Citrobacter rodentium induces rapid and unique metabolic and inflammatory responses in mice suffering from severe disease

Abstract The mouse pathogen Citrobacter rodentium is used to model infections with enterohaemorrhagic and enteropathogenic Escherichia coli (EHEC and EPEC). Pathogenesis is commonly modelled in mice developing mild disease (e.g., C57BL/6). However, little is known about host responses in mice exhibiting severe colitis (e.g., C3H/HeN), which arguably provide a more clinically relevant model for human paediatric enteric infection. Infection of C3H/HeN mice with C. rodentium results in rapid colonic colonisation, coinciding with induction of key inflammatory signatures and colonic crypt hyperplasia. Infection also induces dramatic changes to bioenergetics in intestinal epithelial cells, with transition from oxidative phosphorylation (OXPHOS) to aerobic glycolysis and higher abundance of SGLT4, LDHA, and MCT4. Concomitantly, mitochondrial proteins involved in the TCA cycle and OXPHOS were in lower abundance. Similar to observations in C57BL/6 mice, we detected simultaneous activation of cholesterol biogenesis, import, and efflux. Distinctly, however, the pattern recognition receptors NLRP3 and ALPK1 were specifically induced in C3H/HeN. Using cell‐based assays revealed that C. rodentium activates the ALPK1/TIFA axis, which is dependent on the ADP‐heptose biosynthesis pathway but independent of the Type III secretion system. This study reveals for the first time the unfolding intestinal epithelial cells' responses during severe infectious colitis, which resemble EPEC human infections.

