Background The extent to which numerous strains of genetically engineered mice, including mice lacking Toll-like receptor 5 (T5KO), display colitis is environment dependent. Gut microbiota underlie much of the variation in phenotype. Accordingly, embryonic rederivation of T5KO mice ameliorated their spontaneous colitis despite only partially correcting elevated proinflammatory gene expression. It was postulated that endogenous anti-inflammatory pathways mediated the absence of overt inflammation in these mice when their gut microbiota were reset. Consequently, it was hypothesised that neutralisation of the anti-inflammatory cytokine interleukin 10 (IL-10) might induce uniform colitis in T5KO mice, and thus provide a practical means to study mechanisms underlying their inflammation.
Methods Two distinct strains of non-colitic T5KO mice, as well as mice lacking MyD88, Toll-like receptor 4 (TLR4), IL-1 receptor (IL-1R) and various double knockouts (DKOs) were treated weekly for 4 weeks with 1 mg/mouse of IL-10 receptor neutralising antibody (IL-10R mAb) and colitis assayed 1 week later. The composition of the caecal microbiota was determined by 454 pyrosequencing of 16S rRNA genes.
Results Anti-IL-10R mAb treatment led to severe uniform intestinal inflammation in both strains of T5KO mice. Such neutralisation of IL-10 signalling did not cause colitis in wild-type littermates nor mice lacking TLR4, MyD88 or IL-1R. The susceptibility of T5KO mice to this colitis model was not rescued by absence of TLR4 in that T4/T5 DKO mice displayed severe colitis in response to anti-IL-10R mAb treatment. IL-1β signalling was crucial for this colitis model in that IL-1R/T5 DKOs were completely protected from colitis in response to IL-10R mAb treatment. Lastly, it was observed that blockade of IL-10R function was associated with changes in the composition of gut microbiota, which were observed in mice that were susceptible and resistant to IL-10R mAb-induced colitis.
Conclusion Regardless of whether they harbour a colitogenic microbiota, loss of TLR5 predisposes mice to colitis triggered by immune dysregulation via an IL-1β-dependent pathway.
- IL-10 receptor neutralisation
- gut microbial diversity
- 16S rRNA gene pyrosequencing
- inflammatory bowel disease
- mucosal immunology
Statistics from Altmetric.com
- IL-10 receptor neutralisation
- gut microbial diversity
- 16S rRNA gene pyrosequencing
- inflammatory bowel disease
- mucosal immunology
Significance of this study
What is already known about this subject?
Mutations in innate immunity can increase the risk of developing inflammatory bowel disease (IBD) but are typically not sufficient to drive the disorder.
Some mice lacking Toll-like receptor 5 (referred to as T5KO mice) develop spontaneous colitis while the majority are free of this disorder.
What are the new findings?
T5KO mice that lack colitis displayed altered gene expression that predisposed them to developing colitis upon blockade of signalling by the endogenous anti-inflammatory cytokine interleukin 10 (IL-10).
Colitis that occurred in T5KO mice upon blockade of IL-10 signalling was dependent upon the inflammasome cytokine IL-1β.
Microbiota composition was altered by blockade of IL-10 signalling in wild-type and T5KO mice even though wild-type mice did not exhibit overt colitis.
How might it impact on clinical practice in the foreseeable future?
It might be advisable to monitor patients carrying IBD risk alleles for early markers of inflammation. Innate immune mutations that result in colitis may be treatable by blockade of IL-1β signalling.
Inflammatory bowel disease (IBD) is the collective term for the chronic idiopathic inflammatory diseases of the intestine including Crohn's disease and ulcerative colitis. Development of IBD results from a complex interplay of host genetics, environment and lifestyle.1 Consequently, most persons carrying IBD genetic risk alleles do not ever develop this disease. Accordingly, the majority (73%) of persons with an identical twin with IBD, who can thus be viewed as genetically capable of developing IBD, do not themselves develop IBD despite showing some of the physiological abnormalities associated with the disease such as increased gut permeability.2 While non-genetic risk factors are probably complex, one likely important environmental determinant is the intestinal microbiota, which is the collective term for the 1014 bacteria that normally inhabit the intestinal tract.3–5 The mechanisms by which microbiota may confer susceptibility to IBD remain unclear, but include the possibilities that patients with IBD carry (or previously carried) an opportunistic pathogen and/or that lack of potentially pathogenic organisms results in a dysregulated immune system causing the host to attack its benign microbiota.6
