Article Text

PDF

Original article
Type I interferon signalling in the intestinal epithelium affects Paneth cells, microbial ecology and epithelial regeneration
  1. Markus Tschurtschenthaler1,2,
  2. Jun Wang3,4,
  3. Cornelia Fricke5,
  4. Teresa M J Fritz5,
  5. Lukas Niederreiter1,
  6. Timon E Adolph1,
  7. Edina Sarcevic6,
  8. Sven Künzel3,
  9. Felix A Offner6,
  10. Ulrich Kalinke7,
  11. John F Baines3,4,
  12. Herbert Tilg2,8,
  13. Arthur Kaser1,5
  1. 1Division of Gastroenterology & Hepatology, Department of Medicine, University of Cambridge, Cambridge, UK
  2. 2Department of Medicine I, Innsbruck Medical University, Innsbruck, Austria
  3. 3Max Planck Institute for Evolutionary Biology, Plön, Germany
  4. 4Institute for Experimental Medicine, Christian-Albrechts-University of Kiel, Kiel, Germany
  5. 5Department of Medicine II, Innsbruck Medical University, Innsbruck, Austria
  6. 6Department of Pathology, Academic Teaching Hospital Feldkirch, Feldkirch, Austria
  7. 7Division of Immunology, Paul-Ehrlich-Institute, Langen, Germany
  8. 8Christian-Doppler Research Laboratory for Gut Inflammation, Innsbruck Medical University, Innsbruck, Austria
  1. Correspondence to Professor Arthur Kaser, Division of Gastroenterology and Hepatology, Department of Medicine, University of Cambridge, Addenbrooke's Hospital, Level 5, Box 157, Cambridge CB2 0QQ, UK; ak729{at}cam.ac.uk

Abstract

Objective Intestinal epithelial cells (IECs) at the internal/external interface orchestrate the mucosal immune response. Paneth cells secrete antimicrobial peptides and inflammatory mediators, protect from pathogens and shape the commensal microbiota. Prompted by the genetic association of the locus harbouring the type I interferon (IFN) receptor (IFNAR1) with Crohn's disease, and a transcriptional signature for type I IFN signalling in Paneth cells, we studied the function of IFNAR1 in IECs.

Design Type I IFN signalling was studied in mice with conditional deletion of Ifnar1 in IECs. Phenotype was characterised at baseline, and gut microbiota composition was assessed by 16S rDNA ribotyping. The role of IFNAR1 was also investigated in experimental colitis induced by dextran sodium sulfate (DSS) and colitis-associated cancer induced by DSS in conjunction with azoxymethane (AOM).

Results Ifnar1−/−(IEC) mice displayed expansion of Paneth cell numbers and epithelial hyperproliferation compared with Ifnar1-sufficient littermates. While Ifnar1−/−(IEC) mice did not exhibit spontaneous inflammation or increased severity in DSS colitis compared with Ifnar1+/+(IEC) mice, they exhibited an increased tumour burden in the AOM/DSS model. Both hyperproliferation and tumour promotion were dependent on the microbial flora, as the differences between genotypes were marked upon separately housing mice, but disappeared when Ifnar1−/−(IEC) and Ifnar1+/+(IEC) mice were co-housed. Accordingly, ribotyping revealed marked differences between Ifnar1−/−(IEC) and Ifnar1+/+(IEC) mice that where diminished upon co-housing.

Conclusions IFNAR1 in IECs, and Paneth cells in particular, contributes to the regulation of the host–microbiota relationship, with consequences for intestinal regeneration and colitis-associated tumour formation.

  • INTESTINAL EPITHELIUM
  • INTESTINAL GENE REGULATION
  • EXPERIMENTAL COLITIS
  • INTESTINAL MICROBIOLOGY
  • EPITHELIAL PROLIFERATION

Statistics from Altmetric.com

Significance of this study

What is already known on this subject?

  • A genetic locus on Chr21 is associated with Crohn's disease and computational prediction suggests IFNAR1 or IFNGR2 as the likely causal genes at this locus.

  • Paneth cells have been reported to exhibit a signature of type I interferon (IFN) signalling, but the role of type I IFN signalling in the intestinal epithelium is little understood.

  • Type I IFNs are well known for their antiviral, antiproliferative, antitumour and immunomodulatory functions.

  • The intestinal epithelium, which is uniquely exposed to a complex microbial habitat consisting of trillions of microbes, orchestrates the immune system to discriminate between pathogens and commensals.

What are the new findings?

  • Type I IFN receptor in the intestinal epithelium regulates the intestinal microbial ecosystem.

  • Type I IFN signalling determines Paneth and goblet cell numbers in the ileum.

  • Epithelial hyperproliferation and increased tumour burden in a colitis-associated cancer model are consequences of type I IFN receptor-deficient intestinal epithelium

  • Co-housing of epithelial type I IFN receptor-deficient with receptor-sufficient littermate animals rescues pathological microbial composition and surprisingly reverses the hyperproliferation in mice with an epithelial-specific deletion of type I IFN receptor

  • Gene–microbiota interaction determines phenotypic presentation, and ‘normal microbiota’ can mitigate phenotypes

How might it impact on clinical practice in the foreseeable future?