C3H/HeN, which encodes intact TLR4) develop severe fatal diarrhoeal disease, which arguably provides a more clinically relevant model for human paediatric diarrhoeal disease, where enteropathogenic Escherichia coli (EPEC) infection can be lethal (Liu et al., 2016). However, due to the availability of mouse resources (e.g., knockout mice), most of our current knowledge of C. rodentium infection comes from studies in C57BL/6 mice, whereas little is currently known about pathogen-host interactions in mice developing severe disease.
The infection cycle of C. rodentium in C57BL/6 mice is divided into four defined phases ). An initial establishment phase (1-3 days post infection [DPI]), where C. rodentium resides in the caecal patch, is followed by an expansion phase (4-8 DPI) in which C. rodentium colonises the colon, initially adhering sporadically to the upper surface of the crypts and then expanding along the entire colonic mucosa. C. rodentium reaches peak bacterial load and the beginning of steady-state phase at 8 DPI, and the bacterial clearance phase begins at 12 DPI. During the clearance phase, elimination of C. rodentium from the mucosa is mediated by serum IgG and phagocytosis, whereas in the lumen, the pathogen is outcompeted by commensals (Kamada et al., 2015;Masuda et al., 2008). In contrast, C3H/HeN mice are unable to clear C. rodentium infection, which causes significant weight loss and dehydration, with animals reaching their humane endpoint at 10-12 DPI (Vallance, Deng, Jacobson, & Finlay, 2003).
Infection of both C57BL/6 and C3H/HeN mice is dependent on a Type III secretion system (T3SS), which injects bacterial proteins directly into the cytosol of intestinal epithelial cells (IECs), where they subvert multiple signalling pathways (Pinaud, Sansonetti, & Phalipon, 2018;Wong et al., 2011). One of the hallmarks of C. rodentium infection is induction of attaching and effacing lesions, which are characterised by effacement of the brush border microvilli and intimate attachment of the pathogen to the apical surface of IECs (Mundy et al., 2005). A second hallmark of C. rodentium infection is colonic crypt hyperplasia (CCH), a tissue regeneration response manifested by proliferation of IECs (Berger et al., 2018;Collins et al., 2014). IECs comprise a number of different cell types that form a defensive barrier that is replenished every 3 to 5 days (Barker, 2014). Cell renewal is fuelled by the canonical self-regulating signalling pathways that maintain gut homeostasis including Wnt, ZNRF3, and its paralogue RNF43 (Beumer & Clevers, 2016;Gehart & Clevers, 2019;Hao et al., 2012). At the base of the colonic crypts lie deep crypt secretory cells and LGR5+ stem cells, which give rise to a zone of rapidly dividing transit amplifying cells that terminally differentiate into absorptive enterocytes, goblet cells, enteroendocrine cells, and tuft cells (Gehart & Clevers, 2019;Johansson & Hansson, 2016;Sasaki et al., 2016). C. rodentiuminduced CCH is fuelled by Rspo3-mediated Wnt signalling in both C57BL/6 and C3H/HeN mice. In contrast, Rspo2 is specifically induced in C3H mice, amplifying Wnt signalling and leading to excessive expansion of transit amplifying cells, which are poorly differentiated, characterised by low abundance of the colonocyte markers SLC26A3 and CA4 responsible for chloride exchange and fluid absorption Kang et al., 2018;Kang, Yousefi, & Gruenheid, 2016;Papapietro et al., 2013;Teatero et al., 2011).
In this study, we applied a combination of in vivo imaging, -omics analysis, and cell biology to study host and microbiome responses to C. rodentium infection in C3H/HeN mice. This revealed rapid and uniform colonisation of C. rodentium at 3 DPI, 3 days earlier than in mice suffering from mild disease (C57BL/6). Earlier colonisation coincided with robust induction of host responses as early as 2 DPI and distinct upregulation of a unique repertoire of metabolic and immune related genes, including LDHA, MCT4, NLRP3, and ALPK1, specifically in C3H/HeN IECs.
2 | RESULTS 2.1 | C. rodentium establishes rapid colonic infection in C3H/HeN mice We monitored colonisation and the health status of the host following oral gavage of C3H/HeN mice with the bioluminescent C. rodentium strain ICC180. Enumeration of faecal colony-forming units (CFU) revealed that C. rodentium shedding reached levels of 5 × 10 8 CFU g −1 stool by 3 DPI and peaked at 2 × 10 9 CFU g −1 stool at 6 DPI (Figure 1 a). From 6 DPI, shedding plateaued, coinciding with signs of morbidity, including weight loss and increased faecal water content (Figure 1b,c).
All mice reached the predefined endpoint by 11 DPI. Analysis of temporal colonic colonisation by staining detected low and sporadic levels of C. rodentium on the distal colonic mucosa from 2 DPI and uniform colonisation at 3 DPI, which was maintained until 8 DPI (Figure 1d).
For comparison, uniform colonisation of C57BL/6 mice was only observed from 6 DPI .
We used live in vivo bioluminescent imaging (BLI) to compare infection dynamics of C3H/HeN and C57BL/6 mice. This revealed caecal colonisation of C3H/HeN mice at 1 DPI with extensive colonisation of both the caecum and colon at 3 DPI, which was maintained until 8 DPI (Figure 1e). In contrast, following a longer establishment phase in the caecum (1-4 DPI), colonisation in C57BL/6 mice was largely confined to the colon from 5 DPI with significantly less bioluminescent signal in the caecum and colon (Figure S1A-C; Hopkins et al., 2019;Wiles, Pickard, Peng, MacDonald, & Frankel, 2006). Ex vivo BLI revealed that ICC180 had colonised the caecal tissue and the mid and distal colon, extending into the proximal colon of C3H/HeN mice ( Figure 1f). Colonisation covered 77.0 ± 7.9% of the colon length at 8 DPI (Figure 1f,g), significantly higher than the 48.2 ± 7.3% observed in C57BL/6 mice at the same time point. Overall, these data suggest that compared with C57BL/6 mice, the pathophysiology of C3H/HeN mice is mediated by earlier C. rodentium colonisation with increased coverage of the colonic and caecal tissue.
2.2 | The epithelial proteome of C3H/HeN mice responds swiftly to C. rodentium To study global changes in the biological processes of C3H/HeN IECs in response to C. rodentium infection, we carried out quantitative mass spectrometry-based proteomic analysis. IEC proteins were extracted from at least four mice colonised to levels greater than 1 × 10 8 CFU g −1 of stool at 6 DPI. Protein lysate of two biological repeats were pooled at a 1:1 ratio and used to quantify infectioninduced fold change (FC) in protein abundance compared with an uninfected control ( Figure S2). As a control, we simultaneously FIGURE 1 Citrobacter rodentium infection in C3H/HeN mice. Temporal faecal shedding (a), changes to body weight (b), and increased faecal water content (c) in individual infected mice. Solid connecting line indicates the mean. * P < .05 (Students t test). (d) Representative colonic sections from C3H/HeN mice stained for C. rodentium (white) and DNA (blue) at 0, 1, 2, 3, 6, and 8 DPI (n ≥ 4). Sporadic colonisation is seen at 2 DPI (arrows); uniform colonisation is seen from 3 DPI. Scale bar 200 μm. (e) Representative in vivo images of C3H/HeN mice infected with C. rodentium ICC180 showing caecal colonisation at 1 DPI and robust colonic colonisation from 3 DPI (n ≥ 7). (f) Ex vivo tissue images of three representative C3H/HeN (n = 10) and C57BL/ 6 mice (n = 5) at 8 DPI, showing extended colonic colonisation of the former. The scale bar in e and f indicates signal intensity (photons s −1 cm −2 sr −1 ). (g) Quantification of the colon length colonised by C. rodentium as depicted in (f). **** P < .0001 (Students t test) analysed the C57BL/6 IECs' proteome at 6 DPI. Comparative gene set enrichment analysis of C3H/HeN and C57BL/6 revealed largely conserved changes to IEC processes upon infection ( Figure S3).
The KEGG pathways, Staphylococcus aureus infection, complement and coagulation cascades, and phagosome were found to be opposingly regulated at 6 DPI; however, the differential response of proteins assigned to these pathways converged by 8 DPI (PRIDE identifier: PXD005004; Berger et al., 2017). A comprehensive analysis of the C57BL/6 proteome has recently been reported elsewhere (Berger et al., 2017;Hopkins et al., 2019).
We quantified a total of 11,602 proteins in C3H/HeN mice, of which 9,184 were mapped to Mus musculus and 2,418 to C. rodentium genes (peptide false discovery rate [FDR] <1%). Proteins with altered abundance at or above 1.5-fold (0.59 Log2 FC) compared with the uninfected control samples were considered changed upon infection.
Gene set enrichment analysis revealed mitochondrial associated processes such as fatty acid elongation (enrichment score: −0.74, P value: .003), TCA cycle (enrichment score: −0.59, P value: 8.29E-11), and OXPHOS (enrichment score: −0.46, P value: 2.51E-15) amongst the most downregulated pathways (Figure 2a). Conversely, steroid FIGURE 2 Citrobacter rodentium subverts metabolism and cell proliferation in C3H/HeN IECs. (a) Log2 fold change values of proteins ranked from most downregulated (left) to most upregulated (right) in abundance in infected samples compared with uninfected (top panel). KEGG pathways determined as significantly enriched following 1D enrichment analysis of the infected IEC proteome compared with uninfected samples and ranked from most negatively enriched (i.e., most downregulated; top) to most positively enriched pathways (i.e., most upregulated; bottom). Individual proteins associated with ranked KEGG pathways are highlighted in the heat-map adjacent to the corresponding KEGG pathway (lower panel). (b) Crypt length measurements over time showing increased CCH at 3 DPI followed by a further expansion at 6 DPI (each point represents the mean crypt length of an individual mouse). (c) Quantification of the PCNA-positive zone calculated as a percentage of the total crypt length (each point represents the mean crypt length of an individual mouse). (d) Representative immunostaining of PCNA (green), C. rodentium (pink), and DNA (blue) from C. rodentium-infected colonic sections, showing an increased proliferation zone at 3 DPI followed by a further expansion at 6 DPI (n ≥ 4). Scale bar 50 μm biosynthesis (enrichment score: 0.44, P value: 0.003), DNA replication (enrichment score: 0.61, P value: 1.14E-08), and cell cycle (enrichment score: 0.31, P value: 9.96E-07) associated proteins were found in significantly higher abundance in infected IECs ( Figure 2a).
As pathway analysis revealed significant enrichment of processes involved in cell proliferation, we quantified the temporal induction of CCH in infected C3H/HeN mice. We found that CCH was induced in two stages: an initial expansion was observed at 3 DPI, which was followed by a further increase in crypt length at 6 DPI that remained which is implicated in import of microbiota-generated butyrate that feeds the TCA cycle (Berger et al., 2017).

