Insights into the pathogenesis of ulcerative colitis from a murine model of stasis-induced dysbiosis, colonic metaplasia, and genetic susceptibility

Abstract

Gut dysbiosis, host genetics, and environmental triggers are implicated as causative factors in inflammatory bowel disease (IBD), yet mechanistic insights are lacking. Longitudinal analysis of ulcerative colitis (UC) patients following total colectomy with ileal anal anastomosis (IPAA) where >50% develop pouchitis offers a unique setting to examine cause vs. effect. To recapitulate human IPAA, we employed a mouse model of surgically created blind self-filling (SFL) and self-emptying (SEL) ileal loops using wild-type (WT), IL-10 knockout (KO) (IL-10), TLR4 KO (T4), and IL-10/T4 double KO mice. After 5 wk, loop histology, host gene/protein expression, and bacterial 16s rRNA profiles were examined. SFL exhibit fecal stasis due to directional motility oriented toward the loop end, whereas SEL remain empty. In WT mice, SFL, but not SEL, develop pouchlike microbial communities without accompanying active inflammation. However, in genetically susceptible IL-10-deficient mice, SFL, but not SEL, exhibit severe inflammation and mucosal transcriptomes resembling human pouchitis. The inflammation associated with IL-10 required TLR4, as animals lacking both pathways displayed little disease. Furthermore, germ-free IL-10 mice conventionalized with SFL, but not SEL, microbiota populations develop severe colitis. These data support essential roles of stasis-induced, colon-like microbiota, TLR4-mediated colonic metaplasia, and genetic susceptibility in the development of pouchitis and possibly UC. However, these factors by themselves are not sufficient. Similarities between this model and human UC/pouchitis provide opportunities for gaining insights into the mechanistic basis of IBD and for identification of targets for novel preventative and therapeutic interventions.

the incidence of inflammatory bowel diseases (IBD) has increased over the last century (16), and over 160 gene variants associated with increased risk for IBD (8, 14) have now been identified. However, genetic risk alone is insufficient to account for most cases of IBD. The development of gut dysbiosis due to nongenetic environmental stressors is now believed to be another important factor in either causing or driving IBD (8). Although lifestyle changes, including the use of antibiotics, sanitation, vaccinations, and shifts to “Westernized” diets, certainly contribute, the underlying etiologies for IBD susceptibility remain unclear. Host intestinal homeostasis is strongly influenced by the microbiome, with the epithelium and underlying lymphoid cells normally sensing colonizing microbes and reciprocally establishing an immunologically tolerant mucosal barrier (24). However, in genetically susceptible hosts, altered microbial diversity can lead to aberrant immunological responses and greater IBD risk. The inability to follow patients from a known “time zero” until the development of mucosal inflammation makes it difficult to draw conclusions about the roles of specific communities of bacteria in IBD. Most likely, IBD is caused by the convergence of genetic susceptibility and nongenetic environmental stressors that cause intestinal dysbiosis (1, 13, 18), and the underlying mechanisms remain to be elucidated.

Ulcerative colitis (UC) is a clinical subset of IBD characterized by inflammation occurring exclusively in the colon with sparing of the small intestine. In medically refractory UC, occurring in an estimated 30–50% of patients, the ultimate treatment is total proctocolectomy, followed by the surgical creation of an ileal pouch anal anastomosis (IPAA) (1, 6). The terminal ileum is typically configured into a double-limbed pouch and is connected to the anal canal, so as to serve as a reservoir for stool and permit continence of intestinal contents. An initial ileostomy is constructed to allow the IPAA surgery to heal and after 2–3 mo, if no abnormalities are detected, the ileostomy is taken down, restoring the fecal flow to the ileal pouch. UC patients, however, are at increased risk of inflammation of the pouch, termed pouchitis, compared with non-IBD patients (i.e., those with familial adenomatous polyposis) who undergo similar surgical pouch construction and rarely, if ever, develop pouchitis (1). Although other causative factors continue to be investigated, pouchitis patients exhibit microbial dysbiosis and often respond to antibiotic treatment (5), implicating microbes as causal agents. Since UC-associated disease normally spares the small intestine, understanding the altered microbial community structure and reciprocal mucosal regulation that predisposes the ileal pouches of IPAA to inflammatory complications may provide important clues to our understanding of IBD. Yet model systems have been lacking.

Here, we investigate a clinically relevant murine surgical model of IPAA using wild-type (WT) and IL-10-deficient (IL-10^−/−) mice genetically prone to IBD with self-filling blind ileal loops (SFL), vs. self-emptying loops (SEL) as controls. Our results demonstrate a dramatic microbial community restructuring and altered host mucosal and immune regulation as a consequence of intestinal stasis in SFL loops. The necessary role of microbes in disease pathogenesis is underscored by the lack of IBD in germ-free (GF) animals. The IL-10/TLR4 axis underlies the inflammatory disease in the genetically prone host of this model. Our model defining the essential role for fecal stasis, gut dysbiosis, and genetic susceptibility offers insights into human pouchitis and UC.

MATERIALS AND METHODS

Animals and human samples.

Human biopsies and stool samples were obtained under institutional review board (IRB) approval and all privacy and consent protocols were followed. Human biopsies were taken from UC patients who underwent a total abdominal proctocolectomy with ileal pouch-anal anastomosis at the University of Chicago Medical Center between 2009 and 2011. Patients provided informed, written consent, and the IRB of the University of Chicago approved the study. UC patients with IPAA underwent sequential endoscopic evaluations of their pouches with biopsies and luminal aspirates. Biopsies were obtained from the pouch base (pouch) as well as the normal ileum proximal to the pouch (prepouch).