  • Functional interrogation is an essential step in translating genetic discoveries into personalised medicine and novel therapeutics.

  • Results highlight the complex role of the microbiota on the phenotype induced by genetic constitution of the host.

  • The tri-directional relationship of host genes–microbial habitat–phenotype has critical implications for the design of human microbiota studies.

  • Highlights the importance of proper design and reporting of housing information in mouse experiments.

Introduction

The inner surface of the intestine, which is lined by a single-cell layer of intestinal epithelial cells (IECs) and their specialised subtypes (Paneth, goblet and enteroendocrine cells), all arising from stem cells at the crypt bottom,1 builds the physical border between external environment and internal tissue. It provides the first line of defence against microbial pathogens, toxins and other environmental cues.2 The functional barrier of the intestine that separates the enormous abundance of microbial life within the intestinal microbial habitat, which includes bacteria, Archaea, fungi, viruses, and little characterised forms of life, from the generally sterile host does not only involve IECs as part of the physical barrier, but a large variety of innate and adaptive immune cells. IECs have emerged to play a fundamental role as central organisers of these mucosal immune responses.3

One major family of cytokines fundamentally involved in antimicrobial host defence are the type I interferons (IFN-α, IFN-β and further family members), which exert antiviral,4 ,5 antiproliferative,6 antitumour7 and a multitude of regulatory actions on innate and adaptive immune cells.8–11 Type I IFNs can be considered as first line of defence against viruses4 ,5 and probably also against certain bacterial infections.12 Type I IFNs are induced by stimulation of pattern recognition receptors of the innate immune system, such as Toll-like receptors (TLRs) and retinoic acid-inducible gene (RIG)-like helicases.9 They act through the ubiquitously expressed type I IFN receptor, consisting of IFNAR1 and IFNAR2 subunits, which constitutively associate with Janus kinase 1 (JAK1) and non-receptor tyrosine kinase 2 (TYK2)9 to enhance signal transducer and activator of transcription (STAT) activation, but also signal through mitogen-activated protein kinase (MAPK), phosphoinositide 3-kinase (PI3K) and nuclear factor κ B (NF-κB)-signalling pathways.9 Structurally unrelated type III IFNs (IFN-λ1-3; also named IL-29, IL-28A, IL-28B) signal through a receptor consisting of the heterodimeric IL-28Rα/IL-10Rβ chains and elicit a signal transduction response (IFN-stimulated genes; ISGs) similar to that of type I IFNs and also exert antiproliferative functions.13 ,14

A novel Crohn's disease (CD)-specific genetic susceptibility locus on chromosome 21 that was discovered through the immunochip experiment harbours IFNAR1, and in silico prediction pinpointed IFNAR1 and IFNGR2 as the likely causal risk genes at this locus.15 TLR9-elicited type I IFNs had previously been reported to protect mice from colitis incited by the endoplasmic reticulum stress-inducing16 agent dextran sodium sulfate (DSS),17 and Ifnar1−/− mice exhibited increased sensitivity to DSS colitis.17 Mechanistically, it was suggested that basolateral stimulation of IEC monolayers with type I IFNs elicited barrier protection as deduced from in vitro experiments,17 although in another model system, IFNAR1 signalling was only noted arising from the apical membrane.13 Elegant reporter experiments in Ifnar1−/− and Il28Rα−/− mice, which abrogate type I and type III IFN signalling, respectively, indicated that type I IFN signalling in small IECs had a less prominent role in handling oral infection with rotavirus infection compared with type III IFN signalling.13 However, expression of Ifnα mRNA and IFN-α-related transcripts including those that are involved in the type I IFN feed-forward amplification loop,18 such as Irf7, have been reported specifically in Paneth cells under homeostatic conditions.19 Finally, type I IFN-regulated genes have featured prominently in comparing transcriptomes in the large intestine of germ-free and mice held under specific pathogen-free (SPF) conditions,19 implying that type I IFNs might have an important role in host–microbe mutualism.

Since numerous IBD risk genes are functionally involved in intestinal epithelial function, and several key CD risk genes functionally converge on Paneth cell function in particular,20–24 we chose to study the function of IFNAR1 in the intestinal epithelium.

Materials and methods

Mouse strains

Ifnar1fl/fl mice25 ,26 were crossed with Villin-Cre mice27 (purchased from Jackson Laboratories), both backcrossed onto a C57Bl/6 background, to obtain Ifnar1fl/fl; Villin-Cre (Ifnar1/(IEC)) mice. If not stated otherwise, mice were kept separated by genotype and experiments were performed with gender-matched and age-matched littermate controls. Animals were housed under SPF conditions at ZVTA, Innsbruck Medical University, Innsbruck, and all animal procedures were approved by the Austrian Federal Ministry of Science and Research.

Histology

Formalin-fixed and paraffin-embedded intestines were sectioned and stained with H&E as described previously.23 Baseline inflammation in the small intestine was assessed on H&E sections using a score, which considers four criteria: mononuclear cell infiltrate (0–3), crypt hyperplasia (0–3), epithelial injury (0–3) and polymorphonuclear cell infiltrates (1–3). Extent factor was derived according to the area of inflammation: 1 for <10%, 2 for 10–25%, 3 for 25–50% and 4 for >50% involvement by inflammatory changes. Periodic acid Schiff (PAS) staining was performed according to standard protocols.

