Microbial signals, MyD88, and lymphotoxin drive TNF-independent intestinal epithelial tissue damage

Anti-TNF antibodies are effective for treating patients with inflammatory bowel disease (IBD), but many patients fail to respond to anti-TNF therapy, highlighting the importance of TNF-independent disease. We previously demonstrated that acute deletion of 2 IBD susceptibility genes, A20 (Tnfaip3) and Abin-1 (Tnip1), in intestinal epithelial cells (IECs) sensitized mice to both TNF-dependent and TNF-independent death. Here we show that TNF-independent IEC death after A20 and Abin-1 deletion was rescued by germ-free derivation or deletion of MyD88, while deletion of Trif provided only partial protection. Combined deletion of Ripk3 and Casp8, which inhibits both apoptotic and necroptotic death, completely protected against death after acute deletion of A20 and Abin-1 in IECs. A20- and Abin-1–deficient IECs were sensitized to TNF-independent, TNFR1-mediated death in response to lymphotoxin α (LTα) homotrimers. Blockade of LTα in vivo reduced weight loss and improved survival when combined with partial deletion of MyD88. Biopsies of inflamed colon mucosa from patients with IBD exhibited increased LTA and IL1B expression, including a subset of patients with active colitis on anti-TNF therapy. These data show that microbial signals, MyD88, and LTα all contribute to TNF-independent intestinal injury.

We previously demonstrated that IEC-specific deletion of either A20 or Abin-1 alone does not lead to overt weight loss or intestinal injury, but acute simultaneous deletion of both A20 and Abin-1 leads to spontaneous IEC apoptosis, fulminant enterocolitis, and rapid mouse lethality (9). In this setting, A20 and Abin-1 cooperatively restrict both TNF-dependent and TNF-independent IEC death. TNF-independent IEC death is substantially less well Anti-TNF antibodies are effective for treating patients with inflammatory bowel disease (IBD), but many patients fail to respond to anti-TNF therapy, highlighting the importance of TNF-independent disease. We previously demonstrated that acute deletion of 2 IBD susceptibility genes, A20 (Tnfaip3) and Abin-1 (Tnip1), in intestinal epithelial cells (IECs) sensitized mice to both TNF-dependent and TNF-independent death. Here we show that TNF-independent IEC death after A20 and Abin-1 deletion was rescued by germ-free derivation or deletion of MyD88, while deletion of Trif provided only partial protection. Combined deletion of Ripk3 and Casp8, which inhibits both apoptotic and necroptotic death, completely protected against death after acute deletion of A20 and Abin-1 in IECs. A20-and Abin-1-deficient IECs were sensitized to TNF-independent, TNFR1-mediated death in response to lymphotoxin α (LTα) homotrimers. Blockade of LTα in vivo reduced weight loss and improved survival when combined with partial deletion of MyD88. Biopsies of inflamed colon mucosa from patients with IBD exhibited increased LTA and IL1B expression, including a subset of patients with active colitis on anti-TNF therapy. These data show that microbial signals, MyD88, and LTα all contribute to TNF-independent intestinal injury.
Microbial signals, MyD88, and lymphotoxin drive TNF-independent intestinal epithelial tissue damage To better understand the pathways leading to TNF-independent IEC injury we performed in vivo and in vitro analysis of IECs after acute simultaneous deletion of A20 and Abin-1.
TRIF contains a RIP homotypic interaction motif (RHIM) domain and interacts with receptor-interacting serine/threonine kinase 3 (RIPK3) to induce death signaling downstream of TLR3 in response to the dsRNA analog polyinosinic-polycytidylic acid [poly(I:C)] (43,(57)(58)(59)(60). We tested the susceptibility of A20/Abin-1 T-ΔIEC Tnf -/enteroids to poly(I:C)-induced death, and we observed that deletion of A20 and Abin-1 in IECs dramatically increased susceptibility to poly(I:C)-induced death ( Figure 3A). LPS, in contrast, induced minimal cytotoxicity (Figure 3A). Both poly(I:C) and LPS activate TRIF, but IEC death in response to LPS has previously been reported to depend on TNF, in contrast to poly(I:C) (43). We deleted Trif in A20/Abin-1 T-ΔIEC Tnf -/mice using CRISPR/Cas9 editing and generated a strain of mice with a deletion and premature stop codon at the Trif locus (Supplemental Figure 3, B and C). Splenocytes derived from A20/Abin-1 T-ΔIEC Tnf -/-Trif -/mice exhibited significantly reduced IFN-β production in response to poly(I:C), confirming functional deletion at the protein level (Supplemental Figure 3D). To determine if Trif deletion protected A20-and Abin-1-deficient IECs from TNF-independent death in vitro, we expanded enteroid cultures from these mice. A20/ from spontaneous cell death (Supplemental Figure 1, B-D). This suggests that IEC-extrinsic factors in vivo drive TNF-independent IEC death and mortality in A20/Abin-1 T-ΔIEC Tnf -/mice. Since in vitro IEC enteroid cultures are sterile, we considered that microbial signals might promote death in vivo. While our prior studies suggested that broad-spectrum-antibiotic treatment was insufficient to rescue A20/Abin-1 T-ΔIEC Tnf -/mice (9), we hypothesized that residual microbes in these mice could trigger IEC death. Accordingly, we derived A20/Abin-1 T-ΔIEC Tnf -/germ-free mice by cesarean section. Germ-free Tnf -/mice were largely protected from death upon deletion of A20 and Abin-1 in IECs ( Figure 1B). To control for developmental alterations by germ-free derivation, we conventionalized germ-free A20/Abin-1 T-ΔIEC Tnf -/mice with cecal contents from corresponding specific pathogen-free (SPF) mice in our facility. Germ-free mice conventionalized with cecal contents from SPF A20/Abin-1 T-ΔIEC Tnf -/mice (GF-CONV) exhibited rapid mortality upon deletion of A20 and Abin-1 ( Figure 1B), suggesting the increased survival of germ-free mice was not due to a developmental aberration. Therefore, microbial signals contribute to TNF-independent IEC death in the setting of acute A20 and Abin-1 deletion.
