Article Text
PDF
Abstract
Objective Loss of the Crohn’s disease predisposing NOD2 gene results in an intestinal microenvironment conducive for colonisation by attaching-and-effacing enteropathogens. However, it remains elusive whether it relies on the intracellular recruitment of the serine-threonine kinase RIPK2 by NOD2, a step that is required for its activation of the transcription factor NF-κB.
Design Colonisation resistance was evaluated in wild type and mutant mice, as well as in ex-germ-free (ex-GF) mice which were colonised either with faeces from Ripk2-deficient mice or with bacteria with similar preferences for carbohydrates to those acquired by the pathogen. The severity of the mucosal pathology was quantified at several time points postinfection by using a previously established scoring. The community resilience in response to infection was evaluated by 16S ribosomal RNA gene sequence analysis. The control of pathogen virulence was evaluated by monitoring the secretion of Citrobacter-specific antibody response in the faeces.
Results Primary infection was similarly outcompeted in ex-GF Ripk2-deficient and control mice, demonstrating that the susceptibility to infection resulting from RIPK2 deficiency cannot be solely attributed to specific microbiota community structures. In contrast, delayed clearance of Citrobacter rodentium and exacerbated histopathology were preceded by a weakened propensity of intestinal macrophages to afford innate lymphoid cell activation. This tissue protection unexpectedly required the regenerating family member 3β by instigating interleukin (IL) 17A-mediated neutrophil recruitment to the intestine and subsequent phosphorylation of signal transducer and activator of transcription 3.
Conclusions These results unveil a previously unrecognised mechanism that efficiently protects from colonisation by diarrhoeagenic bacteria early in infection.
- Barrier Function
- Interleukins
- Macrophages
- Colonic Microflora
- Antibacterial Peptide
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
Citrobacter rodentium that mimics human infections with enteropathogenic Escherichia coli and enterohaemorrhagic E. coli is outcompeted, at least in part, by metabolically related commensals.
In response to bacterial muramyl dipeptide, the nucleotide-binding oligomerisation domain containing protein 2 (NOD2) physically interacts with caspase recruitment domain-containing protein 9 (CARD9) for downstream activation of Jun N-terminal kinase and p38 mitogen-activated protein kinase family members. Loss of either NOD2 or CARD9 impairs mucosal immunity against C. rodentium.
Shaping of the composition of the gut microbiota depends on the Crohn’s disease predisposing genes encoding for NOD2 and CARD9, through yet poorly understood mechanisms.
Significance of this study
What are the new findings?
The loss of RIPK2 results in specific alterations in the bacterial composition of the gut microbiota and in delayed clearance of C. rodentium infection in mice.
The transfer of the dysbiotic microbiota of Ripk2-deficient mice in germ-free mice did not affect their colonisation resistance against C. rodentium.
Clearance of C. rodentium requires the regenerating family member 3β (REG3b) that is secreted downstream of RIPK2 signalling.
Loss of either RIPK2 or the C-type lectin REG3b impairs the ability of intestinal phagocytes to trigger IL-17A-mediated neutrophil influx by innate lymphoid cells for efficiently limiting pathogen colonisation and tissue damage at the early phase of the infection.
How might it impact on clinical practice in the foreseeable future?
With the above findings in mind, caution should be used when prescribing RIPK2 inhibitors for limiting outbreaks of diarrhoeal diseases.
Knowledge on the alarmin REG3b (also referred to as pancreatitis-associated protein 1 in humans) will help restore patient’s mucosal barrier protection to normalcy and also limit the rise of antibiotic resistance by engineering enteric delivery of such peptides with immunostimulatory properties on gut barrier function.
