Citrobacter rodentium Infection Induces Persistent Molecular Changes and Interferon Gamma-Dependent Major Histocompatibility Complex Class II Expression in the Colonic Epithelium

ABSTRACT Most studies of infections at mucosal surfaces have focused on the acute phase of the disease. Consequently, little is known about the molecular processes that underpin tissue recovery and the long-term consequences postinfection. Here, we conducted temporal deep quantitative proteomic analysis of colonic intestinal epithelial cells (cIECs) from mice infected with the natural mouse pathogen Citrobacter rodentium over time points corresponding to the late steady-state phase (10 days postinfection [DPI]), the clearance phase (13 to 20 DPI), and 4 weeks after the pathogen has been cleared (48 DPI). C. rodentium, which relies on a type III secretion system to infect, is used to model infections with enteropathogenic and enterohemorrhagic Escherichia coli. We observe a strong upregulation of inflammatory signaling and nutritional immunity responses during the clearance phase of the infection. Despite morphological tissue recovery, chromogranin B (ChgB)-positive endocrine cells remained significantly below baseline levels at 48 DPI. In contrast, we observed an increased abundance of proteins involved in antigen processing and presentation 4 weeks after pathogen clearance. In particular, long-term changes were characterized by a persistent interferon gamma (IFN-γ) response and the expression of major histocompatibility complex class II (MHCII) molecules in 60% of the EpCAM+ cIECs, which were not seen in Ifnγ−/− mice. Nonetheless, both wild-type and Ifnγ−/− mice mounted similar systemic and colonic IgG responses to C. rodentium and were equally protected from rechallenge, suggesting that cIEC MHCII is not necessary for protective immunity against C. rodentium.


RESULTS
Temporal cIEC proteomic responses to C. rodentium infection during the clearance and recovery phases. We previously described cIEC responses to C. rodentium during the expansion phase (4 and 6 DPI) (20) and at the entry to the steady-state phase (8 DPI) (28) of infection. To characterize the temporal changes in cIECs during the late steady-state and clearance phases of the infection, we performed proteomic analysis of cIECs isolated from C57BL/6 mice at 10, 13, 17, and 20 DPI. Additionally, to characterize the nature and duration of the regeneration processes postinfection, we analyzed cIECs isolated 4 weeks after C. rodentium had been cleared (48 DPI). Mockinfected mice harvested at 10 DPI were used as uninfected (UI) controls. Analysis of enriched cIECs showed a purity of around 90 to 95% epithelial cell adhesion moleculepositive (EpCAM 1 ) cells (see Fig. S1A in the supplemental material). As expected, C57BL/6 mice started to spontaneously clear C. rodentium after 10 DPI; the pathogen fell below the level of detection by immunofluorescence at 13 DPI, and the majority of mice no longer had detectable fecal C. rodentium by 20 DPI (Fig. 1A; Fig. S1B and C).
Three biological repeats for each time point (UI and 10, 13, 17, 20, and 48 DPI) were analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS). We quantified a total of 8,439 unique proteins that mapped to the Mus musculus proteome (Table S1), 7,301 of which were identified in at least two biological repeats, and 814 unique proteins which mapped to the C. rodentium proteome (Table S2) (peptide false discovery rate [FDR] of ,1%). The adhesin intimin and the T3SS effectors Tir, EspF, and EspI/NleA were highly enriched at 10 and 13 DPI before their abundance declined (Table S2).
Principal-component analysis (PCA) of mouse proteins showed close clustering of biological replicates from UI and 48 DPI samples (Fig. 1B), demonstrating that most infection signatures in the proteome had returned to preinfection homeostasis by 48 DPI, 4 weeks after C. rodentium clearance at 20 DPI. One-way analysis of variance (ANOVA) identified 4,164 significantly changed proteins across all the investigated time points, and Tukey's postanalysis (FDR of ,0.05) showed that 10 and 13 DPI had the most significantly changed proteins (around 2,800) compared to UI mice ( Fig. 1C;  Fig. S1D). One-dimensional (1D) enrichment analysis showed temporal changes in cIEC metabolic and immune processes as they transitioned between phases of the infection cycle (Fig. 1D).
The colonic crypt exhibits persistent changes 4 weeks after pathogen clearance. Consistent with C. rodentium-induced CCH, the cell proliferation processes "DNA replication," "mismatch repair," and "cell cycle" were significantly enriched at 10 to 13 DPI before recovering by 17 DPI (Fig. 1D). Curiously, proteins of the pentose phosphate pathway, the main source of nucleotide intermediates, were found in lower abundances at 4 to 10 DPI (20,28) (Fig. 1D); therefore, cIECs seem to use alternative sources of nucleotides for cell proliferation. Both the expansion of the crypt length (CCH) ( Fig. 2A;  Fig. S2A) and the abundance of the cell proliferation marker PCNA peaked at 13 DPI before recovering, and immunofluorescence staining of PCNA mirrored and validated the proteomic changes seen in the cIECs (Fig. 2B to D).