| C. rodentium induces conserved changes to cholesterol homeostasis
Cholesterol is an important component of de novo membrane biogenesis during cell proliferation as well as lipid rafts, which amplify downstream TLR signalling through proximity-induced receptor interactions (Litvinov, Savushkin, & Dergunov, 2018;Silvius, 2003). We have recently reported that in C57BL/6 mice infected with C. rodentium, SREBP2 was activated in IECs at 8 DPI, leading to increased abundance of proteins of the cholesterol biosynthetic pathway and import, including HMGCR, LDLR, and PCSK9 (Berger et al., 2017;Hopkins et al., 2019). The proteomic data of the C3H/HeN mice also showed upregulation of the steroid biosynthetic pathway at 6 DPI (Figure 5a), including HMGCR (1.34 Log2 FC), LDLR (1.66 Log2 FC), and PCSK9 (1.63 Log2 FC; Figure 5b). However, using qRT-PCR to follow temporal gene expression revealed that contrary to the proteome, on the transcriptional level, the mRNA levels of Hmgcr, Pcsk9, and Ldlr did not significantly change up to 6 DPI (although Ldlr showed an initial increase in transcription at 2 DPI) and decreased at 8 DPI, specifically for Hmgcr and Ldlr  (c) and Hmgcr (d) transcripts decreases at 8 DPI, respectively, whereas Pcsk9 is unchanged (e). The abundance of Abcg5 decreases from 2 DPI (f), whereas that of Abca1 increases from 6 DPI (g). Each point represents an individual mouse at 0, 1, 2, 3, 6, and 8 DPI. (h) A gradual increase in faecal cholesterol measured at 0, 3, and 6 DPI in individual mice; each point indicates an individual mouse. * P ≤ .05, ** P ≤ .01, *** P ≤ .001, and **** P ≤ .0001 (one-way ANOVA) of HMGCR, PCSK9, and LDLR protein abundance may be regulated posttranscriptionally .
In parallel, the abundance of the basolateral cholesterol transporter ABCA1, which is regulated by the transcription factor LXR/RXR, and the cholesterol binding protein, APOA1 (involved in reverse cholesterol transport), were also higher in the proteomes of infected, compared with uninfected, IECs (3.11 and 1.95 Log2 FC for ABCA1 and APOA1, respectively, Figure 5b; Murthy, Born, Mathur, & Field, 2002). Temporal gene expression profiling revealed that while the level of Abcg5 transcript, which is also controlled by LXR/RXR and was undetected in the proteome, decreased as early as 2 DPI, expression of Abca1 was upregulated at 6 DPI and maintained at 8 DPI (Figure 5f,g). Indeed, increased faecal cholesterol was detectable by 3 DPI, which reached significance by 6 DPI (Figure 5h). Taken together, these results show that similarly to C57BL/6 mice, cholesterol biogenesis and cholesterol efflux are simultaneously activated in the IECs of C3H/HeN mice.