All animal protocols were approved by institutional animal care and use committee at the University of Chicago. Mice (C57Bl/6 background) were originally purchased from Jackson Laboratory (IL-10, stock no. 002250; TLR4, stock no. 007227. Bar Harbor, ME) and bred in-house under standard 12:12 light/dark conditions at the University of Chicago. IL-10 and TLR4 knockouts were maintained as homozygous; double knockouts were bred in-house and maintained as homozygous. Female mice, ∼8 wk old, were fed ad libitum gel diet 76A (catalog no. 72-07-5022, Clear H20, Portland, ME) starting 5 days prior to surgery to prevent obstruction at the anastomosis. Females were used because our pilot studies demonstrated better surgical recovery and survival compared with male counterparts. Animals were deeply anesthetized with ketamine (100 mg/kg) and xylazine (10 mg/kg). Aseptic surgery was performed to resect 2.5 cm of ileum 3 cm proximal to the ileal-cecal value with anastomosis to the ileum using 8-0 suture (Fig. 1A). In contrast to human surgery, colons are not removed in our mouse model, since this procedure would dramatically complicate the surgical procedure, prolong surgical recovery, and decrease animal survival. The abdominal wall was closed with interrupted 4-0 silk suture and skin was closed with staples. Analgesics (2 mg/kg buprenorphine) were provided postoperatively. Dextran sulfate sodium (DSS) experiments were performed as previously described (3), administering 2% DSS (catalog no. 160110, Affymetrix/USB/MP, Cleveland, OH) for 3 days. After 5 wk, mice were humanely euthanized. Intestinal loops and sham tissues were collected for RNA, protein, and histology. Loop, sham ileum, and sham colon contents were collected and snap frozen at −80°C for microbiota analysis.

Fig. 1.

Self-filling loops adopt a colon architecture, transcript, and microbiome. A: gross appearance of the self-filling (SFL) and self-emptying (SEL) loops, where peristalsis (black arrow) drives toward or away from the blind end. B: principal coordinate analysis (PCoA) using weighted UniFrac of microbial communities from wild-type (WT) SFL and SEL demonstrates clustering with sham colon (SH-C) and ileum (SH-I), respectively. 10KO, IL-10^−/−. C: PCoA analysis of IL-10^−/− SFL and SEL microbial community clustering with sham colon and ileum, respectively. D: taxonomic analysis of microbial genera by group. E: Shannon diversity index from microbial communities by group. F: microarray gene signatures demonstrate global transcriptional similarity between SH-I and WT-SEL and between SH-C and WT-SFL, where red = upregulated genes (UP) and blue = downregulated genes (DN). G: representative histology sections from sham ileum, WT-SEL (WT-E), sham colon, and WT-SFL loop (WT-F), where SH-I and WT-SEL appear similar while SH-C and WT-SEL appear similar. H: periodic acid-Schiff (PAS) stain of WT-SEL and WT-SFL. I: Ki-67 immunofluorescence of WT-SEL and WT-SFL, with Ki67 (red) and DAPI (blue). J: mRNA expression for the morphology genes Wnt-5a, SHH, and PDGFRα was significantly altered between SFL and SEL (n = 3). K: Wnt-5a protein levels in SFL and SEL (representative Western blots). Data are expressed as means ± SE. *P < 0.05 vs. WT-SEL unless otherwise stated. Scale bar = 100 μm.

RNA extraction, cDNA synthesis, and quantitative real-time PCR.

Total RNA isolation was performed on mucosal scrapings with the TRIzol (Ambion) and chloroform method. RNA concentration and quality was determined by UV spectrophotometry. RNA was reverse transcribed by using anchored-oligo(dT) and random hexamer primers. Following RT (Transcriptor Reverse Transcriptase Reaction buffer 5×; Protector RNase inhibitor 40 U/μl; deoxynucleotide mix, 10 mM; transcriptor reverse transcriptase 20 U/μl; Roche Applied Science, Indianapolis, IN), RT-PCR was performed on resulting cDNA in triplicate, by using the manufacturer's protocol (Roche Applied Science), in LightCycler capillary. Gene-specific primers from murine GAPDH, PDGFRα, SHH, Wnt5a, IFN-γ, TNF-α, IL-1β, TLR2, and TLR4, in addition to human GAPDH, Wnt-5a, and PDGFRα are shown in Table 1. The optimal concentration of cDNA and primers, as well as the maximum efficiency of amplification, were obtained through five-point, twofold dilution curve analysis for each gene. RT-PCR amplification consisted of an initial denaturation step (95°C for 10 min), 45 cycles of denaturation (95°C for 10 s), annealing (55°C for 20 s), and extension (60°C for 30 s), followed by a final incubation at 55°C for 30 s and cooling at 40°C for 30 s. All measurements were normalized by the expression of GAPDH gene, considered as a stable housekeeping gene. Gene expression was determined by the delta-delta Ct method: 2^−ΔΔCT{ΔΔCt = [Ct(target gene) − Ct(GAPDH)]_patient − [Ct(target gene) − Ct(GAPDH)]_control}. Real-time data were analyzed by use of the Roche LightCycler (Roche Applied Science).

Table 1.

Primers used in this study

qRT-PCR and PCR Primers Used in This Study
Murine Primers
TNFα forward: 5′ - GCC TCC CTC TCA TCA GTT CT - 3′
TNFα reverse: 5′ - CAC TTG GTG GTT TGC TAC GA - 3′
IFNγ forward: 5′ - CAC GGC ACA GTC ATT GAA AG - 3′
IFNγ reverse: 5′ - TTT TGC CAG TTC CTC CAG AT - 3′
IL-1β forward: 5′ - ACC TTT TGA CAG TGA TGA GAA – 3′
IL-1β reverse: 5′ - GAG ATT TGA AGC TGG ATG CT −3′
TLR2 forward: 5′ – GCT GGA GGA CTC CTA GGC T – 3′
TLR2 reverse: 5′ – GTC AGA AGG AAA CAH TCC GC – 3′
TLR4 forward: 5′ – ACC AGG AAG CTT GAA TCC CT – 3′
TLR4 reverse: 5′ – TCC AGC CAC TGA AGT TCT GA – 3′
PDGFRα forward: 5′ – TCC ATG CTA GAC TCA GAA GTC A – 3′
PDGFRα reverse: 5′ – TCC CGG TGG ACA CAA TTT TTC – 3′
SHH forward: 5′ – AAA GCT GAC CCC TTT AGC CTA – 3′
SHH reverse: 5′ – TTC GGA GTT TCT TGT GAT CTT CC – 3′
Wnt-5a forward: 5′ – CAA CTG GCA GGA CTT TCT CAA – 3′
Wnt-5a reverse: 5′ – CAT CTC CGA TGC CGG AAC T – 3′
GAPDH forward: 5′ – GGC AAA TTC AAC GGC ACA GT – 3′
GAPDH reverse: 5′ – AGA TGG TGA TGG GCT TCC C – 3′
Human Primers
PDGFRα forward: 5′– TTT TTG TGA CGG TCT TGG AAG T −3′
PDGFRα reverse: 5′ – TGT CTG AGT GTG GTT GTA ATA GC −3′
Wnt-5a forward: 5′ – TCG ACT ATG GCT ACC GCT TTG – 3′
Wnt-5a reverse: 5′ – CAC TCT CGT AGG AGC CCT TG – 3′
GAPDH forward: 5′ – TTC TAT AAA TTG AGC CCG CA - 3′
GAPDH reverse: 5′ – CGA CGC AAA AGA AGA TGC - 3′
16S Primers
338F: 5′ - GTGCCAGCMGCCGCGGTAA - 3′
806R: 5′ - GGACTACHVGGGTWTCTAAT - 3′

DNA isolation.

