T-cell-derived Interferon-γ programs stem cell death in immune-mediated intestinal damage

Abstract

Despite the importance of intestinal stem cells (ISCs) for epithelial maintenance and the significance of epithelial injury in immune-mediated intestinal diseases, there is limited understanding of how immune-mediated damage impacts ISCs and their niche. We found that stem cell compartment injury is a shared feature of both alloreactive and autoreactive intestinal immunopathology, reducing ISCs and impairing their recovery in T-cell-mediated injury models. While imaging revealed few T cells near the stem cell compartment in healthy mice, donor T cells infiltrating the intestinal mucosa after allogeneic bone marrow transplantation (BMT) primarily localized to the crypt region lamina propria. Further modeling with ex vivo epithelial cultures indicated ISC depletion and impaired human as well as murine organoid survival upon co-culture with activated T cells, and screening of effector pathways identified Interferon-γ as a principal mediator of ISC compartment damage. Interferon-γ induced JAK1- and STAT1-dependent toxicity, initiating a pro-apoptotic gene expression program and stem cell death. BMT with Interferon-γ-deficient donor T cells, with recipients lacking the Interferon-γ receptor (IFNγR) specifically in the intestinal epithelium, and with pharmacologic inhibition of JAK signaling all resulted in protection of the stem cell compartment. Additionally, epithelial cultures with Paneth-cell-deficient organoids, IFNγR-deficient Paneth cells, IFNγR-deficient ISCs, and purified stem cell colonies all indicated direct targeting of the ISCs that was not dependent on injury to the Paneth cell niche. Dysregulated T cell activation and Interferon-γ production are thus potent mediators of ISC injury, and blockade of JAK/STAT signaling within target tissue stem cells can prevent this T-cell-mediated pathology.

One Sentence Summary

T-cell-derived IFNγ can directly target intestinal stem cells to induce their apoptosis in a JAK/STAT-dependent manner.

Introduction

Epithelial stem cells are critical for physiologic self-renewal as well as regeneration after injury (1). The trans-membrane protein leucine-rich repeat-containing G protein-coupled receptor 5 (Lgr5) marks crypt base columnar intestinal stem cells (ISCs) capable of regenerating all the cells of the epithelium in the small intestine (SI) and large intestine (LI) (2). Paneth cells, which are progeny of ISCs, provide an epithelial niche for Lgr5⁺ ISCs in SI by producing growth factors including Wnt3 and epidermal growth factor (EGF) (3, 4). Despite the importance of the stem cell compartment for epithelial maintenance and regeneration after gastrointestinal (GI) damage (5, 6), and despite increasing evidence for immunologic effects on tissue regeneration (7–9), there is little understanding of the effects of immune-mediated damage on tissue stem cells.

The GI tract is a frequent site of tissue damage after allogeneic hematopoietic/bone marrow transplantation (BMT), and injury to intestinal crypt epithelium is a characteristic finding of graft vs. host disease (GVHD) in transplant recipients (10, 11). GVHD is an immune-mediated complication of BMT in which donor T cells attack recipient tissues. The crypts contain the stem cells and progenitors of the intestinal epithelium, and it has been reported that both ISCs and their Paneth cell niche are reduced in mice with GVHD (8, 12–15). However, the mechanisms leading to their loss, the relationship between these cell populations during tissue injury, and the relevance of these findings to tissue damage beyond the transplant setting are all poorly understood.

Cytotoxicity and cytokine production are principal effector functions of T cells, and both functions have been studied considerably in GVHD models (16–29). Although T cells can mediate potent tissue damage in the GI tract, the impacts of cytokine signaling and cytotoxicity on the ISC compartment are not well defined. Inflammatory cytokines such as IFNγ and TNFα have been associated with damage to the Paneth cell niche (30–32), and IFNγ contributes to reduced epithelial proliferation in mice with colitis (33). In contrast to how group 3 innate lymphoid cells and IL-22 can signal to ISCs to protect them and promote epithelial regeneration, it is possible that there are also direct interactions between ISCs and inflammatory cytokines during pathologic immune responses that compromise the ISC compartment. We thus sought to examine the specific cellular interactions and molecular mechanisms underlying ISC loss in immune-mediated GI damage. Using a combination of phenotypic and functional characterizations of the ISC compartment after alloreactive and autoreactive intestinal injury in vivo, coupled with ex vivo modeling of T cell interactions with ISCs and their Paneth cell niche in organoid cultures, we found that ISCs can be directly targeted by T-cell-derived cytotoxic cytokine signaling.