| C. rodentium infection triggers swift IL-22 responses in IECs of C3H/HeN mice
Cholesterol is an essential component of cellular membranes, needed for sustainable cell proliferation. In addition, proliferation of IECs is a tissue repair response triggered by IL-22, which also mediates expression of nutritional immunity proteins, for example, calprotectin and antimicrobial peptides, including REG3γ and REG3β, as well as matrix metallopeptidase 9 (MMP9) and inducible nitric oxide synthase (iNOS).
Consistently, the proteomics analysis revealed that the abundances of S100A8, S100A9, REG3γ and REG3β, MMP9, and iNOS were significantly increased at 6 DPI compared with uninfected controls, corresponding to 5.81, 4.60, 4.71, 6.69, 2.27, and 3.77 Log2 FC, respectively (Figure 6a). Following temporal expression of S100a8, Reg3γ, Nos2 (encoding iNOS), and Cxcl1 (a neutrophil recruiting chemokine undetected in the proteome) in IECs revealed significant transcriptional upregulation as early as 2 DPI, just as the pathogen is sporadically detected in the colon (Figure 6b-e). Of note, transcriptional upregulation of S100a8, Reg3γ, Nos2, and Cxcl1 in C57BL/6 mice was only observed from 6 DPI . In addition, we detected a higher abundance of IDO1 (a reporter for 4.36 Log2 FC) at 6 DPI of C3H/HeN mice (Figure 6a). Elevated Ido1 transcription was observed by 3 DPI, which was maintained at the same level up to 8 DPI (Figure 6f). These results show that while triggering similar innate immune responses in IECs, there are clear temporal differences between C. rodentium-infected C3H/HeN and C57BL/6 mice, with responses occurring much earlier in the former.
Temporal analysis of Alpk1 expression by qRT-PCR revealed gradually increasing expression in C3H/HeN mice from 1 DPI, which reached significance at 6 DPI and was maintained at 8 DPI ( Figure 7f); expression of Alpk1 remained unchanged at 3 and 6 DPI C57BL/6 mice and significantly decreased at 8 DPI (Figure 7f).