Intestinal contents were homogenized in 1 ml extraction buffer [50 mM Tris (pH 7.4), 100 mM EDTA (pH 8.0), 400 mM NaCl, 0.5% SDS] containing 20 μl proteinase K (20 mg/ml, catalog no. 03115887001, Roche). 0.1-mm-diameter zirconia/silica beads (BioSpec Products, Bartlesville, OK) were added to the extraction tubes and a Mini-Beadbeater-8 cell disrupter (BioSpec Products) for 5 min to lyse cells. After overnight incubation at 55°C with agitation, extraction with phenol:chloroform:isoamyl alcohol and precipitation with ethanol were performed. Isolated DNA was dissolved in nuclease-free water and stored at −80°C.

16S rRNA-based polymerase chain reaction, Illumina library preparation, and data analysis.

Polymerase chain reaction was performed as follows: 5 μl of 10× Ex Taq buffer containing 20 mM MgCl₂ (Takara, Tokyo, Japan), 4 μl of 2.5 mM dNTP Mixture (Takara), 1 μl each of forward (27F, 5′-AGA GTT TGA TCC TGG CTC AG-3′) and reverse (1492R, GGT TAC CTT GTT ACG ACT-3′) primer (10 mM each), 0.25 μl of Taq polymerase (Takara), 36.75 μl nuclease-free water, and 2 μl of DNA template. The PCR conditions were 94°C for 5 min followed by 30 cycles of amplification consisting of denaturation at 94°C for 30 s, annealing at 58°C for 1 min, and extension at 72°C for 1.5 min.

PCR primers used were specific for the 515–806 bp region of the 16S rRNA encoding gene (Table 1) and contained Illumina 3′ adapter sequences as well as a 12-bp barcode. This barcode-based primer approach allowed sequencing of multiple samples in a single sequencing run without the need for physical partitioning. Sequencing was performed by using an Illumina MiSeq DNA sequencer at Argonne National Laboratory's Next Generation Sequencing Core. Sequences were then trimmed and classified with the QIIME toolkit (version 1.8.0). By using the QIIME wrappers, Operational Taxonomic Units (OTUs) were picked at 97% sequence identity by using uclust, and a representative sequence (centroid) was then chosen for each OTU by selecting the most abundant sequence in that OTU. These representative sequences were aligned by using PyNAST, and taxonomy was assigned to them with the uclust consensus taxonomy assigner. The PyNAST-aligned sequences were also used to build a phylogenetic tree with FastTree and weighted UniFrac distances were then computed between all samples for additional ecological analyses, including principal coordinates analysis (PCoA).

Microarrays data and pathway analysis.

Total RNA was reverse transcribed for cDNA synthesis, labeled, and hybridized to Illumina (San Diego, CA) MouseRef-8 v2.0. Expression BeadChips with 25k probes and scanned by using Illumina HiScan in the Functional Genomics Core of University of Chicago. Summary data were obtained via the BeadStudio software from Illumina. We evaluated the array quality and processed the data using the R/Bioconductor package “limma” for background correction and quantile normalization (19, 25). RNA samples used in this study passed the established quality criteria including an RNA Integrity Number greater than 7.5, and a 260/280 nm optical density ratio above 1.8. We identified differentially expressed genes using empirical Bayes statistics implemented by eBayes in “limma” package (20). The criteria of significance is set at false discovery rate (FDR) <5% and fold change >2.0. Microarray raw data and normalized data have been deposited in Gene Expression Omnibus database (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE71997.

Significant canonical pathways were identified using Ingenuity Pathway Analysis (IPA) software (Ingenuity Systems, Redwood City, CA). The significance of a canonical pathway was determined by the right-tailed (referring to the overrepresented pathway) Fisher's exact test with P < 0.05.

Histology/histopathology score/PAS/Ki-67 immunofluorescence.

Mucosal tissues were fixed in 4% formalin/PBS overnight. Five-micrometer sections were cut, deparaffinized in xylene, rehydrated, and stained with either hematoxylin and eosin or periodic acid-Schiff (PAS) base (catalog no. 3952016 Sigma, St. Louis, MO) and imaged on a Leica DM2500 microscope (Leica Microsystems, Wetzlar, Germany) through a ×10 lens objective with Image Pro-Plus software (Media Cybernetics, Silver Spring, MD) for image capture. Mucosal inflammation was blindly assessed by using the colitis scoring method previously described (2) and displayed as the disease activity score. To assess cell proliferation, Ki-67 staining was performed. Briefly, following rehydration, slides were heated in sodium citrate buffer (pH 6.0), blocked with blocking solution (catalog no. x0909, Dako, Carpinteria, CA), and stained overnight with rabbit anti-Ki67 primary antibody (catalog no. RM-9106, Thermo Scientific, Waltham, MA) in antibody diluent (catalog no. S3022, Dako). Slides were washed and incubated with donkey anti-rabbit secondary antibody (catalog no. A31572, Alexa Fluor Life Technologies) before being counterstained and coverslipped with ProLong Gold anti-fade with DAPI (P36931, Thermo). Slides were processed and imaged in parallel and signal was semiquantified blindly with ImageJ. Ki-67 signal was measured in 20–25 crypts per animal and displayed as mean fluorescent intensity per crypt.

Western blot.

Immunoblotting was performed as previously described (22) with the following antibodies: anti-TLR4 (sc-293072), anti-TLR2 (sc-10739), from Santa Cruz Biotechnology (Santa Cruz, CA); and anti-TNF-α (no. 3707) from Cell Signaling Technology (Boston, MA). All numerical results are expressed as means ± SE of the indicated number of experiments. The results of blots are presented as direct comparisons of bands in autoradiographs and quantified by densitometry using ImageJ.

