (H+L) secondary antibody conjugated to Alexa Fluor 488 (Cat. No. A-11001; Thermo Fisher Scientific) for 45 min. F-actin was stained with Alexa Fluor 568 phalloidin (Cat. No. A12380). Nuclei and bacteria were stained with DAPI. Slides were mounted with a Prolong Gold antifade reagent, covered with a glass coverslip, and examined under a Zeiss Axio Imager M2 microscope. For determining cell death, cells were harvested with Accutase and resuspended in PBS containing 5% FBS. Cells were treated with 5 µM of CellEvent Caspase-3/7 Green Detection Reagent (Cat. No. C10423; Thermo Fisher Scientific) according to the manufacturer’s recommendations or 1 µL of propidium iodide (10 mg/mL) and measured using Attune Nxt acoustic focusing cytometer (Life Technologies, Carlsbad, USA) and FlowJo software (v10.8).

C. rodentium mouse infections

All animal experiments were performed in accordance with the guidelines of the Canadian Council on Animal Care and the University of British Columbia (UBC) Animal Care Committee (certificate A16–0216). Mice were ordered from Jackson Laboratory (Bar Harbor, ME) and maintained in a specific pathogen-free facility at UBC. Six-week-old female C57BL/6J mice were orally gavaged with 100 µL of overnight culture of C. rodentium grown in LB (containing ~ 3 × 10⁸ CFU of bacteria, as confirmed by retrospective plating). At day 6 post-infection, mice were humanely euthanized by isoflurane anesthesia followed by carbon dioxide inhalation. C. rodentium colonization of the cecum and colon was assessed by homogenizing tissues in PBS and plating dilutions on MacConkey agar.

Six-week-old female C3H/HeJ mice were orally gavaged with 100 μL of overnight culture of C. rodentium (∼3 × 10⁸ CFU) grown in LB. Mice were monitored daily for weight loss and clinical symptoms. C. rodentium shedding was monitored by plating dilutions of fecal samples on MacConkey agar every two days. Upon reaching the humane endpoint (weight loss of 20%; or any one of: bloody diarrhea, severe hunching, severe piloerection, slow or no response to stimuli, labored breathing, or rectal prolapse; or any three of: moderate hunching, moderate piloerection, some lethargy, some change in breathing rate, effort, or pattern), mice were euthanized by isoflurane anesthesia followed by carbon dioxide inhalation.

In vivo gene expression analysis

For RT-qPCR comparison of in vitro and in vivo gene expression in C. rodentium, DBS100 was grown in LB medium at 37ºC with aeration to OD_600 nm~0.5 (log-phase condition) or overnight (stationary-phase condition – the same growth conditions used for inoculum cultures for mouse infections). RNA was extracted from in vitro-grown cultures as described above. In vivo RNA was isolated from the distal colon of 7-week-old female C57BL/6 mice 9 days post-infection with 4 × 10⁸ CFU of DBS100. Upon tissue harvest, tissue sections were flash-frozen in liquid nitrogen. Colon RNA was extracted by placing the frozen tissue sections in 1 mL of TRIzol Reagent (Thermo Fisher) in a Lysing Matrix D tube (MPBio). Tissues were homogenized for 2 × 40 s using a FastPrep-24 homogenizer (MPBio), then RNA was extracted following the TRIzol manufacturer’s instructions. RNA was further purified following the “RNA Cleanup Protocol” in the GeneJet RNA Purification Kit (Thermo Fisher). DNase treatment, reverse transcription, and RT-qPCR were performed as described above, with one exception: rather than using a single endogenous control, the geometric mean of the C_Ts of three endogenous control primers (gyrB, recA, and frr) was used in the ΔΔC_T calculations in order to improve accuracy. No amplification was detected using cDNA prepared from mock-infected mice, confirming that primers were C. rodentium-specific.

Tn-Seq

An EDL933 transposon mutant library suitable for Tn-Seq was generated using the plasmid pSAM-Ec³⁹. The conjugal donor strain MFDpir carrying pSAM-Ec was mixed with EDL933 in a 2:1 ratio and incubated on LB agar containing DAP at 37ºC for 5 h. Bacteria were then scraped off the plate and resuspended in 2.5 mL LB. 100 µL of the cell suspension was reserved for serial dilutions to determine the efficiency of mutagenesis; the remaining 2.4 mL were spread on LB+Kan plates (100 µL/plate). Six independent biparental matings were conducted in this fashion, giving a total of 144 LB+Kan plates. After the incubation at 37ºC for 20 h, colonies were harvested from the plates, pooled, and thoroughly mixed. A “working stock” of the transposon library was prepared by inoculating 1 mL of the pooled cell suspension into 100 mL of LB+Kan broth, incubating at 37ºC with aeration for 2 h, then adding glycerol to 20% and freezing 1 mL aliquots at −70ºC.

For the characterization of the transposon mutant library, a 1-mL aliquot was thawed and genomic DNA was extracted using genomic-tip columns (Qiagen). For the characterization of EHEC genes affecting fitness during growth in microbiota metabolites, a 1-mL aliquot was thawed and inoculated into 50 mL LB broth containing 50 µg/mL kanamycin. The culture was incubated at 37ºC with aeration until reaching an OD_600 nm of 0.8, then 100 µL of this culture was inoculated into four replicate subcultures, each with 50 mL volume, of the same three culture conditions described above for RNA-Seq (M9 minimal medium, rich medium, and microbiota metabolite medium). The subcultures were incubated at 37ºC with aeration until reaching an OD_600 nm of 0.5, then bacteria were pelleted, and genomic DNA was purified using Qiagen Genomic-tips. Sequencing libraries were prepared as previously described⁴⁰. Briefly, purified genomic DNA was digested with MmeI, treated with calf intestinal phosphatase, and purified by phenol-chloroform extraction and ethanol precipitation. Digested DNA was then ligated with phosphorylated double-stranded Tn-Seq adaptors (consisting of annealed PBGSF29 and PBGSF30 oligomers, each containing a unique sequencing index for multiplexing; Table S2). Ligation products were amplified by PCR (25 cycles) using NEBNext High-Fidelity PCR Master Mix (New England Biolabs) and primers PBGSF23 and PBGSF31. PCR products were purified using MinElute PCR purification columns (Qiagen) and sequenced on an Illumina HiSeq 2500 in 1 × 50 bp single-read configuration (Genewiz). Sequencing barcodes were removed using the FastX barcode splitter, and sequencing reads were aligned to the EDL933 genome⁴¹ using the ESSENTIALS pipeline⁴², with 16 bp minimum sequence match for alignment. Sites harboring transposon insertions were distinguished from background “noise” as previously described⁴⁰. To identify genes affecting fitness in the presence of microbiota metabolites, the following criteria were applied: the gene was not considered essential (according to ESSENTIALS analysis); an average of ≥ 100 reads were mapped to the gene in at least one of the two control conditions; the number of reads mapped to the gene in the microbiota metabolite condition was either higher or lower than both controls; and the adjusted p value of both comparisons (microbiota metabolites vs. minimal medium and microbiota metabolites vs. chemostat medium) was <0.01.