Analogous to the case for humans, development of chronic intestinal inflammation in laboratory animals is determined by both genetics and environment. Accordingly, some strains of mice (naturally occurring and genetically engineered) exhibit robust colitis in some vivariums but appear to have little to no disease in others. IL10-deficient mice are perhaps the best-studied example of this phenomenon. The composition of the gut microbiota is a key determinant of the variation in experimental colitis. In accordance with this, embryo transplants that have the effect of providing to pups the microbiota of specific dams have been shown to markedly reduce the extent of colitis in several strains/colonies of colitic interleukin 10 (IL-10)-deficient mice.7 8 Select species of Helicobacter, namely Helicobacter bilis and Helicobacter hepaticus, have been shown to affect disease susceptibility, as these bacteria have been associated with colitis in some strains of genetically engineered mice. Furthermore, specific transfer of such Helicobacter species can induce colitis in some strains of engineered mice but does not cause disease in wild-type (WT) mice.7 However, while IL-10 knockout (KO) mice maintained in some vivariums do not develop spontaneous colitis, they are still prone to developing colitis in response to non-bacterial triggers; for instance they develop robust chronic colitis when exposed to the cyclooxygenase inhibitor piroxicam, which causes only mild transient colitis in WT mice.9 10
We previously reported that in a colony of mice lacking the flagellin receptor Toll-like receptor 5 (TLR5), a subset of mice developed spontaneous colitis.11 In contrast to some strains of mice that develop colitis due to reductions in anti-inflammatory cytokines and/or regulatory T cells, such TLR5 KO (T5KO) mice exhibited elevated levels of the anti-inflammatory cytokine IL-10. This elevation was greatest in mice with severe colitis, suggesting that their anti-inflammatory signalling pathways were overwhelmed rather than dysfunctional. In accordance with the notion that murine colitis is environment dependent, colitis was not reported in a colony of this strain of T5KO mice maintained elsewhere, although this colony exhibited greater susceptibility to dextran sodium sulfate (DSS) colitis.12 The basal phenotype differences were probably related to their gut microbiota, as exchange of the microbiota of our colony of T5KO mice via embryo transplant into mice purchased from Jackson Laboratories (Bar Harbor, Maine, USA) eliminated histopathological evidence of colitis. However, these rederived mice still displayed elevated levels of proinflammatory cytokines that were associated with development of metabolic syndrome.13 We hypothesised that the reason such elevated levels of proinflammatory cytokines in rederived T5KO mice did not result in colitis was that in the absence of a colitogenic microbiota, anti-inflammatory signalling was sufficient to prevent development of colitis. Here we show that, in agreement with our hypothesis, neutralisation of a key anti-inflammatory cytokine, IL-10, resulted in severe uniform colitis in two distinctly generated strains of T5KO mice. We report a series of experiments that elucidate the mechanisms that make these mice more prone to developing colitis upon immune dysregulation. Such experiments reveal a key role for the inflammasome cytokine IL-1β in mediating the susceptibility of T5KO mice to colitis.
Six- to eight-week-old WT, T5KO (from Drs Akira and Flavell), T4KO, IL-1 receptor (IL-1R) KO, T4/T5 double knockout (DKO), T5/IL-1R DKO and MyD88KO mice under the C57BL/6 background (back-crossed to C57BL/6 mice for 10 generations) were bred and maintained in the Emory University animal facility. All experiments involving animals were approved by the Emory University's animal use committee.
Male T5KO mice were bred with WT female C57BL/6J mice purchased from Jackson Laboratory. The embryos were transferred to surrogate C57BL/6J mice to generate TLR5 heterozygotes. These mice were bred to obtain WT and T5KO mice with gut microbiota identical (or similar) to C57BL/6J mice from Jackson Laboratories.13
Anakinra (Kineret), an analogue of secreted IL-1R antagonist (sIL-1Ra) currently in use for rheumatoid arthritis, type 2 diabetes and gout, was purchased from Amgen (Thousand Oaks, California, USA).
IL-10R neutralising monoclonal antibody (mAb; 1B1.3a), as well as mAb to PD-1 and PDL1, were purchased from BioXcell (West Lebanon, New Hampshire, USA).
Total RNA was isolated from colonic tissues using TRIzol (Invitrogen, Carlsbad, CA, USA) and purified using the RNeasy Plus Mini kit (Qiagen, Valencia, California, USA) according to the manufacturer's instructions.14
Microarray analyses were performed at the Emory Biomarker Microarray Core. Briefly, mRNA samples were reverse-transcribed, amplified, labelled and used to probe MouseWG-6 v2 chips purchased from Affymetrix (Santa Clara, California, USA). Samples were assayed using a Molecular Devices Gene Pix (4100A) (Silicon Valley, California, USA), and raw fluorescence readings were processed by an algorithm designed to reduce spurious readouts of gene activation. Microarray data were quantile-normalised using freely available scripts written in R (http://www.R-project.org). Significantly altered genes were identified using significance of analysis of microarray analyses and assessed by hierarchal clustering and principle component analysis using Spotfire Decision Site for Functional Genomics software (TIBCO, Somerville, Massachusetts, USA) to determine relatedness of gene expression patterns resulting from loss of TLR5.