Enzyme-linked immunosorbent assay.

ELISA for TNF-α protein on tissue homogenate was performed per manufacturer instructions (catalog no. 88-7324-22, eBiosciences, San Diego, CA).

Statistics.

Data from the in vivo studies are presented as means ± SE; statistical significance was analyzed by the analysis of variance test (ANOVA) followed by analysis of significance (Tukey-Frame's multiple-comparisons test) or t-test, where appropriate. P < 0.05 was considered statistically significant, unless otherwise stated. 16S and microarray statistical analysis were performed as described above.

RESULTS

Self-filling blind intestinal loops harbor a colon-like microbiome.

Ileal blind SFL or SEL were surgically created in 8-wk-old WT, IL-10^−/−, TLR4^−/−, and IL-10^−/−/TLR4^−/− C57Bl/6 mice and animals were allowed to recover for 5 wk. Gross inspection of WT SFL revealed distension with fecal contents and thickening of the bowel wall compared with the normal ileal appearance of SEL (Fig. 1A). By 16S rRNA gene sequencing analysis, the SFL and SEL microbiota were significantly different (P < 0.0001 by either ANOSIM or PERMANOVA), the former clustering in a manner similar to colonic microbiota of sham-operated colon, and the latter with sham-operated ileal microbiota, as shown by PCoA of weighted UniFrac distances (Fig. 1B). The SFL microbiota showed strong representation of Bacteroidetes and Firmicutes, whereas the SEL microbiota was more similar to that of sham-operated ileum (phylum level assignments in Fig. 1D; family level in Fig. 2).

Fig. 2.

16S family taxonomic analysis for data shown in Fig. 1D. Color coding of relative microbial abundances at the family level is displayed for each sample type. Generally, the SFL and colon samples (WT-SH-C) display greater levels of Bacteroidetes families compared with the Firmicutes and Proteobacteria families, which are more dominant in SEL and ileum samples (WT-SH-I). Similarly, while different bacterial families dominate the human vs. mouse samples, the pouch has a similar phylum composition to SFL.

To determine how the selected microbial communities would be impacted in a model of IBD in a genetically susceptible host, surgery was performed to create SEL and SFL in IL-10^−/− mice, where ∼15–20% develop spontaneous colitis after 12 wk of age (9). Importantly, all IL-10^−/− mice used for these experiments had normal ileal mucosa by gross visualization at the time of surgery. PCoA analysis again showed that the microbial community patterns in IL-10^−/− SFL and SEL were highly comparable to the SFL and SEL observed in WT loops (Fig. 1C) and SFL clustered significantly different from SEL (P < 0.0001 by ANOSIM or PERMANOVA). This can also be shown by differences in taxonomic composition as well as in measures of alpha diversity, specifically the Shannon diversity index (Fig. 1, D and E). The diversity of microbiota in sham-operated colon, WT SFL, and IL-10^−/− SFL microbiota was significantly higher than that in SEL and sham-operated ileum. These changes demonstrate that, regardless of genetic background, SFL stasis, defined as accumulation of fecal contents, is a major driver of microbial structure; equally important, this microbiome is similar to that found in colon.

Changes in mucosal histology of self-filling loops resemble colonic mucosal structure.

Histological examination of SFL mucosa suggested differences with SEL and sham-operated ileum, characterized by elongation of crypt lengths, relative villus shortening, and increased goblet cell numbers that collectively provide resemblance of colonic mucosa (bottom, Fig. 1, G and H). These changes were not always present and are also variably observed in the mucosa of the ileal pouch of UC patients who have undergone IPAA (12). In contrast, the SEL epithelium remained comparable to the sham ileal epithelium (top, Fig. 1, G and H).

Changes in mucosal gene expression in the self-filling ileal loop are consistent with colonic metaplasia.

To characterize molecular differences in mucosal tissues between SFL and SEL, Illumina microarray analysis was performed. WT SFL exhibited suppression of immune response pathways, including regulation of antigen processing and presentation, cell surface receptor-linked signal transduction, and proteolysis (Fig. 3). Despite these observed pathway differences, specific gene changes that would provide insights into the morphometric alterations in crypt depth and villous atrophy were not obvious (Fig. 4: Metaplasia-associated transcription changes in self-filling loops). Therefore, we performed targeted analysis for morphogen-related genes that control intestinal architecture. Since Wnt-5a is required for crypt regeneration following mucosal injury and negatively regulates Sonic hedgehog (SHH), which is required for villous outgrowth, we hypothesized that noncanonical Wnt ligands were involved in our observed phenotype (11, 15). Indeed, Wnt5a gene expression was upregulated while SHH was downregulated in SFL compared with SEL (Fig. 1J). Wnt-5a protein levels were altered in a consistent manner between SFL and SEL (Fig. 1K). Expression of the gene for PDGFRα, a receptor responsible for epithelial proliferation, was significantly elevated in SFL. Consistently, the SFL demonstrated evidence of elevated cell proliferation compared with SEL, assessed by Ki-67 immunofluorescence (Fig. 1I). Finally, analysis of the gene array results revealed unique gene signatures for SFL and SEL, with global similarities observed between SFL and colonic transcriptomes as well as those observed between SEL and ileal transcriptomes (Fig. 1F). Interestingly, these morphogen-related genes are also consistently altered in clinical human pouch biopsies compared with prepouch ileum as determined by RNA-Seq (unpublished data). This study supports the notion that, in the presence of stasis-induced colonic-like microbiota, the SFL develops both colon-like histological features and certain colon-like gene expression changes (differentially expressed genes and pathway analysis from WT SFL and SEL are shown in Fig. 4).

Fig. 3.

Self-filling loop tissues mimic colon transcript patterns. Pathway analysis of microarray results from loop tissues demonstrated that WT SFL mucosa express colon-associated gene patterns compared with WT SEL loops (top). Specifically, these altered genes suggested decreases in pathways associated with immune activation (bottom). Differentially expressed genes are displayed where red = upregulated genes; blue = downregulated genes. Microarray targets (25k probes) were quantified and normalized to the average signal within each sample. The criteria of significance is set at false discovery rate (FDR) <5% and fold change (FC_log2) >2.0.

