Atg14 protects the intestinal epithelium from TNF-triggered villus atrophy

ABSTRACT

Regulation of intestinal epithelial turnover is a key component of villus maintenance in the intestine. The balance of cell turnover can be perturbed by various extrinsic factors including the cytokine TNF, a cell signaling protein that mediates both proliferative and cytotoxic outcomes. Under conditions of infection and damage, defects in autophagy are associated with TNF-mediated cell death and tissue damage in the intestinal epithelium. However, a direct role of autophagy within the context of enterocyte cell death during homeostasis is lacking. Here, we generated mice lacking ATG14 (autophagy related 14) within the intestinal epithelium [Atg14^f/f Vil1-Cre (VC)⁺]. These mice developed spontaneous villus loss and intestinal epithelial cell death within the small intestine. Based on marker studies, the increased cell death in these mice was due to apoptosis. Atg14^f/f VC⁺ intestinal epithelial cells demonstrated sensitivity to TNF-triggered apoptosis. Correspondingly, both TNF blocking antibody and genetic deletion of Tnfrsf1a/Tnfr1 rescued villus loss and cell death phenotype in Atg14^f/f VC⁺ mice. Lastly, we identified a similar pattern of spontaneous villus atrophy and cell death when Rb1cc1/Fip200 was conditionally deleted from the intestinal epithelium (Rb1cc1^f/f VC⁺). Overall, these findings are consistent with the hypothesis that factors that control entry into the autophagy pathway are also required during homeostasis to prevent TNF triggered death in the intestine.

Abbreviations: ANOVA: analysis of variance; Atg14: autophagy related 14; Atg16l1: autophagy related 16-like 1 (S. cerevisiae); Atg5: autophagy related 5; cCASP3: cleaved CASP3/caspase-3; cCASP8: cleaved CASP8/caspase-8; CHX: cycloheximide; EdU: 5-ethynyl-2´-deoxyuridine thymidine; f/f: flox/flox; H&E: hematoxylin and eosin; MTT: 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide; Nec-1: necrostatin-1; Rb1cc1/Fip200: RB1-inducible coiled-coil 1; Ripk1: receptor (TNFRSF)-interacting serine-threonine kinase 1; Ripk3: receptor (TNFRSF)-interacting serine-threonine kinase 3; Tnfrsf1a/Tnfr1: tumor necrosis factor receptor superfamily, member 1a; Tnf/ Tnfsf1a: tumor necrosis factor; VC: Vil1/villin 1-Cre

Introduction

Regulation of epithelial turnover is a key mechanism underlying the protection afforded by mucosal barriers [1]. This is particularly true for the mouse small intestinal epithelium; it is a rapid turnover system based on a single, continuous layer of epithelial cells organized into repeating units of Crypts of Lieberkühn and villi. The crypts contain intestinal epithelial stem and progenitor cells that perpetually and continuously produce multiple differentiated lineages. Most epithelial lineages (other than Paneth cells located at the crypt base) migrate out of crypts onto adjacent villi where upward migration is organized in coherent columns [2]. Epithelial turnover is completed through a process of regulated cell death (apoptosis) that is concentrated at villus tips, where apoptotic cells are extruded into the lumen [3]. However, cell death and extrusion can also occur along the length of the villi as well as in crypts [4]. In order to maintain crypt-villus architecture, the rate of apoptotic cell death that occurs along the crypt-villus axis must balance the rate of proliferation in crypts [5]. Various perturbations to this system such as infection and inflammation can increase the rate of epithelial turnover [6] and in some cases skew towards an increase in epithelial stem and progenitor cells in the crypt [7].

Intestinal epithelial cell death can be modulated by several pathways including pyroptosis [8], genotoxic and endoplasmic reticulum stress [9,10], loss of contact with the basement membrane that induces anoikis [11] and activation of death receptors through high levels of stimulation by TNF (tumor necrosis factor) family cytokines [12]. Loss of function studies of negative regulators of TNF-induced cell death (such as Tnfaip3/A20, Tnip1/Abin-1, and Ripk1) show increased apoptosis and necroptosis in the intestinal epithelium [13–15]. One conclusion from these studies is that host genetic factors can regulate the intrinsic sensitivity of the intestinal epithelium to TNF-mediated death signaling.

The process of autophagy is a candidate pathway to regulate cell death in the intestinal epithelium. Conditional loss of function studies of various autophagy enzymes [notably the conjugation system enzymes ATG5 (autophagy related 5) and ATG16L1 (autophagy related 16-like 1 [S. cerevisiae])] show an important role for autophagy within the intestine [16]. None of these models display a substantial alteration in epithelial cell death during homeostasis. However, a number of recent studies demonstrate that autophagy allows protection against TNF-mediated epithelial cell death in the presence of pertinent environmental exposures: smoking, colonic damage (dextran sulfate sodium, DSS), infection (Helicobacter or Toxoplasma), or combination of infection and DSS [17–20].

Despite these studies, open questions remain. First and foremost, is the process of autophagy required to protect the intestinal epithelium against cell death during homeostasis? Thus far only superimposition of infection/damage on mice with autophagy defects (namely ATG5 or ATG16L1) can increase epithelial cell death. Secondly, recent work suggests that residual autophagy can occur even in the complete absence of ATG5 or ATG16L1. In contrast, factors such as ATG14 (autophagy related 14) demonstrate a complete block in autophagic flux [21]. ATG14 and RB1CC1/FIP200 are well-characterized factors that mediate autophagy initiation through the class III phosphatidylinositol 3-kinase complex I (PIK3C3/VPS34-BECN1-PIK3R4/VPS15-ATG14) or ULK1 complex (ULK1-RB1CC1/FIP200-ATG13-ATG101), respectively [22–24]. Thus, it is critical to test the role of intestinal epithelial cell death and turnover in the context of loss-of function of proteins critical for autophagy initiation.

