TNF Receptor 1 Promotes Early-Life Immunity and Protects against Colitis in Mice

SUMMARY

Neutralization of tumor necrosis factor (TNF) represents a widely used therapeutic strategy for autoimmune diseases including inflammatory bowel disease (IBD). However, the fact that many patients with IBD are non-responsive to anti-TNF therapies suggests the need for a better understanding of TNF signaling in IBD. Here, we show that co-deletion of TNF receptor 1 (TNFR1, Tnfrsf1a) in the Il10^−/− spontaneous colitis model exacerbates disease, resulting in very-early-onset inflammation after weaning. The disease can be interrupted by treatment with antibiotics. The single deletion of TNFR1 induces subclinical colonic epithelial dysfunction and mucosal immune abnormalities, including accumulation of neutrophils and depletion of B cells. During the pre-disease period (before weaning), both Tnfr1^−/− and Il10^−/− Tnfr1^−/− animals exhibit impaired expression of pro-inflammatory cytokines compared with wild-type and Il10^−/− controls, respectively. Collectively, these results demonstrate the net anti-inflammatory functions of TNF/TNFR1 signaling through the regulation of colonic immune homeostasis in early life.

Graphical Abstract

In Brief

Although anti-TNF therapies are used to treat colitis, Liu et al. demonstrate that colitis-susceptible mice deficient for TNF receptor 1 (TNFR1) paradoxically develop severe disease shortly after weaning. TNFR1 function can be traced back to its mediation of pro-inflammatory responses during a critical period of immune development in early life.

INTRODUCTION

The global incidence of inflammatory bowel disease (IBD) has increased steadily, especially in children under the age of 10 years (Benchimol et al., 2014; Kappelman et al., 2013; Schildkraut et al., 2013). A combination of host genetics, environmental exposures, intestinal mucosal dysfunction, and microbial dysbiosis contributes to IBD pathogenesis (Jostins et al., 2012). Although diverse genes and pathways regulate IBD susceptibility, their roles in immune development in early life and the biological mechanisms linking them to intestinal inflammation remain to be elucidated.

Tumor necrosis factor (TNF, formerly known as TNF-α) is a major therapeutic target in IBD. Pediatric patients with IBD have elevated circulating and intestinal levels of TNF (Breese et al., 1994; Murch et al., 1991). Recent genome-wide association studies (GWAS) have identified mutations in the TNF signaling pathway to confer risk for both ulcerative colitis and Crohn’s disease (Bank et al., 2014; Ferguson et al., 2009; Lappalainen et al., 2008; Li et al., 2016; McGovern et al., 2015; Pierik et al., 2004; Sashio et al., 2002; Waschke et al., 2005). The immediate therapeutic benefits of anti-TNF agents have suggested that TNF is a pro-inflammatory cytokine responsible for mucosal damage in IBD. However, the therapeutic effects of anti-TNF agents are short lived for many (Ben-Horin et al., 2014; Colombel et al., 2007; Ford et al., 2011; Schreiber et al., 2007). In mice, deletion of TNF or either of its transmembrane receptors, TNFR1 (p55) and TNFR2 (p75) (encoded by the Tnfrsf1a and Tnfrsf1b genes, respectively), has differing effects on disease susceptibility in established colitis models (Dubé et al., 2015; Ebach et al., 2005; Mizoguchi et al., 2008; Wang et al., 2012; Wang et al., 2013). Strikingly, loss of TNF-signaling-pathway members Tnf (Hale and Greer, 2012) or Tnfr2 (Punit et al., 2015) results in severe disease in the Il10^−/− spontaneous colitis model (Kühn et al., 1993). By ablating a critical tolerogenic signal, interleukin 10 (IL-10), the Il10^−/− model replicates key features of Crohn’s colitis, including the developmental dynamics of disease. However, the role of TNFR1 in the susceptibility to Il10^−/− colitis has not, to our knowledge, been elucidated. It is not known whether TNFR1, on balance, would mediate the beneficial or deleterious effects of TNF signaling.

Here, we report severe, very-early-onset (VEO) colitis in Il10^−/− Tnfr1^−/− mice. Paradoxically, the early disease onset was preceded by impaired expression of pro-inflammatory cytokines during the period of life before weaning. TNFR1-mediated signaling in early life is, therefore, essential for the acquisition of mucosal tolerance.