Introduction
Enterohaemorrhagic Escherichia coli (EHEC) and enteropathogenic E. coli (EPEC) are significant causes of mortality and morbidity worldwide, especially among neonates. The naturally occurring mouse pathogen Citrobacter rodentium is widely used for studying how such human diarrhoeal infections are caused by EHEC/EPEC.1 As observed in humans, C. rodentium infection results in self-limiting colitis that is characterised by crypt hyperplasia and goblet cell loss.2 The locus of the enterocyte effacement pathogenicity island (that codes for a type III secretion system) allows this Gram-negative pathogen to co-opt host cell machinery for forming actin-rich pedestal-like structures that anchor the bacterium to the host membrane.3 The intimate attachment of the bacillus to the membrane of intestinal epithelial cells results in a local effacement of the brush border microvilli that ultimately contributes to the formation of attaching-and-effacing lesions. As a result, some monosaccharide-consuming bacteria compete for nutrients with C. rodentium 4 and contribute to its intraluminal clearance and systemic spread by regulating the induction of specific IgG responses against virulent bacteria for host survival.5
The major Crohn’s disease predisposing NOD2 gene encodes for the nucleotide-binding oligomerisation domain protein 2 (NOD2) that functions as an intracellular sensor for bacterial muramyl dipeptide (MDP). It is composed of a C-terminal leucine-rich repeat region, a centrally located NOD and an N-terminal tandem of caspase recruitment domains that interact within the cell with either the receptor-interacting serine/threonine-protein kinase 2 (RIPK2 also known as RICK, CARDIAK, CCK and RIP2)6 or the caspase-recruitment domain (CARD)-containing adaptor protein CARD9 (7). While CARD9 is needed for MDP-induced activation of the kinases p38 and Jun N-terminal kinase in myeloid cells,7 RIPK2 is expressed in several cell types and required downstream of both NOD1 and NOD2 for induction of autophagy and nuclear factor-kappa B (NF-κB) signalling activation.8–11 Previous studies have shown that the greater susceptibility to C. rodentium among either Nod1/2-deficient or Card9-deficient mice is linked to failure in mounting a robust humoral response against C. rodentium and to a delayed recruitment of inflammatory monocytes that is preceded by early defects in the interleukin (IL) 17 response.12–14 While being involved in neutrophil recruitment, IL-17A is particularly secreted on activation of both T helper cells (Th) 17 and type 3 innate lymphoid cells (ILC3) that both depend on the transcription factor RORγt for their development. Their activation in response to C. rodentium is at least partially attributed to intestinal macrophages arising from inflammatory monocytes through the C-C chemokine receptor type 2.15 In this context, the aim of this study was to interrogate the contribution of RIPK2 signalling on the resolution of C. rodentium infection.
Whereas CARD9 is required for regulating the ability of the microbiota to outcompete C. rodentium, we herein demonstrate that RIPK2 contributed to early phases of host defence against C. rodentium in a microbiota-dependent manner, but independently of microbiota community structure driven by RIPK2 deficiency. When compared with CARD9 deficiency,13 we found that loss of RIPK2 differently alters the composition of the gut microbiota at baseline and the faecal transplantation experiment excluded the possibility that these changes may regulate the virulence of the bacteria. Equally of importance is that loss of RIPK2 lowered the secretion of several Th17-inducing cytokines that are required for protective immunity to such extracellular pathogen.16 17 Such impaired secretion of IL-17A by ILC3 was not a consequence of the dysbiotic microbiota of Ripk2-deficient mice, but was rather associated with a delayed recruitment of neutrophils to the intestine and a blunted activation of the pleiotropic transcription factor STAT3. This defect was caused by an impaired secretion of the regenerating family member 3β (REG3b, also known as pancreatitis associated protein 1 in humans) that is regulated by STAT3 for barrier protection. REG3b is a soluble calcium-dependent carbohydrate-binding protein that displays anti-inflammatory properties.18 We then provided evidence that the colonisation resistance of Ripk2-deficient and Reg3b-deficient mice was improved when using a genetically modified dairy Lactococcus lactis NZ9000 for enteric delivery of pancreatitis-associated protein 1. In this work, we have identified a pathogen-induced regulatory circuit in which REG3b is critical for neutrophil recruitment at the early phase of the infection by acting on macrophages to amplify IL-17A production by ILC3. Altogether, these findings provide an unexpected mechanism on how NOD1/2 signalling contributes to another central aspect of host immune response against such enteric bacterial pathogens, which activates ILC3 for optimal production of IL-17A that is required for neutrophil recruitment at the site of infection.
Results
RIPK2 signalling is required for efficient control of C. rodentium infection
To study in detail how mucosal immunity to enteric bacterial infection is controlled by NOD1/2 signalling independently of CARD9 (figure 1A), we first studied C. rodentium-driven colitis in Ripk2-deficient mice that were reared under conventional specific pathogen-free (SPF) conditions. The eradication of C. rodentium was monitored by determining bacterial numbers in faeces of infected wild type (WT) and Ripk2-deficient (Ripk2-/- ) mice on days 0, 4, 12, and 20 postinfection (pi). Mutant mice displayed higher bacterial titres in their faeces as early as day 12 pi (figure 1B). This severely delayed pathogen clearance was similar to what was observed in mice deficient for either NOD1 or NOD2 (online supplementary figure 1 and figure 1B respectively). It was preceded with a pronounced accumulation of the active form of caspase-3 in intestinal epithelial cells (online supplementary figure 2A) that may subsequently inhibit a protective type 1 interferon response against C. rodentium.19 Consequently, a greater bacterial dissemination to distant spleen (figure 1C) was associated with splenomegaly among mutant mice by day 12 pi (figure 1D). Histological analyses of H&E-stained tissue sections (figure 1E) confirmed that these changes occurred prior to the onset of hyperplasia, disruption of normal epithelial crypt architecture, oedema and ulceration at the distal colon of Ripk2-/- mice (figure 1F and online supplementary figure 2B). Likewise, similar findings were observed in the caecum from infected Ripk2-/- mice (data not shown). Thus, our results revealed that loss of RIPK2 results in a delayed clearance of C. rodentium infection as a likely consequence of a defective pathogen sensing through NOD1/2.