Although germ-free A20/Abin-1 T-ΔIEC Tnf -/mice exhibited increased survival, it was unclear whether this was due to reduced IEC death or merely due to broadly reduced septic sequelae under germ-free conditions. Histologically, acute deletion of A20 and Abin-1 in the intestinal epithelium caused rapid intestinal epithelial denudation, inflammatory infiltrate, cryptitis, and loss of mucosal architecture in both the small intestine and colon within 40 hours in GF-CONV mice ( Figure 1, C and D). In contrast, germfree A20/Abin-1 T-ΔIEC Tnf -/mice exhibited far less histologic injury ( Figure 1, C and D). Since IEC loss in this model is further characterized by massive apoptotic IEC death, we performed cleaved CASP3 (CC3) immunohistochemistry. In parallel to the reduction in histologic disease severity, we observed dramatically reduced CC3 in IECs of germ-free A20/Abin-1 T-ΔIEC Tnf -/mice as compared with GF-CONV counterparts ( Figure 1, E and F).
These results highlight that A20/Abin-1 T-ΔIEC Tnf -/mice provide a window into studying TNF-independent IEC death. A20/Abin-1 T-ΔIEC Tnf -/mice die due to spontaneous fulminant IEC death in vivo, but those IECs survive in vitro in the absence of hematopoietic cells, autocrine TNF, and microbial stimuli. When microbial stimuli are removed under germ-free conditions, the A20-and Abin-1-deficient IECs survive in vivo even with hematopoietic cells and other potential cytotoxic factors present. Interestingly, some germ-free A20/Abin-1 T-ΔIEC Tnf -/mice die ( Figure 1B), suggesting that there may be some sterile inflammatory factors that can contribute to TNF-independent IEC death, but microbial factors are a primary driver of intestinal inflammation in this model.
Deletion of MyD88, and to a lesser extent Trif, rescues A20/ Abin-1 T-ΔIEC Tnf -/mice. Given the dramatic improvement in survival and intestinal epithelial integrity in germ-free A20/ Abin-1 T-ΔIEC Tnf -/mice, we hypothesized that microbial signaling through MyD88 mediated the intestinal inflammation in this model. To facilitate combining of multiple mutant alleles, we targeted MyD88 in A20/Abin-1 T-ΔIEC Tnf -/zygotes as previously described (51). Using 2 guide RNAs (gRNAs) targeting with slightly delayed median survival relative to Trif +/+ controls ( Figure 3B). Interestingly, heterozygous and homozygous deletion of Trif conferred similar survival benefit. We examined the small intestine and colon 40 hours after tamoxifen-induced deletion of A20 and Abin-1 in A20/Abin-1 T-ΔIEC Tnf -/-Trif -/mice and saw a trend toward improvement in histologic severity, but this was not statistically significant when compared to the A20/Abin-1 T-ΔIEC Tnf -/mice ( Figure 3, C and D). Deletion of Trif partially reduced CC3 frequency, but not enough to prevent lethality in the A20/ Abin-1 T-ΔIEC Tnf -/-Trif -/enteroids were almost entirely resistant to poly(I:C)-and LPS-induced cell death ( Figure 3A). Therefore, A20 and ABIN-1 cooperatively restrict TRIF-mediated death in response to TLR3 and TLR4 agonists in vitro.
LTα 3 induces apoptosis and necroptosis downstream of TNFR1 in A20/Abin-1 T-ΔIEC Tnf -/enteroids. As combined deletion of Ripk3 and Casp8 completely prevented IEC death after deletion of A20 and Abin-1, but Trif deletion did not provide equivalent survival benefit, we hypothesized that another TNF superfamily member may be contributing to TNF-independent, IEC-extrinsic death in vivo. We previously reported that the TNF superfamily members TWEAK (TNF-like weak inducer of apoptosis), Fas ligand, and TRAIL (TNF-related apoptosis-inducing ligand) failed to induce significant death of A20/Abin-1 T-ΔIEC Tnf -/enteroids in vitro (9). Here we tested RANKL (receptor activator of NF-κΒ ligand), LIGHT (homologous to lymphotoxin, exhibits inducible expression and competes with HSV glycoprotein D for binding to herpesvirus entry mediator, a receptor expressed on T lymphocytes), and TL1A (TNF-like cytokine 1A), none of which induced significant IEC death ( Figure 5A).