RESULTS

Very-Early-Onset Colitis in Il10^−/− Tnfr1^−/− Mice

To characterize the role of TNFR1 in an animal model of spontaneous colitis, we bred Tnfr1^−/− mice (Pfeffer et al., 1993) to Il10^−/− mice (Kühn et al., 1993) to generate double-knockout animals. Consistent with previous reports (Bristol et al., 2000; Farmer et al., 2001; Mähler et al., 2002), Bl/6 Il10^−/− mice are relatively resistant to colitis. Il10^−/− Tnfr1^−/− mice, however, developed severe, spontaneous, early-onset colitis, with a mixed mucosal infiltrate, cryptitis, abscesses, and epithelial hyperplasia. Although Il10^−/− Tnfr1^−/− animals at 2 (n = 8; Figures 1A and 1B) and 3 (n = 4, not shown) weeks old were spared colitis, whereas, at 4 weeks old, 8/13 (62%) of the Il10^−/− Tnfr1^−/− mice had histologic evidence of colitis (Figures 1C and 1D) (median histologic disease score: 7/15). At 6 weeks old, 4/7 (57%) of the Il10^−/− Tnfr1^−/− mice were colitic, and by 8 weeks old, 11/12 (92%) of the Il10^−/− Tnfr1^−/− mice had colitis (median histologic disease score: 8/15). Littermate Il10^−/− mice had minimal disease (Figure 1E); moreover, heterozygous Il10^−/− Tnfr1^+/− littermates were spared severe disease and early mortality. Only 50% of Il10^−/− Tnfr1^−/− mice survived to 12 weeks old (Figures 1F, S1A, and S1B); their colons demonstrated increased RNA expression of cytokines, including Tnf (Figure 1G, from the NanoString assay) and increased crypt cell proliferation (phosphorylated histone H3 [pH-H3⁺] cells; Figure 1H) and apoptosis (TUNEL⁺ cells; Figure 1I). The disease affected the entire colon and cecum, resulting in serrated adenocarcinoma (Figure S1C), and retarded growth in both males and females (Figure S1D). This disease is, therefore, different from colitis in Il10^−/− Tnfr2^−/− mice, in which the cecum was not involved (Punit et al., 2015).

Figure 1. Il10^−/− Tnfr1^−/− Mice Develop Early-Onset, Severe Colitis at 4 Weeks Old.

(A and B) Hematoxylin-and-eosin (H&E)-stained sections of pre-weaning 2-wk-old Il10^−/− (A) (n = 7) and Il10^−/− Tnfr1^−/− (B) (n = 8) mice show normal distal colon.

(C and D) H&E sections of 4-week-old post-weaning mice reveal normal colon in the Il10^−/− genotype (C) (n = 6) but colitis in the Il10^−/− Tnfr1^−/− genotype (D) (n = 13).

(E) Histological scoring of colitis severity (worst disease score is 15). Horizontal lines indicate median.

(F) Kaplan-Meier survival curve indicates early mortality in Il10^−/− Tnfr1^−/− mice.

(G) Nanostring analysis demonstrates upregulation of Tnf expression in Il10^−/− Tnfr1^−/− animals.

(H and I) Crypts in 12-wk-old Il10^−/− Tnfr1^−/− mice exhibited increased numbers of proliferating (pH-H3⁺) (H) and apoptotic (TUNEL⁺) (I) cells.

(J and K) Flow cytometric analysis of colonic mucosal scrapings demonstrates increased colonic mucosal infiltration of CD45⁺ hematopoietic cells (J) with increased immune cell subsets (K) in 8-wk-old Il10^−/− Tnfr1^−/− mice.

Scale bars: (A)–(D) 100 μm. *p < 0.05. See also Figures S1, S2, and S3. Error bars: standard error.

Because TNF signaling has integral roles in immune cell activation and specialization, we hypothesized that the loss of TNFR1 altered the relative proportions of the immune cells in colitis. We analyzed distal colonic mucosal scrapings by flow cytometry with a panel of 13 immune-cell-targeted antibodies (Figure S2) (Yu et al., 2016). Adult (8-week-old) Il10^−/− Tnfr1^−/− specimens showed elevated numbers of CD45⁺ (hematopoietic origin) cells compared with age- and litter-matched Il10^−/− mice (Figure 1J), suggestive of increased immune cell infiltration into the mucosa. Within the CD45⁺ cell population, the relative proportions of neutrophils and activated CD4⁺ T cells were increased compared with those of the controls. However, several cell types showed a proportional reduction in the Il10^−/− Tnfr1^−/−animals: B cells, intraepithelial lymphocytes, dendritic cells, and macrophages (Figure 1K). Thus, the loss of TNFR1 results in a fundamental imbalance of immune cell subtypes in the context of inflammation.

To determine whether TNFR1 expression protected from colitis in a general setting of IL-10 signaling inhibition, we assessed the susceptibility of Tnfr1^−/− mice to colitis induced by repetitive administration of anti-IL-10 receptor-targeted antibodies (Figure S3A) (Carvalho et al., 2012; Kullberg et al., 2006; Singh et al., 2016). Antibody-injected Tnfr1^−/− (knockout) mice exhibited crypt loss and mucosal immune infiltration, hallmarks of colitis, whereas antibody-injected Tnfr1^+/− (heterozygous) mice were relatively protected from colitis (Figure S3B), as assessed histologically (Figure S3C). Thus, TNFR1 restricts colitis associated with the loss of IL-10 signaling.