Ripk2-deficient mice are susceptible to Citrobacter rodentium infection. Ripk2-/- (n=20), Nod2-/- (n=24) and wild type (WT) mice (n=20) housed under specific pathogen-free (SPF) conditions were orally infected with 1×109 colony forming units (CFU) of C rodentium. Results are representative of at least two independent experiments. (A) Scheme of study design. (B) Pathogen load was quantified by determining CFU counts in faecal samples at the indicated day postinfection (dpi). (C) Number of translocated bacteria was determined in the spleens. (D) Spleen weights were measured. (E) Grade of colonic inflammation was evaluated by histological scoring of H&E-stained tissue sections, representatively shown aside. (F) Representative H&E stained images at 20 dpi. Scale bar represents 200 µm. Data are plotted as mean±SEM. Statistical significance was calculated by non-parametrical Mann-Whitney test. P values <0.05 (*) were considered statistically significant.
The significant alterations of the gut microbiome caused by RIPK2 deficiency are not sufficient for impairing the colonisation resistance against C. rodentium
Loss of either NOD2 or CARD9 signalling renders mice more susceptible to colitis as a consequence of specific changes in the bacterial composition of the gut microbiota.20 21 Likewise, the Crohn’s disease-associated NOD2 variant L1007fsinsC is associated with enhanced mucosal colonisation by the Bacteroidaceae in humans.22 23 These results suggested that the infection severity in Ripk2-deficient mice may result from the failure of its microbiota to efficiently repress pathogen virulence. Indeed, C. rodentium is outcompeted by some commensals that are growing on structurally similar carbohydrates, such as E. coli.4 To assess this eventuality, we performed 16S ribosomal RNA gene sequence analysis by using the caecum content of WT, Nod2-/- and Ripk2-/- mice with respect to days 0, 4, 12 and 20 pi (figure 3A). With regard to taxonomic abundance distributions among non-infected animals, LEfSe (linear discriminant analysis effect size) analysis revealed that several bacteria belonging to the Burkholderiales subdivision of Proteobacteria were significantly enriched in controls when compared with mutant mice that were housed in the same room of our vivarium (online supplementary figure 3A,B). In contrast to what was observed in either Nod2-deficient or Card9-deficient mice,20–22 24 25 some Lachnospiraceae (but not specifically segmented filamentous bacteria) were found more abundant in Ripk2-/- mice at baseline (online supplementary figure 3C). On C. rodentium infection, significant changes in microbial diversity have occurred over time in both mutant mice compared with WT (online supplementary figure 3D). Consequently, the distance between bacterial communities was significantly changed during the infection course according to the genotype (online supplementary figure 3E), which coincided with an earlier drop in α diversity in Ripk2 -/- mice (online supplementary figure 3F). To determine whether RIPK2 signalling modulates the ability of E. coli to regulate the virulence of C. rodentium,4 we generated germ-free (GF) Ripk2-deficient mice and monitored the burden of C. rodentium for the whole infection course (figure 2A). By contrast to what was reported in either Card9-deficient mice13 or in those lacking mature B and T lymphocytes,5 the specific intestinal IgG/IgA response against C. rodentium coincided with similar luminal levels of the inflammatory lipocalin-2 in control and mutant mice (figure 2B,C). Consequently, the number of C. rodentium was similarly reduced in the faeces of mutant and control mice by more than 15-fold (figure 2D). This is in agreement with a comparable spleen weight (figure 2E) and pathogen translocation in the spleen (figure 2F) that were noticed as early as 2 weeks after their inoculation with monosaccharide-consuming E. coli. This indicated that the presence of the microbiota contributed to the severity of the colitis in Ripk2-deficient mice. Consistently, a greater susceptibility of Rag2 -/- γc -/- Ripk2 -/- was observed when compared with mice that are solely deficient for RIPK2 (online supplementary figure 4). We next directly assessed the impact on the colonisation resistance against C. rodentium of specific compositional changes in the gut microbiota that are caused by RIPK2 deficiency under enhanced hygienic conditions. To this end, we performed faecal microbiota transplantation (FMT) by introducing the gut microbiota from WT (F-WT) and Ripk2-/- (F-Ripk2-/- ) mice into GF WT recipients 3 weeks prior to C. rodentium infection (n=3–5 mice per time point). C. rodentium infection was monitored over the indicated time (figure 2G). Unexpectedly, we failed to observe significant differences on pathogen burden in stools (figure 2H) despite some significant compositional changes in the gut microbiota at both baseline and the indicated day pi (online supplementary figure 5). Consequently, spleen weights (figure 2I), pathogen invasion into the spleen (figure 2J) and accumulation of the active form of caspase-3 (online supplementary figure 6A) were found comparable among ex-GF animals. Furthermore, histological analyses of their colon failed to reveal differences in mucosal pathology (figure 2K,L and online supplementary figure 6B). Collectively, these results indicate that the pathogen was effectively outcompeted by the gut microbiome of Ripk2-deficient mice despite some specific alterations in its bacterial composition over time, including persistence of segmented filamentous bacteria that are strong inducers of Th17 response.