Results

Alloreactive and autoreactive immune responses impair the intestinal stem cell compartment

We first evaluated ISC kinetics in a clinically relevant major histocompatibility complex (MHC)-matched allogeneic BMT model. Three days after transplantation, BMT recipients receiving marrow alone (no GVHD) or marrow and T cells (for induction of GVHD) both demonstrated a reduction in SI Lgr5⁺ ISCs compared to normal mice (Fig. 1, A and B, top panels). On day 10 post-BMT, Lgr5⁺ ISC numbers had recovered in recipients transplanted without T cells, but ISC numbers remained reduced in GVHD recipients transplanted with donor T cells, demonstrating impairment of ISC recovery in immune-mediated GI damage occurring after BMT (Fig. 1, A and B, bottom panels). In contrast, lysozyme⁺ Paneth cell numbers remained intact early after transplant, but were reduced by day 10 post-BMT in GVHD mice (Fig. 1C and fig. S1A), indicating that ISCs were reduced prior to Paneth cells after allogeneic BMT. Testing an independent haploidentical MHC-mismatched model also demonstrated rapid Lgr5⁺ ISC reduction followed by substantial recovery in mice without GVHD, but persistent diminution of Lgr5⁺ ISCs in T cell recipients (Fig. 1D). Once again, reduction of Paneth cells in this model only occurred following the reduction of ISCs (Fig. 1E and fig. S2).

Fig. 1. Alloreactive and autoreactive immune responses injure the intestinal stem cell compartment.

(A to C) LP-into-Lgr5-LacZ-B6 MHC-matched BMT. (A) Representative images of SI (ileum) Lgr5-LacZ staining on day 3 and day 10 post-BMT. Scale bars = 500μm (Upper) or 50μm (Lower). (B) SI ISC frequency (n = 8–25 independent sections/group) and (C) SI lysozyme⁺ Paneth cell frequency (n = 6–15 independent sections/group) on day 3 and day 10 post-BMT (D and E) B6-into-Lgr5-LacZ-BDF1 MHC-mismatched BMT, SI ISC frequency (D, n = 5–15 independent sections/group) and SI lysozyme⁺ Paneth cell frequency (E, n = 6–13 independent sections/group) on day 3 and day 10 post-BMT. (F and G) Foxp3-WT and Foxp3-DTR mice treated with DT. SI ISC frequency (F) and SI lysozyme⁺ Paneth cell frequency (G) 5 days after DT treatment (n = 19–20 independent sections/group). (H and I) Representative images and numbers of day 5 SI organoids from recipients day 3 (H) and day 10 (I) after LP-into-B6 BMT. Organoid culture from 150 crypts. Scale bars = 500μm (Upper) or 200μm (Lower). (n = 3 mice/group). (J) Day 5 SI organoid numbers from 150 crypts harvested 9 days after DT treatment (n = 4 mice/group). Data are mean and s.e.m.; comparisons performed with t-tests (two groups) or one-way ANOVA (multiple groups); *P < 0.05, **P < 0.01, ***P < 0.001. Data are representative of at least two independent experiments, or combined from two experiments (A to E)

To determine if these effects were specific to alloreactive damage and BMT, we examined the ISC compartment during systemic autoimmunity by crossing Foxp3-diphtheria toxin receptor (DTR) mice (34) with Lgr5-LacZ reporters. Mutations in the FOXP3 gene in humans result in IPEX (Immune dysregulation, polyendocrinopathy, enteropathy, X-linked) syndrome, which is frequently associated with intestinal pathology (35), and ablation of Foxp3⁺ regulatory T cells (Tregs) in mice leads to rapid systemic and intestinal autoimmune responses as well (36). Induction of systemic autoimmunity by DT-mediated Treg depletion quickly resulted in fewer Lgr5⁺ ISCs, while Paneth cell numbers were maintained (Fig. 1, F and G). Reduction of ISCs was thus a shared feature of alloreactive and autoreactive immune-mediated GI damage, and it occurred prior to epithelial niche impairment as manifested by Paneth cell deficiency.

We next evaluated the effects of immune-mediated damage on the ISC compartment functionally, assessing ex vivo organoid-forming capacity after in vivo challenge. Intestinal crypts are functional units containing epithelial stem, progenitor, and niche cells, and isolated crypts can generate intestinal organoids ex vivo. These organoids recapitulate in vivo intestinal organization with crypt-villus structures and central lumens (37). Consistent with the Lgr5⁺ ISC frequencies (Fig. 1, A–B and D), SI crypts isolated early post-transplant demonstrated significant impairment in organoid-forming capacity compared to normal mice (Fig. 1H). The functional ability to generate organoids rapidly recovered in crypts from mice without GVHD, but organoid formation remained impaired ten days after BMT with allogeneic T cells (Fig. 1I). Likewise, in comparison to cultures from control mice, Treg depletion in vivo significantly impaired ex vivo organoid formation from isolated crypts (Fig. 1J). Furthermore, we also examined the in vivo function of ISCs after BMT using genetic marking of the stem cells and their progeny. In addition to Lgr5, Olfm4 also marks ISCs in mouse SI, and Olfm4-driven Cre expression has been shown to be more robust than that of Lgr5 (38). We thus performed allogeneic BMT into Olfm4-CreERT2xRosa26 reporter mice, treating the BMT recipients with Tamoxifen prior to transplantation to activate Cre-driven lineage tracing (fig. S3). In comparison to normal controls, transplanted mice demonstrated reduced tracing from Olfm4⁺ cells, even in the absence of donor T cells, suggesting ISC damage due to the pre-transplant conditioning. Moreover, BMT with T cells led to significant further reduction in tracing, providing additional functional validation for the loss of ISCs in GVHD (fig. S3). In total, both alloreactive and autoreactive in vivo immune responses resulted in reduction of ISCs and functional impairment of the stem cell compartment.