Antibiotics Induce Remission of Colitis

Because IBD is thought to represent an abnormal immune reaction against gastrointestinal commensals, we tested whether antibiotics would alter the severe colonic injury observed in Il10^−/− Tnfr1^−/− mice. We treated 8-week-old Il10^−/− Tnfr1^−/− mice with neomycin and metronidazole for 2 weeks. Distal colonoscopy performed before and immediately after treatment showed resolution of the disease in the antibiotic-treated group (Figure 2A) compared with the vehicle-treated controls (p = 0.0043; Figure 2B). The difference in outcomes was apparent histologically (p = 0.0042; Figures 2C and 2D) and was associated with reduced numbers of pH-H3⁺ (proliferative) and cleaved caspase-3⁺ (apoptotic) cells (Figure 2E). In a mixed cohort of adult Il10^−/− (n = 7) and Il10^−/− Tnfr1^−/− (n = 2) mice, we found that antibiotic treatment reduced fecal lipocalin, a marker of neutrophil activity, to baseline uninflamed levels (<50 pg/mg stool) within the first week of treatment (Figure 2F). Lipocalin levels rebounded within 1 week after antibiotic withdrawal. We noted diarrhea in Il10^−/− Tnfr1^−/− mice, which precluded the comparison of absolute lipocalin levels between genotypes. Taken together, these results indicate that disease continuation in Il10^−/− Tnfr1^−/− mice requires neomycin/metronidazole-sensitive commensals and that colonic epithelial and neutrophil abnormalities in these mice result from host-microbe interactions.

Figure 2. Broad-Spectrum Antibiotic Treatment Improves Colitis in Il10^−/− Tnfr1^−/− Mice.

(A–E) Improved endoscopic appearance (A), endoscopic score (B), histology (C and D), and crypt cell dynamics (E) were observed in 8-week-old antibiotic-treated Il10^−/− Tnfr1^−/− mice (n = 3) compared with controls (n = 3).

(F) Fecal lipocalin levels diminished to nearly undetectable levels during antibiotic treatment of Il10^−/− (n = 7) and Il10^−/− Tnfr1^−/− (n = 2) mice but quickly rebounded.

(G) 16S analysis of Il10^−/− Tnfr1^−/− and Il10^−/− cecal microbiomes collected from 2-, 4-, 6-, and 8-week-old Il10^−/− Tnfr1^−/− and Il10^−/− mice. Shown are weighted UniFrac distances plotted using multidimensional scaling. Each sample is coded by age (gray facets), genotype (color), and colitis score (size). At 8 weeks old, the genotypes could be qualitatively discriminated (dotted black line). The displayed p value for each age is computed by permuted analysis of variance using littermates, colitis severity, sex, and genotype conditioned on dam identity as model factors.

(H) Bar graph displays the total number of significant operational taxonomic units (OTUs) that are differentially abundant in Il10^−/− versus Il10^−/− Tnfr1^−/− animals by age.

(I) OTUs with taxonomic classifications plotted to indicate their relative abundance in Il10^−/− Tnfr1^−/− versus Il10^−/− samples. The transparency of the bars denotes the total abundance of the OTUs on a log scale (more transparent is less abundant).

(J) Depiction of the multigenerational experiment analyzing the effects of maternal antibiotic treatment on colitis development, with associated colitis scoring shown in the inset.

Scale bars: 200 μm. Error bars: standard error.

We next assessed whether the very-early-onset colitis observed in Il10^−/− Tnfr1^−/− mice was due to a distinct microbial ecosystem in early life. Using 16S sequencing, we profiled the luminal bacterial composition from the previously presented (Figure 1) cohort of littermate Il10^−/− and Il10^−/− Tnfr1^−/− mice at different ages. Analysis of variance (permutational multivariate analysis of variance [PERMANOVA]) showed that microbial composition was significantly associated with age (p = 0.001) and dam (p = 0.01) but not genotype (p = 0.22) (Figure 2G). The abundances of few operational taxonomic units (OTUs) were significantly altered at 2 weeks old, but alterations increased after onset of colitis at 4 weeks old (Figure 2H). We note that at 2 weeks old, low-abundance Coriobacteriaceae were depleted in Il10^−/− Tnfr1^−/− samples (Figure 2I), consistent with this taxon’s reduced representation in pediatric IBD (Maukonen et al., 2015). At 8 weeks old, a qualitative difference in microbiome composition was observed (Figure 2G), but that difference was likely secondary to colitis. Colitic mice were depleted of Lachnospiraceae, a known producer of butyrate (Figure 2I) (Geirnaert et al., 2017; Surana and Kasper, 2017; Vital et al., 2014). Thus, a specific commensal signature did not cause exacerbated disease.

We next tested whether treatment of mice with antibiotics during the perinatal period would affect the onset of disease. Pregnant Il10^−/− Tnfr1^+/− dams were exposed to neomycin and metronidazole in their final week of pregnancy and up to 7 d postpartum. After treatment, F1 pups of all genotypes (Il10^−/−, Il10^−/− Tnfr1^+/−, and Il10^−/− Tnfr1^−/−) remained free of disease to at least 16 weeks old (Figure 2J). Heterozygous Il10^−/− Tnfr1^+/− pups were subsequently interbred, and their F2 pups of the Il10^−/− Tnfr1^−/− genotype exhibited colitis at 16 weeks old. Thus, the early onset of colitis in Il10^−/− Tnfr1^−/− mice can be interrupted by maternal treatment with antibiotics.