Competitive growth assay

Wild-type and mutant EHEC strains were cultured individually overnight in LB broth. The following day, 1 mL of each culture was pelleted and washed once in PBS. Triplicate mixed cultures in the media conditions of interest were inoculated with each strain at a starting OD_600 nm of 0.002 and incubated at 37ºC with aeration for 24 h. Samples of mixed cultures were serially diluted and plated on LB agar and LB agar containing 40 µg/mL kanamycin at 0 h and 24 h. The competitive index was calculated as the proportion of Kan^R colonies at 24 h divided by the proportion of Kan^R colonies at 0 h.

NADPH/NADP⁺ assay

EHEC strains were cultured in triplicate overnight in LB, then subcultured 1:100 into M9-glucose minimal medium or M9-glucose containing 40 mM acetate. Cultures were incubated at 37ºC with aeration to an OD₆₀₀ of 0.5–0.8, then pelleted and washed once in cold PBS. The cellular NADPH/NADP+ ratio was determined using the EnzyChrom NADP+/NADPH Assay Kit (BioAssay Systems) according to manufacturer’s instructions.

Statistical Analysis

Statistical analyses were performed in GraphPad Prism v9.4.1 (www.graphpad.com). Survival curves were analyzed using the log-rank Mantel-Cox test. For comparison of multiple groups, one-way ANOVA was conducted with post-hoc Dunnett’s test or two-way ANOVA with post-hoc Tukey’s as appropriate. The results represent the mean ± SEM as stated in the figure legends, and statistical significance is represented by *p-value <0.05, **p-value <0.01, ***p-value <0.001 and ****p-value <0.0001.

Results

Transcriptome analysis of EHEC’s response to gut microbiota-produced metabolites

To address our first objective of identifying changes in EHEC gene expression induced by exposure to microbiota-produced metabolites, we used an RNA-Seq approach. The gut microbiota metabolites used for this experiment originated from the sterile-filtered supernatant of an in vitro-grown, defined community of 33 human gut isolates called MET-1, which was originally developed as a synthetic stool transplant to treat recurrent C. difficile infections²⁹. For the RNA-Seq experiment, we extracted RNA from the EHEC wild-type strain EDL933 that was grown in three different conditions: (1) the MET-1 supernatant; (2) the rich medium that was used to culture the MET-1 community; and (3) M9-glucose minimal medium. All three media were supplemented with the same amount of concentrated M9-glucose in order to ensure a comparable baseline of essential nutrients in all conditions (Figure 1a). By including both a rich medium control and a minimal medium control, we were able to identify changes in EHEC gene expression resulting specifically from exposure to microbiota metabolites (i.e., genes which were upregulated or downregulated in the microbiota metabolite condition compared to both controls) rather than from nutrient depletion in the microbiota supernatant compared to the rich medium control (i.e., genes whose expression in the microbiota metabolite condition were intermediate between the rich medium and minimal medium controls). A total of 170.0 million sequencing reads were successfully aligned to the EDL933 genome (14.2 million to 23.0 million reads per sample).

Figure 1. RNA-Seq analysis of EHEC’s response to gut microbiota-produced metabolites. (a) Overview of experimental design. (b) Overview of RNA-Seq results. Scatterplot shows transcript abundance in EHEC grown in MET-1 microbiota metabolites compared to the minimal medium control on the x-axis and compared to the rich medium control on the y-axis. Transcripts that were considered significantly upregulated (fold change threshold of ≥ 2 and q < 0.05 in both metabolites vs. minimal medium and metabolites vs. rich medium) are colored red and those considered significantly downregulated (fold change threshold of≤0.5 and q < 0.05 in both metabolites vs. minimal medium and metabolites vs. rich medium) are colored blue. Genes that are discussed further in the text are labeled. (c) Gene ontology (GO) pathway enrichment analysis of differentially expressed genes (DEGs). Gene Ratio refers to the proportion of enriched genes per pathway.

We identified 116 genes that were upregulated at least two-fold and 32 genes that were downregulated at least two-fold during EHEC growth in the MET-1 metabolites compared to both control conditions (q < 0.05; Figure 1b; Dataset S1). The genes that were most strongly upregulated or downregulated in the presence of gut microbiota metabolites largely fell into two major categories: metabolism/biosynthetic pathways (including downregulation of biotin biosynthesis genes and upregulation of nitrogen starvation genes) and stress responses (Figure 1c). Each of these categories of genes, the types of microbiota metabolites that may affect their expression, and the manner in which they may contribute to EHEC growth and survival in the gut environment are discussed in more detail below.

Uptake of microbiota-derived biotin causes repression of EHEC biotin biosynthesis genes

Among the genes most strongly repressed by microbiota metabolites in the RNA-Seq experiment were bioF, bioB, bioC, bioD, and bioA (Dataset S1; Figure 1b). These genes comprise two divergently transcribed transcriptional units (bioBFCD and bioA) that, together, encode the majority of the enzymes required for biosynthesis of biotin, which is an essential coenzyme required for fatty acid biosynthesis⁴³. The expression of these operons is regulated by the dual-function transcriptional regulator and biotin ligase protein BirA in response to intracellular biotin availability⁴³. When biotin is abundant, it acts as a co-repressor with BirA to repress expression of the bio operons; on the other hand, when biotin is limiting, BirA cannot bind to DNA, and therefore bio transcription is de-repressed. We hypothesized that low expression of bioA and bioBFCD during EHEC growth in the MET-1 supernatant could indicate a high abundance of biotin in the microbiota-conditioned medium.