Quantitive reverse transcription–PCR (qRT–PCR)
RNAs were quantified in Realplex4 (Eppendorf, Hauppauge, NY, USA) using a QuantiFast SYBR Green RT-PCR Kit (Qiagen) with specific mouse oligonucleotides. The sense and antisense oligonucleotides used were, respectively: 36B4, 5′-TCCAGGCTTTGGGCATCA-3′ and 5′-CTTTATCAGCTGCACATCACTCAGA-3′; pro-IL-1β, 5′-TTGACGGACCCCAAAAGATG-3′ and 5′-AGAAGGTGCTCATGTCCTCAT-3′; CXCL1, 5′-TCGCGAGGCTTGCCTTGACC-3′ and 5′-AGACGGTGCCATCAGAGCAG-3′; pla2g4c, 5,-acctcgcaggggtctctggg-3′ and 5′-gcctggaaacaattaaataggccc-3′; pnliprp2, 5′-ctgccaccgacccagccacc-3′ and 5′-cgacccgagtgttgtaggaggc-3′; ang4, 5′-tccccagttggaggaaagctgg-3′ and 5′-tggcatcatagtgctgacgtagg-3′; and def3, 5′-gactaaaactgaggagcagccagg-3′ and 5′-ctttctgcaggtcccattcatgcg-3′. Each sample was run in duplicate. All results were normalised to the unaffected housekeeping 36B4 gene.
Caecal microbiota analysis by 16S rRNA sequencing
To avoid the confounding effects of co-housing on the diversity of caecal bacteria, we selected mice from multiple litters that were housed separately. Bulk DNA was extracted from frozen extruded caecal contents using a PowerSoil-htp kit from MoBio Laboratories (Carlsbad, California, USA) with mechanical disruption (bead-beating). 16S rRNA genes were amplified from each sample using a composite forward primer and a reverse primer containing a unique 12-base barcode, designed using the Golay error-correcting scheme, which was used to tag PCR products from respective samples.15 We used the forward primer 5′-GCCTTGCCAGCCCGCTCAGTCAGAGTTTGATCCTGGCTCAG-3′: the italicised sequence is 454 Life Sciences primer B, and the bold sequence is the broadly conserved bacterial primer 27F. The reverse primer used was 5′-GCCTCCCTCGCGCCATCAGNNNNNNNNNNNNCATGCTGCCTCCCGTAGGAGT-3′: the italicised sequence is 454 Life Sciences primer A, and the bold sequence is the broad-range bacterial primer 338R. NNNNNNNNNNNN designates the unique 12-base barcode used to tag each PCR product, with 'CA' inserted as a linker between the barcode and rRNA primer. PCRs consisted of HotMaster PCR mix (Eppendorf), 200 μM of each primer and 10–100 ng of template, and reaction conditions were 2 min at 95°C, followed by 30 cycles of 20 s at 95°C, 20 s at 52°C and 60 s at 65°C on an Eppendorf thermocycler. Three independent PCRs were performed for each sample, combined and purified with Ampure magnetic purification beads (Agencourt, Madison, WI, USA), and products visualised by gel electrophoresis. No-template extraction controls were analysed for lack of visible PCR products. Products were quantified using the Quant-iT PicoGreen double-stranded DNA (dsDNA) assay as described by the manufacturer. A master DNA pool was generated from the purified products in equimolar ratios to a final concentration of 21.5 ng/ml. The pooled products were sequenced using a Roche 454 Titanium pyrosequencer at the University of South Carolina (EnGenCore, Columbia, SC, USA).
16S rRNA gene sequence analysis
Sequences were denoised16 17 and analysed using the open source software package Quantitative Insights Into Microbial Ecology (QIIME18). 16S rRNA gene sequences were assigned to operational taxonomic units (OTUs) using UCLUST19 and a threshold of 97% pairwise identity, then classified taxonomically using Greengenes.20 A single representative sequence for each OTU was aligned using PyNAST,21 then a phylogenetic tree was built using FastTree.22 The phylogenetic tree was used for computing the UniFrac distances between samples.23
Administration of anti-IL-10R mAb and Anakinra
Mice (n=5–6) were treated with 1 mg of IL-10R mAb (intraperitoneally) weekly for 4 weeks as previously described.24 Where indicated, mice (WT and T5KO mice) also received the sIL-1Ra agonist Anakinra, which was administered at doses of 1 or 25 mg/kg. Anakinra was administered via thrice weekly subcutaneous injections or daily intraperitoneal injections. As a control for both treatments, mice that were not receiving mAb or drug were injected with sterile phosphate-buffered saline (PBS). Mouse body weight was measured weekly.
Mice were bled via the retrobulbar intraorbital capillary plexus. Briefly, mice were anaesthetised and bled via the orbital sinus which surrounds the globe. After collection of ∼250 μl, blood was placed in a serum separator tube (Becton-Dickinson, Franklin Lakes, NJ, USA), centrifuged at 8000 rpm for 10 min and the supernatant corresponding to the serum collected.
Following euthanasia, colons (1 cm) were removed, cut open longitudinally, washed in Hank's balanced salt solution (HBSS) and cultured in RPMI 1640 medium containing 1% penicillin and streptomycin.25 After 24 h incubation at 37°C with 5% CO2, the supernatants were centrifuged at 4°C and used for assaying cytokines by ELISA.