Ripk2 signalling is dispensable for regulation of phenotypically virulent bacteria by commensals with similar catabolic preferences for saccharides and delayed clearance of Citrobacter rodentium of Ripk2-deficient mice is not attributed to its significant alterations of their gut microbiome at baseline. (A) Scheme of study design. Wild type and Ripk2-deficient germ-free (GF) mice were infected with the DBS120 strain of C. rodentium. At day 19 pi, 1×109 CFU of Escherichia coli was inoculated. Pathogen load was quantified by determining CFU counts in faecal samples on the indicated days before and after inoculation of E. coli. (B) Production of C. rodentium-specific IgG in the faeces of ex-GF WT and Ripk2 –/– mice (n=7–8). (C) Lipocalin-2 levels in faeces of ex-GF mice. (D) Pathogen load was quantified by determining CFU counts in faecal samples at the indicated day pi (dpi). (E) Spleen weights were measured. (F) Number of translocated bacteria was determined in the spleens. (G) Scheme of study design. faecal microbiota transplantation (FMT) was performed on GF WT mice by using faecal samples from WT (F–WT) and Ripk2 -/- (F-Ripk2-/-) mice reared under conventional specific pathogen-free (SPF) conditions. Mice were recolonised for 3 weeks before being infected with 1×109 CFU of C rodentium. Mice were autopsied on days 0, 4, 12, and 20 pi (n=18–20; 3–5 per time point). (H) Bacterial load was assessed by determining CFU counts in faecal samples at the indicated dpi. (I) Number of translocated bacteria was determined in the spleens at days 0, 4, 12 and 20 pi. (J) Spleen weights were measured on days 0, 4, 12 and 20 pi. (K) Grade of colonic inflammation was evaluated by histological scoring of H&E-stained tissue sections, representatively shown aside. (L) Representative H&E-stained images at 20 dpi. Scale bar represents 200 µm. Data are plotted as mean±SEM. Statistical significance was calculated by non-parametrical Mann-Whitney test. A p value <0.05 (*) was considered statistically significant.
Induction of the C-type lectins REG3b and REG3g is lost in the intestine of Ripk2-deficient mice in response to Citrobacter rodentium. Microarray analysis was performed by using total tissue from the caecum tip of mice on days 0, 4 and 12 postinfection (pi) (n=5). Changes (log2-fold>1 and p<0.01) in gene expression profiles of Ripk2-/- and wild type (WT) mice on day 4 pi. (A) and (B) on day 12 pi were calculated relative to non-infected WT controls and compared by correlation analysis. Comparative expression of genes in the caecum by microarray data analysis on day 4 pi (C) and on day 12 pi (D). (E) Expression levels for REG3b and REG3g were examined by RT-qPCR using total tissue from the caecum tip of mice on days 0, 4, 12 and 20 pi. (F) Immunofluorescence staining for REG3b was performed by using tissue sections from the caecum tip of WT and mutant mice on day 4 pi. Non-infected WT mice served as negative controls. Epithelial lining and cell nuclei were visualised by using antibody against E-cadherin and DAPI staining.
C. rodentium fails to elicit an IL-17A-inducing local circuit in the absence of RIPK2 signalling
To get further insights on how RIPK2 signalling can specifically contribute to the eradication of C. rodentium independently of the bacterial community structure, we next performed microarray analysis by using total RNA from caecal tissue of infected WT and Ripk2-/- mice that were sacrificed on days 0, 4 and 12 pi which corresponds to the peak of infection. Comparison of gene expression profiles revealed that RIPK2 signalling contributes to a large proportion of the mucosal response to C. rodentium infection (figure 3A,B). Venn diagram analysis revealed a unique set of 122 transcripts to be differentially expressed at day 4 pi in WT mice, of which merely 30 were independent of RIPK2 signalling (figure 3A). By contrast, RIPK2 was required for the expression of about 702 genes that were differentially regulated at day 12 pi (figure 3B). Meanwhile, gene set enrichment analysis on upregulated genes by RIPK2 signalling revealed that the latter set of transcripts was involved in inflammatory responses and immune system process at day 4 pi. Of these, we noticed that the downstream genes of RIPK2 signalling with the greater mean difference were the C-type lectin REG3b and its closest family member REG3g (figure 3C on day 4 pi and figure 3D on day 12 pi). While REG3g was previously found dispensable for the clearance of C. rodentium,26 the lack of REG3b induction in the intestine of infected Ripk2-deficient mice was next confirmed by reverse transcription (RT)-qPCR (figure 3E). As expected, a lowered protein expression of REG3b within the intestinal epithelium of Ripk2-deficient mice was noticed by immunofluorescence staining (figure 3F).