Colonic Mucosal Dysfunction in Tnfr1^−/− Mice

To determine the specific contribution of TNFR1 to the colitic disease process, we analyzed the physiological effects of TNFR1 genetic loss in isolation, using Il10^+/+ Tnfr1^−/− mice (or simply Tnfr1^−/− mice). The fecal lipocalin levels of Tnfr1^−/− mice at 12 weeks old were all <50 pg/mg stool (n = 7), suggesting that these mice did not have overt colitis. However, we found crypt fission, dropout, and branching (Figure 3A) in foci in 3/7 (43%) Tnfr1^−/− mice. Moreover, at 4, 8, and 16 weeks old, Tnfr1^−/− mice had significantly increased serum recovery of enemaadministered fluorescein isothiocyanate (FITC)-dextran (Figure 3B). Tnfr1^−/− colonic crypts at 4 and 12 weeks old exhibited modestly elevated numbers of proliferating pH-H3⁺ cells (Figure 3C), and crypt abnormalities could also be identified in 4-week-old mice (Figure 3D; in 4/8 mice). At 12 weeks old, Tnfr1^−/− mice had significantly (p = 0.03) increased numbers of DNA-damaged phosphorylated histone 2A.X (pH2A.X)-positive epithelial cells and phosphorylated STAT3-positive epithelial cells (Figure 3E). In contrast to Il10^−/− Tnfr1^−/− mice, intra-crypt staining for cleaved caspase-3 apoptotic bodies was not observed in Tnfr1^−/− mice (not shown). TNFR1, therefore, regulates colonic epithelial function (e.g., morphology, permeability, and DNA-damage signaling) and repair (e.g., proliferation and STAT signaling) from a young age.

Figure 3. Adult Tnfr1^−/− Mice Exhibit Epithelial Dysfunction and Alterations in Colonic Mucosal Immune Cell Representation.

(A) Image montages show normal crypt structures in wild-type mice (n = 8) and focal abnormalities in 8-wk-old Tnfr1^−/− mice (n = 7).

(B) Tnfr1^−/− mice at 4, 8, and 12 weeks old exhibited increased colonic permeability at 30 min after rectal instillation of 4 kDa FITC-dextran. The graph is representative of three different experiments.

(C) Trending increase in pH-H3⁺ proliferative cells in knockout mice at 4 and 12 weeks old.

(D) H&E images of the colonic mucosa at 4 weeks old demonstrate foci of altered crypt patterning in Tnfr1^−/− mice.

(E) Adult 12-week-old Tnfr1^−/− colonic epithelium also had increased numbers of pH2A.X⁺ and pSTAT3⁺ cells.

(F and G) Flow cytometry reveals a trend toward decreased hematopoietic cell numbers (F) in the colonic mucosa of 8-week-old Tnfr1^−/− mice. This was driven by increased neutrophils and a near-total loss of B cells (G).

(H) Flow cytometric assessment of antibiotic-treated Tnfr1^−/− mice reveals reduced numbers of mucosal immune cells.

(I) Changes to pH-H3 staining after antibiotic treatment were more modest and insignificant.

(J) Qualitative reductions (arrowheads) in the density of inter-crypt cells were noted in H&E-stained sections of antibiotic-treated samples.

Scale bars: (A) 100 μm; (D) and (J) 50 μm, *p < 0.05. Error bars: standard error.

We also found significant immune defects through flow cytometry. There was a trending reduction (p = 0.07) in the total number of mucosal immune cells in 8-week-old Tnfr1^−/− mice (Figure 3F). Although the proportions of T cells were similar between Tnfr1^−/− and wild-type mice, the Tnfr1^−/− mice showed an ~90% loss of B cells, consistent with prior reports of the TNFR1 role in the establishment of germinal centers (Le Hir et al., 1996; Matsumoto et al., 1996; Pasparakis et al., 1996). Knockout mice showed a higher percentage of neutrophils (Figure 3G). Treatment of adult Tnfr1^−/− mice with neomycin and metronidazole antibiotics for 2 weeks reduced the total representation of CD45⁺ hematopoietic cells, neutrophils (CD45⁺ Ly6G⁺), and macrophages (CD45⁺ SSC^hi CD11b/c⁺ IA/IE⁺ CD24⁻) (Figure 3H). However, antibiotic-driven reductions in colonic epithelial pH-H3 staining did not reach significance (Figure 3I). A reduced inter-crypt cellularity was histologically apparent in antibiotic-treated animals (Figure 3J). Thus, live commensals contribute to the mucosal defects associated with the loss of TNFR1.

Loss of TNFR1 Reduces Early-Life Cytokine Expression

The very-early onset of disease in Il10^−/− Tnfr1^−/− mice occurs shortly after the completion of the weaning process. We ordinarily separate pups from the dam at 3 weeks of age. Delaying that separation by 1 week (from 3 weeks old to 4 weeks old) did not prevent disease onset at 4 weeks (data not shown). Thus, the cause of disease cannot be reduced to this single ‘‘trigger’’ of maternal separation.