Although none of those alternative TNF superfamily ligands induced death in A20/Abin-1 T-ΔIEC Tnf -/enteroids, we previously showed that deletion of A20 and Abin-1 increases death sensitivity of TNF receptor 1 (TNFR1) to TNF by 1000-fold in IECs (9). Given that TNFR1 sensitivity is significantly increased in A20-and Abin-1-deficient IECs, we considered alternative TNFR1 ligands. Lymphotoxin α homotrimers (LTα 3 ) are reported to bind to TNFR1 and activate death signaling (62)(63)(64). We tested human LTα 3 and observed significant increased susceptibility to death in A20-and Abin-1-deficient IECs ( Figure 5B). To test mouse LTα 3 , we generated a cell line constitutively expressing a mouse LTα 3 -Fc fusion protein and validated the presence of soluble fusion protein (Figure 5C). LTα 3 -Fc fusion protein-conditioned media (LTα 3 -Fc CM) induced significant death in A20/Abin-1 T-ΔIEC Tnf -/enteroids (Figure 5D). Since LTα 3 -Fc CM could contain other factors that drive IEC death in vitro, we wanted to determine whether an anti-LTα monoclonal antibody reversed this activity in vitro. An LTα-specific monoclonal antibody was previously developed and reported to inhibit collagen-induced arthritis (65). We generated purified . For panel A, significance was assessed by 1-way ANOVA with Dunnett's multiple-comparison test relative to vehicle. For panels B and E, significance was assessed using 2-way ANOVA with Tukey's multiple-comparison test. For panels C, D, F, and G, significance was assessed using 1-way ANOVA with Tukey's multiple-comparison test. Only significant differences are shown. *P < 0.05; ***P < 0.001; ****P < 0.0001. Data represent at least 2 independent experiments.
to LTα 3 -induced death downstream of TNFR1. It is important to highlight that the death assays we performed used primary, nonimmortalized IECs in the absence of any death-sensitizing agents (e.g., cycloheximide), unlike most prior studies investigating TNFand LTα 3 -induced death (62,64).
To further understand whether blocking LTα provided a survival benefit in these mice, we administered a monoclonal blocking antibody against LTα to A20/Abin-1 T-ΔIEC Tnf -/-MyD88 +/+ mice and observed a very small but statistically significant increase in weight and median survival as compared with isotype control, with survival increasing from 4.5 to 5.0 days (Figure 7, E and F). As expected, isolated blockade of LTα, without also neutralizing or deleting TNF or MyD88, was insufficient to protect A20/Abin-1 T-ΔIEC mice (Supplemental Figure 5, A and B). Since both Lta and Il1b transcripts were elevated in A20/Abin-1 T-ΔIEC Tnf -/intestine (Figure 7, C and D), and since both IL-1β and microbial ligands signal through MyD88, we inhibited both LTα and MyD88 by administering anti-LTα monoclonal antibody to MyD88 +/heterozygous A20/Abin-1 T-ΔIEC Tnf -/mice. LTα blockade in MyD88 heterozygous mice reduced weight loss and increased median survival from 4.0 to 9.0 days (Figure 7, G and H). Blocking LTα in vivo reduced weight loss and improved survival, particularly when combined with partial inhibition of MyD88 signaling, supporting a model where LTα and MyD88 agonists contribute to TNF-independent intestinal injury in A20-and ABIN-1-deficient intestinal epithelium.

Figure 8. LTA and IL1B mRNA are increased in inflamed intestinal mucosa from patients with IBD and both TNF and LTα 3 are cytotoxic to human colonoids. (A) qPCR for the indicated mRNA transcripts in colonic mucosal biopsies of inflamed (red) or noninflamed (blue) areas from patients with IBD or
without IBD (non-IBD, black). Biopsy samples were obtained at 1 or 2 locations and processed separately; each data point represents 1 patient (median and interquartile ranges, IL1B shown with log 2 y axis to display full range). If the colon was entirely inflamed or noninflamed, values from the 2 sets of biopsies were averaged. If biopsies were obtained separately from inflamed and noninflamed colon, then they were processed and displayed separately. (B) qPCR for the indicated mRNA transcripts in colonic mucosal biopsies in the subset of patients from A on anti-TNF therapy at the time of biopsy (median and interquartile ranges, IL1B shown with log 2 y axis to display full range). (C) Representative microscopy images of SYTOX dead cell nucleic acid stain (pseudocolor red). Scale bars: 100 μm. (D) Quantitative luminescent cell viability assay of human colonoids derived from non-IBD and IBD patients, treated with indicated stimuli for 48 hours (mean ± SEM; 50 ng/mL recombinant human LTα 3 [hLTα 3 ], 25 ng/mL hTNF). For panel A, significance was assessed using 1-way ANOVA with Tukey's multiple-comparison test. For panel B, significance was assessed by unpaired Student's t test. For panel D, significance was assessed using repeated measures 1-way ANOVA with Dunnett's multiple-comparison test relative to vehicle. dent IEC injury, this study demonstrates that A20 and ABIN-1 preserve intestinal homeostasis and IEC survival by restricting TNF-independent inflammatory injury. Microbial signals, MyD88 activation, and LTα all contribute to TNF-independent intestinal inflammation in the setting of acute A20 and ABIN-1 deletion in IECs. IEC death induced by A20 and ABIN-1 deficiency can also be blocked by simultaneous inhibition of CASP8-dependent apoptosis and RIPK3-dependent necroptosis. Given that A20 and ABIN-1 restrict both TNF-dependent and TNF-independent IEC injury, this study further reveals why these proteins have such potent antiinflammatory functions in the intestine. Although many previous studies focus on the role of A20 and ABIN-1 in hematopoietic cells, this study and others add to our understanding of the important role these proteins play in nonhematopoietic cells to preserve tissue integrity (9,16,37,39).