Using quantitative reverse transcriptase PCR (RT-qPCR), we confirmed that bioF expression was substantially reduced during EHEC growth in MET-1 metabolites compared to growth in either rich or minimal media (Figure 2a). bioF expression was similarly repressed during EHEC growth in the culture supernatant from a different human fecal microbiota community (“Donor 5”; Figure 2a), indicating that accumulation of biotin may be common among gut microbial communities. Indeed, both the MET-1 supernatant and the Donor 5 supernatant contained >50 µM biotin, while the rich media in which the communities were grown contained less than 25 µM biotin (Figure 2b). To determine whether uptake of this microbiota-produced biotin was responsible for repression of bio gene expression during EHEC growth in the MET-1 metabolites, we constructed an EHEC mutant lacking yigM, which encodes the high-affinity biotin transporter in E. coli ^44–46. As expected, we found that the addition of external biotin to wild-type EHEC in concentrations as low as 5 nM caused strong repression of bio gene expression, as measured using a bioB-lux transcriptional reporter, while repression of bioB-lux expression was observed in the EHEC ΔyigM mutant only at substantially higher concentrations of external biotin (~5 µM) (Figure 2c). In accordance with the RNA-Seq and RT-qPCR results (Dataset S1; Figure 2a), we found that expression of the bioB-lux reporter was highest during wild-type EHEC growth in minimal medium, intermediate in rich medium, and lowest in the MET-1 metabolites (Figure 2d). Although the expression of bioB-lux in the EHEC ΔyigM mutant was comparable to wild-type expression during growth in a minimal medium, the expression of the bioB-lux reporter was repressed to a much lesser extent when the ΔyigM mutant was grown in the MET-1 metabolites (Figure 2d). Therefore, the drastic repression of bio gene expression during EHEC growth in the presence of microbiota metabolites is caused by the large amounts of biotin released by the gut microbiota, which EHEC can take up via its YigM transporter.

Figure 2. Microbiota-derived biotin represses expression of EHEC biotin biosynthesis genes. (a) RT-qPCR analysis of bioF expression during EHEC growth in two different microbiota culture supernatants (MET-1 and Donor 5), as well as minimal medium and rich medium controls. (b) Biotin concentration in the microbiota supernatants and control media as measured using the Abcam Biotin Quantitation Kit (Colorimetric). (c) bioB-lux reporter activity in wild-type and ΔyigM EHEC grown in M9 minimal medium with and without addition of external biotin (5 nM to 500 µm). (d) Activity of bioB-lux reporter during growth of wild-type and ΔyigM EHEC in MET-1 metabolites and control media. In panels a, c, and d, data represent mean and standard deviation of three biological replicate cultures; in panel b, data represent mean and SEM of three technical replicates.

Microbiota metabolites activate EHEC stress response genes

A second major category of EHEC genes whose expression is altered in the presence of gut microbiota metabolites is related to stress responses (Dataset S1; Figure 1b). Stress response genes whose expression was increased in the presence of gut microbiota metabolites in the RNA-Seq experiment include the oxidative stress response gene soxS, the envelope stress response genes cpxP and spy, and the stress-induced genes of unknown function bhsA and yhcN. We confirmed by RT-qPCR that these five genes are strongly upregulated during EHEC growth in MET-1 metabolites (Figure 3a). All five genes were also activated during growth in Donor 5 metabolites (Figure 3b). In particular, bhsA and yhcN were strongly upregulated (>5-fold) in response to both MET-1 and Donor 5 metabolite conditions. We therefore chose to study bhsA and yhcN in more detail in order to understand how they contribute to EHEC growth in the presence of competing gut bacteria.

Figure 3. EHEC stress response genes are activated during growth in the presence of microbiota metabolites. RT-qPCR analysis of stress response gene expression during EHEC growth in two different microbiota culture supernatants [MET-1 (a) and Donor 5 (b)], as well as minimal medium and rich medium controls. Data represent mean and standard deviation of three biological replicate cultures per condition.

Both BhsA (YcfR/ComC) and YhcN belong to the YhcN family of proteins, which are predicted to be small (generally <100 amino acids), envelope-localized proteins containing the domain of unknown function DUF1471^47,48. These proteins have been found exclusively in Enterobacteriaceae, typically with multiple paralogues per genome – nine paralogous YhcN family proteins are found in E. coli ⁴⁷. Previous studies in E. coli K-12 have linked both bhsA and yhcN to stress response and biofilm formation. For example, expression of both bhsA and yhcN is induced by exposure to hydrogen peroxide in E. coli K-12⁴⁹. In addition, bhsA expression is induced during biofilm growth⁵⁰ and during exposure to a variety of stressors including cadmium⁵¹, copper⁵², and heat shock⁵³. Furthermore, E. coli K-12 mutants lacking bhsA or yhcN have altered biofilm formation and are more sensitive to hydrogen peroxide and cadmium^49,54, suggesting that these proteins may play a role in a general stress response. To determine whether bhsA and yhcN play a similar role in EHEC, we measured biofilm formation of wild-type, ΔbhsA, ΔyhcN, and ΔbhsA ΔyhcN strains using the crystal violet assay, but did not find any significant difference between strains (Figure S1a). Similarly, we did not find any difference in hydrogen peroxide sensitivity between EDL933 and the ΔbhsA ΔyhcN mutant (Figure S1b), suggesting that these proteins may play a different role in EHEC than they do in E. coli K-12.