To test the role of ATG14 in epithelial turnover, we generated a mouse strain with a conditional deletion of Atg14 within the intestinal epithelium. Five-week-old mice spontaneously developed widespread small intestinal villus atrophy. CASP3/caspase-3 mediated apoptosis was the major driver of villus atrophy in these mice. Atg14-deficient intestinal epithelial cell lines were highly sensitive to TNF-induced apoptosis. Correspondingly, neutralization of TNF ligand and deletion of Tnfrsf1a/Tnfr1 (TNF Receptor 1) rescued spontaneous villus atrophy in these mice. Taken together, our results suggest that ATG14 protects intestinal epithelial cells from TNF-mediated programmed cell death.

Results

Deleting Atg14 in mouse intestinal epithelial cells elicits spontaneous villus atrophy and failure to thrive

The major unexpected phenotype of mice with deletion of Atg14 in the intestinal epithelium [Atg14^f/f Vil1-Cre (VC)⁺] [25] as compared to littermate controls (includes Atg14^f/f VC⁻ and Atg14^f/+ VC⁺) was divergent body weights during post-weaning development. This phenotype was not anticipated as similar conditional deletions of autophagy genes in the intestinal epithelium such as Atg5, Atg7 and Atg16l1, do not show this spontaneous phenotype [26–28]. Tracking of body weights over time showed that Atg14^f/f VC⁺ mice displayed a failure to thrive during a period that spanned maturation to early adulthood (3–6 wk of age) (Figure 1(a-b)). During this period of time, Atg14^f/f VC⁺ mice did not show increased mortality as compared to controls, enabling further studies to understand the underlying mechanism of this phenotype.

Failure of weight gain in Atg14^f/f VC⁺ mice correlated with villus loss. Whole mount analysis of the intestinal mucosal surface of 5-week-old Atg14^f/f VC⁺ mice showed a discrete area in the mid-jejunum that did not contain villi in each mouse (Figure S1(a)). Histological analysis confirmed that only rudimentary villi were present in this sharply demarcated area of the small intestine (Figures 1(c-e) and S1(b-d)). The underlying crypts in areas of villus loss showed expansion and epithelial hyperproliferation as determined by quantification of M-phase bodies per crypt (Figure 1(f-g)). The finding that crypt proliferation was maintained or even expanded in Atg14^f/f VC⁺ mice suggests that villus loss was not secondary to loss of the regenerative potential of the intestinal crypts as previously described in mouse models with targeted deletion of cell cycle genes [29].

Figure 1.

Deletion of Atg14 in the mouse intestinal epithelium leads to spontaneous villus loss. (a) Measured weights of Atg14^f/f Vil1-Cre (VC)⁺ and littermate control mice (Atg14^f/f VC⁻ and Atg14^f/+ VC⁺) grouped by indicated age at time of sacrifice; n = 13–24 mice/group, >4 litters/group, two-way ANOVA with Sidak’s multiple comparisons test. (b) Percent weight gain ± SEM of Atg14^f/f VC⁺ and littermate control mice; repeated measures two-way ANOVA. (c-h) Histological analysis of H&E and immuno-stained small intestine sections of Atg14^f/f VC⁺ and littermate controls from 3, 5, and 6-week-old mice; n = 5–7 mice/group, ≥ 100 crypt-villus units quantified/mouse; two-way ANOVA with Sidak’s multiple comparisons test. (c) Representative H&E staining of jejunal sections demonstrating villus loss starting at 5 wk of age and cystic crypts at 6 wk of age within Atg14^f/f VC⁺ mice; bars: 500 μm. (d) Mean villus height ± SEM. (e) Percent intestinal length affected ± SEM. (f) Mean crypt height ± SEM. (g) Mean M-phase figures per 100 crypts ± SEM. (h) Mean lamina propria neutrophils ± SEM identified by LY6G immunohistochemical staining from 5- and 6-week-old mice; n = 6 mice/group from n = 2 independent experiments. **P < 0.01, ***P < 0.001, ****P < 0.0001, ns: not significant.

The jejunal villus phenotype tracked with the divergence of weight loss of Atg14^f/f VC⁺ mice and controls that occurred as the mice aged. Three-week-old Atg14^f/f VC⁺ mice did not show any obvious villus or crypt defects by whole mount or histological analysis and overall weights were comparable with controls at this time (Figure 1(c-e)). Conversely, at 6 wk of age, the affected area of villus loss in Atg14^f/f VC⁺ mice showed extensive expansion both proximally and distally beyond the area of the mid-jejunum (Figures 1(e) and S1(b-d)). Histological analysis of intestinal sections from Atg14^f/f VC⁺ mice at 6 wk of age showed a mix of hyperproliferative and hypoproliferative cystic crypts that were associated with areas of villus loss (Figure S1(e)). Despite the loss of villi, we observed no significant differences in the serum levels of glucose, cholesterol and triglyceride as compared to controls (Figure S2(a-c)). This finding may explain the survival of the Atg14^f/f VC⁺ mice during this period, suggesting that the remaining villi in the proximal and distal small intestine were sufficient for adequate nutrient absorption.

Focal villus blunting was the major alteration in the intestine of Atg14^f/f VC⁺ mice. This phenotype was not associated with complete breakdown of the epithelial barrier, as we detected no ulcerations in the intestinal mucosa. In addition, despite the loss of villi in Atg14^f/fVC⁺ mice, we found only a small increase in lamina propria neutrophils in areas of villus loss (Figure 1(h)); the highest value in any sample was one neutrophil per 100 crypts. As a reference, values >10 neutrophils/crypt can be detected in models with severe acute inflammation [30]. In addition, the number of neutrophils in the epithelial compartment of Atg14^f/f VC⁺ mice were comparable to littermate controls (Figure 1(h)). These data suggest that acute inflammation was not a primary driver of villus loss.