We next asked whether the effects of TNFR1 loss could be discerned in the pre-disease, early-life period. By high-throughput expression profiling, we examined the full-thickness colons of wild-type, Tnfr1^−/−, Il10^−/−, and Il10^−/− Tnfr1^−/− mice at ages corresponding to the pre-colitic (2 weeks) and colitic (8 weeks) states. In RNA sequencing (RNA-seq) studies comparing Il10^−/− Tnfr1^−/− and Il10^−/− mice at 2 weeks old, we observed a profound downregulation of classical inflammatory markers, including Nos2, Cxcl1, Saa1, and calprotectin (S100a8 and S100a9), in the double-knockout mice (Figure 4A). We analyzed those cytokines as a group, the ‘‘TNFR1-associated early-life immune module,’’ and calculated a score (−1.7 ± 0.35, their mean logarithmic expression ratio between TNFR1-deficient and TNFR1-expressing conditions), which quantified their overall change (Figure 4B). That score was significantly less than zero (p = 3 × 10⁻⁵), supporting the overall reduced expression of those cytokines in Il10^−/− Tnfr1^−/− mice at 2 weeks of age. At 8 weeks old, these immune markers had a positive (p = 0.003) expression score of 3.4 ± 1.0 in Il10^−/− Tnfr1^−/− (versus age-matched Il10^−/−) animals, consistent with their broad upregulation in colitis. This pattern of changes was preserved in Tnfr1^−/− mice that had genetically functional IL-10, although the magnitude of the changes was less. Adult Tnfr1^−/− mice had elevated expression of neutrophil-associated genes, such as the calprotectin subunits (S100a8, S100a9) and Duox2. In contrast, pre-weaning Tnfr1^−/− mice at 2 weeks old exhibited reduced transcription of Il1b (IL-1b) (Figure 4A). qPCR studies showed that adult heterozygous Tnfr1^+/− mice had similar calprotectin expression as wild-type littermates (Figure S4A).

Figure 4. TNFR1-Deficient Animals Show Early-Life Defects in Colonic Immune Response.

(A) Heatmaps summarize cytokine-expression levels obtained from RNA-seq/NanoString analysis of colons from wild-type (WT), Tnfr1^−/−, Il10^−/−, and Il10^−/− Tnfr1^−/− animals at 2 weeks (pre-disease) and 8 weeks (post-disease onset) old. Overall cytokine expression was decreased in TNFR1-deficient animals at 2 weeks of age but increased in those animals at 8 weeks old. Note that reads for Il10 were still detected in Il10^−/− animals because of continued production of mRNA encoding of an early-termination codon.

(B) The boxplot shows a summarized score of cytokine expression changes across comparisons of genotypes and ages. The lower and upper hinges represent the 25th and 75th percentiles. Whiskers denote the range of the data, excepting outlying points beyond 150% of the interquartile range. All values were obtained from RNA-seq except for the comparison of Il10^−/− Tnfr1^−/− versus Il10^−/− mice at 8 weeks old (NanoString). Significance of scores was tested using the t test against 0 (no change).

(C) Il10^−/− Tnfr1^−/− animals at 2 weeks old show reduced pH-H3⁺ cell counts per crypt.

(D and E) Flow cytometric analysis of whole colons collected from 2-week-old mice shows overall similarity of colonic hematopoietic (CD45⁺) cell counts (D) but loss of B cells in TNFR1-deficient animals (E).

Significance was evaluated using ANOVA. *p < 0.05. See also Figure S4. Error bars: standard error.

Cytokine changes were associated with functional defects. There was more epithelial pH-H3 (proliferative) staining in 2-week-old Il10^−/− colons versus Il10^−/− Tnfr1^−/− colons (Figure 4C), in marked contrast to results in 8-week-old mice (Figure 1H). On flow-cytometric analysis, we found similar proportions of CD45⁺ cells (Figure 4D) and neutrophils (Figure 4E) in 2-week-old animals across all genotypes. However, reduction in B cell numbers in TNFR1-deficient animals was already evident at 2 weeks of age (Figure 4E). Thus, loss of TNFR1 results in profound immune and epithelial defects in early life.

We ruled out that transcriptional changes in Tnfr1^−/− mice were compensated for by TNFR2. We compared transcriptomic profiles of adult Tnfr1^−/− mice and Tnfr2^−/− mice (Punit et al., 2015). As shown in Figure S4B, there was little overlap (4%) of differentially expressed transcripts in Tnfr1^−/− and Tnfr2^−/− colons. Among the few co-regulated transcripts, there was a significant negative correlation in their direction of regulation (Figure S4C). The loss of Tnfr1 did not affect expression of Tnfr2 and vice versa. Notably, Saa2 and Duox2, associated with abnormal mucosal immune responses to microbiota (Eckhardt et al., 2010; Grasberger et al., 2015), were both upregulated in the absence of either TNFR1 or TNFR2.