| C. rodentium activates the ALPK1 pathway in a T3SS-independent manner
Activation of the ALPK1 immune pathway has been linked to the LPS  The abundance of Nlrp3 transcript increases in C3H/HeN mice from 6 DPI but was undetectable in IECs from C57BL/6. (e) The protein abundance of ALPK1, TIFA, and TRAF6 increases upon C. rodentium infection in C3H/HeN, whereas only TIFA is increased in C57BL/6 mice. Bars indicate standard deviation. (f) The abundance of Alpk1 transcript increases in C3H/HeN from 6 DPI, whereas expression of Alpk1 in C57BL/6 is unchanged upon infection. For all qRT-PCR data, each point represents an individual mouse at 0, 1, 2, 3, 6, and 8 DPI. * P ≤ .05, ** P ≤ .01, *** P ≤ .001, and **** P ≤ .0001 (one-way ANOVA) membrane, we tested if this impacted on growth. Using both rich (LB) and minimal (M9) media, we found that both mutations caused significant growth attenuation in minimal media ( Figure S5B,C) and thus could not be analysed in vivo.
We therefore determined if C. rodentium can activate ALPK1 by analysing formation of TIFA oligomers, also called "TIFAsomes," following infection of TIFA-GFP expressing reporter HeLa cells with wild type (WT), C. rodentium ΔescN (T3SS deficient), ΔhldE, and ΔrfaC; treatment with ADP-hep (10 −7 M) was used as a positive control. This revealed that infection with WT and ΔescN similarly activated ALPK1, determined by quantification of "TIFAsomes," whereas no activation was seen by the ΔhldE, and as expected (due to accumulation of LPS metabolites), overactivation was seen by the ΔrfaC mutant (Figures 8a,b and S6). Taken together, these results suggest that C. rodentium triggers expression of Alpk1 in vivo and can activate the ALPK1-TIFA signalling axis in a T3SSindependent manner in vitro.