Tissue myeloperoxidase (MPO) assay
Neutrophil influx in tissue was accessed by assaying the enzymatic activity of MPO, a marker for neutrophils. Briefly, tissue (50 mg/ml) was thoroughly washed in PBS and homogenised in 0.5% hexadecyltrimethylammonium bromide (Sigma-Aldritch, St Louis, MO, USA) in 50 mM PBS (pH 6.0), freeze-thawed three times, sonicated and centrifuged. MPO was assayed in the clear supernatant by adding 1 mg/ml of dianisidine dihydrochloride (Sigma) and 5×10−4% H2O2, and the change in optical density was measured at 450 nm. Human neutrophil MPO (Sigma) was used as standard. One unit of MPO activity was defined as the amount that degraded 1.0 μmol of peroxide/min at 25°C.26
All ELISA kits are Duoset kits from R&D Systems (Minneapolis, Minnesota, USA) and assays were performed according to the manufacturer protocol.
Following euthanasia, caecum and colon were fixed for 24 h in 10% buffered formalin at room temperature and then subjected to H&E staining on tissue sections of 5 μm thickness.
H&E-stained slides were scored by a pathologist (IN) blinded to the study protocol as previously described.27 Briefly, slides were scored for the presence/absence of active inflammation, the intensity of inflammation (average number of neutrophils and the number of fields that were involved), the extent of inflammation (mucosa, submucosa or serosa), the presence/absence of ulceration, architectural disarray and the pattern of involvement.
Significance was determined using the Mann–Whitney t test or one-way analysis of variance (ANOVA; GraphPad Prism software). For the comparisons of mean phylotype abundances in WT and T5KO mice, significance levels were adjusted for multiple comparisons using Bonferroni's correction (supplementary tables 3 and 4). Differences were noted as significant *p<0.05, **p<0.01, ***p<0.001.
Transcriptome analysis of TLR5-deficient mice
We previously observed that a subset of mice engineered to lack TLR5, referred to as T5KO mice, developed spontaneous colitis, whereas the non-colitic T5KO mice in the colony developed metabolic syndrome.10 11 Analysis of intestinal gene expression in these T5KO mice revealed qualitatively similar patterns of gene expression in colitic and non-colitic T5KO mice, with both groups of mice exhibiting elevated expression of a broad panel of genes related to inflammation/host defence, yet the extent to which the expression of these genes was elevated was significantly greater in colitic T5KO mice.11 Rederivation of T5KO mice by embryo transplantation into mice purchased from Jackson Laboratories markedly attenuated their spontaneous colitis, but such rederived T5KO (Re-T5KO) mice still displayed evidence of low-grade inflammation and metabolic syndrome.13 To better understand the phenotype of these mice in general and to gain insight into why the mice no longer exhibited spontaneous colitis, we performed mRNA microarray analysis to examine intestinal gene expression in Re-T5KO mice relative to their rederived-WT (Re-WT) littermates (the rederivation was performed on heterozygous mice which were then bred to generate Re-WT and Re-T5KO mice). This analysis was performed in biological triplicates on age- and gender-matched mice. Loss of TLR5 resulted in altered expression of hundreds of genes in the colon (figure 1A). Furthermore, loss of TLR5 resulted in significant upregulation of 73 genes and downregulation of 37 other genes in the colon of all Re-T5KO mice assayed (supplementary tables 1 and 2). Although microarray results from distinct studies can be difficult to compare directly, the overall pattern of gene expression observed in these microarray experiments was not dramatically different from that in microarray studies we previously performed on WT and colitic and non-colitic T5KO mice. Specifically, our previous and current studies observed that loss of TLR5 resulted in increased expression of genes broadly related to inflammation and host defence including several genes with direct antibacterial function. The expression of select genes was validated by qRT-PCR (figure 1B–E). Thus, although the extent to which loss of TLR5 results in spontaneous colitis is environment dependent, with the microbiota being a key determinant, regardless of whether their microbiota is colitogenic, loss of TLR5 results in altered elevated gene expression related to host defence and inflammation.
IL-10R mAb treatment induces homogeneous colitis in T5KO mice
We hypothesised that lack of overt colitis in our Re-T5KO mice might reflect that their endogenous anti-inflammatory pathways are sufficient to prevent inflammation in this context. In support of this notion, we previously observed that anti-inflammatory signalling appeared to function effectively in these mice as they display elevated expression of IL-10 in proportion to the severity of their colitis. Moreover, T5/IL-10 DKO mice exhibited severe colitis with 100% penetrance.11 However, the early and severe onset of this colitis precluded the establishment of a breeding colony of T5/IL-10 DKO mice, thus making it difficult to study the mechanisms driving colitis. Thus, we hypothesised that blockade of IL-10 signalling via neutralising antibodies might also result in robust colitis in T5KO mice and, accordingly, provide a tractable model to mechanistically study disease determinants. Maloy and colleagues have shown that, in WT C57BL/6 mice, 4 weekly administrations of a neutralising mAb to IL-10R resulted in colitis only in mice that had been colonised with H hepaticus.24 In accordance, we observed that administering this antibody to our Re-WT mice did not result in any clinical-type indications of colitis and that the mice continued to exhibit a seemingly normal age-related weight gain (figure 2A). In contrast, Re-T5KO mice subjected to this treatment gained almost no weight during this period (a time in which we have previously observed excessive weight gain in these mice), and exhibited colonoscopic evidence of colitis (figure 2B). Detailed postmortem examination of these mice, performed 1 week following the fourth administration of the IL-10R mAb, confirmed that, indeed, neutralising IL-10 signalling resulted in colitis in Re-T5KO, but not Re-WT, mice. Specifically, IL-10R mAb-treated T5KO mice, but not similarly treated WT mice, exhibited colomegaly and, moreover, displayed thickened bloody colons that lacked well-formed stools (figure 2B). Histopathological analysis confirmed that, indeed, IL-10R mAb-treated T5KO mice had developed severe uniform colitis characterised by crypt abscesses and crypt distortion/loss (figure 2C). The histopathological evidence of inflammation correlated well with mRNA levels of the neutrophil chemokine CXCL1 (KC, the murine equivalent of IL-8) and the master proinflammatory cytokine IL-1β (figure 2D). Elevated IL-1β expression was also manifested at the protein level (figure 2E), indicating inflammasome activation. The severe colitis in IL-10R mAb-treated T5KO mice was accompanied by a robust increase in levels of colonic MPO in Re-T5KO mice, although, analogous to our previous observations, some non-colitic mice had elevated colonic MPO levels despite the lack of observable increases in neutrophils or CXCL1 (figure 2F). Lastly, we observed that IL-10R mAb-treated T5KO mice exhibited splenomegaly (figure 2A).