Induction of DSS colitis

At 8 week of age, mice received 4.5% DSS (MP Biomedicals) in their drinking water for 5 or 7 days, respectively. Mice weight and rectal bleeding was assessed daily. Formula used to calculate % weight change=(100 × weight day X)/day 0–100. Histological scoring was performed as described previously.28

Induction of AOM/DSS colitis

At 6–8 weeks of age, mice were injected intraperitoneally with 12.5 mg/kg azoxymethane (AOM) (Sigma) according to established protocols.29 ,30 Colitis was induced by two cycles of 2.5% DSS (MP Biomedicals) in drinking water for 5 days, followed by a 16-day tap water period. The final DSS cycle (2%) was administered for 4 days, followed by a 10-day tap water period. Tumour count and tumour area were determined at day 61 on serial sections of paraffin-embedded and H&E-stained ‘Swiss rolls’ using Image J and AxioVision Release V.4.8 software (Zeiss).

Bromodeoxy-uridine (BrdU) incorporation

2.5 mg BrdU (BD Pharmingen) was injected 24 h before sacrificing mice. BrdU+ cells were stained by BrdU In Situ Detection Kit (BD Pharmingen). BrdU+ cells per total cells along the crypt-villus axis were counted in 5–10 randomly selected crypt-villus axes per single sample.

16S rDNA ribotyping

Bacterial DNA from caecal contents was extracted using the PowerSoil Kit (MoBio, Carlsbad, California, USA). Approximately 200 mg of contents were transferred to the Power Bead tubes containing 60 µL of C1 solution and 20 µL of 20 mg/mL Proteinase K. Samples were incubated at 50°C for 2 h at 850 rpm, and the remaining steps were performed according to the manufacturer. Amplification of the V1 and V2 hypervariable regions of the bacterial 16S rRNA gene, sequencing and data processing were subsequently performed as described previously.31 Briefly, samples were processed on a 454-FLX sequencer, and sequencing output was filtered according to quality (average >25) and length (minimum length 200, maximum length 400). Chimeras were removed and bacterial community analyses were carried out on classified sequences (ribosomal database project (RDP)) and operational taxonomic units (OTUs) at the species-level threshold (97% similarity) using the ‘Vegan’ R package (R Development Core Team 2011). Bray–Curtis dissimilarity, Jaccard dissimilarity and the unweighted and weighted Unifrac indices32 were analysed using constrained analysis of principal coordinates (‘capscale’ function in Vegan for plotting and ‘anova.cca’ for significance test). Further, the significance of clustering for β diversity indices was assessed using analysis of dissimilarity (‘adonis’) implemented in Vegan. To quantify the presence of segmented filamentous bacteria (SFB), a gene-centric strategy is required as ‘SFB’ is not taxonomical but rather a phenotypical term based on the published genome data.33 ,34 Based on a detailed BLAST analysis of bacterial genomes, FliC2 and FliC3 genes are specific and confined to SFB.33 ,35 Therefore, qPCR was performed with FliC2 and FliC3-specific primers (reported in Prakash et al,34 see online supplementary table S1) and normalised to total 16S rDNA (UniF340 and UniR514 primers).

Statistical analysis

Data are presented as mean±SEM. Statistical analysis was performed using Microsoft Excel (Microsoft), GraphPad Prism (Graph Pad Software) and SPSS Statistics. Statistical significance was calculated using a two-tailed Student t test or Mann–Whitney U test. A p value <0.05 was considered significant. Where more than two groups were compared, one-way analysis of variance (ANOVA) with Bonferroni post-hoc testing or Kruskal–Wallis test was performed.

Results

IEC-specific Ifnar1-deficiency results in increased Paneth and goblet cell numbers

To examine the role of type I IFN signalling in murine IECs, we crossed Ifnar1fl/fl mice25 ,26 with Villin-Cre transgenic mice to generate Ifnar1fl/fl;Vil-Cre mice (‘Ifnar1−/−(IEC)), where the Villin promoter directs constitutive Cre expression and consequent Ifnar1 deletion specifically to the small and large intestinal epithelium.23 ,27 As predicted from conventional Ifnar1−/− mice,4 Ifnar1−/−(IEC) mice were viable and their breeding followed a Mendelian ratio (data not shown). Quantitative RT-PCR confirmed efficient deletion of the floxed Ifnar1 allele in ileal and colonic epithelium (see online supplementary figure S1A,B).