Since expression of bhsA and yhcN in EHEC is induced by exposure to microbiota metabolites in our synthetic communities that may also be encountered in the gut environment, we examined whether BhsA and YhcN play a role in host infection. Using a HeLa cell infection model, we observed that the ΔbhsA and ΔyhcN single and double mutants were all able to adhere to human cells and form pedestals characteristic of EHEC infection (Figure 4a). In fact, quantification of adherent EHEC cells from the microscopy images revealed that the ΔbhsA, ΔyhcN, and ΔbhsA ΔyhcN mutants all had significantly more adherent bacteria per HeLa cell than the wild-type strain (p < 0.0001, Dunnett’s multiple-comparison test), with an approximate two-fold increase in adherence in the ΔyhcN and ΔbhsA ΔyhcN mutants (Figure 4b). However, this increase in host cell adherence did not correlate with an increase in host cell death. We observed no difference in propidium iodide staining of HeLa cells infected with wild-type, ΔbhsA, ΔyhcN, and ΔbhsA ΔyhcN strains of EHEC (Figure 4c). Furthermore, the proportion of infected HeLa cells with activated caspase-3, indicating entry into apoptosis, did not differ significantly between wild-type or mutant-infected cells (Figure 4d). These results suggest that mutation of bhsA and yhcN primarily affects EHEC’s adherence properties rather than its cytotoxicity. As further evidence that mutation of bhsA and yhcN does not affect cytotoxicity, we found no increase in expression of the type III-secreted effector genes tir and espFU and the regulatory gene ler in the ΔbhsA and ΔyhcN mutant strains (Figure S2a), nor was there any difference in their profiles of type III secreted proteins when grown in vitro (Figure S2b). The reason for the increased cell adherence of the mutants is currently unknown, as the expression of pilus genes previously shown to have a role in host infection (ecpA, lpfA1, Z2200, and hcpA)⁵⁵ was similar between wild-type and mutant strains (Figure S2c).

Figure 4. Mutations in microbiota-activated genes bhsA and yhcN affect EHEC adherence to HeLa cells. (a) Representative immunofluorescence microscopy images of uninfected and EHEC-infected HeLa cells. Scale bar; 20 m. b) Quantification of adherent bacteria per HeLa cell, based on the immunofluorescence microscopy images. Data represent mean and SEM of 300 HeLa cells per strain. ****, p < 0.0001, Dunnett’s multiple comparison test. (c) Propidium iodide staining of EHEC-infected HeLa cells. Data represent mean and SEM from three biological replicates. (d) Caspase-3 activation in EHEC-infected HeLa cells. Data represent mean and SEM of three biological replicates.

To further examine the role of bhsA and yhcN in host infection, we turned to the C. rodentium model of mouse infection. Since EHEC is a human-specific pathogen that does not cause a representative infection in mice⁵⁶, C. rodentium – a natural murine pathogen – is frequently used to model A/E pathogenesis in an animal host⁵⁷. We assessed the ability of C. rodentium strains lacking bhsA and/or yhcN to colonize the cecum and colon of C57BL/6 mice. Similar to our finding that EHEC ΔyhcN mutants had increased adherence to HeLa cells, we found that the C. rodentium ΔyhcN and ΔbhsA ΔyhcN mutants had a slightly increased ability to colonize the gut of C57BL/6 mice (Figure 5a). Both the ΔyhcN and the ΔbhsA ΔyhcN mutant colonized the cecum at levels 2.0-fold higher than wild-type C. rodentium; in the colon, the ΔyhcN mutant colonized at a level 1.5-fold higher and the ΔbhsA ΔyhcN mutant 2.7-fold higher than the wild-type strain, although these differences were not statistically significant after correction for multiple comparisons (p > 0.05, Dunn’s multiple-comparison test). We next evaluated whether C. rodentium mutants have increased lethality in the susceptible C3H/HeJ mouse strain. We observed a significantly higher fecal bacterial burden in mice infected with the C. rodentium ΔyhcN mutant at 6 days post-infection (Figure 5b). Moreover, we found that mice infected with the ΔyhcN mutant succumbed to infection significantly faster than those infected with the wild-type C. rodentium (Figure 5c; p < 0.01). We also took advantage of the C. rodentium mouse infection model to determine whether A/E pathogens express bhsA and yhcN during infection of the host gut, where they are exposed to gut microbiota-produced metabolites. Using RT-qPCR, we found that both bhsA and yhcN were expressed at substantially higher levels during growth in the colon of C57BL/6 mice compared to in vitro growth in LB, in either the log phase or stationary phase (Figure 5d). This high expression of bhsA and yhcN in the host environment corresponds with their high expression in EHEC cultured in the presence of gut microbiota metabolites. The differences in adherence or colonization of host surfaces by the bhsA and yhcN mutants suggests that our RNA-Seq approach was successful in identifying microbiota-activated genes that are relevant to host infection by A/E pathogens.

Figure 5. Role of bhsA and yhcN in C. rodentium infection of the mouse gut. (a) Enumeration of C. rodentium colonizing the cecum and colon of C57BL/6 mice, 6 days post-infection. Each group represents the mean and standard deviation of N = 8 mice. LOD, limit of detection. (b) Fecal shedding of C. rodentium throughout the course of infection of C3H/HeJ mice. Minimum and maximum values are represented by short vertical lines of whiskers; the box signifies the upper and lower quartiles, and the short line within the box signifies the median. N = 10 mice per strain.   (c) Survival of C3H/HeJ mice infected with wild-type and mutant strains of C. rodentium. Mice were monitored daily and euthanized upon reaching the humane endpoint. **, p < 0.01, Mantel– Cox test.   (d) RT-qPCR analysis of expression of C. rodentium bhsA and yhcN during growth in vitro (in log phase or stationary phase, both in LB medium) or in vivo (in the distal colon of C57BL/6 mice, 9 days post-infection). Horizontal lines represent the mean of N = 3 replicate in vitro cultures or N = 4 mice.

Tn-Seq analysis of EHEC genes that affect fitness during growth in microbiota metabolites

The second major objective of our study was to identify EHEC genes that are important for growth in the presence of gut microbiota-produced metabolites. This objective was addressed using a Tn-Seq approach. We generated a Tn-Seq-compatible mariner transposon mutant library in EHEC strain EDL933 using the transposon donor plasmid pSAM-Ec³⁹. Tn-Seq analysis showed that the transposon mutant library contained 88,154 unique transposon insertion sites (85,707 in the chromosome and 2,447 in plasmid pO157). No obvious insertional hotspots were evident in either the chromosome or the plasmid (Figure S3). We noted two large (~66 kb) chromosomal regions lacking insertions (Figure S3a); these regions represent a duplication in the EDL933 chromosome to which no reads were mapped because we elected to map reads only to unique genomic locations.