Atg14^f/f VC⁺ mouse small intestinal epithelial cells exhibit spontaneous apoptosis

Intestinal epithelial deletion of Atg14 resulted in spontaneous and progressive small intestinal villus atrophy that was most prominent in the mid-jejunums of 5-week-old mice. Comparable models of villus atrophy within the small intestine include mice with an intestinal epithelial deletion of RIPK1 (Ripk1^f/f VC⁺) [13,14]. Interestingly, villus atrophy within the Ripk1^f/f VC⁺ model is accompanied by increased epithelial cell death in the intestinal epithelium [13,14]. Thus, we next determined if villus loss in Atg14^f/f VC⁺ mice correlated with epithelial cell death.

Analysis of the crypt epithelial cells from hematoxylin and eosin (H&E) stained intestinal sections of Atg14^f/f VC⁺ mice and controls showed elevated numbers of apoptotic bodies (as defined by nuclei undergoing karyorrhexis) in Atg14^f/f VC⁺ mice that were 5 and 6 wk of age (Figure 2(a)). These findings represent a substantial amount of spontaneous cell death within the intestinal epithelium of Atg14^f/f VC⁺ mice. Immunohistochemical analysis of a marker of active apoptosis, cleaved CASP3/caspase-3 (cCASP3), suggested an increase in apoptotic cell death within Atg14^f/f VC⁺ epithelial cells at weeks 5 and 6 (Figure 2(b-d)). Increased abundance of cCASP3 was found in both crypts and villi. At week 3, prior to the development of villus or weight loss, there was a significant increase in the number of cCASP3 positive cells in crypts and a trend towards an increase in cCASP3 positive cells in villi of Atg14^f/f VC⁺ mice (Figure 2(b-d)). This increase in apoptotic cell death correlates with the villus loss and body weight defects seen within this model.

Figure 2.

Deletion of Atg14 within the mouse intestinal epithelium results in increased programmed cell death within crypts and villi. (a-d) Histological analysis of H&E and immune-stained small intestine sections from Atg14^f/f VC⁺ and littermate control mice at 3, 5, and 6 wk of age; n = 5 mice/group, >100 crypt-villus units/mouse; two-way ANOVA with Sidak’s multiple comparisons test (variable = genotype). (a) Mean apoptotic bodies ± SEM. (b-c) Percent cleaved CASP3 (cCASP3) positive cells (b) within crypts ± SEM, and (c) within villi ± SEM. (d) Representative immunohistochemical staining for cleaved CASP3 (cCASP3) from 3- and 5-week-old Atg14^f/f VC⁺ and control mice; bars: 100 μm. Insets show one crypt unit at higher power; bars: 25 μm. All statistically significant pairwise comparisons are displayed *P < 0.05, ***P < 0.001, ****P < 0.0001.

TNF triggers caspase-mediated apoptosis within Atg14^f/f VC⁺ intestinal epithelial cells

With the finding that Atg14^f/f VC⁺ mice undergo robust and spontaneous cell death in the intestinal epithelium, we next sought to assess the proliferation and survival of primary intestinal epithelial cells in vitro. To test this hypothesis, we isolated and cultured small intestinal stem cells from Atg14^f/f VC⁺ and control mice [31]. We first compared intestinal epithelial cell growth and survival in culture over time using MTT and EdU assays (see methods) and found no intrinsic differences in these parameters (Figure 3(a-b)).

Based on these results, we hypothesized that an extrinsic factor triggers cell death in Atg14^f/f VC⁺ epithelial cells in vivo. We performed a screen on Atg14^f/f VC⁺ and control epithelial cells using sub-lethal doses of various cytokines and chemicals that are expressed in the intestine (Figure S3). We cultured these epithelial cells under conditions that enriched for enterocytes [32] as these cells are the predominant cell type on villi which are lost in this model. We found that recombinant mouse TNF was the most potent inducer of cell death; this cytokine reduced cell viability in a dose-dependent manner in Atg14^f/f VC⁺ enterocytes relative to littermate control derived cells (Figure 3(c)). These data suggest that TNF is the trigger for apoptotic cell death within Atg14-deficient epithelial cells.

Figure 3.

Atg14-deficient intestinal epithelial cells are sensitized to TNF-induced apoptosis. (a-h) Mouse jejunal spheroids from Atg14^f/f VC⁺ and Atg14^f/f VC⁻ control mice maintained as stem cells (a and b) or differentiated to enterocytes (c-h) n = 3–6 independent experiments. (a) Cellular metabolic activity ± SEM determined by MTT assay on day 1, 2 and 3 after passage (normalized by genotype to day 1); Repeated measures two-way ANOVA with Sidak’s post-test. (b) Cellular proliferation ± SEM determined by EdU assay on day 2; Unpaired Student’s t-test. (c) Relative ATP cell viability ± SEM in cells treated for 12 h with indicated dose of mouse TNF; luminescence normalized to untreated cells. Unpaired Student’s t-test. (d) Representative immunohistochemical stained section and (e) quantitation of cCASP3 positive cells within the wall of spheroids treated with TNF or vehicle; n = 8–10 spheroids/sample; bars: 20 μm. Insets show higher power image of cells; bars: 5 μm. (f) Representative immunoblot for cleaved CASP3 (cCASP3), cleaved CASP8 (cCASP8), and ACTB/β-ACTIN in cells treated with 10 ng/ml mouse TNF for indicated duration. (g) Representative immunoblot for cCASP3 and cCASP8 in cells treated with 10 ng/ml mouse TNF and/or 20 µg/ml Z-VAD-FMK for 12 h; Z = Z-VAD-FMK, T = TNF, TZ = TNF + Z-VAD-FMK. (h) Relative ATP cell viability ± SEM after 12 h treatment 1 ng/ml mouse TNF, 20 µg/ml Z-VAD-FMK, 10 µM nec-1, and/or 50 µg/ml CHX; two-way ANOVA, Sidak post-test. *P < 0.05, **P < 0.01, ****P < 0.0001.