On H&E staining, small intestine and colon of Ifnar1−/−(IEC) mice were histologically indistinguishable from those of Ifnar1+/+(IEC) littermates (see online supplementary figure S1C,D). In order to characterise consequences of Ifnar1 deletion on the different IEC types, we performed lysozyme (Lys)-IHC staining (figure 1A,B) and lysozyme immunofluorescence (figure 1C,D) and noted that the number of Paneth cells was significantly increased by 22% and 25%, respectively. This was associated with markedly increased mRNA expression of lysozyme in small intestinal crypts (figure 1E) of Ifnar1−/−(IEC) compared with Ifnar1+/+(IEC) mice. The number of Paneth cell granules per cell (see online supplementary figure S1E) and their size distribution within Paneth cells according to Cadwell et al22 (see online supplementary figure S1F) remained unaltered between Ifnar1+/+(IEC) and Ifnar1−/−(IEC) mice. Interestingly, amplification with pan-α-defensin primers, or with primers specific for α-defensin Defcr5, did not reveal differences dependent on genotype (see online supplementary figure S1G,H). In colonic crypts, the expression of the antibactericidal cathelicidin antimicrobial peptide (Camp) exhibited no significant difference between Ifnar1+/+(IEC) and Ifnar1−/−(IEC) mice (see online supplementary figure S1I). Along with Paneth cell expansion, PAS staining also revealed an increase in the number of goblet cells in the small but not the large intestine in Ifnar1−/−(IEC) compared with Ifnar1+/+(IEC) mice (see online supplementary figure 1F–H). Consistent with this, expression of Muc2, the major mucin in goblet cells, exhibited a trend towards increased in Ifnar1−/−(IEC) ileal, but not colonic scrapings (see online supplementary figure S1J,K). Finally, since IECs serve as central organisers of mucosal immune mechanisms,3 we surveyed the intraepithelial and lamina propria lymphocyte compartments for alterations due to epithelial Ifnar1 deficiency, which revealed no significant differences in the number of CD4+ and CD8+ T cells, B cells and NK cells between Ifnar1−/−(IEC) and Ifnar1+/+(IEC) mice (see online supplementary table S2 and figure S1L).

Figure 1

Intestinal epithelial IFNAR1 regulates Paneth and goblet cell numbers. (A, B) Determination of Paneth cell number per crypt in Ifnar1−/−(IEC) and Ifnar1+/+(IEC) mice by lysozyme (Lys)-IHC staining (n=6–8; scale bar, 20 µm). (C) Quantification of lysozyme+ Paneth cells based on lysozyme-immunofluorescence staining (n=6). (D) Representative overview (upper six panels; scale bar, 100 µm) and detail pictures (lower six panels; scale bar, 20 µm) of lysozyme-immunofluorescence staining (green) of Ifnar1+/+(IEC) and Ifnar1−/−(IEC) intestinal crypts. Counterstaining was performed with 4′,6-diamidino-2-phenylindole (blue). (E) Expression of lysozyme in small intestinal crypts of Ifnar1+/+(IEC) and Ifnar1−/−(IEC) mice (n=3–4). (F,H) Scoring of goblet cells on Periodic acid Schiff (PAS)-stained sections of ileum and colon (ileum and colon, n=5–6); scale bar, 100 µm. Mean±SEM, Student t test.

Ifnar1−/−(IEC) mice exhibit increased epithelial turnover, regulated by a transmissible component

Alteration in turnover of IECs is a sensitive marker for direct or indirect alteration in IEC function.3 We therefore determined migration of bromodeoxy-uridine (BrdU)-labelled IECs after a 24 h pulse of BrdU. As demonstrated in figure 2A,B, we observed a trend towards increased turnover of IECs in the ileum and colon of mice with either one or two Ifnar1 alleles deleted in their IECs. Within the Ifnar1−/−(IEC) genotype, we noted on post-hoc analysis that the number of BrdU+ cells per crypt appeared to segregate according to whether mice were housed together with, or separately from, their Ifnar1+/+(IEC) littermates. This prompted us to test in an independent experiment the hypothesis that a transmissible component may affect IEC turnover initiated by IEC-specific Ifnar1 deficiency. As demonstrated in figure 2C–F, littermate Ifnar1−/−(IEC) and Ifnar1+/+(IEC) mice that were separated immediately upon weaning exhibited profound differences in IEC turnover, with Ifnar1−/−(IEC) mice displaying a 77% and 48% increase in BrdU+ cells per crypt-villus axis in ileum and colon, respectively. In contrast, upon co-housing of Ifnar1−/−(IEC) and Ifnar1+/+(IEC) mice, increased IEC turnover was abolished to levels observed in Ifnar1+/+(IEC) mice (figure 2C–F). This dependency of hyperproliferation on a transmissible component was unexpected; type I and III IFNs are well known to exert inhibitory effects on cellular proliferation, primarily via activation of STAT1, in other cell types.14 ,36 Immunoblot analysis of total lysates from IEC scrapings revealed a slight increase, rather than decrease, in total and phosphorylated STAT1 in Ifnar1−/−(IEC) compared with Ifnar1+/+(IEC) mice (see online supplementary figure S2A,B, E), while STAT3, which is critically involved in the regulation of intestinal proliferation,37 ,38 was not affected (see online supplementary figure S2C–E).