DISCUSSION

Here, we show that the genetic loss of TNFR1 results in increased susceptibility to colitis. The early age of disease onset in Il10^−/− Tnfr1^−/− mice, after weaning, may be functionally equivalent to infantile or toddler colitis in humans (assuming full weaning at 1–2 years of age). In rodents, at that age, the colonic mucosa undergoes histological maturation (Walthall et al., 2005). During the simultaneous immune ‘‘weaning reaction,’’ pro-inflammatory signals act against intestinal luminal contents to ultimately promote tolerance through the establishment of immunosuppressive cell populations (Al Nabhani et al., 2019; Olszak et al., 2012; Pié et al., 2004; Redhu et al., 2017; Scheer et al., 2017). The results of our study can be interpreted within that framework and demonstrate that TNFR1-mediated pro-inflammatory signaling, including upregulation of Il1b, in early life is essential for the weaning reaction (Pié et al., 2004). These results demonstrate how immunodeficiency during early life can predispose animals toward later autoimmunity, as hinted at by studies of other genetic immunodeficiencies (Glocker and Grimbacher, 2012; Mombaerts et al., 1993; Sadlack et al., 1993; Salzer et al., 2014; Tegtmeyer et al., 2017).

The exacerbation of colitis with TNFR1 loss in mice could partially model the uneven clinical efficacy of anti-TNF antibodies in IBD (Gratz et al., 2002; Kullberg et al., 2001; Scheinin et al., 2003), increased risk for human IBD associated with genetic perturbation of TNF receptors, and new-onset IBD in autoimmune patients treated with anti-TNF (Korzenik et al., 2019; Toussirot et al., 2012; Üsküdar Cansu et al., 2019). Relevant to this study, very-early-onset IBD is frequently linked to the loss of the IL-10 receptor (Bianco et al., 2015) (Begue et al., 2011; Beser et al., 2015; Kotlarz et al., 2012; Moran et al., 2013; Pigneur et al., 2013). Elucidating key functions in TNF receptor signaling in intestinal development and inflammation may have therapeutic benefits.

Limitations of Study

First, it is premature to generalize our findings directly to human anti-TNF therapies, which may work through distinct mechanisms (e.g., Atreya et al., 2011). Second, we have had only limited success in pharmacological manipulation of the weaning reaction in colitis. Third, we do not know whether immune cell alterations and epithelial dysfunction in Tnfr1^−/− animals are caused by one or the other. Future studies will need to elaborate on these aspects of TNF signaling in colitis.

STAR★METHODS

RESOURCE AVAILABILITY

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Brent Polk (dbpolk@chla.usc.edu).

Materials Availability

This study did not generate new unique reagents.

Data and Code Availability

The RNA-seq datasets generated during this study are available at NCBI Gene Expression Omnibus (accession numbers): GSE107933, GSE155654, GSE155626).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Mice

Mice were maintained humanely, in accordance with the rules and regulations of the Institutional Animal Care and Use Committee (IACUC) of Children’s Hospital Los Angeles (CHLA). This study was approved by the CHLA IACUC under the internal protocol number 288. Mice were anesthetized with isoflurane and euthanized via cervical dislocation prior to dissection.

Wild-type (WT), Il10^−/− (stock #002251), and Tnfr1^−/−(stock #002818) mice on the C57BL/6J background were obtained from Jackson Laboratory. Il10^−/− Tnfr1^−/− mice and control Il10^−/− littermates were generated from Il10^−/− Tnfr1^+/− parents. Experimental and control groups of mice were co-housed for > 4 weeks prior to experimentation, and co-housing was maintained during experimentation. Both male and female mice were included in the study. Details regarding the ages of mice used in experiments are given in the description of Results.

METHOD DETAILS

Tissue collection

Colons and ceca were carefully freed from the abdominal cavity of euthanized mice, opened longitudinally, and cleaned of fecal contents. Luminal contents of the ceca were collected for bacterial 16S rRNA sequencing. Colons were opened and splayed; a thin longitudinal strip representing the full proximal-distal length of each colon was removed and preserved for RNA isolation. The remaining tissue was fixed overnight in 10% neutral buffered formalin at room temperature. For histological assessment, colons were washed with 50% ethanol, dehydrated in an ascending ethanol series, cleared with xylenes, and embedded in molten paraffin. Thin (5 μm) sections were prepared on a microtome and stored until their usage in staining procedures.

Histochemistry

Blinded to sample origin, assessment of colitis severity was made on hematoxylin and eosin-stained sections of murine colon by MKW using previously validated and published methodologies (Dubé et al., 2012; Zhang et al., 2012). Histology was scored based on 5 categories: enterocyte loss, crypt inflammation, lamina propria mononuclear cells, polymorphonuclear cells, and hyperplasia. Each category is awarded a score from 0 to 3, with higher score indicating more severe disease. A maximum score of 15 is assigned. For quantification of epithelial cell markers, label-positive cells in 30–100 crypts adjacent to the anal verge were counted and divided by the total number of counted crypts to yield a per-crypt number.

Immunohistochemistry was performed using standard protocols. Primary antibodies were incubated at 4^○C overnight. Primary antibodies used were rabbit anti-pH2A.X (Ser139) (Cell Signaling Technology #9718S, 1:2,000 dilution), rabbit anti-pHH3 (Ser10) (Cell Signaling Technology #9701S, 1:500 dilution), rabbit anti-pSTAT3 (Tyr705) (Cell Signaling Technology #9145, 1:100 dilution), and rabbit anti-cCSP3 (Cell Signaling Technology #9661S, 1:200 dilution).