Phenotypes observed in genetically engineered mice are always at risk of having resulted not from the targeted modification but, rather, from an effect on a nearby gene that might have occurred during a recombination event. Thus, we next sought to determine the extent to which independently generated T5KO mice exhibited increased susceptibility to developing colitis upon neutralisation of IL-10R. Specifically, we obtained an independently generated strain of T5KO mice from Flavell and colleagues28 that we bred to WT mice, followed by crossing their offspring, to generate T5KO mice (termed 'Yale-T5KO') and matched controls. Yale-T5KO mice were subjected to the above-described regimen of IL-10R mAb treatment. Analogous to the results from Re-T5KO mice, administration of IL-10R mAb resulted in robust colitis in Yale-T5KO but not their WT littermates. Specifically, administration of IL-10R mAb to Yale-T5KO mice resulted in weight loss, splenomegaly and colomegaly (supplementary figure 1A), which correlated with the gross altered appearance of the colon (supplementary figure 1B). Histopathological examination showed that IL-10R mAb-treated Yale-T5KO mice had extensive inflammation, particularly neutrophil infiltration extending into the serosa, and exhibited altered crypt morphology, whereas similarly treated WT mice had only mild neutrophil infiltration to the submucosa (supplementary figure 1C). In accordance with this, we observed a marked elevation in colonic MPO in IL-10R mAb-treated Yale-T5KO mice (supplementary figure 1D). This suggests that the increased susceptibility of T5KO mice towards developing colitis in response to this treatment results from loss of TLR5 per se rather than an anomalous recombination event that occurred in the generation of our T5KO mice. We next sought to verify that the colitis resulted from neutralisation of IL-10 signalling and did not simply reflect an increased susceptibility to antibody toxicity. Specifically, we treated Re-T5KO mice with similar amounts of isotype-matched antibodies, namely PD-1 and PDL1, following a similar regimen of administration. No colitis was observed in response to these treatments, indicating that the colitis in T5KO mice had indeed resulted from IL-10 neutralisation (data not shown).
Next, in light of observations that Helicobacter species are present in many mouse vivariums and can influence susceptibility to colitis, we examined if the sensitivity of T5KO mice to development of colitis in response to IL-10R mAb treatment correlated with levels of Helicobacter strains. qRT–PCR, using previously described methodology,29 revealed similar but very low levels of Helicobacter in WT and T5KO mice, arguing against this possibility (supplementary figure 1E). Thus, in the absence of a microbiota that causes them to exhibit basal colitis, loss of TLR5 renders mice highly prone to developing colitis when their endogenous anti-inflammatory mechanisms, particularly IL-10 signalling, are ablated.
Blocking IL-10 signalling induces alteration in gut microbiota
To better understand the role of the microbiota in T5KO colitis, we examined the extent to which rederivation and IL-10R mAb treatment altered the caecal bacterial composition in both WT and T5KO mice. The caecal bacterial community composition was characterised by pyrosequencing of the bacterial 16S rRNA gene. First, we tested whether rederivation, having attenuated their colitis,13 had in fact altered their gut microbiota composition. At the time samples were collected, the rederivation had been performed over 2 years earlier, allowing sufficient time for such 'Jacksonised' mice to equilibrate to the diet and environment of our vivarium. As shown in figure 3A, rederivation did not alter the relative abundances of dominant phyla comprising the microbiota of either WT or T5KO mice (note that non-colitic mice were selected from the original T5KO colony to avoid differences resulting from the inflammation state). However, rederivation did indeed alter the overall diversity of the bacterial communities of both WT and T5KO mice (unweighted UniFrac analysis, figure 3B). These observations are similar to previous characterisation of T5KO and WT mice microbiota13 performed when the colony was younger, indicating a stable pattern of dysbiosis over time. Next, we examined if blockade of IL-10 signalling affected the caecal microbial diversity of WT and T5KO mice. Blockade of IL-10 signalling resulted in an increase in the relative abundance of Firmicutes and, concomitantly, a decrease in the relative abundance of Bacteroidetes (figure 3 and supplementary figure 2). Such changes in Bacteroidetes/Firmicutes ratios have been previously associated with inflammatory states30 31 and thus it is not surprising they were observed in IL-10R mAb-treated T5KO mice, which have robust colitis. In contrast, the occurrence of a similar phylogenic shift in IL-10R mAb-treated WT mice, which lacked colitis, is somewhat more surprising.