Fig. 4.

Metaplasia-associated transcription changes in SFL. Significantly altered metaplasia-associated genes in WT SFL loops compared with WT SEL. WT SFL induced differential gene patterns compared with WT SEL that included significantly up- and downregulated genes (A). Specific genes differentially expressed in WT-SFL loops compared with WT-SEL loops (B). Red = upregulated genes; blue = downregulated genes. Microarray targets (25k probes) were quantified and normalized to the average signal within each sample. The criteria of significance is set at false discovery rate (FDR) <5% and fold change (FC_log2) >2.0.

Active mucosal inflammation and metaplasia occur in the self-filling loop, but not in the self-emptying loop, of the IL-10^−/− mouse.

To examine the impact of stasis-induced colon-like microbiota on a background of genetic susceptibility, we examined the colonic mucosa and inflammatory state of the SFL and SEL in the IL-10^−/− mouse. In a manner similar to WT, IL-10^−/− SFL adopt a colon-like mucosal architecture compared with SEL (Fig. 5A) and showed similar changes in Wnt-5a, SHH, and PDGFRα morphogen-related gene expression (Fig. 5F). IL-10^−/− SFL also displayed elevated proliferation compared with SEL, assessed by Ki-67 (Fig. 5, C and D). However, there were at least three distinct differences between WT and IL-10^−/− mice. First, although WT mice do not develop active mucosal inflammation, IL-10 ^−/− SFL exhibit evidence of lymphoplasmacytic infiltrates and histological inflammation (Fig. 5, A and E). Second, IL-10^−/− SFL showed a loss of goblet cells compared with SEL or WT SFLs (Fig. 5B), consistent with an inflamed mucosa phenotype. Finally, the IL-10^−/− mice with SFL exhibited severe weight loss, cachexia, and anorexia. In contrast, the SEL mucosa of IL-10^−/− mice remained similar to sham ileum and demonstrated little histologically observable inflammation.

Fig. 5.

IL-10^−/− SFL are susceptible to mucosal inflammation. A: representative histology sections of the IL-10^−/− SEL (10KO-SEL) and SFL (10KO-SFL), where SFL demonstrate mucosal inflammation and inflammatory infiltrates. B: PAS stain of 10KO-SFL and 10KO-SEL. C: Ki-67 immunofluorescence of 10KO-SEL and 10KO-SFL, with Ki67 (red) and DAPI (blue). D: quantified Ki-67 intensity per crypt. E: blinded histology score of the SFL and SEL from WT and IL-10^−/− mice. F: expression of morphology genes Wnt-5a, SHH, and PDGFRα was significantly altered between IL-10^−/− SFL and SEL. G: cytokine expression of IL-1β, TNF-α, and IFN-γ was significantly elevated in IL-10^−/− SFL compared with all other groups. H: numerous inflammatory-associated genes were differentially expressed in IL-10^−/− SFL compared with WT SFL, where red = upregulated and blue = downregulated. I: expression and protein levels of TLR4 were significantly elevated in SFL compared with SEL in both genotypes, however, total levels were significantly greater in IL-10^−/−. TLR2 expression and protein levels were significantly elevated in IL-10^−/− SFL loops compared with SEL, but no changes were observed in WT. Data are expressed as means ± SE. *P < 0.05 vs. 10KO-SEL unless otherwise stated. Scale bar = 100 μm.

Dramatic changes in key proinflammatory cytokines accompanied the histological inflammation and clinical deterioration in IL-10^−/− SFL. Thus transcript levels for IFN-γ, TNF-α, and IL-1β (Fig. 5G) were significantly elevated compared with SEL and WT animals. In contrast to IL-10^−/− SEL or WT SFL, pathway analysis of gene microarray data in IL-10^−/− SFL showed increases in antigen processing, host defense, and immune response genes (Fig. 6). Individual genes exhibiting significant differences (defined as >2.0-fold change) between IL-10^−/− SFL and SEL are displayed in Supplemental Table S1 (Supplemental Material for this article is available online at the Journal website).

Fig. 6.

IL-10^−/− SFL vs. SEL and IL-10^−/− vs. WT gene pathway changes. Pathway analysis of microarray results from IL-10^−/− loop tissue demonstrated upregulation of immune associated pathways in SFL compared with SEL tissue (top). Similarly, compared with WT SFL, IL-10^−/− SFL demonstrated upregulation of numerous immune pathways and attenuation of others (bottom). Red = upregulated genes; blue = downregulated genes. Microarray targets (25k probes) were quantified and normalized to the average signal within each sample. The criteria of significance is set at false discovery rate (FDR) <5% and fold change (FC_log2) >1.5.

To verify our histological findings in a second relevant model of colonic inflammation, we used the well-characterized model of DSS administration to nongenetically susceptible WT mice, where inflammation is normally only observed in the colon after providing 2% DSS in drinking water for 3 days. Interestingly, DSS causes inflammation of the WT SFL, but not the SEL, providing additional parallels between the SFL and colon (Fig. 7).

Fig. 7.

DSS induces inflammation in WT SFL. Compared with control WT SFL, the administration of 2% DSS in drinking water for 3 days significantly induced mucosal injury and TNF-α protein levels in WT SFL, but not WT SEL despite slightly elevated TNF-α. Colonic injury was also visible and colonic TNF-α protein levels were significantly elevated in all animals administered DSS. *P < 0.05. Scale bar = 100 μm.

Together, these data strongly indicate that the stasis-induced changes in gut microbiota, although sufficient for inducing intestinal metaplasia, were insufficient alone to cause inflammatory disease, suggesting that a “second hit” provided by genetically (IL-10^−/−) or environmentally induced (DSS) changes are required to induce inflammation.

Innate immunity pathways involving TLR-4 mediate intestinal metaplasia and mucosal inflammation in self-filling loops.

Since the SFL displays epithelial metaplasia and altered inflammatory pathways in the presence of greater bacterial loads and diversity compared with SEL, we assessed mucosal Toll-like receptors, which are responsible for microbial recognition and downstream inflammation. TLR4 mRNA transcript and protein expression were significantly elevated in SFL of both genotypes, yet were higher in IL-10^−/− than WT (Fig. 5I). There were no significant differences in TLR2 in WT SEL or SFL; however, TLR2 was also significantly elevated in SFL compared with SEL in IL-10^−/− mice. Consistent with these observations, gene microarray demonstrated elevated expression of proinflammatory gene pathways in IL-10^−/− SFL compared with IL-10^−/− SEL (Fig. 5H). These observations are consistent with the pathway analysis and differentially expressed inflammatory genes shown in Fig. 6 and Supplemental Table S1, where numerous immune pathways are overexpressed in IL-10^−/− SFL.