Combined acute deletion of A20 and ABIN-1 in the intestinal epithelium is notable in both severity and TNF independence when compared with deletion of other genes in IECs. For example, the fatal enteritis and colitis that develop after simultaneous deletion of A20 and ABIN-1 in IECs rivals the severity reported for intestinal epithelial deletion of NEMO (44,78), FADD (45,79), CASP8 (45,71), RIPK1 (42,80), ATG16L1 (8), or combined deletion of XBP1 and ATG16L1 (7). Acute deletion of SETDB1 (81) in IECs induces mouse lethality with similar kinetics to the combined deletion of A20 and ABIN-1, although Setdb1 -/-IECs die primarily from necroptotic cell death. The TNF-independent IEC death observed in A20/Abin-1 T-ΔIEC Tnf -/mice is also unusual when compared with other models. The colitis in mice with conditional IEC knockout of NEMO, FADD, CASP8, and RIPK1 mice is largely reversed by TNF or TNFR1 deletion (42,44,45,(78)(79)(80). Similarly, the ileitis induced after knockout of XBP1 in IECs is rescued by TNFR1 deficiency (7). Among these models, deletion of A20 and ABIN-1 is unique in that TNF deletion does not confer any significant survival benefit, and it does not appreciably reduce intestinal injury.
The ileal enteropathy associated with IEC knockout of NEMO, FADD, or CASP8 is TNFR1 independent, even though colitis in those mice is TNFR1 dependent. Those models provide some perspective on the TNF-independent IEC death observed in A20/Abin-1 T-ΔIEC Tnf -/mice. Paneth cell loss after IEC knockout of FADD and CASP8 is partially reduced by TNFR1 deletion but is further reduced by deletion of Z-DNA-binding protein 1 (ZBP1), suggesting that ZBP1 contributes to TNFR1-independent death in the ileum (45).
ZBP1 is an IFN-inducible gene product that interacts with RIPK3 via its RHIM domain and activates necroptotic cell death in response to type I and II IFNs (82) and viral nucleic acids (83). It is possible that ZBP1 contributes to TNF-independent IEC death after deletion of A20 and ABIN-1, especially since deletion of TRIF only modestly increased median survival. However, ZBP1 has been shown to activate RIPK3 in IECs when FADD-CASP8 apoptotic death is inhibited (45), or in response to RNA from reactivated endogenous retroviruses after SETDB1 deletion (81). In A20/Abin-1 T-ΔIEC Tnf -/-IECs, both FADD and CASP8 are present to activate apoptotic death, and CC3 is detected in the intestinal epithelium and in intestinal organoids, favoring a central role for CASP8-dependent apoptosis rather than potential ZBP1-dependent necroptosis as the dominant TNF-independent death path-ry pathways could contribute to intestinal injury in patients with IBD, we measured A20 and ABIN-1 mRNA and protein levels in colonic mucosal biopsies from non-IBD and IBD patients, stratifying the IBD samples by whether they were obtained from endoscopically inflamed or noninflamed areas (Supplemental Table  1). A20 (TNFAIP3) is a TNF-and NF-κB-inducible gene, and mRNA levels were predictably increased in inflamed areas of IBD patients as compared with noninflamed IBD biopsies and non-IBD controls (Supplemental Figure 6A). ABIN-1 (TNIP1) mRNA levels were not significantly different among these groups (Supplemental Figure 6A). However, A20 and ABIN-1 undergo posttranslational regulation and degradation in response to a variety of inflammatory stimuli (72)(73)(74), and A20 has been reported to be decreased at the protein level in the mucosa of IBD patients (75). Immunoblot analysis of a subset of mucosal biopsies demonstrated a trend toward lower A20 protein levels in inflamed biopsies and an inverse correlation between A20 mRNA and protein levels (Supplemental Figure 6, B-D). ABIN-1 protein levels were significantly reduced in areas of inflammation in IBD patients (Supplemental Figure 6, B, E, and F). These data suggest that A20 and ABIN-1 proteins are relatively decreased in the intestinal mucosa of IBD patients with active inflammation. Decreased A20 and ABIN-1 levels sensitize IECs to TNFR1-induced apoptosis and necroptosis, so these data suggest that the inflamed mucosa in IBD patients is more susceptible to TNF-dependent and TNF-independent cytotoxic factors.