The colonisation of C. rodentium is restrained by REG3b secretion downstream of RIPK2 signalling
We then examined whether mucosal immunity to C. rodentium may require REG3b secretion downstream of RIPK2 signalling. To address this question, Reg3b-deficient and WT mice were orally inoculated with C. rodentium and the intraluminal eradication of the pathogen was monitored for a period of 3 weeks (figure 4A). In contrast to what was observed in Reg3g-deficient mice,26 the intestine of Reg3b-deficient mice remained heavily colonised by day 12 pi (figure 4B) which coincides with the induction of pathogen-specific IgG responses.4 Consequently, this resulted in a greater spleen weight in Reg3b-deficient mice (figure 4C) that were prone to bacterial dissemination to the spleens (figure 4D). On day 21 pi, the histopathological evaluation of H&E-stained, colonic tissue sections revealed increased crypt lengths (figure 4E) and overt inflammation in Reg3b-deficient mice when compared with controls (figure 4F). In view of these results, we first sought to investigate whether REG3b possesses direct antimicrobial activity against C. rodentium as what was observed against several Gram-positive bacteria.27 28 This antibacterial function was first ascribed to REG3b because it contains a conserved C-terminal C-type lectin carbohydrate recognition domain that specifically binds to bacterial peptidoglycan structure.27 28 To explore this possibility directly, increasing concentrations of recombinant REG3b (0.5–2 µM) were incubated with bacterial culture as previously described.27 Bacterial survival rate expectedly remained unchanged when REG3b-treated bacterial cultures of C. rodentium were compared with untreated controls (online supplementary figure 7A). We also examined whether REG3b could facilitate macrophage binding and engulfment of C. rodentium by evaluating the phagocytic activity of macrophages in response to REG3b. To this end, we applied recombinant REG3b on either Ripk2-deficient or WT primary bone marrow-derived macrophages, which were subsequently infected with C. rodentium at a multiplicity of infection of 10 for 2 hours (online supplementary figure 7B). Bacterial engulfment and killing was evaluated after infection, but comparable bacterial count of internalised C. rodentium was observed following plating of lysates from bone marrow-derived macrophages (online supplementary figure 7C). Overall, our findings highlight the requirement of REG3b for appropriate barrier protection against C. rodentium independently of any direct bacterial killing activity against this Gram-negative bacterium.
Clearance of Citrobacter rodentium is delayed in Reg3b-deficient mice. Reg3b-/- and controls (n=12–15) were orally inoculated with 109 CFU of C. rodentium. Data are representative of at least two independent experiments. (A) Scheme of study design. (B) Bacterial load per gram faeces was determined over a period of 3 weeks. At the end of infection, bacterial translocation to spleen (C) and spleen weight (D) was seen. (E) Histological scoring was determined on 21 days postinfection (dpi). (F) Representative pictures of H&E-stained tissue sections at day 21 pi are shown aside. Scale bar represents 200 µm. Data are plotted as mean±SEM. Statistical significance was evaluated by non-parametrical Mann-Whitney test. Values of p<0.05 (*), p<0.01 (**) and p<0.001 (***) were considered statistically significant. WT, wild type.
The RIPK2/REG3b circuit enhances the propensity of intestinal phagocytes to trigger IL-17A-mediated secretion by ILCs at the site of infection
As we had excluded the possibility of bactericidal and/or phagocytic activity of REG3b on C. rodentium (online supplementary figure 7), we next tested the hypothesis that REG3b may fundamentally serve as an alarmin for restraining the expansion of C. rodentium through induction of neutrophil recruitment in response to Th17-inducing cytokines. In line with our microarray data depicted in figure 3, the colon of Reg3b-deficient mice showed significant changes at the early stages of C. rodentium infection with respect to the expression level of the genes encoding for IL-22 and for tumour necrosis factor α (TNF-α) that indirectly functions as a Th17-inducing cytokine through IL-6 and IL-1β secretion by monocytes29 (figure 5A and data not shown). A significantly lowered protein level of TNF-α was confirmed among CD11c- CD11b+ F4/80+ MHCIIhi that were sorted from the intestine of Ripk2-deficient mice when compared with controls (figure 5B). While IL-10 remained below detectable levels (data not shown), a similar decrease was observed for IL-1b (data not shown), which is another essential cytokine for protecting the tissue from severe infection by the pathogen30 and for the secretion of IL-17A by ILC3 within the intestinal mucosa.31 Strikingly, a lowered protein level of IL-17A temporally coincided with a lowered activation of the transcription factor STAT3 within the intestine of both Ripk2-deficient and Reg3b-deficient mice at day 4 pi when compared with WT controls (figure 5C and online supplementary figure 8). This effect on STAT3 activation may account for the lowered expression of REG3b in the absence of RIPK232 and subsequently facilitates the systemic spread of C. rodentium as was observed in mice with specific deletion of STAT3 in intestinal epithelial cells.17 In this context, we suspected that the weakened production of Th17-inducing cytokines by intestinal phagocyte macrophages from infected Reg3b-deficient mice may subsequently compromise the activation of ILC3s15 that is the major source of IL-17A within the gut mucosa.33 This defect may subsequently be amplified through lowered phosphorylation of STAT3 that is regulating either REG3b or REG3g.32 To directly assess this hypothesis, RORγt+CD3- ILCs were isolated from the intestinal mucosa of RORγtGFP/+ mice and were cocultured with CD11b+ F4/80+ MHCIIhi macrophages that were sorted from the intestine of Ripk2-deficient mice at either day 4 or day 12 pi. While IL-17A secretion was barely detectable from individually cultured ILCs and macrophages (data not shown), this was further enhanced when ILCs were cocultured with intestinal macrophages from WT mice that were orally infected by the pathogen (figure 5D). In contrast, IL-17A production by ILCs was blunted when they were cocultured with purified intestinal macrophages of infected Ripk2-deficient mice (figure 5D), while having no effect on the IL-22-secreting subset of ILCs (data not shown). Similar results were observed with the mononuclear phagocytes from infected Reg3b-deficient mice (figure 5E), providing a powered mechanism for the lowered activation of STAT3 and decreased secretion of IL-17A in both mutant mice. Collectively, these results unexpectedly indicate that RIPK2 deficiency specifically impairs activation of IL-17A-secreting ILCs by intestinal phagocytes in response to C. rodentium and this process is regulated downstream of REG3b.