IL-10RmAb-induced colitis in T5KO mice is TLR4 independent and MyD88 dependent
We previously observed that the tendency of T5KO mice to develop spontaneous colitis was dependent upon expression of TLR4: breeding T5KO mice onto a TLR4-deficient background eliminated their spontaneous colitis.11 This might be because many of the basal alterations in T5KO mice resulted from increased activation of TLR or, in light of the above-described data, because removing TLR4 from T5KO mice simply tips the balance such that endogenous anti-inflammatory signalling prevents overt inflammation despite elevated levels of host defence gene expression. The latter possibility would be in accord with our recent observation that, like T5KO mice, mice lacking both TLR5 and TLR4 (T4/T5 DKO) developed metabolic syndrome.13 To investigate these possibilities, T4/T5 DKO and control mice that lacked only TLR4 (T4KO) were subjected to blockade of IL-10 signalling via administration of IL-10R mAb as described above. Like WT mice, T4KO mice did not show clinical-type signs of colitis and continued to gain weight at a normal rate (figure 4A). In contrast, like mice lacking only TLR5, T4/T5 DKO mice began to lose weight 2 weeks following initiation of IL-10R blockade, suggesting they might be developing colitis. Moreover, postmortem analysis confirmed that these mice had colitis that was easily appreciable by the gross appearance of the colon that correlated with colomegaly, splenomegaly and altered appearance of the colon (figure 4A,B). Moreover, relative to similarly treated T4KO mice, IL-10R mAb-treated T4/T5 DKO mice exhibited crypt distortion, ulceration and extensive neutrophil infiltration into the serosa (figure 4C), which correlated with elevated levels of MPO and IL-1β (figure 4D,E). Thus, while ablation of TLR4 signalling may reduce proinflammatory signalling to an extent sufficient to prevent spontaneous colitis in mice lacking TLR5, T4/T5 DKO mice were still prone to developing colitis particularly upon interference with their endogenous anti-inflammatory mechanisms.
TLR5 signalling is thought to be wholly dependent upon the adaptor protein MyD88 and thus mice lacking MyD88 are completely deficient in TLR5 signalling. However, in contrast to T5KOs, mice lacking MyD88 (MyD88KO) do not develop spontaneous colitis nor do they display evidence of low-grade inflammation that can manifest as metabolic syndrome.32 Thus, we hypothesised that MyD88KO mice might lack the predisposition to developing colitis upon blockade of IL-10 signalling. In agreement with this notion, and work of others indicating that colitis resulting from IL-10 deficiency is MyD88 dependent,33 34 treatment of MyD88KO mice with IL-10R mAb did not alter their growth rate (ie, weight gain) over the course of treatment, nor did postmortem analysis reveal any evidence of colitis (figure 5A–C). Thus, the predisposition of T5KO to developing IL-10R mAb-induced colitis requires MyD88.
IL-10R mAb-induced colitis requires IL-1R signalling
In addition to mediating TLR signalling, MyD88 is required for signalling by IL-1β and IL-18 receptors, suggesting that lack of colitis in IL-10R mAb-treated MyD88KO mice might reflect a role for these cytokines in the susceptibility of T5KO to developing colitis upon IL-10R mAb treatment. In support of this possibility, we previously observed that basal levels of colon IL-1β remain elevated in T4/T5 DKO mice11 and increased markedly upon IL-10R mAb-induced colitis. Moreover, we recently reported that TLR5 activation drives expression of the natural antagonist of IL-1β, sIL-1Ra, and thus loss of TLR5 results in reduced basal levels of sIL-1Ra wherein the ratio of IL-1β/sIL-1Ra correlates with severity of colitis in T5KO mice.14 Together, these results led us to hypothesise that the predisposition of T5KO mice to develop colitis upon IL-10 blockade might require the action of IL-1β. As an initial test of this hypothesis, we examined the ability of the drug Anakinra, a synthetic analogue of sIL-1Ra developed to treat various inflammatory diseases such as rheumatoid arthritis,35 to prevent the colitis that resulted in T5KO mice upon treatment with IL-10R mAb. We tried doses of 1 and 25 mg/kg delivered subcutaneously, regimens roughly analogous to those used to treat humans with rheumatoid arthritis. Anakinra was administered thrice weekly throughout the course of the IL-10R mAb treatment. Such treatment with Anakinra did not alter the severity of colitis in this model (supplementary figure 3), suggesting that IL-1β activity was not required or that the treatment did not achieve a sufficient blockade of IL-1β activity. Similarly, administration of Anakinra by intraperitoneal injection (as described in the 'Methods' section) also failed to attenuate colitis in this model (data not shown). Because administration of higher doses of Anakinra appeared to cause toxicity in all mice,14 we next sought a non-pharmacological means to address the role of IL-1β in this colitis model.