Figure 2

The antiproliferative function of IFNAR1 in large and small intestine is affected by housing condition. (A,B) Quantification of BrdU+ cells of ileal (A) and colonic epithelia (B) in Ifnar1+/+(IEC), Ifnar1+/−(IEC) and Ifnar1−/−(IEC) mice (ileum, n=4; colon, n=4–6; mean±SEM; one-way analysis of variance (ANOVA), Bonferroni post-hoc test). (C–F) Representative images of BrdU stainings of ileal (C) and colonic epithelia (E) from separately housed Ifnar1+/+(IEC) and Ifnar1−/−(IEC) mice after 24 h of BrdU incorporation (scale bar, 100 µm) and quantitative analysis of proliferation (BrdU) on ileal and colonic sections of mice housed separately or co-housed according to genotype (D and F, separately housed mice, n=13–15; co-housed animals, n=5–6; mean±SEM; one-way ANOVA, Bonferroni post-hoc test). (G, H) Representative images of an in situ hybridisation for the stem cell marker Olfm4 (G) on ileal sections of separately housed animals and their quantitative analysis (H, n=4–5; mean±SEM; one-way ANOVA, Bonferroni post-hoc test); scale bar, 50 µm.

Interestingly, it was only hyperproliferation of the intestinal epithelium that was due to deficiency of Ifnar1 in the intestinal epithelium and that required a transmissible component. No difference was noted in the rate of intestinal cell death (see online supplementary figure S2F–I) or in the number of intestinal stem cell cells1 ,39 (figure 2G,H). We therefore asked whether the increase in lysozyme+ Paneth cells per crypt in Ifnar1−/−(IEC) compared with Ifnar1+/+(IEC) mice (figure 1C) was retained when both genotypes were co-housed, and this was indeed the case (see online supplementary figures S2J).

These data demonstrate that increased IEC turnover in Ifnar1−/−(IEC) mice is affected by a transmissible component and suggested a role of Paneth cells.

Epithelial Ifnar1 determines structure of the caecal microbial ecosystem

Paneth cell function has a significant effect on the composition of the intestinal microbiota,21 ,40 and microbial components have previously been shown to affect epithelial proliferation.41 ,42 Mice are coprophagic and therefore horizontal transmission of microbial ecosystems can easily occur, which has been shown to be at the basis of transfer of disease via a microbiota rendered pathogenic by host genetic defects.28 ,43 ,44 However, it has not previously been demonstrated that a ‘normal’ microbiota might be capable of abrogating a phenotype caused by a host genetic defect. Since the number of Paneth cells was increased in Ifnar1−/−(IEC) compared with Ifnar1+/+(IEC) mice irrespective of housing conditions (figure 1C, see online supplementary figure S2J), but hyperproliferation in Ifnar1−/−(IEC) mice abrogated upon co-housing with Ifnar1+/+(IEC) mice (figure 2D,F), we studied the impact of Ifnar1 genotype and housing conditions on intestinal microbial ecology via 454 pyrosequencing of the V1–V2 region of the bacterial 16S rRNA gene in caecal contents of 10–12-week-old separately or co-housed Ifnar1−/−(IEC) and littermate Ifnar1+/+(IEC) mice.

We mainly used two β-diversity measures, that is, differences between samples, to compare bacterial communities: Bray–Curtis dissimilarity based on species-level OTUs, which quantifies abundance of taxa shared between samples versus uniqueness to one sample31; and the weighted Unifrac index, which is a quantitative measure of phylogenetic distinctiveness of bacterial communities.32 Analysis of dissimilarity (adonis) was first used to test whether genotype or housing conditions influence bacterial communities. A significant contribution of genotype (r2=0.1397, p=0.034) and housing condition (r2=0.1648, p=0.007, figure 3A) was revealed for Bray–Curtis, whereas only the combination of both factors revealed a significant separation based on weighted Unifrac (r2=0.5207, p=0.005, figure 3B).

Figure 3

Deletion of Ifnar1 in intestinal epithelial cells results in altered bacterial communities in the caecum. (A–D) Constrained analysis of principal coordinate plot of the abundance of bacterial communities in caecal faeces of Ifnar1+/+(IEC) and Ifnar1−/−(IEC) mice based on Bray–Curtis dissimilarities (A) and weighted Unifrac index (B), with genotype (Bray–Curtis, p=0.034; weighted Unifrac, p>0.05) and housing condition (Bray–Curtis, p=0.007; weighted Unifrac, p>0.05) as constraining factors. Circles depict the 95% confidence interval (CI) of the communities of separately housed (SH, n=5; Bray–Curtis, p=0.047; weighted Unifrac, p=0.044) and co-housed mice (CH, n=3; Bray–Curtis, p=0.112; weighted Unifrac, p=0.103). (C, D) Constrained analysis of principal coordinate plot of Jaccard dissimilarity (C) and the unweighted Unifrac index (D), with genotype (Jaccard, p=0.039; unweighted Unifrac, p>0.05) and housing condition (Jaccard, p=0.011, unweighted Unifrac, p=0.001) as constraining factors. Circles depict the 95% CI of the communities. (E,F) Quantification of bacterial phyla (E) and genera (F) with respect to genotype (Ifnar1+/+(IEC) vs Ifnar1−/−(IEC)) and their housing condition (SH, separately housed vs CH, co-housed). The average value of each taxon is displayed (separately housed mice, n=5; co-housed mice, n=3).