We next evaluated the effects of TNF on cleaved CASP3 (cCASP3) and cleaved CASP8/caspase-8 (cCASP8), markers of apoptosis. We found that the relative abundance of cCASP3 was increased in cryo-sections of TNF-treated Atg14^f/f VC⁺ enterocytes as compared to control enterocytes (Figure 3(d-e)). By immunoblot analysis, cCASP3 and cCASP8 were first detected after 4 h of TNF-treatment of Atg14^f/f VC⁺ enterocytes but not similarly treated controls (Figure 3(f)). We next tested the effects of a pan-caspase inhibitor Z-VAD-FMK in this system. After 12 h of TNF stimulation, cCASP3 and cCASP8 remained elevated within Atg14^f/f VC⁺ enterocytes as compared to controls. The addition of Z-VAD-FMK reduced the abundance of the active form of cCASP3 as well as the p18 active form of cCASP8 (Figure 3(g)). Under both these conditions, we neither detected phosphorylation of RIPK1 on S166, nor of RIPK3 on T231/S232, markers that correlate with necroptosis, a caspase-independent form of cell death. As a positive control for this assay, these markers were detected when jejunal enterocytes were challenged with a combination of TNF, the cell death inducer GDC-0152, and the pan-caspase inhibitor Z-VAD-FMK (Figure S4(a-b)). Additionally, we measured the viability of Atg14-deficient spheroids through the Cell Titer Glo ATP assay when exposed to TNF in conjunction with Z-VAD-FMK (Z-VAD), a pan-caspase inhibitor and/or necrostatin-1 (Nec-1), a chemical inhibitor of RIP Kinase 1-mediated forms of cell death. Treatment with Z-VAD, but not Nec-1, rescued TNF-triggered death within Atg14^f/f VC⁺ cells (Figure 3(h)). Together, these data suggest that Atg14^f/f VC⁺-deficient cells are specifically and potently sensitive to a caspase-dependent form of cell death triggered by the cytokine TNF.

TNF blockade prevents intestinal pathology in Atg14^f/f VC⁺ mice

The differential response to TNF treatment of Atg14-deficient enterocytes versus controls supports the hypothesis that ATG14 inhibits sensitivity to TNF mediated cell death. To connect these in vitro findings to the in vivo mouse model, we measured serum TNF levels in Atg14^f/f VC⁺ mice and controls. At 3 wk of age, a time point that precedes villus loss but still showed evidence of increased apoptosis, TNF serum levels were at or below levels of detection on both groups of mice; by 5 and 6 wk of age, Atg14^f/f VC⁺ mice showed only a modest increase in TNF levels that ranged from 3–10 pg/ml (Figure 4(a)). Interestingly, TNF levels did not show a further increase at week 6 (as compared to week 5 values), a time when the area of villus loss was substantially increased. For comparison, mouse models with severe spontaneous intestinal inflammation (e.g. mice with deletion of Il10rb/Il10r2 (interleukin 10 receptor, beta) and T cell specific mutation of Tgfbr2 (transforming growth factor, beta receptor II), have serum levels of TNF that reach 400 pg/ml [30]. Thus, we propose that the primary effect of ATG14 loss of function in intestinal epithelium is increased sensitivity to TNF mediated cell death.

Figure 4.

TNF neutralization rescues the Atg14-deficient mouse intestinal epithelium from intestinal pathology. (a) Serum TNF levels ± SEM of Atg14^f/f VC⁺ and control mice at 3, 5, and 6 wk; n = 4 mice/group; Two-way ANOVA with Sidak’s multiple comparisons test. (b-e) Atg14^f/f VC⁺ and control mice administered intraperitoneal TNF blocking or IgG isotype control antibody; n > 5 mice/group from n = 2 independent experiments. (b) Percent weight gain from 3 to 4.5 wk of age; repeated measures three-way ANOVA. (c) Representative H&E-stained jejunal sections from 5-week-old littermates; bars: 500 μm. (d and e) >100 crypt-villus units analyzed/mouse; two-way ANOVA with Sidak’s multiple comparisons test (variable = genotype). (d) Mean apoptotic bodies ± SEM within crypts. (e) Percent cCASP3 positive cells ± SEM. All statistically significant pairwise comparisons are displayed; ***P < 0.001, ****P < 0.0001.

To functionally test the role of TNF in vivo, we administered TNF blocking and isotype control antibodies [33] to groups of mice twice a week from 3 to 5 wk of age (Figure S5(a)). Atg14^f/f VC⁺ mice injected with TNF blocking antibodies exhibited similar weight gain as their littermate controls, while Atg14^f/f VC⁺ mice treated with isotype control showed failure to thrive (Figure 4(b)). In addition, histological sections of small intestines of Atg14^f/f VC⁺ mice treated with anti-TNF showed complete rescue of intestinal pathology as these intestines were devoid of villus loss and crypt atrophy as well as disruption of intestinal architecture (Figure 4(c)). Furthermore, Atg14^f/f VC⁺ mice treated with anti-TNF had similar numbers of apoptotic bodies and cCASP3 positive cells as their littermate controls (Figures 4(d-e) and S5(b)). Conversely, Atg14^f/f VC⁺ mice injected with isotype control showed severe intestinal pathology and increased cCASP3 staining (Figures 4(d-e) and S5(b)). Accordingly, Atg14^f/f VC⁺ mice treated with anti-TNF showed significant restoration in crypt and villus height as well as M-phase cells, compared to that of controls; Atg14^f/f VC⁺ animals treated with isotype control demonstrated crypt hyperplasia, villus loss, and elevated numbers of M-phase bodies (Figure S5(c-e)). Taken together, systemic TNF neutralization shows that this cytokine is necessary to alter the pathology of Atg14^f/f VC⁺ mice.