T5KO mice were bred to mice lacking IL-1R. Offspring of these mice were then bred to make T5/IL-1R DKO and matched T5KO, IL-1RKO and WT control mice. Loss of TLR5 on an IL-1R-deficient background did not result in a detectable phenotype as these mice lacked overt colitis (a colitic outcome seemed unlikely since Re-T5KO were used to generate these T5/IL-1R DKOs) (figure 6A). Moreover, qRT–PCR analysis indicated that some of the genes whose expression was basally elevated in Re-T5KO mice were not elevated upon deletion of TLR5 in an IL-1R-deficient background, in particular host defence genes such as Ang4 and Def3, although some genes whose functions are not well understood (eg, Pla2g4c and Pnliprp2) remained elevated to a similar extent to Re-T5KO mice (figure 6B). Next, we investigated whether these mice were prone to developing colitis upon IL-10 blockade. Like WT mice, IL-1RKO and T5/IL-1R DKO mice did not exhibit clinical-type indicators of colitis such as weight loss during the course of the 4 week IL-10R mAb treatment (figure 7A). Moreover, postmortem evaluation, which included gross examination, histopathological analysis and assay of MPO levels, indicated that, in contrast to T5KO mice, T5/IL-1R DKO mice did not develop colitis upon treatment with IL-10R mAb (figure 7A–C). In accordance with this, T5/IL-1R DKO mice did not display the increase in colonic expression of CXCL1, IL-6 and tumour necrosis factor α (TNFα) that was seen in IL-10R mAb-treated T5KO mice (figure 7D). Moreover, T5/IL-1R DKO mice lacked the induction of CXCL1 and IL-1β mRNA that was observed in T5KO mice (figure 7E). T5/IL-1R DKOs also lacked the induction of IL-1β protein (figure 7F). The absence of IL-10R mAb-induced IL-1β expression in the T5/IL-1R DKO mice is in agreement with the observation that IL-1β can be regulated in a feed-forward manner upon activation of IL-1R. Lastly, we observed that the increase in IL-1β in T5KO mice was accompanied by a decrease in expression of its endogenous inhibitor, sIL-1Ra (figure 7F), further supporting the notion that elevated IL-1β activity plays a key role in underlying the susceptibility of T5KO mice to developing colitis.
The substantial discordance of genetically identical individuals for developing IBD highlights the importance of environmental factors in driving this disorder. Such variability in development of IBD can be envisaged to result from the presence or absence of a specific key cofactor (eg, absolute presence or absence of specific microbiota) or more generally reflect that persons predisposed to developing IBD will be very sensitive to a variety of inflammatory challenges. TLR5-deficient mice have been shown to fit this paradigm in that their tendency to develop spontaneous colitis seemed to require element(s) of the microbiota not present in mice purchased from Jackson Laboratories and, moreover, when not spontaneously colitic, exhibited more severe inflammation relative to WT mice when subjected to Salmonella-induced gastroenteritis or DSS colitis.12 36 The notion that the extent to which KO mice exhibit spontaneous inflammation can vary substantially in different vivariums is in agreement with the widely observed variance in severity of colitis exhibited by IL-10-deficient mice7 8 and also consistent with the failure of a recent study to detect basal inflammation in T5KO mice.37 However, such differences also highlight the importance of comparing KO mice with WT mice that have been bred together to normalise their microbiotas. In this regard, the recent observation that a colony of T5KO mice lacked colitis and had only moderate evidence of metabolic disorders (increased fat mass and irregular patterns of weight gain) is limited by the important caveat that the study compared KO mice bred in their facility with WT mice from commercial suppliers.37 Herein, we report that severe colitis in T5KO mice was triggered not only by providing challenges that directly activate proinflammatory gene expression but also by interfering with endogenous anti-inflammatory signalling, in particular via antibody-mediated neutralisation of the IL-10R. Such treatment resulted in colitis in 100% of T5KO mice even following rederivation (‘Jacksonisation’), which eliminated their incidence of robust spontaneous colitis. Such robust colitis upon IL-10 blockade correlated with elevated basal proinflammatory gene expression. These results suggest that non-colitic T5KO mice may avoid colitis because they have a substantial ability to ‘buffer’ the microbial-induced activation of proinflammatory gene expression that results from loss of TLR5.