The rapid cell proliferation of the epithelial barrier in response to infection is fueled by aerobic glycolysis (the Warburg effect) (20,28). Consistently, mitochondrial processes involved in ATP production, including OXPHOS, fatty acid metabolism, and the TCA cycle, were inhibited during the clearance phase (10 to 20 DPI) and fully restored by 48 DPI (Fig. 1D). The basolateral glucose importer SLC5A9, which seems to fuel aerobic glycolysis, was highly upregulated at 10 and 13 DPI and remained modestly elevated at 48 DPI, while glycine amidinotransferase (GATM), which catalyzes creatine production for energy dispersal, was enriched at 10 DPI before being downregulated from 17 DPI onward ( Fig. 2D and E).
CCH is associated with a loss of mature cIEC types, including absorptive enterocytes and goblet cells. Consistently, goblet cell (CLCA1 and TFF3) and deep secretory cell (REG4) markers were significantly reduced before beginning to recover to preinfection levels from 17 DPI onward (Fig. 2F). SLC26A3, a marker for differentiated enterocytes, was significantly decreased in abundance at all time points (although a trend toward recovery was observed from 20 DPI) (Fig. 2F). As previously reported (29), we observed a reduction in the density of alcian blue and periodic acid-Schiff (AB/PAS) stain (which stains all intracellular mucin glycoproteins)-positive cells at 10 to 13 DPI, after which it gradually returned to nearly UI levels, reflecting what was observed in the proteomics analysis ( Fig. 2F and G; Fig. S2B). Interestingly, the goblet cell-secreted antimicrobial protein WFDC2, which plays an important role in preventing epithelial barrier breakdown (30), was present at a lower abundance from 10 to 48 DPI (Fig. 2H) and was transcriptionally downregulated specifically at 10 DPI (Fig. 2I). Together, these data show that despite a recovery of goblet cell density, these cells may remain functionally altered 4 weeks after pathogen clearance. Moreover, the abundance of the panendocrine cell markers chromogranin A (CHGA) and CHGB, which are weakly upregulated during the early stages of infection ( Fig. 2F) (20), was decreased from 17 to 48 DPI (Fig. 2F), and ChgB was significantly transcriptionally downregulated from 13 to 48 DPI compared to UI mice (Fig. 2J). Staining of CHGB-positive cells in the colonic crypts showed a similar trend ( Fig. 2K to L). Accordingly, C. rodentium-induced perturbation of the cellular composition of the colonic crypt fails to fully recover 4 weeks after infection clearance.
Innate immune signaling pathways are upregulated during the clearance phase. We have shown previously that C. rodentium induces a strong cIEC inflammatory response during the expansion phase and at the entry to the steady-state infection phase. Analyzing the proteomic responses during the later infection phases revealed that immune processes continued to be enriched (Fig. 1D), and many innate immunity proteins were identified in the 100 most highly changed proteins at 10 to 20 DPI (Fig. S3A). Immune-related proteins, including NOS2; the antimicrobial peptides REG3b, REG3g, and CAMP; as well the cytosolic nucleic acid sensors ZBP1 (DAI), cGAS, TMEM173 (STING), IFIH1 (MDA5), DDX58 (RIG-I), and AIM2, were significantly upregulated during the late steady-state and clearance phases ( Fig. 3A and B), with Zbp1 and Tmem173 being transcriptionally upregulated at 10 to 20 DPI and 10 to 13 DPI, respectively ( Fig. 3C and D), while Aim2 expression was unchanged at the transcriptional level (Fig. 3E). Moreover, the levels of several factors involved in nutritional immunity, including DMBT1, S100A8/A9, lipocalin-2 (LCN-2), lactotransferrin (LTF), haptoglobin (HP), hemopexin (HPX) and the transferrin receptor (TFRC), were strongly elevated during the clearance phase of infection, peaking at 13 DPI before recovering by 48 DPI (Fig. 3A). In agreement with these results, enzyme-linked immunosorbent assay (ELISA) measurements of LCN-2 excreted into the stool mirrored this pattern (Fig. 3F). Conversely, metallothionein-2 (MT2), a secreted protein that mediates zinc uptake across the apical membrane, and ferritin (FTL1/FTH1), which stores free cellular iron, were both transcriptionally downregulated and found in lower abundances during the late infection phases, failing to recover by 48 DPI (Fig. 3A, G to I), suggesting that changes to cIEC metal ion homeostasis have not recovered by 48 DPI. Overall, these results show that cIEC innate immune responses to C. rodentium infection peak at around 13 DPI, coinciding with the clearance of mucosa-associated C. rodentium, and are largely resolved by 48 DPI, although some remnants of infection-associated immune and nutritional responses remain.
The recovery phase is characterized by activation of antigen presentation processes. We next focused on processes and proteins that are altered during infection and remain significantly changed at 48 DPI, 4 weeks after the infection has effectively been eliminated, thus representing long-term, infection-induced perturbations in the colonic epithelium (Fig. S3B). The most striking change in protein expression during and after the clearance phase of C. rodentium infection was the upregulation of proteins involved in antigen processing and presentation, which was among the most highly enriched processes (Fig. 1D). Indeed, 27 of the 33 identified proteins assigned to the KEGG process "antigen processing and presentation" were considered significantly changed, and many major histocompatibility complex class I (MHCI) and MHCII molecules, and the intracellular proteins required for peptide processing, were among the most highly changed proteins ( Fig. S3A and B).