Atg14^f/f VC⁺ tnfrsf1a^-/- mice do not show villus loss and epithelial apoptosis

We next used a genetic model to test the role of the TNF receptor in driving the pathology of Atg14^f/f VC⁺ mice. Under basal conditions, TNF signals through two distinct receptor signaling complexes, TNFRSF1A/TNFR1 and TNFRSF1B/TNFR2, to activate NF-κB -mediated transcription and promote proliferation and inflammation [34]. However, functional [35] and molecular [36] studies have solely implicated TNFRSF1A for apoptosis. Therefore, we generated Atg14^f/f VC⁺ tnfrsf1a^-/- double knockout mice to confirm the requirement of TNF signaling through TNFRSF1A for the development of the Atg14^f/f VC⁺ phenotype.

We found that Atg14^f/f VC⁺ tnfrsf1a^-/- showed divergent weight gain as compared to Atg14^f/f VC⁺ Tnfrsf1a^+/- littermates (Figure 5(a)). Furthermore, when analyzing the histology sections of 5-week-old Atg14^f/f VC⁺ tnfrsf1a^-/- small intestines, villus and crypt height was unremarkable (Figures 5(b) and S6(a-b)). While Atg14^f/f VC⁺ Tnfrsf1a^+/- mice showed loss of villi, Atg14^f/f VC⁺ tnfrsf1a^-/- showed complete protection from villus loss and crypt hyperplasia (Figure 5(b)). Furthermore, quantitative analysis of histology, proliferation and cell death showed that Atg14^f/f VC⁺ tnfrsf1a^-/- were comparable to Atg14 sufficient controls while Atg14^f/f VC⁺ Tnfrsf1a^+/- showed increased cell death in crypts and increased epithelial proliferation (Figures 5(c-e) and S6(c)). Levels of TNF in the serum were also reduced in all Atg14^f/f VC⁺ tnfrsf1a^-/- mice below 5 pg/ml (Figure 5(f)). Furthermore, Atg14^f/f VC⁺ tnfrsf1a^-/- enterocytes treated with TNF showed no alterations in survival consistent with an epithelial cell specific role for TNFRSF1A (Figure S6(d)). As a control, we confirmed that the transcript abundance of Tnfrsf1a was similar in Atg14^f/f VC⁺ and Atg14^f/f VC^- intestinal epithelial cells (Figure S7). Collectively, these results indicate Atg14^f/f VC⁺ intestinal pathology is dependent on TNFRSF1A signaling and reveal a novel role of Atg14 in protection of TNF-induced apoptosis.

Figure 5.

TNF Receptor 1 deletion in Atg14-deficient mice rescues intestinal pathology. (a) Percent weight gain between 3 to 5 wk of age; Repeated measures two-way ANOVA. (b-f) Histological analysis in H&E stained small intestine sections from Atg14^f/f VC⁺ tnfrsf1a^-/- and control mice (all littermates) among 5-week-old mice; n = 6–10 mice/group, >100 crypt-villus units quantified/mouse; two-way ANOVA with Sidak’s multiple comparisons test. (b) Representative H&E stained jejunal sections from 5-week-old mice; bars: 500 μm. (c) Mean apoptotic bodies ± SEM. (d) Percent cCASP3 positive cells ± SEM. (e) Mean M-phase figures ± SEM. (f) Serum TNF levels ± SEM from indicated genotypes at 5 and 6-weeks. All statistically significant pairwise comparisons are displayed, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Rb1cc1^f/f VC⁺ mice show spontaneous villus atrophy and epithelial apoptosis similar to Atg14^f/f VC⁺ mice

Our hypothesis is that ATG14 prevents intestinal epithelial apoptosis due to its role in autophagy initiation. To test this hypothesis, we evaluated the effects of loss of function for RB1CC1 in the intestinal epithelium; RB1CC1 functions in a separate, more proximal step of autophagy initiation as part of the ULK1 complex [24]. The small intestines of mice with an intestinal epithelial specific deletion of Rb1cc1/Fip200 (Rb1cc1^f/f VC⁺) demonstrated a pattern of villus loss in the mid-jejunum beginning at 5 wk of age similar to the phenotype of Atg14^f/f VC⁺ mice (Figure 6(a-c)). The villus loss and crypt hypertrophy was accompanied by increased apoptosis and crypt cell proliferation (Figure 6(d-g)). Primary small intestinal spheroids isolated from Rb1cc1^f/f VC⁺ mice and differentiated to the enterocyte lineage demonstrated a dose-dependent decrease in viability when challenged by recombinant mouse TNF for 12 h (Figure 6(h)). Taken together, loss of function of RB1CC1 in the intestinal epithelium led to similar effects as loss of function of ATG14, both in vivo and in isolated intestinal epithelial cells.

Figure 6.

Deletion of Rb1cc1/Fip200 in the mouse intestinal epithelium results in increased programmed cell death. (a-e) Histological analysis of H&E stained small intestine sections from Rb1cc1^f/f VC⁺ and littermate control mice between 5 and 9 wk of age; n ≥ 4 mice/group, bar: 100 µm; (b-g) Two-tailed Student’s t-test. (a) Representative images. (b) Villus height ±SEM. (c) Crypt height ±SEM. (d) Quantification of pyknotic and karrorhexic nuclei ±SEM within crypts. (e) Quantification of M-phase nuclei ±SEM within crypts. (f and g) Immunohistochemical analysis of cCASP3-stained intestine sections from Rb1cc1^f/f VC⁺ and littermate control mice between 5 and 9 wk of age; n ≥ 4 mice/group, bar: 100 µm (high power inset: bar: 50 µm). (f) Representative images. (g) Percent of cells positive for cCASP3 ± SEM within crypts and villi, mean of 10 crypts or villi/mouse. (h) Relative cell viability determined by Cell Titer Glo ATP viability assay of indicated genotypes treated with indicated doses of 12-h TNF treatment. TNF+CHX: 100 ng/ml TNF + 50 µg/ml Cycloheximide; n = 6 assays; 2-way ANOVA with Dunnett’s post-test. **P < 0.01, ***P < 0.001, ****P < 0.0001.