Requirement for secretion of Th17-specifying cytokines and ILCs activation by intestinal phagocytes of the RIPK2/REG3b circuit at the site of infection. Reg3b-/- mice and controls (n=10) were orally inoculated with 1×109 CFU of C. rodentium. (A) Total tissue of distal colon from wild type (WT) and mutant mice (n=3–5) was analysed by RT-qPCR for the expression of tumour necrosis factor (TNF)-α-encoding gene on days 0, 4, 12 and 20 postinfection (pi). (B) Colonic lamina propria mononuclea cells (LPMCs) were isolated on day 4 pi. CD11b+F4/80+ cells were sorted by flow cytofluorometry and cultivated overnight. Supernatants were used to evaluate secretion levels of TNF-α. (C) The IL-17A level from the supernatant of colonic explants of WT and mutant mice was specifically determined on day 4 pi by ELISA. (D) The IL-17A level from the supernatant of ILCs that were cocultured with sorted macrophages of either infected WT or mutant mice. Individual values were plotted with mean±SEM. Statistical significance was assessed by non-parametrical Mann-Whitney test. A value of p<0.05 (*) was considered statistically significant. Results are representative of at least two independent experiments.
The defect in STAT3-driven Th17-like response is not attributed to specific alterations in the gut microbiome of Ripk2-deficient mice
To evaluate the impact of specific alterations in the bacterial composition of the gut microbiota that are caused by RIPK2 deficiency, we next performed microarray analysis by using total RNA from caecal tissue of F-Ripk2-/- and F-WT mice that were sacrificed on days 0, 4 and 12 pi. Changes in gene expression profiles were assessed by comparing gene regulation of infected mice to controls under steady-state conditions. Comparison of gene expression profiles revealed numbers of significantly regulated genes in F-Ripk2-/- mice when compared with F-WT controls despite comparable pathogen loads (online supplementary figure 9A,B). Highest differences among gene expression levels were observed for F-Ripk2-/- mice on day 4 pi (online supplementary figure 9A), while numbers of regulated genes on day 12 pi were equally distributed between ex-GF mice showing approximately 75%–85% similarity in gene expression profiles (online supplementary figure 9B). However, no difference for either Reg3b, Reg3g, Tnfa, IL1b, IL6, IL17a or Il22 expression was observed among ex-GF mice (online supplementary figure 9C and data not shown). Consistently, STAT3 was similarly activated at day 4 pi in the caecum of F-Ripk2-/- mice when compared with F-WT (online supplementary figure 9D). In conclusion, our microbiota transplantation experiments ruled out the possibility that the altered host response against enteric bacterial infection in the absence of RIPK2 signalling may result from effects of infection on the resilience of bacterial community structures.