Flow Cytometry

Colons were freshly dissected from mice. If the colon was isolated from an adult (8-wk-old) animal, the mucosal layer was separated from the muscle layer using fine forceps. The mucosal piece was minced into 2-mm² pieces. If the colon was isolated from an unweaned (2-wk-old) mouse, the full-thickness organ was immediately minced into 2-mm² pieces.

Tissue was incubated with prewarmed digestion solution (0.2 Wunsch units/ml Liberase TM + 200 Kuntz units/ml DNase I in DMEM/F12 + 15 mM HEPES) for 30 min at 37^○C with continuous agitation at 180 rpm. After trituration of the tissue and its passage through a 70-μm-pore cell strainer, the tissue was washed with DMEM:F12 supplemented with 10% FBS, and then also with HEPES-buffered saline (Liu et al., 2012) supplemented with 0.5% BSA. All subsequent washes and stainings were performed in HBS+0.5% BSA. The tissue was stained with 0.1 μM DAPI, blocked for 15 min at 4^○C with a solution composed of 5% mouse/rat serum supplemented with mouse Fc block (anti-CD16/32 antibody, Biolegend ‘‘truStain fx’’), and probed with a mixture of preconjugated antibodies for 30 min at 4^○C. Antibodies (working dilutions) were targeted to CD45 (1:250), Ly6G (1:200), CD11b (1:200), CD11c (1:100), IA/IE (1:1,000), CD64 (1:400), CD24 (1:500), Ly6C (1:200), CD69 (1:200), CD8 (1:100), CD4 (1:200), TCRd (1:200), and CD3 (1:200). After washes, cells were analyzed on a BD LSR II. Compensation was adjusted using references obtained by the analysis of antibodies bound to Ultracomp eBeads (ThermoFisher Scientific).

Analysis of flow cytometric data was performed using FlowJo. Gates were adjusted for each experimental day but were consistent between experimental and control samples. Data were pooled across 3–4 experimental days per genotype/age comparison.

Collection and analysis of stool

Stool pellets were collected from mice by manual restraint which typically induced defecation within 30 s. A clean Eppendorf tube was held beneath the anal opening. Pellets were frozen at −80^○C and analyzed for fecal lipocalin-2 levels using a commercial ELISA kit and according to the manufacturer’s instructions (R&D Systems, cat# DY1857).

Antibiotic treatment

Il10^−/− Tnfr1^−/− mice at 8 weeks of age were assigned to either ‘Vehicle’ or ‘Antibiotics’ treatment group (n = 3 in each group). Mice received either water (‘Vehicle’) or broad-spectrum antibiotics for 2 weeks (Neomycin sulfate 500mg tablets purchased from Hi-Tech Pharmacal, Metronidazole 500mg/100ml purchased from Claris Lifescience) ad libitum. Both antibiotics were mixed with drinking water at 1 g/L. Distal colonoscopy on deeply anesthetized mice was performed pre- and post-treatment using a flexible instrument (Karl Storz). Mice were euthanized post-treatment at 10 weeks of age, and colon histology was scored as previously described (Dubé et al., 2012; Punit et al., 2015) to determine colitis severity.

16S sequencing

16S rRNA libraries were prepared from flash-frozen cecal contents. Cecal content was collected from co-housed littermates of Il10^−/− and Il10^−/− Tnfr1^−/− mice euthanized at 2, 4, 6, and 8 weeks. Cecal content was frozen immediately after collection and stored at −80^○C. Cecal microbiome composition was determined by sequencing the 16S ribosomal RNA gene. Sequencing was performed on the Illumina MiSeq platform (Diversity assay 2x300bp 20K bTEFAP® at MR DNA lab (Shallowater, TX USA)).

Expression analysis of immune-related transcripts

RNA was extracted from colon tissues using standard RNA extraction techniques (PureLink® RNA Mini Kit from Ambion) and assayed for mRNA expression of targeted immune genes using the Nanostring technology (nCounter Mouse Immunology Panel, 561 genes).

RNA-Seq

mRNA transcripts were purified from total colonic RNA using oligo-dT coated beads, sheared, and prepared for 2x75 bp sequencing on a HiSeq 4000 (Illumina). Library preparation and sequencing were performed by SeqMatic, LLC (Fremont, CA USA).

Barrier function assay

WT and Tnfr1^−/− mice were co-housed for minimum of 1-week prior to experiment. Mice were sedated and a 0.2 mL volume of fluorescein (FITC)-dextran (MW 3000, Invitrogen, 50 mg/ml solution in PBS) was instilled into the rectum with a 3.5 Fr catheter. Blood was obtained 30 minutes later through a retro-orbital blood draw. Plasma fluorescence was measured in a plate reader (excitation 490nm; emission 520nm).

Anti-IL10R model of colitis

Co-housed Tnfr1^+/− and Tnfr1^−/− mice were intraperitoneally injected, on alternating sides of body, with 1 mg/mouse/dose of anti-IL10R antibody once per week, for a total of 3 injections. Injections were begun when mice were 3 wks old. Euthanasia and postmortem analyses were performed at 6 wks of age.

QUANTIFICATION AND STATISTICAL ANALYSIS

Gene abundance estimates

Transcript quantification and significance testing from RNA-Seq data were performed using the paired-end read option on kallisto/sleuth (Bray et al., 2016). Differential expression of expression data obtained using the Nanostring platform was computed using nSolver version 3.0 software (Nanostring).