In order to identify EHEC genes affecting fitness during growth in the presence of gut microbiota metabolites, we performed Tn-Seq on cultures of the EDL933 transposon mutant library grown in MET-1 metabolites or in control rich or minimal medium (similarly to the RNA-Seq experiment, all cultures were supplemented with concentrated M9 minimal medium). We defined genes that promote fitness in the presence of microbiota metabolites as those to which fewer transposon insertion site reads mapped in cultures grown in MET-1 metabolites compared to either of the control cultures (FDR <0.01); conversely, genes that reduce fitness in the presence of microbiota metabolites were defined as those to which significantly more reads mapped in the MET-1-grown EHEC cultures than in either of the controls (full selection criteria described in Materials and Methods). We identified 23 genes that promote fitness in the presence of microbiota metabolites and 5 genes that reduce fitness in these conditions (Figure 6; Dataset S2).

Figure 6. Identification of EHEC genes affecting fitness during growth in microbiota metabolites using Tn-Seq. Scatterplot shows differences in Tn-Seq reads mapped to each EHEC gene for EHEC cultures grown in MET-1 microbiota metabolites compared to the minimal medium control on the x-axis and compared to the rich medium control on the y-axis. Genes that were considered to promote fitness in microbiota metabolites are colored blue while genes that reduce fitness in microbiota metabolites are colored red (criteria described in Materials and Methods). Genes that are discussed further in the text are labeled.

Short-chain fatty acids are a key driver of EHEC’s response to microbiota-produced metabolites

In order to address our overall goal of understanding EHEC genetic adaptations to growth in the presence of the microbiota, we compared our lists of EHEC genes that were differentially expressed during growth in MET-1 metabolites according to RNA-Seq (objective 1) with those that affected EHEC fitness in these conditions according to Tn-Seq (objective 2) in order to identify any functional overlap between the datasets. Such overlaps might highlight genes or pathways that are particularly important for EHEC survival in the presence of the gut microbiota. We identified only one gene that appeared in both the RNA-Seq and the Tn-Seq datasets: udhA, which encodes a soluble pyridine nucleotide transhydrogenase that is responsible for oxidation of NADPH to maintain redox balance in the cell⁵⁸. Expression of udhA was increased in the presence of MET-1 metabolites, according to both RNA-Seq and RT-qPCR (Figure 7a; Dataset S1). In the Tn-Seq experiment, mutations in udhA were found to be highly deleterious to EHEC fitness during growth in microbiota metabolites (Dataset S2). In order to confirm this finding, we constructed an EDL933 ΔudhA::kan^R mutant by λ Red mutagenesis³⁸. When grown in pure culture, the ΔudhA::kan^R mutant had an obvious growth defect in MET-1 metabolites, but not in control rich or minimal media (Figure S4). We also performed a competitive growth experiment where wild-type and ΔudhA::kan^R strains of EHEC were inoculated into microbiota metabolite-containing or control media at a 1:1 ratio. Upon reaching stationary phase, cultures were plated on LB and LB+Kan to determine the relative proportions of wild-type and mutant cells. This competitive growth experiment confirmed the Tn-Seq finding that mutation of udhA is extremely detrimental to EHEC fitness in the presence of microbiota metabolites, with the mutant comprising ~50% of cells in control cultures but only ~ 0.1% of the cells grown in MET-1 metabolites (Figure 7b).

Figure 7. UdhA promotes EHEC growth in the presence of microbiota metabolites including acetate. (a) RT-qPCR analysis of udhA expression during EHEC growth in MET-1 microbiota culture supernatant and minimal and rich medium controls. Data represent mean and standard deviation of three biological replicate cultures per condition. (b) and (c) Competitive growth of EDL933 and ΔudhA:kan^R in MET-1 microbiota culture supernatant and minimal and rich medium controls (b), or M9-glucose minimal medium with and without 40 mM acetate (c). Competitive index is calculated as the proportion of mutant cells at the end of growth divided by the proportion of mutant cells at the beginning of growth. (d) Ratio of intracellular reduced NADPH to oxidized NADP+, measured using the EnzyChrom NADP+/NADPH Assay Kit. Wild-type and ΔudhA::kan^R EHEC strains were grown in M9-glucose minimal medium with and without 40 mM acetate. Data represent the mean of two to three biological replicate cultures per strain and medium condition.

We next turned to the question of why UdhA is so important for EHEC growth in the presence of microbiota metabolites. The primary role of UdhA is to oxidize NADPH to restore the cellular pool of NADP+ when cells are grown in conditions that generate an excess of reduced NADPH relative to anabolic needs⁵⁸. Previous research has identified several growth substrates that cause an accumulation of NADPH in E. coli; notably, this includes growth on acetate⁵⁸. Since acetate is a well-known end-product of fermentation by various gut microbiota species⁵⁹ that is present in the MET-1 supernatant at a concentration of 40.03–44.03 mM⁶⁰, we reasoned that EHEC utilization of acetate present in the MET-1 supernatant could necessitate UdhA activity to restore the cellular NADP+ pool under these conditions. As predicted, addition of 40 mM acetate to M9-glucose minimal medium had little effect on the growth of wild-type EDL933, but reduced the growth of the ΔudhA::kan^R mutant (Figure S4d). Competitive growth experiments showed that the ΔudhA::kan^R mutant is less fit than wild-type EHEC in medium containing acetate (Figure 7c); however, the fitness defect was not as large as the one observed during growth in MET-1 metabolites, indicating that there are likely other microbiota-produced metabolites in addition to acetate that inhibit the growth of the ΔudhA::kan^R mutant. To test our hypothesis that reduced growth of the ΔudhA::kan^R mutant in the presence of acetate is the result of an accumulation of reduced NADPH, we measured the intracellular NADPH/NADP+ ratios of wild-type and ΔudhA::kan^R strains grown in minimal medium with and without acetate (Figure 7d). Addition of acetate to wild-type EDL933 cultures did not significantly affect the intracellular NADPH/NADP+ ratio (p > 0.05, Tukey’s multiple- comparison test). However, the intracellular NADPH/NADP+ ratio was significantly higher in ΔudhA::kan^R cultures grown in the presence of acetate than those grown without acetate (p < 0.01, Tukey’s multiple-comparisons test). Together, these data suggest that UdhA improves EHEC growth in the presence of microbiota metabolites by oxidizing the excess NADPH that accumulates when EHEC takes up and metabolizes microbiota-derived acetate, and possibly other metabolites.