The RIPK2/REG3b circuit locally prompts IL-17A-mediated neutrophil recruitment for controlling colonisation resistance
Considering these results, we sought to investigate whether this defective Th17-like response that was preceded by a lowered expression of the aforementioned C-type lectins is a consequence of a defective adherence of C. rodentium to intestinal epithelial cells.34 To this end, we performed fluorescent in situ hybridization (FISH) analysis by using tissue sections from caecum of mice sacrificed on day 4 pi. As observed in WT mice, we found bacteria that were intimately attached to the epithelial lining of mutant mice, which further indicates ongoing epithelial shedding in response to C. rodentium infection (online supplementary figure 10). While being part of an innate Th17-inducing gene expression programme, we also noticed similar transcript levels of IL-23 and serum amyloid A1 that is regulated by IL-23 between WT and mutant mice (online supplementary figure 11). These findings suggested that REG3b induction downstream of RIPK2 signalling may promote clearance of C. rodentium through STAT3 activation and recruitment of polymorphonuclear leucocytes by Th17-inducing cytokines at the site of infection.35 To test this, we performed non-invasive monitoring of the neutrophil influx that is required for clearance of C. rodentium.36 To this end, faeces were collected from infected mice and we performed ELISA measurements of lipocalin-2 (also referred to as siderocalin, 24p3 and NGAL for neutrophil gelatinase-associated lipocalin) that is primarily secreted by neutrophils. While levels of faecal lipocalin-2 were found to be similar among ex-GF mice (figure 6A), it was found markedly lowered in the absence of REG3b (figure 6B). Consistently, this reduced faecal amount of lipocalin 2 positively correlated with neutrophil recruitment to the intestine in response to C. rodentium (figure 6C). Likewise, a decreased level of faecal lipocalin-2 was observed in Ripk2-deficient mice in the early phase of the infection even if it was close to statistical significance (data not shown). We next assessed whether it may result as a consequence of IL-17A secretion in response to local production of REG3b downstream of RIPK2 signalling. Cytospin analysis of peritoneal cell composition showed a significant influx of neutrophils as early as 6 hours after intraperitoneal administration of recombinant endotoxin-free REG3b (figure 6D), while the level of macrophages, eosinophils and basophils remained unchanged with REG3b administration (data not shown). Expectedly, quantitative assessment of the infiltrating leucocytes in the peritoneum cavity by flow cytometric analysis showed a 7–10-fold increase in the CD11b+ GR1+ F4/80- neutrophil population when compared with that in mice treated with a control saline solution (figure 6E,F). In accordance with our previous findings,35 the proportion of CD11b+GR1-F4/80+ macrophages was unchanged on administration of REG3b (data not shown). In contrast, no significant neutrophil influx was observed in REG3b-treated Rag2 -/- γc -/- mice (1.40%±0.26%) when compared with saline-injected animals (figure 6E,F), suggesting that activation of some lymphoid cells is involved in REG3b-induced neutrophil influx. Likewise, the neutrophil population was significantly reduced in Il17a -/- mice that were treated with recombinant REG3b as compared with saline-injected controls (figure 6E,F). Additionally, IL-17A was found dispensable for recruitment of the GR1-F4/80+ macrophage population in vivo (data not shown). In agreement with our findings, further studies using a genetically modified dairy Lactococcus lactis NZ9000 for enteric delivery of pancreatitis-associated protein 1 revealed that such daily treatment improves colonisation resistance against the pathogen (online supplementary figure 12). We conclude that IL-17A signalling is required for triggering influx of IL-22-producing neutrophils in response to REG3b.
Delayed IL-17A-mediated influx of polymorphonuclear neutrophils to the intestine of Reg3b-deficient mice. (A) Lipocalin-2 levels in faeces of ex-GF mice at day 4 postinfection (pi). (B) Correlation between lipocalin-2 measurements and associated numbers of neutrophils in the caecum and colon of wild type (WT) and Reg3b-deficient mice at day 4 pi. (C) Representative fluorescence activated cell sorting (FACS) plots are shown for CD11b and Ly6G expression. (D) Associated numbers of neutrophils in the caecum and colon of WT and Reg3b-deficient mice at day 4 pi. (E–G) Peritoneal exudates of WT, Rag2 -/-γc-/- and Il17a -/- mice were harvested 6 hours after 100 µg/kg of purified, endotoxin-free REG3b intraperitoneally and analysed by flow cytometry for Gr-1 and F4/80 staining. Results are representative of four independent experiments. (E) Quantification of neutrophil influx was first made by cytospin analysis on phosphate buffer saline (PBS)-treated and Reg3b-treated mice. (F) Representative flow cytometric analysis of the peritoneal cell population from PBS-treated (top) and REG3b-treated (bottom) WT and mutant mice. (G) Values represent the mean±SEM. Asterisks indicate statistically significant differences, p<0.05.