To further determine whether lymphotoxin or IL-1β contributes to intestinal injury in patients with IBD, we measured LTA and IL1B in colonic mucosal biopsies from non-IBD and IBD patients, again stratifying by whether they were obtained from endoscopically inflamed or noninflamed areas. LTA and IL1B were elevated in areas of inflammation ( Figure  8A). Another IL-1 family member, IL18, was not elevated in inflamed tissue ( Figure 8A). These trends were confirmed in a recent meta-analysis (76). Interestingly, in a subset of patients on anti-TNF therapy, both LTA and IL1B were elevated in areas of active inflammation ( Figure 8B), suggesting that local LTα and IL-1β contribute to TNF-independent intestinal injury even when TNF is functionally neutralized. Finally, TNF and LTα 3 exhibited cytotoxicity in primary human colonoid cultures derived from both non-IBD and IBD patients measured qualitatively by dead cell nucleic acid stain ( Figure 8C) and quantitatively using ATP-based luminescent cell viability ( Figure 8D). Notably, the human IEC colonoid cytotoxicity induced by both TNF and LTα 3 was observed in the absence of any death-sensitizing agents (e.g., cycloheximide or second mitochondria-derived activator of caspases [SMAC] mimetics).

Discussion
Genome-wide association studies have implicated hundreds of genes in CD and UC pathogenesis, but it is unclear how these genes increase disease susceptibility or influence treatment response (77). Since IBD is primarily a complex polygenic disease, except for rare monogenic causes, it follows that model systems often incorporate multiple genes simultaneously to decipher these complex epistatic genetic interactions. While we previously showed that A20 and ABIN-1 cooperatively restrict TNF-depen-body-dependent cell-mediated cytotoxicity or apoptosis of pathogenic TNF-expressing inflammatory cells to the same extent as anti-TNF antibodies (98,99). Intriguingly, the combination of an anti-TNF monoclonal antibody to neutralize TNF plus etanercept, which additionally neutralizes LTα 3 , was effective in a case report of a patient with severe HLA-B27-associated arthropathy who had failed treatment with either agent alone (100).
The other TNF-independent injury pathway highlighted in this study is MyD88. Microbial signals activating MyD88 in the hematopoietic compartment contribute to intestinal injury, suggesting that investigations into which microbes are most pathogenic in A20/Abin-1 T-ΔIEC Tnf -/gnotobiotic mice would be instructive. Interestingly, the fact that some germ-free A20/ Abin-1 T-ΔIEC Tnf -/mice die suggests that sterile inflammation is sufficient to induce death. Dietary or residual bacterial antigens in sterile chow could potentially contribute to intestinal injury in germ-free A20/Abin-1 T-ΔIEC Tnf -/mice (101, 102), but another possibility is that MyD88 activation by endogenous IL-1β and/ or IL-18 drives sterile TNF-independent intestinal injury. Both IL-1β and IL-18 have been proposed as potential mediators of anti-TNF nonresponse in patients with IBD (103-106). Furthermore, whether MyD88 activation leads directly to increased LTα 3 production by innate lymphoid cells or peripheral B cells, for example, remains to be seen (107,108). Ultimately, having animal models to investigate how MyD88, microbial signals, IL-1, LTα 3 , and various intestinal inflammatory cells cooperatively drive TNF-independent intestinal injury could prove useful in understanding anti-TNF nonresponse in subsets of patients. In addition, TNF-and LTα 3 -induced cytotoxicity in primary human intestinal organoid cultures in the absence of death-sensitizing agents could serve as an in vitro approach to further understand anti-TNF refractory disease.
In summary, we show that microbial signals, MyD88 signaling, and LTα 3 drive severe TNF-independent enterocolitis after acute deletion of 2 IBD-associated genes, A20 and Abin-1, which can be prevented by combined deletion of Ripk3 and Casp8. This study highlights that A20 and ABIN-1 cooperatively maintain intestinal homeostasis by inhibiting both TNF-dependent and TNF-independent inflammatory intestinal injury. Understanding the genetic determinants of intestinal epithelial health, their epistatic relationships, and how they could influence therapeutic response in models of IBD will provide important mechanistic insights to inform future translational studies.

Methods
Mice. A20 fl and Abin-1 fl mice were generated in the Ma laboratory and were described previously (9,27,61,109). Tnf -/mice were purchased from Jackson Laboratories, X. Wang (National Institute of Biological Sciences, Beijing, China) provided Ripk3 -/mice, and transgenic mice with a tamoxifen-inducible Cre recombinase under the control of the villin promoter (Vil-cre-ER T2+ ) mice were a gift from Sylvie Robine (Institut Curie-CNRS; ref. 110). These alleles were backcrossed to A20 fl and Abin-1 fl transgenic mice for more than 8 generations, as previously described (9). Acute deletion of floxed A20 and Abin-1 exons in vivo by oral gavage was performed using tamoxifen (2 mg/day; Mil-liporeSigma, T5648) for 3 consecutive days as indicated for survival analysis. MyD88 +/and Trif +/breeders were maintained on antibiotic way. In A20/Abin-1 T-ΔIEC Tnf -/-IECs, RIPK1 is also present and would be predicted to inhibit ZBP1-mediated activation of RIPK3 (84). We cannot entirely exclude a role for ZBP1 in TNF-independent IEC death after deletion of A20 and Abin-1, but the dominant death pathway in A20/Abin-1 T-ΔIEC Tnf -/mice is most consistent with CASP8-mediated apoptosis.