Proteins involved in MHCI antigen processing and presentation, which typically present endogenous antigens, were strongly upregulated at 10 to 20 DPI (Fig. 4A). Six out of nine identified MHCI alpha chains and b2-microglobin were significantly changed, with peak log 2 fold change (Log 2 FC) values of between 1.6 and 4.5. Likewise, calnexin (the heavy chain chaperone) was also significantly upregulated between 10 and 17 DPI. Additionally, other members of the peptide-loading complex, TAP1, TAP2, tapasin (TAPBP), calreticulin (CALR), ERP57 (PDIA3), and ERAP1 (required for peptide trimming), were all significantly upregulated during the clearance phase of infection (Fig. 4A); Tap1 was significantly transcriptionally upregulated at 10 to 20 DPI (Fig. 4B), while the MHCI protein H2-D1 showed a modest transcriptional increase at 10 to 13 DPI (Fig. 4C). The elevation of MHCI antigen processing and presentation proteins was resolved between 20 and 48 DPI, with no proteins being significantly upregulated at 48 DPI compared to the UI controls (FDR of ,0.05 by Tukey's posttest) (Fig. 4A). Intracellular peptides are degraded by specialized subunits of the proteasome for presentation by MHCI molecules; consistently, we observe a robust switch from the constitutive proteasome to the immunoproteasome during the clearance phase of infection. The core subunits PSMB1 to -4 were modestly, but significantly, elevated (peaking at 17 to 20 DPI), while the constitutive proteasome catalytic subunits PSMB5, PSMB6, and PSMB7 were significantly decreased in abundance (Fig. 4D); Psmb6, as a representative gene, was also transcriptionally downregulated (Fig. 4E). Concomitantly, we observe a strong enrichment of the immunoproteasome catalytic subunits PSMB8, PSMB9, and PSMB10 (Fig. 4D); Psmb9 was significantly transcriptionally upregulated from 10 to 20 DPI (Fig. 4F). Furthermore, PSMB8, PSMB9, and PSMB10 remained elevated at 48 DPI compared to the UI controls.
All identified MHCII proteins, which typically present exogenous antigens, as well as the invariant chain (CD74), were strongly elevated throughout the clearance phase of infection, with a peak Log 2 FC of between 3.1 and 4.9 (Fig. 4G). Interestingly, in contrast to the MHCI antigen presentation pathway, MHCII presentation proteins remained strongly elevated after C. rodentium clearance, with Log 2 FC values of 2.7 to 3.2 at 48 DPI (Fig. 4G). Indeed, of the proteins considered significantly changed at 48 DPI compared to the UI controls (FDR of ,0.05 by Tukey's posttest) (Fig. S3B), 5 of the top 10 most highly elevated proteins are involved in MHCII presentation. The strong transcriptional upregulation, persisting until 48 DPI, of H2-Aa was validated by quantitative reverse transcription PCR (qRT-PCR) (Fig. 4H). These data demonstrate that strong local MHCII expression persists throughout the host recovery phase.
Many of the proteins found to remain significantly altered in abundance at 48 DPI, including MHCII proteins, PSMB8, PSMB10, and WFDC2, are IFN-g regulated; moreover, the IFN-g-regulated proteins, IDO1 and selected guanylate-binding proteins (GBPs), showed increased protein abundances at 48 DPI (Fig. 4I), with Ido1 remaining significantly transcriptionally upregulated at 48 DPI (Fig. 4J). Importantly, measurement of IFN-g in whole colonic tissue explants showed a significant increase in IFN-g at 8 DPI, Persistent Mucosal Responses to C. rodentium Infection ® which remained elevated until 48 DPI (Fig. 4K), while IL-22 levels returned to baseline levels ( Fig. S4A), suggesting that IFN-g underpins the long-term changes to cIECs following C. rodentium infection.
The role of cIEC MHCII presentation in protective immunity against reinfection. We next investigated the role of IFN-g and the expression of MHCII molecules on cIECs in recovery from C. rodentium infection and protection against reinfection. IFN-g is   Fig. S4B), with the exception of conventional dendritic cell (cDC) numbers (CD11b 1 DCs), which showed minimal changes over the infection course (Fig. S4B). As expected, most cell types, including B cells and neutrophils, returned to preinfection levels during the host recovery phase. Conversely, CD4 1 T cells, which are reported to secrete IFN-g during and after C. rodentium infection (31), remained elevated above mock-infected levels, albeit not significantly, at 48 DPI (Fig. 5A).
To characterize the progression of MHCII expression by cIECs during C. rodentium infection at the single-cell level, we analyzed cIECs using flow cytometry. This revealed that around 2% of UI cIECs (EpCAM 1 cells) expressed MHCII, which increased gradually to 70% by 20 DPI and remained similarly elevated (at around 60% MHCII 1 cIECs) at 48 DPI ( Fig. 5B and C). Conversely, there was no increase in the cIEC expression of MHCII in IFN-g KO (Ifng 2/2 ) mice, demonstrating that IFN-g is necessary for the induction of MHCII expression on cIECs during C. rodentium infection ( Fig. 5D and E). Ifng 2/2 mice were unable to fully clear the initial C. rodentium infection and persistently harbored C. rodentium at around 10 6 CFU/g stool from 20 to 48 DPI (Fig. 5F), suggesting that IFN-g plays either a direct or an indirect role in full C. rodentium clearance.