Although udhA was the only gene that was both significantly differentially expressed in the presence of microbiota metabolites according to RNA-Seq and also affected EHEC fitness under these conditions according to Tn-Seq, we also identified another case of functional overlap between the two datasets. We found that both genes encoding the NtrBC two-component system (glnL and glnG) promoted fitness in microbiota metabolites (Figure 6; Dataset S2). While expression of these two genes did not meet the cutoff to be considered upregulated by microbiota metabolites in the RNA-Seq screen, the expression of several NtrBC-regulated genes, such as glnK and amtB was strongly upregulated (Dataset S1), suggesting that this two-component system plays an important role during EHEC growth in these conditions.

The NtrBC two-component system has been described as a nitrogen starvation response, since many NtrC-regulated genes are involved in scavenging nitrogen-containing compounds from the environment and increasing the intracellular pools of glutamine and glutamate⁶⁰. The role of NtrBC during nitrogen starvation does not immediately explain its importance during growth in the presence of microbiota metabolites, since all of the media conditions used in our RNA-Seq and Tn-Seq experiments are nitrogen-replete (all media were supplemented with M9-glucose, giving a final ammonium concentration of at least 18.7 mM). Therefore, we aimed to confirm that the NtrBC system was involved in fitness in media containing microbiota metabolites, even under nitrogen-replete conditions. We confirmed by RT-qPCR that expression of the NtrBC-regulated genes glnK, amtB, glnA, and nac ⁶⁰ was increased in the presence of MET-1 metabolites (Figure 8a), indicating that the NtrBC two-component system is activated under these conditions. We also confirmed by competitive growth experiments that a-ΔglnLG::kan^R mutant constructed by λ Red mutagenesis had reduced fitness during growth in MET-1 metabolites, but not during growth in control minimal and rich media (Figure 8b). Although our media conditions were nitrogen-replete, recent studies have shown that the NtrBC two-component system also senses perturbations of carbon metabolism⁶¹. Therefore, we next examined whether the high concentrations of short-chain fatty acids (SCFAs) present in the MET-1 supernatant (acetate, 40.03–44.03 mM; propionate, 19.49–23.22 mM; butyrate, 13.87–21.25 mM)⁶² might be responsible for activating the NtrBC two-component system via altered carbon nutritional status. The addition of acetate and propionate, both singly and especially in combination, at concentrations similar to those in the MET-1 supernatant, activated the EHEC NtrBC two-component system, as measured by the expression of a glnK-lux transcriptional reporter (Figure S5). However, the ΔglnLG::kan^R mutant displayed no fitness defect in the medium containing these SCFAs (data not shown). Therefore, although microbiota-produced SCFAs might be responsible for activation of the NtrBC two-component system, it remains to be determined how this system promotes fitness in the presence of microbiota metabolites.

Figure 8. The NtrBC two-component system promotes EHEC growth in the presence of microbiota metabolites. (a) RT-qPCR analysis of expression of NtrBC-regulated genes during EHEC growth in MET-1 microbiota culture supernatant and minimal and rich medium controls. Data represent mean and standard deviation of three biological replicate cultures per condition. (b) Competitive growth of EDL933 and ΔglnLG::kan^R in MET-1 microbiota culture supernatant and minimal and rich medium controls. Competitive index is calculated as the proportion of mutant cells at the end of growth divided by the proportion of mutant cells at the beginning of growth.

Discussion

As evidenced by the thousands of EHEC infections per year in the United States alone¹, EHEC is clearly able to colonize the human gut even in the presence of an established gut microbial community. However, the mechanisms by which EHEC senses and responds to the presence of these competing microbes are currently not well understood. In this study, we characterized EHEC’s response to gut microbiota-produced metabolites in order to better understand its genetic and metabolic adaptations to growth in the presence of the gut microbiota. By combining RNA-Seq, Tn-Seq, and classical loss-of-function studies, we provide a comprehensive look at how EHEC responds to an established therapeutic microbial community, MET-1.

Our first objective was to identify EHEC genes that are differentially expressed during exposure to microbiota metabolites. Using RNA-Seq, we found two major categories of microbiota metabolite-responsive genes: those genes related to metabolism (including biotin biosynthesis) and those encoding stress responses (Figure 1b; Dataset S1). We will discuss each of these categories of genes, and how they may affect EHEC’s ability to infect the host gut, in turn in the following paragraphs.

Metabolic cross-feeding and competition for nutrients between pathogens and gut microbiota have been well established in the case of carbohydrates^16,63, but not as well studied in the context of other nutrients, such as vitamins. The dramatic downregulation of biotin biosynthetic enzymes in the presence of metabolites from two different gut microbiota communities establishes biotin as a key microbiota metabolite affecting pathogen behavior (Figure 2). Accumulation of multiple B vitamins in the supernatant of a fecal microbial community has previously been observed⁶⁴, although it remains unclear whether the biotin and other vitamins in the microbiota culture supernatant are secreted by live cells or simply released upon cell lysis. Even when mice are fed chow with low vitamin content, B vitamin auxotrophic members of the mouse gut microbiota are able to persist, which strongly suggests that sharing of B vitamins between gut microbes also occurs in vivo ⁶⁴. Interestingly, the presence of biotin biosynthesis is a metabolic feature of Bacteroides-enriched microbiomes, further indicating that the makeup of the gut microbiota community can shape EHEC’s response to the gut environment⁶⁵. Our data indicate that EHEC can take up microbiota-produced biotin from the gut environment via YigM, thereby avoiding de novo biotin synthesis and providing an energetic advantage to the pathogen. In addition, it has been reported that high biotin concentration is an environmental cue that causes downregulation of the LEE genes in EHEC⁶⁶. Since biotin production by the microbiota would be expected to lead to higher biotin levels in the lumen of the colon compared to the epithelial surface, this regulation may also help to ensure that EHEC represses expression of its virulence factors until it is in close contact with host cells.