There are multiple potential translational implications for the findings that LTα 3 and MyD88 contribute to TNF-independent intestinal damage in this model of severe enterocolitis. Deletion of A20 and Abin-1 unveils a role for LTα 3 -induced TNF-independent IEC death in primary cells both in vivo and in vitro. To our knowledge, this is the first description of TNF-independent LTα 3 -induced cytotoxicity in IECs in the absence of death-sensitizing agents. There are many potential mechanisms by which patients with IBD may fail to respond to anti-TNF therapy, including activation of alternative inflammatory pathways, neutralizing antibodies, subtherapeutic levels, or inability to block autocrine IEC-derived TNF, but it is tempting to consider whether LTα 3 -mediated TNFR1-induced IEC death could play a role in a subset of patients. Polymorphisms in LTA were associated with anti-TNF nonresponse in a small cohort of patients (95), but this was not reproduced in a larger cohort (96). Moreover, etanercept, a soluble TNFR2-Fc receptor that blocks TNF and LTα 3 , was ineffective for CD (97). On the other hand, subsequent data suggest that etanercept is ineffective in part because it does not induce anti-(IDT) according to the manufacturer's instructions. RNPs were electroporated into mouse zygotes as previously described (51), with the main exception that the standard square curve electroporation (2 pulses at 30 V for 3 ms with an interval of 100 ms) was performed once or 4 times with an interval of 3 seconds using a Gene Pulser Xcell (Bio-Rad). Zygotes were cultured to 2-cell embryos and transferred into pseudopregnant females. All mutations were confirmed by cloning and sequencing. Mice with a confirmed deletion or premature stop codon were then backcrossed 3 generations to the parental strains to breed out off-target edits before analysis.
IEC enteroid culture and confocal microscopy. For murine enteroid culture, intestinal crypts were isolated from the small intestine as previously described (111), with the modifications of substituting 10% R-spondin-1-conditioned medium for recombinant R-spondin-1, and the addition of Normocin (100 μg/mL, Invivogen). R-spondin-1expressing 293T cells were a gift from Noah Shroyer (Baylor College of Medicine, Houston, Texas, USA). For all enteroid experiments, enteroids were derived from at least 2 mice on separate occasions and representative data are shown. Deletion of A20 or Abin-1 in vitro was performed via treatment with 4-OHT for 24 hours (200 nM; Milli-poreSigma). For human colonoid culture, biopsies were cryopreserved, thawed, and digested as previously described (112). Human colonoids were expanded as previously described (113), with WNT surrogate-Fc fusion (ImmunoPrecise) and supplemented with 25% FBS (VWR), CHIR99021 (Cayman Chemical Company), valproic acid (MilliporSigma), [leucine 15 ] gastrin-1 human (MilliporeSigma), human R-spondin-3 instead of R-spondin-1 (Peprotech), human stem cell factor (STEM-CELL Technologies), human FGF-basic (Peprotech), and human IGF-1 (BioLegend). Confocal imaging of organoids was performed on a Leica SP5 laser scanning confocal system using a 10× dry objective. Images were acquired in a 512 × 512 format, with a line average of at least 3, scan speed of 400 Hz, and pinhole airy unit 1. Excitation for both PI or SYTOX was done with the 488 nm laser line at 30% power with a detection band of 550 to 732 nm (PI) or 520-550 nm (SYTOX). Image analysis was performed using the Leica Application Suite.
Cell death assays. Enteroid death assays were performed by resuspending in Matrigel (Corning) and plating 25 μL per well in 96-well flat-bottom opaque plates (Nunc). Deletion of A20 or Abin-1 in vitro was performed via treatment with 200 nM 4-OHT for 24 hours, at which point the enteroids were stimulated as indicated. Viability was measured using the CellTiter Glo 3D assay (Promega) according to the manufacturer's specifications, with the exception that 100 μL of reagent was added to 200 μL of culture for a final volume of 300 μL prior to reading. For colonoids, stimulations were performed as indicated for 48 hours and viability was measured using the CellTiter Glo 3D assay according to the manufacturer's specifications. Luminescence was read on a SpectraMax M5 (Molecular Devices), analyzed using SoftMax Pro (Molecular Devices), and expressed as percentage viability relative to vehicle control.