In contrast, small intestinal sections from Atg5^f/f VC⁺ mice demonstrated no significant alternations in either villus or crypt height relative to littermate control animals (Figure S8(a-b)). ATG5 functions at a distinct step of the autophagy pathway. It is not involved in the entry steps of the pathway and there are no spontaneous defects in villus homeostasis, nor cell death [37]. Meanwhile ATG14 and RB1CC1 share a common function as part of two distinct complexes to regulate entry into the autophagy pathway [38]. These data suggest that Atg14 and Rb1cc1 loss of function models share a common defect that sensitize the small intestinal cell to TNF-triggered apoptosis.

Discussion

This work establishes an essential role for Atg14 within the intestinal epithelium as its conditional loss of function in these cells leads to spontaneous and rapidly progressive villus atrophy. Furthermore, we found that the villus atrophy is a consequence of increased intestinal epithelial cell death that is mediated by TNF–triggered apoptosis. As the loss of function of Rb1cc1 in the intestinal epithelium also results in a similar phenotype, we propose that initiation of autophagy is required to regulate intestinal epithelial cell turnover and homeostasis.

Our findings expand upon the known roles of autophagy within the intestinal epithelium. Polymorphisms in Atg16l1 are associated with Crohn disease and result in defective morphology of secretory cells in the intestinal epithelium [27]. Mouse loss of function studies of Atg5, Atg7 and Atg16l1 in the intestinal epithelium identify that the autophagy pathway is required for the secretory function of Paneth and goblet cells [26,39]. Increased cell death of the absorptive enterocyte lineage is not noted in these experiments; however increased cell death in Paneth cells occurs when Atg16l1 hypomorphic mice are crossed to Xbp1 (x-box binding protein 1) knockout mice due to an enhanced unfolded protein response [10,40].

Mice that are deficient or mutant for autophagy proteins at ATG5 and ATG16L1 in the intestinal epithelium show enhanced intestinal epithelial cell death when challenged by intestinal infection (Toxoplasma Gondii, Helicobacter hepaticus, mouse Norovirus) or cell damage (DSS, GVHD, smoking) [18,19,41]. This is in contrast to the effects of loss of function of the autophagy initiation factors RB1CC1 and ATG14 that we have demonstrated to have spontaneous, more severe intestinal pathology. It is interesting that for all of these models, the cytokine TNF was central to the development of pathology.

TNF signaling is well known to mediate two specific outcomes: survival through the stimulation of NF-κB stimulated genes and death through CASP3 and CASP8 mediated apoptosis or alternative forms of programmed cell death [36,42,43]. Matsuzawa-Ishimoto, et al show that Atg16l1^f/f VC⁺ organoids demonstrate programmed cell death upon long-term TNF treatment, modeling the susceptibility of Atg16l1^f/f VC⁺ to TNF-triggered intestinal epithelial death secondary to a combination of intestinal viral infection and damage (i.e. murine norovirus and DSS). Within the context of our model of Atg14 intestinal deficiency, we demonstrated an increase in CASP3 mediated apoptosis. Furthermore, we demonstrated that within 8 h, there is a potent activation of the CASP8, CASP3 axis that is not observed in littermate control cells. Cell death is rescued though the addition of the caspase-inhibitor Z-VAD in vitro, validating the role of caspases in mediating loss of cell viability. This work provides a robust spontaneous model to study the interaction between the intestinal autophagy pathway and the TNF-triggered apoptotic pathways that govern cellular turnover.

In other cell lineages, defects in factors that mediate the initial steps of the autophagy pathway have been linked to the TNF triggered apoptotic cascade. Atg13-deficient mice undergo spontaneous apoptotic cell death within the heart and the liver [44]. Mouse embryonic fibroblasts (MEFs) derived from atg13^-/- and rb1cc1^-/- animals are sensitive to apoptotic cell death when challenged with TNF [41]. Furthermore, Becn1 deletion within a subset of neurons results in defective autophagy and increased apoptosis [45]. These reports all highlight phenotypes where defects in the initiation steps of the autophagy pathway, not terminal steps of conjugation or egress, lead to cellular lethality.

Numerous reports highlight a functional distinction among factors that regulate initiation into the autophagy pathway as compared to termination factors regulating LC3 conjugation (ATG5, ATG16L1, and ATG7) or lysosomal fusion [21,46–48]. In particular, the LC3 conjugation complex, ATG5, ATG16L1, and ATG7, is required to suppress pathogenicity of the parasite Toxoplasma gondii, while ATG14 and other components of the initiation and fusion steps of autophagy are not required for this function [46,48]. Furthermore, recent reports demonstrate that cells deficient in components of the autophagy conjugation complex, such as ATG5, remain able to undergo a basal level of autophagic flux, while loss of ATG14 completely blocks autophagic fusion with lysosomes to degrade cargo [21]. Therefore, we propose a model in the intestinal epithelium in which initiation factors such as ATG14 and RB1CC1 have unique and critical roles to protect against TNF-induced apoptosis. However, the molecular mechanism that distinguishes the function of initiation and conjugation/termination complexes still needs to be resolved.