Discussion
In this study, we examined how the RIPK2/REG3B circuit reciprocally regulates mucosal immunity with emphasis on its effect on the eradication of a Gram-negative enteropathogenic bacterium that models human infections with EPEC and EHEC. We herein provide evidence that the delayed eradication of such Th17-inducing bacterial pathogen in Ripk2-deficient mice is caused by a defective recruitment of IL-22-secreting neutrophils that is locally induced by REG3b, whereas having no effect on the secretion of IL-22 by ILCs. While RIPK2 signalling was found to modulate the composition of the gut microbiota, this defect in clearance of the pathogen was herein demonstrated to occur independently of bacterial community structure under carefully controlled environmental conditions. By contrast to what was observed in Reg3g-deficient mice,26 loss of either RIPK2 or REG3b subsequently impaired downstream activation of STAT3 that is required in intestinal epithelial cells for limiting expansion of C. rodentium.37 The early induction of IL-17A by RORγt+CD3- ILCs was lowered when cocultured with macrophages from infected Ripk2-deficient mice that secreted less Th17-specifying cytokines, including TNF-α and IL1b.31 Likewise, similar results were observed when coculturing RORγt+CD3- ILCs with macrophages that were isolated from infected Reg3b-deficient mice. Consequently, the secretion of IL-17A was required for REG3b-inducing recruitment of neutrophils that confers host protection against C. rodentium.32 One possibility for this dispensable role of REG3b on the secretion of IL-22 by ILCs is that the macrophages from mutant and control mice were secreting a similar amount of IL-23, which subsequently induces the production of IL-22 by such an ILC compartment.38 That knowledge on the alarmin REG3b will help in understanding the pathogenesis of IBD and in restoring mucosal barrier protection of patients with IBD to normalcy. One may also presume that our findings will limit the rise of antibiotic resistance by engineering peptides with immunostimulatory properties on gut barrier function. While CARD9 mediates susceptibility to intestinal pathogens through microbiota modulation and control of bacterial virulence,13 loss of the IBD-predisposing gene encoding for RIPK2 that signals downstream of NOD1/2 molecules modulates the susceptibility to intestinal infection by controlling the neutrophil recruitment to the intestine in a microbiota-dependent manner within carefully controlled environmental conditions. In this context, we suspected that this mechanism downstream of RIPK2 signalling may be important for ultimately limiting the expansion of C. rodentium as has been observed in response to other adherent bacteria such as segmented filamentous bacteria.39 Besides the possibility that the activation of RIPK2 signalling in epithelial cells may contribute to colonisation resistance, further work is now eagerly awaited for evaluating alternatives scenarios, including a role of REG3b in macrophages.40 Together this also raises the question whether the secretion of REG3b downstream of RIPK2 signalling may function as a crucial host defence mechanism against fungal infections as has been observed in mice that are deficient in either CARD9 or IL-17A.41 42 With this in mind, caution should be used when prescribing RIPK2 inhibitors for limiting outbreaks of diarrhoeal diseases.
Material and methods
Additional protocols and complete procedures are described in the online supplementary material and methods section.
Supplemental material
Mice
Age-matched and gender-matched Ripk2-deficent mice (Ripk2 -/-), Nod2-deficent mice (Nod2 -/-), Nod1-deficent mice (Nod1 -/-), Reg3b-deficient mice (Reg3b -/-) and C57BL/6J WT mice had free access to standard laboratory chow and were housed under SPF conditions.
Bacterial infection
Age-matched and gender-matched mice were orally inoculated with the kanamycin-resistant C. rodentium strain DBS120 (1×109 CFU). Infection course was monitored by counting CFU per gram faeces.
Faecal microbiota transplantation
GF WT mice were transplanted with donor microbiota from age-matched and sex-matched Ripk2-/- mice and WT controls by oral gavage. After 3 weeks of microbial recolonisation, ex-GF mice were orally inoculated with the DBS120 strain (1×109 CFU) under enhanced hygienic conditions.
Supplemental material
Acknowledgments
The authors thank F Dejardin, L Trottereau and T Durand for excellent technical assistance in managing the colony of mutant mice. The authors also thank N Jouy and H Le Roy Bauderlique for technical assistance and their help and support for the cell sort at the Flow Core Facility - BioImaging Center Lille (F-59000 Lille, France).
References
Footnotes
SK and TS contributed equally.
Contributors NW and MC performed study design, acquisition of data, analysis and interpretation of data and statistical analysis. NW, SK, TS, VZ, OB and LFP acquired, analysed and interpreted data with the help of EF, OB, MD and BL. H-PP, AS, MLR, PL and HS contributed to interpretation of microarray data, pyrosequencing data and statistical analysis. J-MC, J-CD, GE, BR, NK and JLI provided critical materials. NW and MC wrote the manuscript and all authors discussed the results and commented on the manuscript.
Funding This work was supported by the Fondation pour la Recherche Médicale grant (grant number DEQ20130326475) to MC and by the Agence Nationale de la Recherche grant (grant number ANR-13-BSV3-0014) to MC, AS and HS. NW is a recipient of a postdoctoral fellowship from the Agence Nationale de la Recherche (grant number ANR-13-BSV3-0014) and VZ is a recipient of a postdoctoral fellowship from the Agence Nationale de la Recherche (grant number ANR-13-PRTS-0006).
Competing interests None declared.
Patient consent Experimental studies using mice only.
Ethics approval The local investigational review board approved all animal studies (CEEA – “75 Comité d’Ethique en Expérimentation Animale Nord - Pas de Calais” (CEEA232009R). Animal experiments were performed in an accredited establishment (N° B59–108) according to FELASA and governmental guidelines N°86/609/CEE.
Provenance and peer review Not commissioned; externally peer reviewed.