Canonical antigen-presenting cells (APCs) typically present exogenous antigens via MHCII to CD4 1 T cells, which subsequently induce antibody (including IgG) production by plasma B cells and the generation of memory B cells (32,33). To test whether prolonged secretion of IFN-g, and the resulting expression of MHCII molecules on cIECs, is required for protection against reinfection, WT and Ifng 2/2 mice were infected with C. rodentium or mock infected with phosphate-buffered saline (PBS) as a control. All mice received ciprofloxacin treatment at 42 to 45 DPI, to clear persistently colonizing C. rodentium in Ifng 2/2 mice, and then rechallenged with C. rodentium at 52 DPI. As expected, control WT and Ifng 2/2 mice that were initially mock infected developed a normal C. rodentium infection after the second challenge (UI1CR) (Fig. 5G; Fig. S4C). Conversely, rechallenged WT and Ifng 2/2 mice (CR1CR) displayed protective immunity against C. rodentium reinfection ( Fig. 5G; Fig. S4C). This was correlated with comparable immune cell profiles (Fig. S4D) and levels of anti-C. rodentium serum and colonic IgG in both WT and Ifng 2/2 mice (Fig. 5H). Together, these data demonstrate that the long-term presence of IFN-g following infection and the ensuing cIEC MHCII antigen presentation are not required for protection against reinfection with C. rodentium.

DISCUSSION
Studies of mucosal barrier perturbations have mainly focused on the acute phase of the disease at a single time point, but little is currently known about the dynamics of the resolution and regeneration phases or the molecular processes that underpin the recovery process and return to homeostasis in epithelial cells. We addressed this knowledge gap by characterizing the temporal cellular responses of cIECs using quantitative proteomics across the late steady-state (10 DPI), clearance (13 to 20 DPI), and subsequent host recovery (20 to 48 DPI) phases using C. rodentium infection as a model. We found that the cIEC innate inflammatory responses to C. rodentium and infection-induced changes to metabolism and bioenergetics are largely resolved following pathogen clearance, at between 20 and 48 DPI, which we define as the "host re- Surprisingly, we find that while morphologically (i.e., CCH and PCNA staining) the mucosa has recovered 4 weeks following C. rodentium clearance, on the molecular and cellular levels the cIECs have not fully reset to the preinfection homeostatic state. The panendocrine cell markers CHGA and CHGB sharply decrease from 8 to 20 DPI and remain decreased at 48 DPI. In healthy tissue, enteroendocrine cells are located throughout the colonic crypt and secrete hormones; they consist of several specialist cells types, of which enterochromaffin (EC) cells, which secrete serotonin (5-hydroxytryptamine ), are the most abundant (34,35). EC cells play a functional role in C. rodentium pathogenesis as serotonin has been shown to repress C. rodentium virulence in vivo (36), and we have previously shown CHGA/CHGB to be weakly upregulated during early infection (20), when C. rodentium adapts to the host environment and turns on virulence gene expression. Levels of WFDC2, a goblet cell-secreted antibacterial protein that is negatively regulated by IFN-g (30), also remained significantly decreased at 48 DPI. WFDC2 was found to be downregulated in human ulcerative colitis (UC) colonic samples compared to healthy patients, and Wfdc2 1/2 mice show abnormalities in colonic epithelial intercellular junctions (30). Although goblet cell numbers recovered by 20 DPI, WFDC2 remained at a reduced abundance at 48 DPI. Overall, these data suggest that the integrity and cellular composition of the colonic epithelial cell barrier may not be fully restored.
Interestingly, we observed an increased abundance of proteins involved in cytosolic nucleic acid sensing, mainly characterized as antiviral sensors, during the clearance phase of C. rodentium infection (13 to 20 DPI). While the upregulation of these sensors may be a bystander effect of the increased levels of IFN-g, they could also play an active role in the clearance of the pathogen and the onset of tissue recovery via the induction of type I IFN responses, which have been linked to the maintenance of gut homeostasis and tissue repair (37) and inflammatory cell death pathways. For example, necroptosis likely plays a role in pathogen clearance as C. rodentium lacking the T3SS effector EspL, which inhibits necroptotic cell death pathways, is cleared quicker from the colon (38). The nature of the ligands activating the sensors needs further investigation, but it is likely that both endogenous mitochondrial DNA (39), liberated in response to C. rodentium-induced mitochondrial damage, and exogenous nucleic acids from the microbiota (40) can activate these sensors. Additionally, EHEC nucleic acids have been shown to lead to cytosolic NLRP3 inflammasome activation, opening the possibility of C. rodentium DNA/RNA being the ligand itself (41).
We find that the host recovery phase of C. rodentium infection is characterized by a strong IFN-g response, which persists for 4 weeks after the pathogen has been cleared. The most striking IFN-g-induced change to the cIECs was the continued expression of MHCII antigen presentation proteins at 48 DPI. While MHCII molecules are constitutively expressed in small intestinal IECs (siIECs), cIECs have been shown to express MHCII in response to IFN-g (42), and we find that Ifng 2/2 mice do not show an increase in MHCII 1 cIECs during or after C. rodentium infection, confirming that IFN-g is essential in this context. The role of cIEC MHCII expression during and after infection is not well understood. The induction of protective immunity against C. rodentium is dependent on the induction of local (gut) immune responses, while mice with high levels of systemic (but not local) C. rodentium-specific IgG alone remain susceptible to reinfection (43), raising the possibility that cIECs may play a role as nonconventional antigen-presenting cells, promoting immune memory responses in the colon. While we cannot fully rule out a contributing role, we find that Ifng 2/2 and WT mice are equally protected against rechallenge, demonstrating that IFN-g-dependent cIEC MHCII expression is not necessary for IgG-mediated protective immunity against C. rodentium.