The second major category of genes that were differentially expressed in the presence of gut microbiota metabolites relates to stress responses. We found that numerous stress response genes were upregulated during EHEC exposure to gut microbiota metabolites, including the oxidative stress response gene soxS and the envelope stress response genes cpxP and spy (Figure 3). Activation of multiple stress responses in the presence of competing bacteria is in agreement with the hypothesis of competition sensing, which states that a major function of bacterial stress responses is to sense nutrient limitation or cell damage caused by the presence of competitors, and to mount adaptive responses, which can include both defensive functions (e.g. repairing cellular damage) and counter-attack (e.g. producing toxins)⁶⁷. In our experiments, EHEC is likely sensing the presence of harmful metabolites released by competing gut bacteria. More specifically, the oxidative stress response could be activated due to the presence of redox-cycling compounds, which have been shown to increase expression of soxS by oxidizing the iron-sulfur cluster in the transcription factor SoxR^68,69. Although the gut microbiota’s ability to produce redox-cycling compounds has not yet been examined, a variety of such compounds are known to be secreted by both plants and bacteria; for example, Pseudomonas species secrete numerous redox-active phenazine compounds that inhibit the growth of competitors⁶⁹. Expression of the envelope stress response gene cpxP is activated by the CpxAR two-component system⁷⁰, while spy expression is activated by both the CpxAR and BaeSR envelope stress responses⁷¹. These envelope stress responses are known to be activated by a variety of envelope-damaging agents; however, one metabolite that activates both responses is indole⁷¹. Numerous members of the gut microbiota produce tryptophanase enzymes capable of generating indole; indeed, indole is present in high concentrations in the lumen of the mouse colon and also in human feces²⁵. The ability to sense harmful metabolites released by the gut microbiota and activate appropriate stress responses is likely beneficial to A/E pathogens in the challenging environment of the mammalian gut. For example, it has been reported that C. rodentium mutants lacking the Cpx envelope stress response have a greatly reduced ability to colonize the mouse gut^72,73.

The EHEC stress response genes stimulated by exposure to gut microbiota metabolites appear to extend beyond the well-characterized oxidative and envelope stress responses. bhsA and yhcN were two of the genes most strongly upregulated during EHEC exposure to microbiota metabolites in vitro (Figures 1b and 3) and were also highly expressed by C. rodentium in the mouse colon (Figure 5b), indicating that upregulation of these genes may constitute a conserved response to competing bacteria. BhsA and YhcN have been described as potential stress response proteins due to their differential expression and mutant phenotypes related to oxidative stress and biofilm formation in E. coli K-12^49,54. However, in our study EHEC ΔbhsA and ΔyhcN mutants did not appear to share these altered biofilm formation and hydrogen peroxide sensitivity phenotypes (Figure S1). The differing phenotypes of bhsA and yhcN mutants in different E. coli strain backgrounds could result from genetic redundancy; the EHEC EDL933 genome encodes more than 1300 genes not found in E. coli K-12 MG1655⁴¹ and may therefore contain additional stress resistance genes. One common trait among bhsA and yhcN mutants in various E. coli genetic backgrounds appears to be altered adherence properties. Here, we found increased adherence to HeLa cells by EHEC ΔbhsA and ΔyhcN mutants (Figures 4a,b) and a possible increase in adherence of C. rodentium ΔyhcN and ΔbhsA ΔyhcN mutants to the mouse cecum and colon (Figure 5a), which parallels the previously described increased biofilm formation in E. coli K-12 ΔbhsA and ΔyhcN ^49,54. The mechanism underlying the increased adherence of EHEC ΔbhsA and ΔyhcN is currently unknown and does not appear to result from altered expression of pili or T3SS genes (Figure S2). Mutation of bhsA in E. coli K-12 increases cell surface hydrophobicity⁵⁴, although changes in hydrophobicity were not observed in bhsA mutants of EHEC strains Sakai or TW14359⁷⁴. Further study is needed to determine whether BhsA and YhcN affect surface properties of EHEC EDL933 cells, and how these changes might affect growth or survival in the gut environment.

Our second objective was to identify genes that promote EHEC survival and growth in the presence of microbiota-produced metabolites. Using a Tn-Seq approach, we found that EHEC’s growth under these conditions requires several genes (udhA, glnLG) that may help it to cope with the presence of high amounts of microbiota-produced SCFAs. SCFAs are excreted as a fermentative end-product by many gut microbiota species. The most abundant SCFA in the gut is acetate, the concentration of which is estimated to reach ~40–90 mM in the human colon⁷⁵. E. coli is known to both produce and consume acetate, being traditionally believed to excrete acetate (dissimilation) during growth on glycolytic carbon sources, with consumption of acetate (assimilation) occurring only after glucose has been depleted and catabolite repression relieved⁷⁶. However, it was shown that E. coli can take up and metabolize extracellular acetate when present at a concentration of at least 8 mM, even in the presence of excess glucose or other glycolytic substrates such as fucose⁷⁷. Since the concentration of acetate in the human colon is believed to be well in excess of this threshold, it is likely that EHEC undergoes net consumption of acetate in this environment, even if other carbon substrates are available. However, high concentrations of acetate have growth-suppressive effects, such as inhibition of glycolysis and the TCA cycle⁷⁸, and therefore EHEC has likely evolved strategies to mitigate the negative effects of acetate consumption.

One gene that may facilitate EHEC growth in the presence of microbiota-produced acetate is udhA, which encodes a soluble pyridine nucleotide transhydrogenase. We found that expression of udhA was increased during growth in MET-1 metabolites (Figures 1b, 7a, Dataset S1); in addition, mutations in or deletion of udhA sharply reduced fitness during growth in these conditions (Figures 6, 7b, Dataset S2). UdhA is responsible for regenerating the cellular NADP+ pool when the supply of NADPH exceeds the anabolic demand, a condition that is known to occur during growth on acetate as a carbon source⁵⁸. Accordingly, we found that the addition of acetate to its growth medium reduced the fitness of the udhA mutant (Figure 7c), and that the udhA mutant, but not wild-type EHEC, had an increased intracellular NADPH:NADP+ ratio when grown in the presence of exogenously added acetate (Figure 7d). However, it should be noted that addition of acetate alone to the growth medium did not reduce the fitness of the udhA mutant to the same extent as the MET-1 metabolites, which suggests that UdhA may play a role in responding to other microbiota-derived metabolites that can cause a redox imbalance, in addition to acetate.