Given the colon-like epithelium associated with upregulated TLR4 in SFL, we examined the role of TLR4 in mediating the observed changes in the SFL using TLR4^−/− mice. Surprisingly, the histological features of SFL and SEL of TLR4^−/− mice were not obviously different from those of sham-operated ileum (Fig. 8A vs. Fig. 1G). Along with the lack of metaplastic changes in TLR4^−/− SFL mice, no significant differences were observed in SHH and Wnt5a gene expression compared with WT SFL mice (Fig. 8B) Similarly, no differences in IFN-γ or TNF-α expression were found (Fig. 8C) compared with IL-10^−/− SFL. In contrast to WT and IL-10^−/− animals, there were no differences in crypt Ki-67 signal between SFL and SEL in TLR4^−/− animals (data not shown). The SFL of one of three TLR4^−/− animals showed variability in the metaplasia gene signature and the colon gene signature, with profiles more akin to WT SFL and sham colon rather than ileum (Fig. 9, A and B). In contrast to the lack of host adaptation, it is important to note that the relative microbiota profiles of TLR4^−/− mice were significantly different between SFL and SEL (P < 0.014 and P < 0.007 by ANOSIM and PERMANOVA, respectively), similar to the microbiota profiles of WT and IL-10^−/− mice (phylum level shown in Fig. 8D; family level shown in Fig. 10; WT phylum level shown in Fig. 1D). Collectively, these results suggest TLR4 signaling plays an important role in mediating the metaplastic changes that occur in response to stasis and that stasis remains the dominant driver of microbial community structure over the TLR4^−/− host genotype.

Fig. 8.

Self-filling loop morphology and inflammation requires TLR4 signaling. A: representative histology of TLR4^−/− SFL (T4KO-F) and SEL (T4KO-E) demonstrated normal mucosal architecture. B: compared with WT SEL, TLR4^−/− demonstrate no differences in morphology genes, SHH and Wnt-5a. C: compared with IL-10^−/− SFL, TLR4^−/− demonstrate no differences in proinflammatory genes, TNF-α and IFN-γ. D: the composition of loop microbial genera in TLR4^−/− mice remained similar to that of WT mice and cluster with SFL and SEL samples, respectively. E: representative histology of TLR4/IL-10^−/− SFL and SEL also demonstrated normal mucosal architecture. F: no differences were observed in morphology gene expression between TLR4/IL-10^−/− SFL and SEL. G: no differences were observed in proinflammatory cytokine expression between TLR4/IL-10^−/− SFL and SEL. Data are expressed as means ± SE. *P < 0.05 vs. T4KO-SFL. Scale bar = 100 μm.

Fig. 9.

Self-filling loop gene signatures are abated without TLR4. Metaplasia and colonic gene signatures. Microarray gene targets (25k probes) were quantified and normalized to the average signal within each sample. Altered genes are displayed as red = upregulation; blue = downregulation. Criteria of significance is set at false discovery rate (FDR) <5% and fold change (FC_log2) >2.0. Analysis of differentially expressed microarray genes demonstrated a blunted response in the TLR4^−/− SFL loops compared with the more robust change in WT SFL samples (A). Similarly, colon-associated gene signatures that are evident in the WT SFL loops (Figs. 1F and 6) are less robust in TLR4^−/− SFL loop samples (B).

Fig. 10.

16S family taxonomic analysis for data shown in Fig. 8D. Color coding of relative microbial abundances at the family level is displayed for each sample type. In contrast to minimal histological changes, compared with the T4KO-SEL the T4KO-SFL displayed an increased relative abundance of Bacteroidetes and fewer Firmicutes, suggesting the microbial populations are shifting in composition in response to the stasis environment, regardless of host mucosal phenotype.

To further examine the TLR4 dependence of the colonic-like metaplasia and inflammation in mice genetically prone to inflammation, SFL and SEL were studied in IL-10^−/−/TLR4^−/− double knockout (KO) mice. Similar to the TLR4^−/− mice, SFLs in IL-10^−/−/TLR4^−/− showed no signs of metaplasia (Fig. 8E); no changes in expression of the morphology-related genes Wnt-5a, SHH, and PDGFRα (Fig. 8F); and no changes in inflammatory genes IFN-γ and TNF-α (Fig. 8G) compared with SEL mucosa. This strongly indicates that the observed changes in the IL-10^−/− SFL rely heavily upon TLR4 signaling.

Microbes from SFL cause heightened colonic morphology in GF WT mice and colitis in GF IL-10-deficient mice.

To determine the differential effects of SFL and SEL microbiota on the colonic epithelia, we examined morphological and inflammatory outcomes of colonic tissue in conventionalized GF mice. It is well documented that in GF animals colonic epithelial development is incomplete, with shortened crypt depth, atrophied villi, and fewer goblet cells. We conventionalized WT GF mice with SFL or SEL contents (Fig. 11A). After 4 wk, GF mice conventionalized with SFL contents induced hypertrophy of the muscle layer, increased neuronal fibers, and altered gastrointestinal motility compared with GF mice conventionalized with SEL luminal contents (unpublished Chang laboratory data). In addition, morphogen-related gene expression (Wnt5a, SHH) was altered in a pattern similar to that observed in SFL and SEL mucosae (Fig. 12), with greater Wnt-5a and decreased SHH following colonization with SFL microbiota.

Fig. 11.