Operational taxonomic units (OTUs) were generated from 16S sequences by clustering at 3% divergence (97% similarity) in mothur (Schloss et al., 2009). Final OTUs were taxonomically classified against a curated database derived from the Ribosomal Database Project (Cole et al., 2014) and NCBI. Additional analyses were performed using the ‘phyloseq’ (v1.19.1) (McMurdie and Holmes, 2013) and ‘vegan’ (v2.4–2) (Dixon, 2003) packages in R (version 3.3). Random forest regression was used to predict colitis score with OTU relative abundances as covariates (‘randomForest’ package v4.6). Only OTUs with > = 1% relative abundance in at least 5 samples were used. Linear discriminant analysis was performed using LEfSe (Segata et al., 2011).

Statistical tests

Data are presented as individual points with summary statistics reporting the mean and standard error of the mean. Significance was evaluated using the t test, unless otherwise noted.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
FITC rat anti-CD45	Biolegend	103107; RRID: AB_312972
Alexa Fluor 700 rat anti-Ly6G	BD Biosciences	561236; RRID: AB_10611860
APC-Cy7 rat anti-CD11b	BD Biosciences	561039; RRID: AB_2033993
APC hamster anti-CD11c	BD Biosciences	561119; RRID: AB_10562405
PE rat anti-IA/IE	BD Biosciences	562010; RRID: AB_10893194
PE-Cy7 mouse anti-CD64	Biolegend	139313; RRID: AB_2563903
BV421 rat anti-CD24	BD Biosciences	562563; RRID: AB_2737002
PerCP-Cy5.5 rat anti-Ly6C	ThermoFisher Scientific	45-5932-80; RRID: AB_2723342
BV786 hamster anti-CD69	BD Biosciences	564683; RRID: AB_2738890
BV510 rat anti-CD8a	Biolegend	100751; RRID: AB_2561389
BV605 rat anti-CD4	Biolegend	100451; RRID: AB_2564591
PE-CF594 hamster anti-gdTCR	BD Biosciences	563532; RRID: AB_2661844
BUV395 hamster anti-CD3ε	BD Biosciences	563565; RRID: AB_2738278
Rabbit anti-pH2A.X (Ser139)	Cell Signaling Technology	9718S; RRID: AB_2118009
Rabbit anti-pH-H3 (Ser10)	Cell Signaling Technology	9701S; RRID: AB_331535
Rabbit anti-pSTAT3 (Tyr705)	Cell Signaling Technology	9145; RRID: AB_2491009
Rabbit anti-cleaved CSP3	Cell Signaling Technology	9661S; RRID: AB_2341188
truStain fcX (rat anti-CD16/32)	Biolegend	101320; RRID: AB_1574975
Rat anti-mouse IL-10R (CD210), 1B1.3A	Bio X Cell	BE0050; RRID: AB_1107611
Chemicals, Peptides, and Recombinant Proteins
Neomycin sulfate – 500 mg tablets	Hi-Tech Pharmacal	50383-565-10
Metronidazole – 500 mg/100 mL	Claris Lifesciences	36000-001-24
4 kDa FITC-dextran	ThermoFisher Scientific	D3305
Liberase TM	Sigma-Aldrich	5401119001
DNase I	Sigma-Aldrich	D5025
Ultracomp eBeads	ThermoFisher Scientific	01-2222-41
Critical Commercial Assays
16S library preparation	MR DNA (Molecular Research)	bTEFAP
mRNA-Seq library preparation (Illumina)	SeqMatic	20020594
nCounter Mouse Immunology Panel	Nanostring	XT-CSO-MIM1-12
Mouse Lipocalin-2/NGAL DuoSet ELISA	R&D Systems	DY1857
Deposited Data
RNA-Seq (Tnfr1−/− versus WT, 8 wks of age)	Gene Expression Omnibus (GEO)	GSE107933
RNA-Seq (Tnfr1−/− versus WT, 2 wks of age)	Gene Expression Omnibus (GEO)	GSE155654
RNA-Seq (Il10−/− Tnfr1−/− versus Il10−/−, 2 wks of age)	Gene Expression Omnibus (GEO)	GSE155626
Experimental Models: Organisms/Strains
Mice: C57Bl6/J	Jackson Labs	000664
Mice: Il10−/−	Jackson Labs	002251
Mice: Tnfr1−/−	Jackson Labs	002818
Software and Algorithms
Kallisto/sleuth v0.44.0	Pachter Lab	https://pachterlab.github.io/kallisto/
nSolver v3.0	Nanostring	https://www.nanostring.com/products/analysis-software/nsolver
mothur v1.40.1	Schloss Lab	https://mothur.org/
FlowJo v10.4	FlowJo	https://www.flowjo.com/

Highlights.

• Il10^−/− Tnfr1^−/− mice exhibit severe colitis beginning shortly after weaning

• Colitis is dependent on the microbiome but is not taxa specific

• Tnfr1^−/− mice exhibit colonic immune dysregulation and abnormal epithelium

• Tnfr1^−/− mice have reduced cytokine expression during a critical weaning period