The tendency of T5KO mice to exhibit spontaneous colitis could be ameliorated by altering their microbiota or by deleting TLR4.11 Yet, herein, we observed that neither of these approaches prevented colitis upon IL-10 blockade, indicating that prevention of spontaneous colitis is not necessarily equivalent to restoring a WT phenotype. In accordance, we observed qualitatively similar changes in gene expression in Re-T5KO mice relative to their WT littermates. Rather, we speculate that rederivation or loss of TLR4 in T5KO mice prevented them exceeding the ability of endogenous anti-inflammatory mechanisms to buffer inflammation but did not restore WT 'equilibrium'. In contrast to mice lacking both TLR4 and TLR5, mice lacking MyD88 or IL-1R were not primed to develop colitis upon IL-10 blockade. The lack of colitis in IL-10R mAb-treated MyD88KO mice is in agreement with results indicating that loss of MyD88 protected mice against the spontaneous colitis that can develop in IL-10KO mice.34 However, our observation that intestinal levels of IL-1β expression consistently correlated with colitis severity and that loss of IL-1R also protected mice against IL-10R mAb-induced colitis suggests that the protection afforded by loss of MyD88 may reflect loss of IL-1β signalling rather than indicate a role for TLR signalling. In any case, these results indicate an important role for IL-1β in mediating the increased susceptibility of T5KO mice to developing colitis. We speculate that the inability of our pharmacological attempts to prevent IL-10R mAb-induced T5KO colitis reflects inadequate dosing and delivery of the synthetic sIL-1Ra analogue Anakinra.
The potential role of IL-1β signalling in driving colitis in our mice is consistent with the notion that IL-1β is perhaps the most potent proinflammatory cytokine yet described.38 In light of its potency, IL-1β activity is regulated at multiple levels including transcription of the pro-IL-1β gene, the inflammasome-mediated maturation of pro-IL1β into mature IL-1β and modulation of IL-1β activity via regulated expression of sIL-1Ra.39 40 Loss of TLR5 might contribute to elevated IL-1β activity at all of these levels.14 For example, loss of TLR5 upregulates several other innate immune receptors that might induce IL-1β transcription. Moreover, loss of TLR5 upregulates intestinal expression of the intracellular flagellin receptor NLRC4,11 which mediates inflammasome-mediated maturation of pro-IL-1β. Lastly, we recently reported that TLR5-mediated activation of intestinal epithelial cells induces robust expression of sIL-1Ra, which probably normally plays a role in rapidly turning off IL-1β signalling. Thus, while TLR5 signalling in select populations of leucocytes may, directly or indirectly, induce IL-1β, the net activity of TLR5 signalling in vivo is to reduce IL-1β activity. The notion that regulation of the inflammasome plays a role in colitis is highlighted by recent studies that indicate that mice deficient in Nalp3 are highly susceptible to DSS41–43 colitis although, interestingly, a recent study reported opposite results, again highlighting the possibility that differences in a mouse colony's microbiota can have strong effects on phenotype.44
We envisage that the approach of using IL-10R neutralisation to reveal a predisposition to developing colitis will be helpful in further mechanistic studies in T5KO mice. Indeed, while T5KO mice may exhibit enhanced inflammation in several models, the striking difference between T5KO and WT mice upon IL-10R mAb treatment makes this approach much more amenable to the elucidation of additional mechanistic determinants. We speculate that greatly enhanced sensitivity to this treatment reflects that T5KO mice are far more reliant on IL-10 signalling than WT mice due to their basally elevated proinflammatory gene expression. Moreover, we anticipate that the approach of antibody-mediated IL-10R neutralisation will probably be broadly applicable to many strains of mice that might be predisposed to this disorder. Indeed, mice engineered to have IBD-associated genotypes, such as Nod2 deficiency, and IBD-associated phenotypes, such as modestly enhanced gut permeability, have not exhibited spontaneous colitis.45 46 Interestingly, while IL-10R mAb treatment did not cause robust colitis in WT mice, it did cause clear changes in gut microbial diversity analogous to those associated with inflammation and did cause modest elevations in one inflammatory marker, namely MPO. One possible explanation of these data is that loss of IL-10 signalling, by itself, is sufficient to drive changes in proinflammatory gene expression but other endogenous compensatory mechanisms are able to prevent robust inflammation. In this scenario, the diversity changes we observed in the microbiota would be viewed as a consequence of host proinflammatory gene expression but not necessarily a consequence of histological inflammation per se. In conclusion, we speculate that lack of overt spontaneous colitis in T5KO mice, and perhaps other mice with select innate immune deficiencies, may reflect their considerable ability to counter-regulate any changes that may result from these engineered changes and that reliance on such anti-inflammatory pathways might also make them sensitive to IL-10 blockade. Applying this concept to IBD suggests that discordance of the disease among genetically similar individuals might not result from the absolute presence or absence of specific confounding factors per se but reflect the extent to which cumulative environmental factors have overwhelmed endogenous anti-inflammatory mechanisms thus crossing a 'tipping point' from which recovery is very difficult. Defining the extent to which development of IBD indeed manifests in this manner and the specific factors that control whether this threshold is achieved remain key challenges in IBD research.
Funding This work was supported by NIH grants DK061417 and DK083890 to ATG. FAC and MVK are recipients of, respectively, Research Fellowship and Career Development awards from the Crohn's and Colitis Foundation of America. We also acknowledge NIH Digestive Disease Research and Development Center (DDRDC) grants to Emory University (DK064399), and support to OK (HMP DACC), REL (The Hartwell Foundation and the Arnold and Mabel Beckman Foundation) and RK (NIH HG004872, Crohns and Colitis Foundation of America, and the Howard Hughes Medical Institute).
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
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.