Two other genes that strongly promoted fitness in the presence of gut microbiota metabolites were glnL and glnG, encoding the nitrogen starvation response of the two-component system NtrBC (Figures 6, 8b, Dataset S2). Moreover, several genes in the NtrBC regulon, including glnK and amtB, were strongly upregulated during growth in these conditions (Figures 1b, 8a, Dataset S1), emphasizing the importance of the NtrBC system in EHEC’s response to gut microbiota metabolites. Although NtrBC is primarily known for its role in responding to nitrogen-limiting conditions⁶⁰, it has also been involved in promoting antibiotic-tolerant persister cells⁷⁹. We hypothesized that nitrogen limitation was not responsible for the activation of NtrBC in the presence of MET-1 metabolites, since all of our cultures were supplemented with ammonium chloride to a sufficient concentration to be considered nitrogen-replete⁶¹. Therefore, we explored other possible explanations for why NtrBC was activated by MET-1 metabolites. We found that acetate and propionate, and especially both SCFAs combined, activated the expression of the NtrBC-dependent gene glnK (Figure S5). There are several possible mechanisms by which SCFAs might cause activation of NtrBC. First, the enzyme AckA can convert acetate to acetyl phosphate⁷⁶, which can directly phosphorylate NtrC independently of NtrB and thereby activate transcription of the NtrBC regulon⁸⁰. In addition, elevated intracellular α-ketoglutarate can activate NtrBC even when cells are not nitrogen-starved, suggesting that NtrBC senses both the carbon and nitrogen nutritional status of the cell⁶¹. Since α-ketoglutarate is a key metabolite of the TCA cycle, any changes in flux throughout this cycle due to alterations in substrate or coenzyme availability may have the potential to affect NtrBC activity. Notably, catabolism of both acetate and propionate feeds into the TCA cycle, which is a second way that these SCFAs may contribute to NtrBC activation. Interestingly, although SCFAs activated NtrBC, we did not observe any loss of fitness of the glnLG mutant during growth in a minimal medium containing SCFAs at concentrations similar to those found in the MET-1 metabolites (data not shown), whereas the fitness of this mutant was consistently reduced when grown in the full MET-1 supernatant (Figure 8b). These observations suggest that activation of NtrBC and the decreased fitness of the glnLG mutant may be caused by different components of the MET-1 metabolite mixture.

Our study relied primarily on an in vitro approach of culturing EHEC in the presence of metabolites from a gut microbiota community. A major strength of this approach is that it allowed us to avoid the confounding effects of microbiota-host and pathogen-host interactions that necessarily occur in in vivo studies. For this reason, we can directly attribute the changes in EHEC gene expression or the fitness of EHEC mutants that we observed in our RNA-Seq and Tn-Seq experiments to the effects of microbiota-produced metabolites, and in some cases, we were able to identify specific metabolites that were likely mediating these effects (e.g. biotin and acetate). Conversely, this approach fails to capture all of the effects of the host and the diet on the metabolic milieu of the colon; moreover, it does not capture any direct effects of live microbiota cells upon EHEC. For example, several examples of microbiota species using a type VI secretion system to target and eliminate pathogens have now been described^81–83. An important future direction for this work will be to examine the importance of each of the identified microbiota-responsive genes in vivo during A/E pathogen-host infection.

In summary, this study demonstrates that EHEC’s response to gut microbiota-produced metabolites extends well beyond the previously observed changes in virulence gene expression to also encompass metabolic adaptations and induction of stress responses that may be beneficial to EHEC as it colonizes the colon. Understanding EHEC’s response to the gut microbiota is critically important because microbiota metabolites hold enormous potential as novel anti-infective therapeutics. For example, increasing the colonic concentration of biotin has already been proposed as a method for reducing EHEC virulence gene expression, thereby preventing infection⁶⁶. Since oral consumption of biotin may not sufficiently increase colonic biotin levels due to its absorption in the small intestine⁶⁶, our study suggests that a probiotic approach making use of biotin-excreting microbiota strains may represent an alternative means to achieve this goal. New strategies for treatment or prevention of EHEC infection are urgently needed, since treatment of EHEC infection with antibiotics is not recommended due to the ability of some antibiotics to induce the SOS response and thereby increase the release of Shiga toxins and the risk of hemolytic uremic syndrome⁸⁴. Ultimately, an improved understanding of the intricate interplay between host, pathogen, and microbiota will be essential for reducing the disease burden imposed by gastrointestinal pathogens.

Author contributions

S.L.V, A.S.P., B.B.F. designed the study. M.C.D. and E.A.V. prepared and provided materials for the study. S.L.V., A.S.P., A.S.S., L.V.B. designed and performed the in vitro experiments. S.L.V., A.S.P., S.E.W. conducted the animal experiments. S.L.V., S.P.W.V. performed the bioinformatics analyses. S.L.V., A.S.P., S.E.W. performed the data analysis. S.L.V, A.S.P., S.E.W., A.S.S., S.P.W.V., M.C.D., L.V.B., A.J.G., D.J.M., E.A.V., B.B.F. discussed the data and wrote the manuscript. E. A.-V. is co-founder and CSO of NuBiyota, a company seeking to create ‘microbial ecosystem therapeutics’ for the treatment of disease in humans.

Acknowledgments

We are grateful to Tracy Raivio (University of Alberta) for sharing the pJW15 luminescence reporter plasmid and to Matthew Mulvey (University of Utah) for sharing the pSAM-Ec transposon donor plasmid. We would like to thank Wanyin Deng for his valuable feedback on the manuscript. S.L.V. was the recipient of a Killam Postdoctoral Research Fellowship, a Michael Smith Foundation for Health Research Postdoctoral Fellowship, and a CIHR Postdoctoral Fellowship. A.S.P was the recipient of a CIHR Postdoctoral Fellowship [MFE-164659]. B.B.F. is a University of British Columbia Peter Wall Distinguished Professor.

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability statement

The datasets generated during the current study are available in the NCBI Sequence Read Archive (SRA). The RNA sequencing data is available under BioProject ID PRJNA894183 (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA894183) and the Tn-seq data is available under BioProject ID PRJNA894187 (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA894187). Additional data that support the findings of this study are available from the corresponding author [B.B.F.] upon reasonable request.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19490976.2023.2190303.

Additional information

Funding

This work was supported by grants from the Canadian Institutes of Health Research (MOP-136976; FDN-159935) to B.B.F.