We further considered other β-diversity measures, which correspond to Bray–Curtis and weighted Unifrac, but emphasise the presence/absence of major taxa/lineages, namely, the Jaccard index and unweighted Unifrac index. Jaccard displayed high correspondence to Bray–Curtis (Mantel test, r2=0.9938, p=0.001) and consequently similar adonis results (figure 3C). Unweighted Unifrac, on the other hand, showed significant separation according to housing condition (r2=0.1673, p=0.001) and the combination of genotype and housing condition (r2=0.3391, p=0.001, figure 3D).

In summary, these analyses reveal that housing status and genotype profoundly affect microbial communities. Under separate housing conditions, Ifnar1+/+(IEC) and Ifnar1−/−(IEC) mice showed significant differences (adonis; Bray–Curtis r2=0.4012, p=0.047; weighted Unifrac adonis r2=0.4113, p=0.044), which were lost under co-housing conditions (adonis; Bray–Curtis r2=0.3291, p=0.112; weighted Unifrac r2=0.4456, p=0.103).

At the phylum level, Bacteroidetes were most abundant (average 72±19%), followed by Firmicutes (27±19%), Proteobacteria and Actinobacteria, accounting for 0.85±0.48% and 0.08±0.12% of all reads, respectively (figure 3E). At the genus level, seven classified genera accounted for >1% of total reads, with Barnesiella (43±15%) being most abundant, followed by Robinsoniella (19±21%), Anaerophaga (8±4.6%), Alistipes (4±3.52%), Prevotella (2±0.12%) and Paraprevotella (1±0.29%, figure 3F). Differences in bacterial communities were observed dependent on genotype and housing conditions, as shown in figure 3E. Bacteroidetes were highest in separately housed Ifnar1−/−(IEC) mice (88±11%), and lowest in separately housed Ifnar1+/+(IEC) mice (47±33%), while the co-housed mice of both genotypes exhibited intermediate relative abundances (ANOVA, p=0.036). Firmicutes, in turn, followed an opposing pattern and were highest in separately housed Ifnar1+/+(IEC) mice (53±33%), and lowest in separately housed Ifnar1−/−(IEC) mice (11±10%), while co-housed mice of both genotypes exhibited intermediate relative abundances (ANOVA, p=0.034). Housing condition or genotype alone does not show significant differences in major phyla abundances (ANOVA, p>0.05).

Barnesiella, a representative genus of Bacteroidetes, and Robinsoniella, a representative genus of Firmicutes, followed the pattern described above (figure 3F). Differences in abundance with regard to genotype and housing were significant for Robinsoniella (ANOVA, p=0.0124). A further genus that exhibited a significant difference in abundance across genotype and housing condition is Anaerophaga (ANOVA, p=0.048), which was higher in co-housed mice (average 12±6.6%) compared with mice that were separately housed (average 4.7±2.7%). Of note, SFB,45 which profoundly affect mucosal homeostasis, were not significantly affected by genotype or housing condition (see online supplementary figure S3A).46 ,47

Epithelial Ifnar1 deficiency does not affect severity of DSS-induced colitis

Germ-line Ifnar1−/− mice have previously been reported to exhibit increased severity of DSS colitis, and a barrier-protective role of type I IFN stimulation of IECs has been proposed as an underlying mechanism.17 Further, absence of NOD2 or NLRP6 had previously been reported to render the microbiota colitogenic in conjunction with DSS administration,43 ,44 highlighting the major contribution of dysbiosis to exogenously induced intestinal inflammation. We therefore tested the hypothesis that epithelial type I IFN signalling might affect intestinal inflammation induced by DSS in littermate Ifnar1−/−(IEC) that were separately housed from Ifnar1+/+(IEC) mice to maximise the contrast in microbial composition in-between these two genotypes. Although Ifnar1−/−(IEC) mice exhibited increased weight loss at days 4 and 5 of DSS colitis compared with Ifnar1+/+(IEC) mice (figure 4A), no difference in histological severity between genotypes was noted in the proximal (figure 4B–D) or distal colon (figure 4E–G) at the end of the experiment at day 7. This was also reflected by similar colon length (figure 4H) and haematocrit value (figure 4I) at the end of the experiment in either genotype. Accordingly, mRNA expression patterns of inflammatory cytokines in Ifnar1−/−(IEC) and Ifnar1+/+(IEC) mice were also similar (figure 4J). To specifically address a potential role of IFNAR1 in the early phase of colitis as indicated by the significant difference in weight loss at days 4 and 5, we investigated histological changes and gene expression at day 5 in an independent experiment. However, Ifnar1−/−(IEC) and Ifnar1+/+(IEC) mice showed similar weight loss (see online supplementary figure S4A,B), similar colon length and weight (see online supplementary figure S4C,D) and no difference in quantitative histological analysis of proximal (see online supplementary figure S4E–G) and distal colon (see online supplementary figure S4H–J) at this early time point. Accordingly, protein secretion (see online supplementary figure S4K) and mRNA expression (see online supplementary figure S4L) of inflammatory cytokines was also similar between both genotypes.