| DISCUSSION
Our investigation into host responses to C. rodentium infection in susceptible C3H/HeN mice revealed accelerated colonisation and rapid IEC responses, starting from 2 DPI, and upregulation of ALPK1 and NLRP3. In contrast, the IECs of C57BL/6 mice are nonresponsive to C. rodentium until 4 DPI  with no detectable upregulation of ALPK1 and NLRP3. A summary of the main differences between mice suffering from mild and severe C. rodentium infection is shown in Table 1.
Global analysis of changes to the infected C3H/HeN and C57BL/6 IEC proteomes revealed downregulation of central metabolism processes occurring in parallel to upregulation of pathways involved in cell proliferation as well as DNA replication and DNA damage repair response. We have recently shown in C57BL/6 mice an increased PCNA-positive proliferation zone at 4 DPI, which is followed by an increase in total crypt length at 6 DPI . In that LDHA, which mediates pyruvate conversion to lactate, and MCT4, which shuttles lactate across the cell membrane, were significantly upregulated at 6 and 8 DPI, respectively. However, unlike S. Typhimurium, C. rodentium was unable to utilise L-lactate as a carbon source. Nonetheless, in both C3H/HeN and C57BL/6 mice, the apical butyrate transporter MCT1, which feeds OXPHOS, is downregulated whereas SGLT4, which supports aerobic glycolysis, is upregulated. This suggests that oxygenation of the colon, facilitated by the metabolic switch to glycolysis in IECs, may benefit both C. rodentium (which has a preference for aerobic metabolism) and the host (e.g., by boosting growth of L-lactate catabolising Enterobacteriaceae; Garvie, 1980;Lopez et al., 2016). In contrast, it appears that S. Typhimurium has adapted to the L-lactate-rich ecological niche (Gillis et al., 2018).  LXR/RXR appears to be activated, suggesting differences in gene regulation between macrophages and epithelial cells. We postulate that C. rodentium-induced colitis leads to establishment of a novel balance between cholesterol biogenesis and efflux that could impact on immune responses and the microbiome and thereby influence disease progression irrespective of the host genetic background.

Cholesterol biogenesis is a key pathway altered in IECs upon
We observed increased abundance of the pattern recognition receptors NLRP3 and ALPK1 specifically in IECs of C3H/HeN mice from 6 DPI, both at the transcriptional and translational levels. Activation of SREBP2 triggers expression and activation of NLRP3 via TIFA, whereas a SCAP-SREBP2 complex in the ER has recently been shown to directly activate the NLRP3 inflammasome (Guo et al., 2018;Lin et al., 2016;Xiao et al., 2013). Whereas ALPK1 was induced specifically in C3H/HeN mice, TIFA was found in higher abundance in both C3H/HeN and C57BL/6 mice. This was expected given the reported TLR4/MYD88-dependent TIFA upregulation in response to hypoxiareoxygenation of the liver; a phenomenon also undergone in the colon during C. rodentium infection regardless of host genetics (Ding et al., 2013).
Tri-parental conjugation was carried out as previously described (Berger et al., 2017). All plasmids and deletion mutants were confirmed by sequencing (Eurofins). Deletion mutants were tested for growth attenuation by measuring OD600 during growth in LB at 37°C with 200 rpm shaking.

| Immunostaining
Cells were permeabilised with 0.1% Triton×100 (Sigma) for 5 min at room temperature and incubated with a primary rabbit polyclonal anti-C. rodentium antibody (1/200) for 1 hr. They were then washed in PBS and incubated with a donkey anti-rabbit Alexa Fluor 467 antibody (1/500, Invitrogen) for C. rodentium staining. DNA was stained with Hoescht 33342. Images were acquired with an ImageXpress Micro (Molecular Devices, Sunnyvale, USA). Each data point corresponds to duplicate wells; nine images were taken per well. Image analysis was performed using the custom module editor of MetaXpress as previously described (Garcia-Weber et al., 2018). For quantification of TIFA-GFP oligomerisation, TIFAsomes were detected by applying the "Find blobs" function based on size and intensity above background parameters. Cell nuclei were identified with the "Find round objects" module. Cells were defined by extending each nucleus by 10 pixels with the "Grow objects without touching" function, and TIFA punctates were quantified within this mask.

| Ethics statement
All animal experiments complied with the Animals Scientific Procedures Act 1986 and U.K. Home Office guidelines and were approved by the Animal Welfare and Ethical Review Body at Imperial College London. Experiments were designed in agreement with the ARRIVE guidelines for the reporting and execution of animal experiments, including sample randomisation and blinding (Kilkenny et al., 2010).