This work lays a foundation for future studies of the interaction between autophagy and TNF signaling in the intestinal epithelium. We propose that the intestine is an amenable system to study this question. There are predictable spatial and temporal effects in the small intestine for loss of autophagy (i.e. phenotype begins in the mid small intestine and spread proximally and distally). The relative sensitivity of intestinal epithelial cells in different regions of the intestine suggests that there may be additional as of yet unknown factors involved in the spatial control of sensitivity to TNF mediated cell death. Future work is also needed to assess in human the extent to which polymorphisms in autophagy mediators also sensitize the intestine and other organs to TNF mediated cell death.

Materials and methods

Mice

Atg14^flox/flox (Atg14^tm1Aki), Rb1cc1^flox/flox (Rb1cc1^tm1.1Guan ), Atg5^flox/flox (Atg5^tm1Myok) mouse strains [39] were bred to Vil1-Cre mice [25](Jackson labs) to conditionally delete Atg14, Fip200 and Atg5, respectively, from the intestinal epithelium. tnfrsf1a^-/- (Tnfrsf1a^tm1Imx) mice [49] were obtained from Jackson Laboratories and crossed to Atg14^flox/floxVC⁺ to generate Atg14^flox/floxtnfrsf1a ^-/- VC⁺ mice. All mice were maintained at the SPF animal facilities of Washington University in St. Louis and kept in enhanced autoclaved cages with autoclaved regular chow diet and water. All lines were maintained on a C57BL/6 genetic background. All experimental animals were handled according to the institutional and national American animal regulations. Animal protocols were approved by the ethics committee of Washington University.

Serum cytokine analysis

Mouse blood was collected through terminal cardiac puncture in accordance with all animal protocols and regulations. Serum was isolated by centrifugation of blood at 18,000 x g and collecting the supernatant. Serum was stored at -80°C before analysis. Serum TNF concentrations were measured using a sandwich ELISA utilizing mouse TNF-α ELISA MAX Standard kit (Biolegend, 430902).

Primary intestinal epithelial cell culture

Small intestinal jejunal crypts were isolated and cultured from mice as previously described [50]. Briefly, stem cell media containing 50% L-WRN (WNT3A, RSPO3 and NOG) conditioned medium: was used to culture epithelial spheroids enriched for stem cells [31]. To differentiate spheroids into enterocytes, spheroids were incubated in Advanced DMEM/F-12 (Invitrogen, 12634028), with 2 mM L-glutamine, 100 units/ml penicillin, and 0.1 mg/ml streptomycin, freshly supplemented with 50 ng/ml EGF (Peprotech, 315–09), 10 µM ROCK1/ROCK2 inhibitor Y-27632 (R&D Systems, 1254) and 10 µM PTGER4 inhibitor L-161,982 (R&D Systems, 2514).

All enterocyte differentiated spheroids were cultured in differentiation media for 24 h followed by differentiation media ± (1–100 ng/ml) of recombinant mouse TNF (Biolegend, 575206) for another 12 h for optimal cell death assay (unless indicated otherwise). Cycloheximide (50 μg/ml; Enzo, ALX-380–269-G001) along with 100 ng/ml of TNF (Biolegend, 575206) were used as positive control for cell viability assays. Additionally, 20 µg/ml Z-VAD-FMK (APExBIO, A1902), 10 µM necrostatin-1 (MedChemExpress, HY-15760), 1 µM Staurosporine (MedChemExpress, HY-15141), and 20 µM GDC-0152 (APExBIO, A4224) were used as indicated to rescue or induce cell death, respectively.

Intestinal spheroid viability assays

Small intestinal spheroids were collected in cell recovery solution (Corning, 354253), fixed in 4% paraformaldehyde in PBS for 16 h at 4°C, immersed in 20% sucrose solution for 16 h at 4°C and then processed in O.C.T. compound (Fisher, 23-730-571). 5 µm sections were additionally fixed with 4% PFA prior to immunohistochemical staining. Spheroid imaging was performed using published methods [51]

Small intestinal epithelial stem cells were cultured as previously described and plated in a 96 well plate for 3 d for MTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide) cell proliferation assay (American Type Culture Collection, 30-1010K). MTT reagent was added to each well in 1:10 dilution and the cells were incubated at 37°C for 2 h. Next, 100 µL of MTT detergent reagent was added to each well and incubated in the dark at room temperature for 2 h. Absorbance reads were measured at 570 nm using Cytation 5. Cell proliferation was measured as fold change with values at day 3 of culture normalized to day one.

For EdU proliferation assays, small intestinal epithelial cells were cultured for 24 h then stained for EdU for 2 h. Next, the cells were trypsinized into single cells, washed and collected for fixation and EdU labeling. The cells were fixed in 10% formalin for fifteen minutes, washed and stained with 488-Azide Fluorophore for thirty minutes in the dark. The cells were then washed and sorted through Fluorescence activated cell sorting (FACS) and analyzed for % live EdU positive cells. EdU assay was performed using Click-iT EdU Alexa Fluor 488 Imaging Kit (Invitrogen, C10639).

Small intestinal spheroids were expanded as previously described and plated in a 96 well clear bottom white-walled plate (Corning, 3610). Spheroids were differentiated to enterocytes in the presence of PTGER4 inhibitor [32] and treated for 12 h with recombinant mouse TNF or other listed compounds. Dose-dependent viability was measured in jejunal enterocyte differentiated spheroids using Cell Titer Glo 3D ATP viability reagent (Promega Life Sciences, G9682), and luminescence was detected using a Cytation 5 multi-mode plate reader (Biotek). Luminescence values were averaged by treatment group and normalized to the vehicle control group per genotype.

Immunoblot analysis

Protein lysates were prepared from differentiated enterocytes cultured from small intestinal crypts as previously described [50]. Differentiated enterocytes were collected in cell recovery solution (Corning, 354253), centrifuged at 4°C at 200 x g for 10 min, lysed using 200 μl of RIPA buffer (Sigma, R0278) containing protease and phosphatase inhibitor (Cell Signaling Technologies, 5872) and sonicated (QSonica). Proteins were separated by SDS-PAGE (Bio-Rad), transferred to nitrocellulose, and analyzed by immunoblotting. Microsoft Office was used to convert scanned immunoblot images to grayscale and to crop images.