Figure 4

Ifnar1 deletion in the intestinal epithelium does not influence severity of dextran sodium sulfate (DSS)-induced colitis. (A) Acute colitis in mice held separately according to genotype was induced by 4.5% DSS in the drinking water for 7 days. Weight curve of Ifnar1+/+(IEC) and Ifnar1−/−(IEC) mice displaying the course of colitis induction (n=11, mean±SEM, Student t test). (B–G) Severity of colitis was determined by a histological score of H&E sections of the proximal (B–D) and distal (E–G) colon as described previously28 (n=11; mean±SEM, Mann–Whitney U test). (H,I) Determination of colon length (H) and blood haematocrit levels (I) of Ifnar1+/+(IEC) and Ifnar1−/−(IEC) mice (n=10–11; mean±SEM, Student t test). (J) Expression of pro-inflammatory and anti-inflammatory cytokine mRNAs in Ifnar1+/+(IEC) and Ifnar1−/−(IEC) mice (n=9; mean±SEM, Student t test); scale bar, 100 µm.

In summary, these experiments indicate that neither IEC-specific Ifnar1 deletion nor the accompanying profound changes in microbial ecology result in increased susceptibility to DSS colitis.

Epithelial Ifnar1 deletion causes increased colitis-associated tumourigenesis

Intestinal hyperproliferation as a response to intestinal injury can associate with subsequent tumourigenesis,37 ,48 ,49 which is thought to contribute to the increased propensity to develop tumours over time in patients with IBD.50 We therefore wondered whether increased IEC turnover observed in Ifnar1−/−(IEC) compared with Ifnar1+/+(IEC) mice, when housed separately, may associate with an increased propensity to develop colitis-associated tumours. Colitis-associated tumourigenesis was induced by the administration of the mutagen AOM, followed by three cycles of DSS (figure 5A, see online supplementary figure S5A,B).29 ,30 As depicted in figure 5B–D, Ifnar1−/−(IEC) compared with Ifnar1+/+(IEC) mice developed a more than twofold larger tumour burden as assessed by tumour area and number of lesions. Interestingly, STAT1 and STAT3 phosphorylation remained unchanged between separately housed Ifnar1+/+(IEC) and Ifnar1−/−(IEC) mice (see online supplementary figure S5C,D). Remarkably, only mice that were separated according to genotype immediately after weaning exhibited a genotype-dependent difference in tumour burden, whereas housing littermates of opposing genotypes together after weaning abrogated genotype-specific differences (figure 5C,D). These results suggest that epithelial Ifnar1-deficiency modifies the intestinal microbial habitat in a way that is transmissible and permissive to the development of colitis-associated tumours.

Figure 5

Intestinal epithelial cell-specific Ifnar1 genotype and housing conditions determine propensity for experimental tumour formation. (A) Colitis-associated tumours were induced by administration of the carcinogen azoxymethane (AOM) followed by three cycles of dextran sodium sulfate. (B) The upper images depict H&E-stained paraffin sections of ‘Swiss-rolls’ of the colon. High-power magnifications show representative regions of lesions in colon of separately housed Ifnar1+/+(IEC) and Ifnar1/(IEC) mice (scale bar, 200 µm). (C, D) The number of lesions (C) as well as tumour area (D) was determined in separately housed and co-housed Ifnar1+/+(IEC) and Ifnar1/(IEC) mice (separately housed mice, n=6–8; co-housed animals, n=5; mean±SEM, Student t test).

Discussion

Here we report that the type I IFN receptor on the intestinal epithelium plays an important role in intestinal epithelial proliferation via a bidirectional host–microbiota relationship. Despite expecting that loss of the antiproliferative effect due to genetic deletion of the type I IFN receptor might be ‘hard-wired’ on the host side, we unexpectedly discovered that a ‘normal’ microbiota present in wild-type mice can abrogate hyperproliferation caused by Ifnar1 deletion. We show that deletion of Ifnar1 in IECs results in profound alterations in intestinal microbial ecology compared with Ifnar1-sufficient littermates. We suggest that the latter is caused by an increase in Paneth cell numbers and lysozyme expression observed in Ifnar1−/−(IEC) compared with Ifnar1+/+(IEC) mice. Paneth cells secrete a variety of antimicrobial peptides, such as lysozyme and α-defensins, and have previously been shown to profoundly regulate the intestinal microbial habitat.21 ,40 ,51

Increased proliferation and turnover of IECs can have a variety of underlying causes,14 but commonly represents a uniform reaction to epithelial injury that, in case it is sustained, lays the ground for an increased propensity to tumour formation.48 ,49 Indeed, we observed increased tumour burden in Ifnar1−/−(IEC) compared with Ifnar1+/+(IEC) mice in the AOM/DSS model, a well-established murine model of colitis-associated cancer.29 ,30 Tumour formation in the AOM/DSS model has previously been shown to be profoundly affected by transmissible microbial communities that arise from host genetic defects, such as demonstrated for NOD2 and NLRP6,43 ,44 ,52 which has important implications for our understanding of cancers arising in long-standing IBD.50 Of note is that in the case of Ifnar1−/−(IEC) mice, this increased tumour burden compared with Ifnar1+/+(IEC) mice cannot be attributed to increased severity of DSS-induced colitis, which drives tumour propagation in this model43 ,52 as the severity of DSS colitis was indistinguishable between Ifnar1−/−(IEC) and Ifnar1+/+(IEC) mice.