The benefit to the host of continued IFN-g expression in the colon for weeks after pathogen clearance is currently unknown; however, it is possible that the resulting cIEC MHCII may have a role in mucosal regeneration or a noncanonical immunomodulatory effect. Small intestinal epithelial stem cells act as nonprofessional antigen-presenting cells via MHCII, which assist in epithelial regeneration following infection (44). Although the high percentage (60 to 70%) of cIECs that express MHCII during and after C. rodentium infection suggests that this is not confined to colonic stem cells, it is plausible that cIEC MHCII may also play a similar role in colonic epithelium renewal. Additionally, a possible role for cIEC MHCII in favoring an immunosuppressive response during intestinal inflammation via the activation of regulatory T cells has been suggested, as shown by a study showing that mice lacking MHCII in nonhematopoietic cells developed more severe T cell-mediated colitis (45). Taken together with the fact that Ifng 2/2 mice develop increased mucosal damage 14 days after infection with C. rodentium (25), it is reasonable to consider that cIECs could play a similar role during and after C. rodentium infection. Accordingly, we rule out a role for cIEC MHCII in the generation of B cell-dependent protective immunity, suggesting that the focus should now be shifted to understanding what role they play postinfection in the recovery of tissue homeostasis.

MATERIALS AND METHODS
Bacterial growth and maintenance. C. rodentium strains ICC169 (46,47) and ICC180 (19) were used in this study. C. rodentium was grown in lysogeny broth (LB; VWR) at 37°C with shaking at 200 rpm unless otherwise specified. For CFU quantification, C. rodentium was grown on LB solidified with 1.5% agar (LBA; VWR). Where necessary, the following antibiotics were added: nalidixic acid at 50 mg/mL and kanamycin at 50 mg/mL.
Mouse experiments. Mouse experiments were performed in accordance with the Animals Scientific Procedures Act of 1986 and were approved by the local Ethical Review Committee according to UK Home Office guidelines. Specific-pathogen-free female C57BL/6 mice (18 to 20 g) were purchased from Charles River, London, United Kingdom, and housed in cages of five. For each experiment, five mice were randomly assigned to experimental groups. Investigators were not blind to the allocation. C57BL/ 6J IFN-g knockout mice (B6.129S7-Ifngtm 1Ts/J ; Jackson Laboratories) (referred to as Ifng 2/2 throughout) and C57BL/6J WT control mice (Jackson Laboratories) were bred at Imperial College London and housed in cages of one to four. Male and female mice (8 to 12 weeks old) were allocated to experimental groups to ensure even distributions of age and sex. Investigators were not blind to the allocation. All mice used in this study were housed in individually HEPA-filtered cages with bedding and nesting and given free access to food and water. Mice that lost 20% of their initial weight or became moribund during the experimental period were humanely culled and excluded from all analyses (see Table S3 in the supplemental material). Where mice were purchased from a supplier, we used female mice only to avoid the logistical issues and stress to animals associated with male in-cage fighting. Where mice were locally bred, both sexes were used due to the limited availability of single-sex animals.
C. rodentium infection. Mice were orally gavaged with approximately 1 Â 10 9 CFU C. rodentium (strain ICC169 unless otherwise specified) in 200 mL sterile PBS as previously described (21). Mockinfected ("uninfected") mice received 200 mL sterile PBS. The inoculum CFU was retrospectively confirmed by CFU quantification as previously described (21). C. rodentium stool shedding was determined by collecting stool samples at the specified days postinfection, homogenizing the samples in PBS, and plating serial dilutions onto LBA containing selective antibiotics. For rechallenge experiments, mice in all groups were orally gavaged with 2 mg ciprofloxacin at 42, 43, and 45 DPI and rechallenged by oral gavage with approximately 1 Â 10 9 CFU C. rodentium (strain ICC180) at 52 DPI. Any mouse that did not reach a C. rodentium colonization threshold of 1 Â 10 8 CFU/g stool at 6 DPI (first C. rodentium infection) was excluded from all analyses (Table S3); no C. rodentium colonization threshold was implemented for rechallenge with ICC180. After the exclusion of cIEC samples based on these criteria, all proteomics samples consisted of cIECs pooled from at least 4 mice per group in each biological repeat.
Histopathological analysis of colonic tissue. Formalin-fixed 0.5-cm distal colon samples were processed, paraffin embedded, and sectioned at 5 mm. Formalin-fixed, paraffin-embedded (FFPE) sections were then stained with either hematoxylin and eosin (H&E) or alcian blue and periodic acid-Schiff (AB/ PAS) stain using standard techniques. Images were acquired using a Zeiss AxioVision Z3 microscope with a 20Â lens objective and processed using Zen 2.3 (Blue version; Carl Zeiss MicroImaging GmbH, Germany). CCH and goblet cell density measurements were obtained from at least 10 well-oriented crypts per mouse. Colonic sections from which fewer than 10 well-orientated crypts were observed were excluded from all analyses. CCH measurements were performed on H&E-stained sections. The goblet cell density was measured on AB/PAS-stained sections. The goblet cell density (goblet cells/100 mm crypt) for each mouse was calculated by dividing the total number of AB/PAS-positive cells counted exclusively from the crypts by the total crypt length of the crypts from which they were counted.