| Infection of mice
Pathogen-free female 8-to 10-week-old C3H/HeNCrl mice (Charles River Laboratories) were housed in high-efficiency particulate airfiltered cages, with food and water given ad libitum. Mouse experiments were performed with a minimum of four mice per group. Mice were orally gavaged as described (Crepin, Collins, Habibzay, & Frankel, 2016). Uninfected mice were given PBS (200 μl). Enumeration of viable bacteria (CFU) per gram of stool sample was determined by plating onto selective agar as previously described (Crepin et al., 2016). Faecal water content was determined by weighing stools pre and post vacuum drying.
Images were processed using Living Image software (Perkin Elmer).
For BLI of tissue at necropsy, the colon and ceacum were removed and cut longitudinally, and the ceacal contents and stools were removed. The colonic and ceacal tissue were washed briefly in PBS and the tissues imaged individually, with the mucosa exteriorised in an IVIS Spectrum CT.

| Extraction of IECs
IECs were extracted from 4-cm distal colon as previously described (Berger et al., 2017) and frozen at −80°C until processing for proteomic analysis or RNA extraction.  (Livak & Schmittgen, 2001). Expression was normalised to the housekeeping gene Gapdh. Primer pairs are as previously reported . Additional primer pairs are listed in Table S4.

| Colorimetric cholesterol assay
Stools were collected from mice, vacuum dried, and weighed before being crunched to powder and lipids extracted in an 800-μl mixture of chloroform : isopropanol : NP-40 (7:11:0.1). Faecal cholesterol was quantified using a colorimetric reaction, as per the manufacturer's recommendations (Cell Biolabs, STA-384). Colorimetric signal was analysed using the FLUOstar Omega spectrophotometric microplate reader (BMG Labtech) in the 540-to 570-nm range.

| 16S rRNA gene sequencing
Upon necroscopy, 1-cm distal colon was immediately frozen and stored at −80°C. DNA was extracted using a Power Soil kit (MoBio).

| Statistical analysis
Graphpad Prism 7 was used for all statistical calculations unless otherwise specified. FC values were log transformed prior to analysis of variance being carried out. P values <.05 were considered significant ( * P ≤ .05, ** P ≤ .01, *** P ≤ .001, and **** P ≤ .0001). Statistical analysis of the microbiome was performed at individual taxon levels. For each taxon, the Mann-Whitney U statistical test was carried out between uninfected and infected groups, correcting P values for FDR using the Benjamini and Hochberg method. Unpaired t test was used for statistical analysis of faecal water content and quantification of total bioluminescent flux from ex vivo tissues. Pathway analysis was performed with the 1D enrichment method in the Perseus software (Cox & Mann, 2012;Tyanova et al., 2016). The enrichment score shows whether the proteins of a pathway tend to be upregulated or downregulated based on Wilcoxon-Mann-Whitney test and is normalised within min = −1 and max = 1. Comparison to C57BL/6 proteins at 8 DPI was carried out using data deposited in PRIDE with identifier PXD005004 (Berger et al., 2017). Proteins from enriched pathways of interest (Benjamini-Hochberg FDR < 0.05) were selected based on previous literature on C57BL/6 mice. Proteins were considered significant if abundance changed 1.5-fold or higher upon infection.
supported by a Wellcome Trust studentship. This work was supported by grants from the Medical Research Council, the Wellcome Trust, and the Royal Society.

AUTHOR CONTRIBUTIONS
DC performed the experiments, analysed the data, and wrote the manuscript. EGDH performed the proteomics aspects, analysed the data, and edited the manuscript. TIR was involved in preparing and running the mass spectrometry samples, data analysis, and editing the manuscript. CMS performed the microbiome analysis, with the help of EE, and edited the manuscript. RB supervised the project, analysed the data, and edited the manuscript. DGW and CA designed and performed the in vitro cell infection assays and edited the manuscript. GF and JSC were responsible for overseeing the overall running of the project, data analysis, and writing the paper.