Cell signaling assays and immunoblot analysis. For enteroid lysates, cultures were resuspended in Cell Recovery Solution (Corning) supplemented with 10 μM Y-27632 (Calbiochem) and incubated for 15 minutes on ice, followed by centrifugation at 500g for 5 minutes. Cell pellets were lysed in ice-cold NP40 lysis buffer (1% NP40 [v/v], 50 mM Tris HCl pH 7.4, 150 mM NaCl, and 10% glycerol [v/v]) supplemented with complete EDTA-free Protease Inhibitor Cocktail (Roche), phosphatase inhibitors (1 mM Na 3 VO 4 and 10 mM NaF), and 10 mM N-ethylmaleimide (MilliporeSigma). Colonic water ad libitum with 1 mg/mL ampicillin, 0.5 mg/mL vancomycin, and 1 mg/mL neomycin (all MilliporeSigma), and offspring were transitioned to regular water at weaning. For in vivo monoclonal antibody treatment, mice were injected with 370 μg of antibody on days -2, 0, 1, 2, 5, 7, and 9. For all experimental mice, genotypes were confirmed twice. When possible, littermates were used as controls. Both males and females were included in all experiments, with no observable sex differences. Mice were analyzed between 7 and 12 weeks of age for all experiments. All mice were on the C57BL/6 background.
Study participants and specimen collection. Patients undergoing colonoscopy or sigmoidoscopy for standard-of-care indications were screened for study eligibility. All patients gave written informed consent and approval. Cold-forceps biopsy samples were obtained from patients with CD or UC (IBD), and individuals without IBD (non-IBD). Baseline demographic and clinical data for the study participants are provided in Supplemental Table 1. Options were defined by the investigators and participants chose their classifications. Non-IBD patients were patients without known or suspected IBD undergoing elective colonoscopy or sigmoidoscopy for various indications (e.g., colorectal cancer screening). Biopsy samples were categorized as coming from an area that was endoscopically inflamed or noninflamed. Biopsies were obtained from the right (proximal) and/or left (distal) colon. Biopsies were placed in basal organoid media consisting of advanced DMEM/F12 with nonessential amino acids and sodium pyruvate (Thermo Fisher Scientific) supplemented with HEPES (10 mM; Corning), Glutamax (2 mM; Thermo Fisher Scientific), Normocin (100 μg/mL; Invivogen, ant-nr-2), penicillin-streptomycin-neomycin (Thermo Fisher Scientific), and N-acetylcysteine (1 mM; MilliporeSigma), supplemented with Y-27632 (10 μM; MedChem Express). Samples were immediately placed on ice and transported to the laboratory for processing.
Germ-free derivation, maintenance, and conventionalization. Germfree A20/Abin-1 T-ΔIEC Tnf -/mice were derived by cesarean delivery into the UCSF Gnotobiotic Core Facility (gnotobiotics.ucsf.edu) and maintained in germ-free isolators. Stool pellets from isolators in the Gnotobiotic Core Facility were screened every 2 to 3 weeks by culture and qPCR, whenever mice were transferred from breeding to experimental isolators, and at the beginning and end of each experiment. 515F and 806R universal primers were used to amplify the V4 region: 515F-GTGCCAGCMGCCGCGGTAA and 806R-GGACTACHVGG-GTWTCTAAT. For cecal content conventionalization experiments, cecal contents from SPF mice were resuspended in 10% (w/v) Bacto brain heart infusion (BHI) (VWR) supplemented with 0.05% (w/v) L-cysteine (MilliporeSigma) in aerobic conditions, passed through a 100-μm filter, and 200 μL was administered to germ-free mice by oral gavage 5-7 days prior to tamoxifen treatment.
CRISPR/Cas9 editing of MyD88, Trif, and Casp8. Two gRNAs for each target gene were designed by Benchling. The following gRNAs were selected for editing A20/Abin-1 T-ΔIEC Tnf -/mice: MyD88 guide 1 CCCACGTTAAGCGCGACCAA and MyD88 guide 2 AAGGAGCT-GAAGTCGCGCAT targeting exon 1 of MyD88; and Trif guide 1 TCT-GGAACGCTAATTTCGTG and Trif guide 2 CAAGCTATGTAACA-CACCGC targeting exon 2 of Trif. The following gRNAs were selected for editing A20/Abin-1 T-ΔIEC Ripk3 -/mice: Casp8 guide 1 TAGCTTCT-GGGCATCCTCGA and Casp8 guide 2 CTTCCTAGACTGCAAC-CGAG targeting exon 2 of Casp8. The corresponding crRNAs targeting each gene and tracrRNA were annealed and ribonucleoprotein (RNP) complexes were prepared with HiFi Cas9 nuclease ribonucleoprotein per group. When comparing every mean to every other mean by ANO-VA, Tukey's multiple comparison test was used. When comparing each mean to a control, Dunnett's multiple comparison test was used. When comparing only a subset of means, Bonferroni's multiple-comparison test was used. Categorical variables were analyzed by Fisher's exact or χ 2 tests where appropriate. A P value of less than 0.05 was used as the threshold for statistical significance.
Histology and immunohistochemistry, antibodies, reagents, and LTα 3 -Fc CM are described in Supplemental Methods.
Study approval. All animal studies were conducted in accordance with the UCSF Institutional Animal Care and Use Committee (AN183350). The study was conducted according to Declaration of Helsinki principles and was approved by the Institutional Review Board of UCSF (15-17757 and 19-27302). Written informed consent was received from participants prior to inclusion in the study.