Self-filling loop microbes induce colitis in IL-10^−/− GF mice and SFL resemble human samples. A: representative histology of colons from WT germ-free (GF) and IL-10^−/− germ-free (GF 10KO) mice conventionalized with contents from SFL or SEL. GF 10KO colons exhibited inflammation with abscess (inset), crypt dysplasia (dashed arrow), and luminal neutrophils and mucosal lymphoplasmacytic cells (solid arrow). B: blinded histology of colon mucosa from conventionalized WT or IL-10 mice. C: IFN-γ, TNF-α, and TLR4 expression was significantly higher in GF IL-10^−/− mice conventionalized with SFL contents compared with all other groups. D: microbial community composition of samples from murine and human samples, where human pouch samples were similar to murine sham colon (SH-C) and WT and IL-10 KO SFL contents compared with sham ileum (SH-I) or WT and IL-10 KO SEL contents. E: compared with human prepouch ileal biopsies, human pouch biopsies differentially expressed the morphology genes Wnt-5a and PDGFRα, similar to the pattern observed in murine SFL vs. SEL mucosa. Data are expressed as means ± SE. Scale bar = 100 μm.

Fig. 12.

Morphology-related genes in conventionalized germ-free colons. Conventionalization of GF mice with SFL or SEL loop contents significantly altered the morphology genes Wnt-5a and SHH in a similar pattern to what is observed in SFL and SEL loop tissues, respectively. Conventionalization (CONV) of germ-free WT animals with microbes from SFL resulted in greater Wnt-5a and lower SHH compared with conventionalization with microbes from SEL; n = 4/group. *P < 0.05.

As described, microbial stasis induced severe inflammation in the IL-10^−/− SFL and led to a mucosa with some degree of colon-like resemblance. To test whether SFL microbes alone drive the IL-10^−/− inflammatory response, we conventionalized GF IL-10^−/− and WT mice with microbes from WT SFL and SEL. Following conventionalization, IL-10^−/− colons displayed greater mucosal thickness than WT mice, making the colonic cross section larger. Conventionalization of GF IL-10^−/− mice with SFL microbes induced severe active colitis, characterized by increased lymphoplasmacytic infiltrates and crypt abscesses, while SEL microbes failed to induce disease (Fig. 11, A and B); in contrast, conventionalization of WT mice with SFL contents resulted in no active colonic inflammation as determined by blinded histology inflammation score assessment (Fig. 11B). Consistent with histology, colonic epithelial expression of proinflammatory markers, including IFN-γ, TNF-α, and TLR4 gene expression were significantly greater in IL-10^−/− GF mice conventionalized with SFL microbes compared with all other groups (Fig. 11C).

These results support the observation that select and specific microbial community composition modifies colonic morphology in GF animals and causes aberrant mucosal inflammatory response in genetically susceptible hosts.

Human correlations to morphogenic and microbial data is consistent with results from the mouse model.

To correlate our findings in the mouse model of pouchitis with human samples, we obtained pouch biopsies and fecal samples from patients with previous total abdominal colectomies and IPAA. We first compared human microbiota samples to murine loops using 16S rRNA gene analysis. Whereas microbial genera from ileum, cecum, and distal colon in sham mice remain somewhat distinct from human pouch samples, microbial communities from WT and IL-10^−/− SFL appear compositionally most similar to human pouch samples (phylum level assignments shown in Fig. 11D; family level shown in Fig. 2). PCoA analysis further demonstrated microbial community clustering between SFL samples and human samples (Fig. 13), whereas the other samples cluster more similarly. Because we did not have human prepouch 16S rRNA gene samples that could be included in this analysis, our interpretation is limited to the conclusion that the human pouch samples are more similar to murine SFL than they are to murine SEL communities. Finally, human ileal pouches exhibit morphology-related gene responses that are similar to those observed in the murine SFL, with elevated Wnt-5a and PDGFRα expression, compared with normal human prepouch ileum (Fig. 11E).

Fig. 13.

Human and mouse microbial principal coordinate analysis. Principal coordinate analysis (PCoA) based on weighted UniFrac of murine and human microbiota samples. Compared with human pouch samples (n = 22), murine WT SFL (n = 18) and IL-10^−/− SFL (n = 3) samples globally cluster most closely while WT SEL (n = 18) and IL-10 SEL (n = 3) samples and all sham samples (ileum, n = 5; cecum, n = 5; and distal colon, n = 5) cluster more similarly.

DISCUSSION

UC never involves the small intestine and total colectomy is curative. However, as nature would have it, at least half of patients who undergo IPAA, a procedure that involves the creation of a pseudorectum from the terminal ileum, will develop within 1–2 years an inflammatory condition called UC (1, 6). Many believe that the development of condition is a recapitulation of some of the same pathophysiological processes that caused the original UC because it has many of the endoscopic and histological features of the original disease and it occurs much less commonly in non-IBD [familial adenomatous polyposis (FAP)] patients who undergo the same procedure. Even in these patients, the FAP pouchitis looks different and often has a milder clinical course compared with UC pouchitis. However, if UC only involves colonic tissue, why would UC pouchitis develop in the ileal mucosa of the pouch? This question remains unanswered although the notion that the pouch mucosa undergoes a colonic-type metaplasia has often been theorized. This study now provides molecular evidence to support this hypothesis. Using a murine model that reproduces many of the same conditions in UC, we show that the cooccurrence of a colonic epithelial transcriptome, the assemblage of a colonic-like microbiome, and genetic susceptibility are all key factors that can result in the development of a UC-like inflammatory response. Each process by itself, however, is insufficient in causing disease.

The development of a colonic-like microbiome characterized by greater bacterial load and increased membership by Bacteroidetes and Firmicutes (Fig. 1D) is almost certainly a consequence of intestinal stasis, a condition shared between normal murine colon and murine blind SFL, and is similar to the microbiome profile of healthy humans (23). The similar composition between our murine SFL microbial community and human pouch samples, compared with SEL samples, further supports that this is a general phenomenon of stasis rather than being species or disease specific. The fact that the SEL blind loop, which is identical in everyway with the SFL except for the propagative direction of intestinal content, does not develop a colonic-like microbiota strongly supports this notion. In addition, antibiotic treatment prevents the development of colonic-like metaplasia in the SFL (data unpublished).

In response to the colonic-like microbiota, the general SFL mucosal morphology and gene expression diverge away from normal small intestine and toward a colonic-like phenotype (Fig. 1G). This observation is consistent with a previous report suggesting the presence of fecal anaerobic bacteria is associated with colonic-like morphology in human pouches (10). The changes in gene expression of key morphogens such as Wnt5a and SHH in SFL provide molecular explanation for the development of the observed colon-like metaplasia. Together these data illustrate the highly adaptable nature of the intestinal mucosa to alter with shifting microbial communities.