Immunofluorescence staining of colonic tissue. FFPE sections were processed and stained as previously described (20). Images were acquired and processed as described above for histopathological analysis. The following primary antibodies were used: rabbit polyclonal anti-C. rodentium antibody (1:50) (Statens Serum Institute, Copenhagen, Denmark), mouse anti-PCNA antibody (1:500) (catalog number ab29; Abcam), and rabbit anti-CHGB antibody (1:50) (catalog number 14968-1-AP; Proteintech). The following secondary antibodies were used: Alexa Fluor 555 anti-rabbit and rhodamine Red-X donkey antimouse (both 1:100) (Jackson Immunoresearch). DNA was stained with Hoechst 33342 or 49,6-diamidino-2-phenylindole (DAPI) (1:1,000). PCNA measurements were obtained from at least 10 well-oriented crypts per mouse. Colonic sections from which fewer than 10 well-orientated crypts were observed were excluded from all analyses. The endocrine cell density (cells per square millimeter of crypt area) for each mouse was calculated by dividing the total number of CHGB-positive cells within the crypt area by the total crypt area. The total crypt area was calculated by tracing around the edges of colonic crypts and the lumen of the gut using the Contour(polygon) tool in Zen 2.3 (Blue version; Carl Zeiss MicroImaging GmbH, Germany).
cIEC purification. cIECs were isolated from 3.5-cm distal colonic tissue samples (following the removal of the most distal 1-cm tissue for histological analysis), as previously described (20). Briefly, the 3.5-cm colonic tissue sample was opened longitudinally and briefly washed in 1Â Hanks' balanced salt solution (HBSS) without Mg and Ca. The tissue sample was incubated at 37°C with shaking for 45 min in enterocyte dissociation buffer (1Â HBSS without Mg and Ca, containing 10 mM HEPES, 1 mM EDTA, and 5 mL/mL 2-b-mercaptoethanol). The remaining tissue was removed, and lifted enterocytes were subsequently collected by centrifugation (2,000 Â g for 10 min), followed by three PBS washes at 4°C. cIEC pellets were stored at 280°C.
Sample preparation and TMT labeling. cIEC pellets isolated from 4 to 5 mice per group (time point) per biological replicate were individually solubilized using probe sonication in lysis buffer (100 mM triethylammonium bicarbonate [TEAB], 1% sodium deoxycholate [SDC], 10% isopropanol, 50 mM NaCl) supplemented with Halt protease and phosphatase inhibitor cocktail (Thermo Scientific), boiled for 5 min at 90°C, and sonicated once more. The protein concentration was determined with a Coomassie Plus assay (Thermo Scientific) according to the manufacturer's protocol. Equal amounts of protein from each mouse from each group were combined. A total of 100 mg of protein per group was reduced with 5 mM Tris-2-carboxyethyl phosphine (TCEP) for 1 h, followed by alkylation with 10 mM iodoacetamide (IAA) for 30 min, and then digested by adding trypsin (Pierce) at a final concentration of 75 ng/mL to each sample and incubating the samples for 18 h at room temperature. Peptides were labeled with tandem mass tag (TMT) multiplex reagent (Thermo Scientific) for 1 h before quenching with a final volume of 5% hydroxylamine for 15 min (Sigma). TMT-labeled peptides were combined at equal amounts, and SDC was precipitated with formic acid (FA) at a final concentration of 2% (vol/vol) and centrifugation for 5 min at 10,000 rpm. The supernatant containing TMT-labeled peptides was dried with a centrifugal vacuum concentrator.
High-pH reversed-phase peptide fractionation. TMT-labeled peptides were reconstituted in 0.1% ammonium hydroxide and fractionated using a high-pH reversed-phase (RP) Waters XBridge C 18 column (2.1 by 150 mm, 3.5 mm) on a Dionex Ultimate 3000 high-performance liquid chromatography (HPLC) system. Mobile phase A was 0.1% ammonium hydroxide, and mobile phase B was 100% acetonitrile and 0.1% ammonium hydroxide. A gradient elution at 200 mL/min was used to separate the peptides in the following steps: isocratic for 5 min at 5% phase B, gradient for 40 min to 35% phase B, gradient to 80% phase B in 5 min, isocratic for 5 min, and reequilibration to 5% phase B. A total of 65 fractions were collected every 42 s into a 96-well plate, dried, and concatenated into 28 fractions upon reconstitution in 0.1% formic acid.