Author contributions
MGK and A Ma conceived of the project. MGK oversaw the project and designed and performed the experiments with assistance from co-first authors IR and EM. Co-first author order was determined by duration of time spent on the project. IR performed in vivo experiments, in vitro death assays, and helped generate LTα 3 -Fc CM. EM performed enteroid immunoblot analysis and in vitro assays. EM, IR, and MN performed RNA-ISH analysis with advice and guidance from ODK. JLB consented patients and collected and processed patient biospecimens. ZL performed CRISPR/ Cas9 zygote editing with guidance from A Marson. JLB, XS, YYR, ZW, RA, and PA assisted with mouse breeding, genotyping, and histology quantitation. KML, PJT, and JAT derived germ-free mice and performed germ-free experiments in the UCSF Gnotobiotics Core Facility. BR provided a key observation that lymphotoxin could contribute to TNF-independent death in this model. BAM, BR, and LS critically reviewed the manuscript and provided helpful discussion. MGK performed the statistical analyses and wrote the manuscript with editing from IR, EM, BAM, A Ma, and input from all co-authors. mucosal biopsy samples were lysed in Lysing Matrix D tubes (MP Biomedicals) containing RIPA lysis buffer (1% NP40 [v/v], 0.5% sodium deoxycholate [w/v], 0.1% sodium dodecyl sulfate, 50 mM Tris HCl pH 7.4, 150 mM NaCl, and 10% glycerol [v/v]) supplemented with complete EDTA-free Protease Inhibitor Cocktail, phosphatase inhibitors (1 mM Na 3 VO 4 and 10 mM NaF), 10 mM N-ethylmaleimide, and antifoaming Reagent DX (Qiagen). Samples were homogenized using a FastPrep-24 homogenizer (MP Biomedicals). RKO (ATCC, CRL-2577) lysate was used as a reference. Biopsy samples with low lysate protein concentration (below 1000 μg/mL) were excluded. After lysis, samples were centrifuged for 20 minutes at 21,130g to remove debris, and the supernatants were quantified using the BCA Protein Assay Kit (Pierce). Lysates were normalized and denatured in LDS Sample Buffer (Invitrogen), followed by resolution in NuPage precast 4% to 12% Bis-Tris gels (Invitrogen) and transferred to PVDF for immunoblotting. Quantitation was performed using Image Lab (Bio-Rad).
qPCR. Mouse small intestine or colon (1 cm) was flushed with saline solution and placed in RNAlater (Thermo Fisher Scientific) and stored for 16 to 72 hours at 4°C. For colonic mucosal biopsies, samples were placed in RNAlater and stored for 16 to 72 hours at 4°C. The solution was then aspirated, and samples were stored at -70°C. Samples were then thawed and homogenized in Buffer RLT (Qiagen) supplemented with β-mercaptoethanol using Lysing Matrix D tubes (MP Biomedicals) on the FastPrep-24 (MP Biomedicals), followed by QIAshredder homogenization (Qiagen). RNA was prepared using the RNeasy Mini Kit (Qiagen) with on-column DNase digestion according to the manufacturer's instructions. Quality was confirmed using the NanoDrop ND-1000 (Thermo Fisher Scientific) and Agilent RNA 6000 Nano Kit on the Agilent 2100 Bioanalyzer, according to manufacturers' instructions. cDNA was synthesized from 2000 ng of total RNA using the High-Capacity RNA-to-cDNA kit (Thermo Fisher Scientific). qPCR on murine samples was performed using TaqMan probes for mouse Lta (Mm00440228_gH), mouse Il1b (Mm00434228_m1), and mouse Actb (Mm02619580_g1), and the TaqMan Universal Master Mix II with UNG on the QuantStudio 6 (Thermo Fisher Scientific). qPCR on human samples was performed using TaqMan probes for human LTA (Hs99999086_m1), IL1B (Hs01555410_m1), IL18 (Hs01038788_m1), TNFAIP3 (Hs00234713_m1), TNIP1 (Hs00374581_m1), and GAP-DH (Thermo Fisher Scientific, 4485713) with the Taqman Multiplex Master Mix (Thermo Fisher Scientific 4461884) on the QuantStudio 6 (Thermo Fisher Scientific). Relative gene abundance was normalized to the mean expression of the housekeeping gene Actb or GAPDH, and 2 -ΔΔCt was calculated relative to a reference sample. All samples were run in duplicate. For mucosal biopsies, if the colon was entirely inflamed or noninflamed, values from the 2 sets of biopsies were averaged. If biopsies were obtained separately from inflamed and noninflamed colon, then they were processed and displayed separately. Samples with low RNA concentration and amplification of GAPDH at Ct greater than 24 were excluded. Data were analyzed using Quant-Studio Real-Time PCR Software (Thermo Fisher Scientific).
Statistics. Statistical analysis was performed with GraphPad Prism 9. Kaplan-Meier survival curve comparisons were performed using the log-rank Mantel-Cox test. Comparisons between 2 groups were performed by 2-tailed, unpaired Student's t test. Multigroup comparisons were performed by 1-way analysis of variance (ANOVA) if comparing 1 variable per group or 2-way ANOVA if there were multiple variables