In health, small intestinal epithelial cells and lamina propria lymphoid cells maintain low expression of TLR4 and TLR2 to prevent overresponsive immune activation. However, elevated TLR4 and TLR2 levels have been reported in active human pouchitis samples (17, 21). Although TLR4 expression was elevated in WT SFL, it was not associated with alterations in molecular or histological markers of inflammation. On the contrary, WT SFL display downregulated gene expression of immune signaling pathways, most notably those suppressing antigen processing and presentation. Since Bacteroides are elevated in the SFL microbiota, and these commensal genera are known stimulators of IL-10 producing CD-4 T cells (13), we posit that an IL-10-mediated mechanism for WT tolerance. Murine IL-10 is known to inhibit Th1-mediated pathways by supporting Th2 responses. Polymorphisms of the IL-10 and IL-1p receptor have been associated with increased risk for IBD and very early onset UC. In IL-10^−/− mice, we typically observe an incidence of 15–20% for the development of spontaneous chronic colitis, although this rate can be highly influenced by environmental, dietary, and microbial factors. For these experiments, it is a useful model of genetic susceptibility that is also considered to be a conditional requirement for the development of IBD (9).

Our data demonstrate that in marked contrast to WT SFL, the SFL of IL-10^−/−-deficient mice manifest elevated signaling of numerous innate and adaptive immune pathways, including bacterial sensing and processing pathways, and elevated proinflammatory cytokines. On a background of IL-10^−/− deficiency, severe histological injury in the SFL mucosa develops whereas the SEL tissue remains largely noninflamed (Figs. 5 and 6 and Supplemental Table S1). However, there are a number of limitations to our interpretation. First, the present study examined loops at 5 wk postsurgical placement and it is possible that WT mice would eventually develop inflammatory responses, which are more immediately robust under IL-10^−/− susceptibility. We have, however, collected WT SFL tissues at 50 days postsurgery and find little evidence of infiltrates or epithelial erosion at that time point (unpublished). IL-10^−/− animals, on the other hand, cannot tolerate SFL beyond 35 days, so the 5-wk end point was chosen in this report. Another potential confounder is that the increased relative abundance of the Proteobacteria family Enterobacteriaceae observed in IL-10^−/− SFL compared with WT SFL mice could play a role in the disease pathogenesis (4), although this family is still observed in IL-10^−/− SEL that do not have severe disease.

Since inducible TLR4 levels were observed in SFL of both WT and IL-10^−/− mice, we tested the role of TLR4 signaling in mucosal hyperplastic remodeling and inflammation using TLR4-deficient mice. The absence of morphological remodeling at the histological and transcriptional level suggests mucosal sensing of microbes through this TLR pathway drives the metaplasia transcriptional responses changes in the SFL ileum (Fig. 8, A and B, and Fig. 9). Interestingly, IL-10^−/−/TLR4^−/− mice failed to exhibit both metaplasia and inflammatory disease in SFL compared with SEL (Fig. 8, F and G). We had anticipated the possibility of inflammation since others previously described spontaneous colitis associated with adenoma-carcinoma formation in IL-10^−/−/TLR4^−/− mouse colons as early as 3–6 mo of age (7, 26). However, in our work, no colonic inflammation (thickening, opaqueness) or adenomas were visible at the time of harvest (3 mo). In the ileum of IL-10^−/−/TLR4^−/− mice, the lack of inflammation or adaptation in SFL suggests TLR4 has fundamentally important regulatory roles in driving ileal inflammation (and colonic-like metaplasia) following stasis.

The microbial communities of the SFL are also functionally different in that they proved to be colitogenic in IL-10^−/− GF mice. Further examination revealed that SFL contents induced hypertrophy of the muscle layer, increased neuronal fibers and altered gastrointestinal motility compared with SEL contents (unpublished Chang laboratory data). Additionally, the colonic mucosal expression of the morphology-related genes, Wnt-5a and SHH, were altered in a pattern similar to the murine SFL vs. SEL expression (Fig. 12).

In conclusion, this mouse model recapitulates many of the features of the pouch in UC patients following total proctocolectomy with IPAA. An important correlate is that much like the mouse SFL, the IPAA pouch in UC sometimes acquires “colon-like” characteristics, including the development of a colonic-like microbiota. This, in turn, in part via a TLR4-dependent pathway, leads to alterations in mucosal morphology that resemble colon and, indeed, transcriptional analysis supports that notion that the SFL has functionally and immunologically adapted to become colon mucosa. In genetically susceptible hosts, here represented by IL-10^−/−-deficient mice, the colonic microbial community triggers an inflammatory response in the “colon-ized” SFL. Our studies suggest that these changes are related in that the development of a colon-like phenotype occurs in both humans and mice and is a result of the stasis-induced conditions that favor the assemblage of a colonic-like microbiota. These findings provide a new paradigm for our understanding of IBD-associated pouchitis and the mechanisms responsible for disease pathogenesis in a region of intestine never affected by UC prior to total colectomy and IPAA.

GRANTS

Ths research was supported by NIDDK DK42086 (DDRCC), UH3 DK083993, Leona and Harry Helmsley Trust (SHARE), R37 DK47722, T32 DK07074, F32 DK105728, Gastrointestinal Research Foundation of Chicago, and a Peter and Carol Goldman Family Research grant.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

M.A.W., N.H.H., J.C.A., and E.B.C. conception and design of research; M.A.W., J.F.P., R.F.L., B.D.S., S.R.D., and M.W.M. performed experiments; M.A.W., J.F.P., R.F.L., Y.H., S.R.D., C.R.W., V.A.L., M.W.M., and D.A.A. analyzed data; M.A.W., V.A.L., G.C.A., L.E.R., M.L.S., N.H.H., J.C.A., and E.B.C. interpreted results of experiments; M.A.W., J.F.P., R.F.L., Y.H., B.D.S., S.R.D., C.R.W., V.A.L., M.W.M., G.C.A., M.C.R., D.T.R., L.E.R., D.A.A., M.L.S., N.H.H., J.C.A., and E.B.C. approved final version of manuscript; J.F.P., C.R.W., M.W.M., and D.A.A. prepared figures; J.F.P., M.C.R., and J.C.A. drafted manuscript; J.F.P., G.C.A., D.T.R., M.L.S., N.H.H., J.C.A., and E.B.C. edited and revised manuscript.