Histology

Freshly isolated small intestine and colon tissue were pinned and fixed in 10% neutral buffered formalin (Sigma, HT501128). Tissues were then processed for paraffin embedding and hematoxylin and eosin staining using published protocols [52].

For immunohistochemistry, paraffin embedded sections were deparaffinized by incubating the sections in xylene and isopropanol for 3 washes (5 min each). The sections were then incubated in 10% hydrogen peroxide in methanol, rinsed in PBS, and boiled in Trilogy solution (Cell Marque, 920P-09) for 20 min. The sections were incubated in blocking buffer, (10 mg/ml bovine serum albumin/0.1% Triton-X 100) for 30 min and incubated with primary antibody overnight at 4°C. After overnight primary antibody incubation, sections were rinsed in PBS, incubated with species specific biotinylated secondary antibodies (Invitrogen, 31820) and Vectastain ELITE ABC kit was used to label streptavidin-HRP (Vector Laboratories, PK-6100). Staining were then visualized by DAB (2, 2’-diaminobenzidine) peroxidase substrate (Vector laboratories, SK-4100). Olympus BX51 microscope and DP Controller software were used to obtain bright-field images.

Antibodies and reagents

Immunohistochemistry was performed using rabbit monoclonal anti-cleaved CASP3 Asp175 (Cell Signaling Technologies, 9664), rabbit monoclonal anti-cleaved CASP8 Asp387 (Cell Signaling Technologies, 8592), mouse polyclonal anti-ACTB (Abcam, ab6276), rat monoclonal LY6G (BioLegend, 127602), and corresponding species-specific secondary antibodies.

Immunoblotting was performed using rabbit monoclonal anti-cleaved CASP3 (Cell Signaling Technology, 9664), mouse polyclonal anti-ACTB (Abcam, ab6276), rabbit monoclonal anti-cleaved CASP8 (Asp387) (D5B2) (Cell Signaling Technology, 8592). All antibodies were diluted 1:1000 in Blocking One solution (Nacalai, 03953–95) and incubated 16 h at 4°C. Species appropriate HRP-conjugated secondary antibodies (ThermoFisher, 32260; Abcam, ab102448) were diluted 1:15,000 in Blocking One solution (Nacalai, 03953–95) and incubated 1 h at 24°C. Clarity ECL (Bio-Rad, 1705060) and Prosignal Dura ECL (Prometheus, 20–301) was used for detection with autoradiography film.

Mouse TNF neutralization studies

For TNF neutralization in mice, low endotoxin rat anti mouse TNF (clone MP6-XT22) 1.0 mg/ml; (Biolegend, 506332) or rat IgG1 Isotype, κ control (Biolegend, 400432) was administered by 500 µl intraperitoneal injections twice per week during the entire experimental period. Littermates were used as controls in all experiments. Groups of mice for the experiment include male and female animals.

PCR genotyping and qPCR

The Atg14^f/f strain was genotyped by PCR of tail DNA using a forward primer, 5’-CCC ATC TCC ATT CCT GGA TTA CTG GAC-3’, a reverse primer #1, 5’-ACA AGA TGC AGA ACT GAT GGC AGG-3’, and a reverse primer #2, 5’-ACA GAG TTA GTT CCA GGA CAG CCA GG-3’ to generate a 430 base pair PCR fragment for the wild-type allele and a 530 base pair PCR fragment for the Atg14 floxed allele. The genotyping PCR protocol for Atg14^f/f mice is as follows: 1 min at 94°C, 30 cycles of annealing of 1 min at 94°C, 1 min at 65°C and 2 min at 72°C, elongation period of 10 min at 72°C, and final 4°C incubation. Vil1-Cre mice were genotyped using a forward primer 5’- GTG TGG GAC AGA GAA CAA ACC -3’ and a reverse primer 5’- ACA TCT TCA GGT TCT GCG GG -3’. Tnfrsf1a was genotyped by PCR using a forward primer 5’- GCT ACT TCC ATT TGT CAC GTC C -3’, a reverse primer #1, 5’- ATG GGG ATA CAT CCA TCA GG -3’, and a reverse primer #2, 5’ GGG GAA CAT CAG AAA CAA GC -3’ to generate a 362 base pair PCR fragment for the wild-type allele and a 270 base pair PCR fragment for the TNF receptor 1 allele. Vendor PCR protocol for TNF receptor was used to genotype TNF receptor 1 allele. Tnfrsf1a mRNA was measured through qPCR using forward primer, 5’- GCT CTG CTG ATG GGG ATA CAT C-3’, and a reverse primer, 5’-ACC TGG GAC ATT TCT TTC CGA C-3’.

Statistical analysis

All statistical analysis were performed using GraphPad Prism software (version 7.01) by either unpaired Student’s t test or 2-way ANOVA Sidak’s multiple T-test or Tukey’s T-test, as indicated.

Supplementary information on serum metabolite analysis, immunoblot analysis, and PCR genotyping and qPCR, is provided in SI Materials and Methods.

Funding Statement

This work was supported by the Crohn’s & Colitis Foundation ; Kenneth Rainin Foundation; National Heart, Lung, and Blood Institute [T32 HL007317]; National Institutes of Health [P30DK052574].

Acknowledgments

We thank all lab members for discussion and helpful comments especially Chin-Wen Lai and Aaron Ver Heul for support with critical experiments and Darren Kreamalmeyer for animal care and breeding. The Washington University Digestive Disease Research Center (Morphology Core) was funded by NIH P30DK052574. J.S.L-E. was funded by NIH/NHLBI T32 HL007317.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplemental data for this article can be accessed here.