Intriguingly, the marked differences in intestinal epithelial proliferation and colitis-associated tumour formation between Ifnar1−/−(IEC) and Ifnar1+/+(IEC) mice were completely abrogated when littermates of both genotypes were co-housed. Our breeding regimen involved crossing of Vil-Cre;Ifnar1fl/fl [‘Ifnar1−/−(IEC)’] with Ifnar1fl/fl [‘Ifnar1+/+(IEC)’] mice, which generated 50% Ifnar1−/−(IEC) and 50% Ifnar1+/+(IEC) offspring. The direct comparison of microbial communities in 10–12-week-old littermates that had either been separated according to genotype or co-housed reflects both ends of the spectrum, namely, (1) the impact that host genetics has on the composition of microbial communities, and (2) the partial (presumably coprophagic) adaptation of the microbiota despite genotype-specific differences. Both the accentuation of genotype-dependent differences and their diminution, conditional on housing arrangements, are important factors to be taken into account in the design of experimental protocols. These effects, along with the inherent differences in microbial communities in animal facilities due to differences in housing and dietary conditions, could potentially contribute to the observation of seemingly contrasting results in host–microbiota relationships reported from different laboratories.53–56 Our results imply that transmissible microbial factors that impact on the phenotype might not only arise from genetically affected hosts, but conversely, transmitting a ‘healthy’ microbiota may prevent host-genotype-induced pathology as demonstrated for IEC hyperproliferation. Importantly, these experiments lend further experimental support for the intricate gene–microbiota relationship that determines host physiology.57

Regulation of proliferation of IECs is complex and is primarily determined at two specific locales within the crypt-villus axis: at the crypt bottom, where intestinal stem cells are intermingled with Paneth cells, and the transit amplifying zone.1 ,51 Numerous factors affect the proliferative output at both locales under homeostatic conditions, including nutrient exposure,58 circadian rhythm59 and the presence or absence of a commensal flora.42 Notably, metabolites derived from specific bacterial species within the intestinal microbiota can profoundly impact on the rate of IEC proliferation, as for example, demonstrated by lactate derived from Lactobacillus murinus in starvation-refed mice.42 In this model, the concomitant administration of a mutagen resulted in increased occurrence of aberrant crypt foci upon re-feeding.42 We therefore speculate that IEC-specific Ifnar1 deletion might propagate specific bacterial species or metabolic pathways within the microbiota that promote intestinal hyperproliferation that manifests in the context of microbiota–gene interaction.

Although the induction of an antiviral immune response is commonly considered the most important function of type I IFNs, bacteria and their ligands similarly induce type I IFNs, predominantly via TLR-dependent, but also NOD1/2-dependent pathways.9 ,12 Type I IFNs are indeed constitutively expressed at low level in various organs, including the small intestinal mucosa,7 ,60 and, for example, Paneth cells have been reported to express Ifna1 transcripts under baseline conditions.19 Low-level type I IFN secretion might account for constitutive engagement of the type I IFN receptor on the intestinal epithelium,13 resulting in the IFNAR1-dependent biological function in IECs as revealed through genetic deletion of Ifnar1 as reported herein.

The mechanism whereby Ifnar1 deletion in IECs results in increased numbers of lysozyme+ Paneth and PAS+ goblet cells remains to be determined. Interestingly, 6 h after intraperitoneal administration of 70 IU/g human IFNα2b, Paneth and goblet cells of Wistar rats exhibited evidence of depletion of their apical granules with features suggesting exocytic activity as observed through transmission electron microscopy.61 Although it is unclear whether this pharmacologically induced secretion is directly or indirectly mediated, we speculate that low-level signalling through the type I IFN receptor might contribute to the tonic, constitutive secretion from Paneth and goblet cells.51 ,62

In summary, we report that the type I IFN receptor, encoded by a predicted risk gene at a CD susceptibility locus,15 functions on IECs to restrain Paneth and goblet cells and to shape the microbial composition, with implications on epithelial regeneration and colitis-associated tumour formation when impaired. As such, IFNAR1 joins several other genetic risk factors of IBD whose products profoundly affect the intestinal microbiota and/or Paneth cell function.21–24 ,35 ,44

Acknowledgments

The authors thank Dr Sieghart Sopper and Dr Tim Raine for assistance and advice with FACS. CF would like to dedicate this work in memory of Dr C.-B. Chien.

References

View Abstract
  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:

Footnotes

  • Contributors MT, CF and AK conceived and designed the experiments. MT, JW, CF, TMJF, LN, TEA, ES and SK performed the experiments. MT, JW, CF, HT and AK analysed the data. MT and AK wrote the paper with input from all other authors.

  • Funding This work was supported by the Austrian Science Fund project grant P21530-B18, with additional support from the European Research Council under the European Community’s Seventh Framework Programme (FP7/2007-2013)/ERC Grant agreement no. 260961 (AK); the National Institute for Health Research Cambridge Biomedical Research Centre (AK); the Addenbrooke's Charitable Trust (AK); fellowships from the European Crohn's and Colitis Organization (MT and TEA); and the Crohn's in Childhood Research Association (LN) and German Research Foundation (DFG) Excellence Cluster ‘Inflammation at Interfaces’ (JFB).

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.