Liquid chromatography-mass spectrometry analysis. All data were acquired using an Orbitrap Fusion mass spectrometer (Thermo Scientific) coupled to a Dionex Ultimate 3000 system. The mobile phases were 0.1% FA (phase A) and 80% acetonitrile in 0.1% FA (phase B). Samples were loaded onto a C 18 trapping column (Acclaim PepMap 100, 100 mm by 2 cm, 5 mm, 100 Å) at a 10-mL/min flow rate, followed by low-pH gradient elution on a nanocapillary reversed-phase column (Acclaim PepMap C 18 , 75 mm by 50 cm, 2 mm, 100 Å) at 45°C. A flow rate of 300 nL/min was applied to separate the peptides over a 90-min gradient from 5% to 38% phase B followed by 10 min up to 95% phase B, isocratic for 5 min at 95% phase B, reequilibration to 5% phase B in 5 min, and isocratic for 10 min at 5% phase B. The instrument method included Orbitrap MS1 scans (automatic gain control [AGC] of 4 Â 10 5 , maximum injection time of 50 ms, and resolution of 120,000), ion trap MS2 scans (top-speed mode [3 s], collision-induced dissociation [CID] energy of 35% with a quadrupole isolation width of 0.7 Th, AGC target of 1 Â 10 4 , and maximum injection time of 50 ms), and MS3 scans of the top seven most abundant CID fragments isolated with synchronous precursor selection (resolution of 50,000, mass range of m/z 100 to 500, higher-dissociation collision energy of 65%, AGC target of 1 Â 10 5 , and injection time of 105 ms).
Protein identification and TMT-based quantification. Proteome Discoverer 2.3 (Thermo Scientific) using the SEQUEST-HT algorithm was employed to search MS/MS spectra against UniProt annotated reference proteomes of Mus musculus and Citrobacter rodentium. Searches were performed with a 20-ppm precursor mass tolerance, a 0.5-Da fragment ion mass tolerance, and trypsin digestion with up to two missed cleavages. The static modifications were specified as carbamidomethylation of Cys residues (157.02146) and TMT modification of peptide N-terminal and Lys residues (1229.16293), while dynamic modifications were specified as oxidation of Met (115.99491) and deamidation of Asp and Glu residues. The confidence of peptide discovery was estimated at a 1% FDR with the percolator node based on the q value and decoy database search. The reporter ion quantifier node included a TMT 10-plex quantification method with an integration window tolerance of 15 ppm and an integration method based on the most confident centroid peak at the MS3 level. Protein groups with only unique peptides were used for quantification.
LCN-2 ELISA. Stool samples were homogenized in PBS with 0.1% Tween 20 using a vortex machine for 15 min. Samples were centrifuged at 16,000 rpm for 10 min, and the supernatant was extracted and stored at 280°C. The LCN-2 concentration was determined using a DuoSet mouse lipocalin-2 ELISA (R&D Systems), according to the manufacturer's instructions.
IFN-cand IL-22 ELISAs. A 0.5-cm distal colonic tissue sample without feces was weighed and washed for 2 h in RPMI 1640 medium with glutamine (Sigma) with 100 mg/mL of streptomycin and 100 mg/mL of penicillin to remove residual cytokines present on the tissue surface. Explants were then rinsed and cultured in complete RPMI medium (RPMI 1640 medium with glutamine [Sigma] supplemented with 10% heat-inactivated fetal bovine serum [FBS] [Gibco], 1 mM sodium pyruvate, 100 mg/mL penicillin, 100 mg/mL streptomycin, and 10 mM HEPES [all from Sigma]) for 24 h at 37°C with 5% CO 2 . The supernatant was extracted, centrifuged for 10 min at 3,000 Â g to remove cellular debris, and stored at 280°C. IFN-g levels were determined using an IFN-g Femto-HS high-sensitivity mouse uncoated ELISA kit (Thermo Fisher Scientific), and IL-22 levels were determined using an IL-22 mouse uncoated ELISA kit (Thermo Fisher Scientific), in both cases according to the manufacturer's instructions.
Anti-C. rodentium IgG ELISA. For measurement of colonic IgG, 0.5-cm distal colon explants were prepared as described above for the IFN-g ELISA. For serum measurements, blood was extracted from mice after cervical dislocation and allowed to coagulate at room temperature for at least 3 h. Blood was then centrifuged at 2,000 rpm for 15 min to pellet the blood clots, and the supernatant (serum) was collected and stored at 280°C. To prepare heat-killed C. rodentium, ICC169 was grown to saturation in LB at 37°C at 200 rpm, diluted 1:50 in low-glucose Dulbecco's modified Eagle's medium (DMEM) (Sigma), and grown overnight at 37°C with 5% CO 2 to stimulate T3SS expression. The culture was then normalized to an optical density (at 600 nm) of 0.4, heat inactivated in the presence of EDTA-free Pierce protease inhibitor cocktail (Thermo Fisher Scientific) at 60°C for 1 h, and finally diluted 20-fold in PBS. Immuno 96-well plates (Thermo Fisher Scientific) were incubated with 50 mL/well heat-killed C. rodentium for the samples or 100 mL/well goat anti-mouse IgG antibody for the standards (2 mg/mL) (Poly4053; BioLegend) overnight at 4°C. Plates were washed three times with PBS plus 0.1% Tween 20, blocked for 1 h at room temperature with PBS plus 0.1% Tween 20 and 5% bovine serum albumin (BSA), and incubated with samples or serial dilutions of purified mouse serum IgG as the standard (Merck) overnight at 4°C. Plates were washed as described above and incubated with horseradish peroxidase (HRP)-conjugated goat antimouse IgG antibody (Jackson Immunoresearch) (1:2,000 dilution) for 2 h at room temperature. Plates were washed as described above and developed by the addition of 100 mL 3,39,5,59-tetramethylbenzidine (TMB; Thermo Fisher Scientific) for 5 to 10 min before quenching with 2 N H 3 PO 4 . The absorbances at 450 and 540 nm were measured in a plate reader (FLUOstar Omega; BMG Labtech), and anti-C. rodentium IgG was quantified in samples by comparison to a mouse IgG standard curve.
Quantitative reverse transcription-PCR (qRT-PCR). RNA was isolated from extracted cIECs using the RNeasy minikit (Qiagen) according to the manufacturer's instructions. RNA was treated with RQ1 RNase-free DNase (Promega) for 30 min, and subsequently, cDNA was synthesized using a Moloney murine leukemia virus reverse transcription kit (Promega). Amplification from cDNA was performed using Power SYBR green PCR master mix (Thermo Fisher Scientific). Assay reactions were performed in a 20-mL volume with 2Â Power SYBR green PCR master mix, efficiency-optimized primers to a final concentration of 0.05 mM each, and 1 to 10 ng total cDNA. All reactions were carried out in technical duplicate. The DDC T method of quantification was performed to give Log 2 FC values relative to the expression of the averaged baseline measurement (48). Expression was normalized to the expression of the housekeeping gene Gapdh. Primer pairs used are listed in Table S4.
Single-stain controls for compensation were prepared using VersaComp beads (Beckman Coulter) and ArC amine-reactive compensation beads (Thermo Fisher Scientific). Flow cytometry analysis on 50,000 single cells for cIEC analysis and 50,000 live cells for immune cell analysis was performed on a BD LSRFortessa cell analyzer (BD Biosciences). Data were analyzed using FlowJo v10.7.1. Any immune cell samples with fewer than 40% live cells were excluded from further analysis.
Statistical analysis and data presentation of proteomics data. Column normalization was used to correct for different protein loading in each channel, and each TMT set was scaled within itself to give relative abundance values. Where proteins were absent in some of the TMT channels, missing values were replaced with the minimum relative abundance value in the TMT set. For each infection time point, Log 2 FC values were calculated compared to the mock-infected mice in each experiment. Log 2 FC values were averaged across the three biological repeats for data representation. Proteins identified in only one biological repeat were excluded from all analyses. 1D enrichment, PCA, and statistical analysis of proteomic data sets were performed in Perseus (version 1.6.12) (49). Proteins significantly changed among all groups were determined by one-way ANOVA with permutation-based FDR correction (FDR of ,0.05) for multiple testing, performed on the relative abundance values. For comparison of individual groups, Tukey's honestly significant difference (HSD) posttest was applied to proteins considered significantly changed by one-way ANOVA (FDR of ,0.05). The "Hallmarks" annotation was extracted from the msigdb package in R (50). 1D enrichment values in Fig. 1D represent the average enrichment scores, where the process was considered significantly enriched in at least 2 out of 3 biological repeats. Heat maps were generated and hierarchal clustering was performed using Morpheus (https://software .broadinstitute.org/morpheus). For comparison of the data reported here with our previously published proteomic data sets from the establishment and early steady-state stages of C. rodentium infection (20,28), data were normalized using z scores [z score = (Log 2 FC 2 mean sample Log 2 FC)/standard deviation of the sample Log 2 FC]. The "top changed" proteins at 10 to 20 DPI (presented in Fig. S3A) indicate the proteins with the greatest Log 2 FC when the average Log 2 FC values of time points from 10 to 20 DPI of all proteins considered significantly changed (FDR of ,0.05 by one-way ANOVA) are ranked from highest to lowest in terms of the absolute Log 2 FC.
Statistical analysis and data presentation of nonproteomic data. Data were plotted and statistically analyzed in Prism (version 8). Data that passed a Shapiro-Wilk normality test (quantitative reversetranscription PCR [RT-qPCR] data) or the Kolmogorov-Smirnov normality test (all other data) were analyzed using ordinary or repeated-measures (RM) one-way ANOVA with Dunnett's posttest for multiple comparisons to a control group or unpaired two-tailed Student's t test, as specified in the figure legends. P values of ,0.05 were considered significant. Data that did not pass the specified normality test were log transformed, and, if necessary, the robust regression and outlier removal (ROUT) method (where Q, which reflects the probability of falsely identifying an outlier, was kept at 1%) was then applied to identify any outliers that were removed prior to statistical analysis. Log-transformed data with outliers removed that did not pass the specified normality test were not subjected to further statistical analysis, as stated in the figure legends. For statistical analysis of IFN-g ELISA data for colonic explants (Fig. 4K), the data point that fell below the limit of detection (31.25 pg/g tissue) was given an imputed value of 30.
Data replicates. For temporal proteomic and qRT-PCR analyses, four biological repeats (three biological repeats for proteomics analysis and one biological repeat for qRT-PCR analysis), each using five mice/group (time point), were performed. Histopathological and immunofluorescence analyses of the distal colon at 10 to 48 DPI were performed on the same mice as the ones used for proteomics and qRT-PCR analyses. For all other experiments, biological replicates are detailed in the figure legends.
Data availability. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (51) partner repository with the data set accession number PXD022230.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.

ACKNOWLEDGMENTS
We thank Izabela Glegola-Madejska for assistance with mouse experiments, Theo Roumeliotis for assistance with proteomic data analysis, Sarah Jordan for technical assistance, and Lorraine Lawrence for performing histological staining. We declare that no competing interests exist.