Allogeneic T cells preferentially invade crypt region lamina propria after BMT

In order to examine the intramucosal localization of T cells mediating epithelial injury and ISC reduction after BMT, we next performed 3-D confocal microscopy of intact whole-mount intestinal tissue. This approach has recently identified preferential infiltration of crypt region mucosa after allogeneic BMT (39), but such localization has not been distinguished between the intraepithelial and lamina propria components of the mucosa, and it has not been defined in homeostasis either. After staining for CD3, nuclei, and cellular membranes, imaging of full-thickness ileum allowed for accurate determination of mucosal architecture as well as precise localization and quantification of T cells in the epithelial and lamina propria regions of SI crypt and villus compartments. Normal B6 mice demonstrated similar T cell densities in crypt and in villus regions of ileal lamina propria at steady state (Fig. 2, A and B). In contrast, intraepithelial T cells were much more abundant in villus epithelium than in crypt epithelium at steady state (Fig. 2, A and B).

Fig. 2. Donor T cells infiltrate the epithelial layer and lamina propria in the crypt region after bone marrow transplantation.

3-D whole-mount immunofluorescent confocal imaging of mouse ileum. (A and B) T cells in normal B6 mice were identified by anti-CD3 immunofluorescence. (A) Left panels: Representative 3-D projection images of full-thickness SI tissue divided into villus and crypt regions, with cellular membrane staining (DiD lipophilic dye; blue) indicating the tissue architecture utilized for distinguishing IELs and LPLs within the ileum; yellow, CD3⁺ IELs; red, CD3⁺ LPLs. Right panels: 3-D projections of CD3⁺ IELs and CD3⁺ LPLs in the villus and crypt regions, with cellular membrane staining removed and tissue orientation (and thus T cell localization within the 3-D tissues) indicated by 2-D slices shown on the posterior projection walls. (B) Quantification of CD3⁺ IEL and CD3⁺ LPL densities in normal B6 (n = 6 independent 3-D views/group). (C to F) B6-into-LP allogeneic BMT was performed using wild-type B6 marrow and purified GFP⁺ B6 T cells, with donor T cells in the epithelium shown in purple and donor T cells in the lamina propria shown in green. (C) Representative 3-D projections and (D) quantifications of donor T cells in the villus and crypt regions 4 days post-BMT (n = 24 independent 3-D views/group combined from 2 transplants). (E) Representative 3-D projections and (F) quantifications of donor T cells in the villus and crypt regions 7 days post-BMT (n = 12 independent 3-D views/group). Tissue orientation and T cell localization are again indicated by 2-D slices shown on the posterior walls of the 3-D projections. Graphs indicate mean and s.e.m.; comparisons performed with t-tests; **P < 0.01, ***P < 0.001. Data are representative of two independent experiments unless otherwise mentioned.

We next evaluated the location of donor T cells in recipient intestinal mucosa after allogeneic BMT, using B6-GFP mice as the source of donor T cells. Four days post-transplant, the early infiltration of GFP⁺ donor T cells in recipient ileum was primarily located in the lamina propria of the crypt region, and few donor T cells could be identified in the villi (Fig. 2, C and D). Three days later, donor T cell invasion of the mucosa was much more substantial, and again most donor T cells were located in the crypt region lamina propria (Fig. 2, E and F). In contrast to intraepithelial T cells in B6 ileum at steady state, intraepithelial donor T cells one week after BMT were significantly more frequent in the crypts than in the villi (Fig. 2, E and F). These results indicated that donor T cells mediating GVHD primarily infiltrated the lamina propria of the crypt region, in proximity to the ISC compartment, and most donor T cells invading the intestinal epithelium after BMT were present in the crypt region as well.

Activated T cells produce IFNγ that targets the intestinal epithelium, reduces ISCs, and eliminates organoids ex vivo

Investigation of T-cell-derived molecules mediating tissue injury can be complicated by redundancies in effector pathways and by the numerous potential targets that may exhibit divergent responses to similar molecules (16–29, 40, 41). Additionally, some conflicting experimental results may be due in part to the lack of models for studying specific interactions between immune effectors and primary cells as well as to challenges in deciphering responses against specific cellular subsets within a tissue. As such, we sought to establish a model for studying interactions between T cells and the ISC compartment by culturing intestinal organoids with T cells ex vivo. While co-culture with naive allogeneic T cells had no effect on regeneration from dissociated mouse organoid cells, allo-activated T cells significantly reduced allogeneic SI and LI organoid numbers in a concentration-dependent fashion (Fig. 3A and fig. S4A). Co-culture with polyclonally-activated allogeneic T cells also impaired organoid formation, and both CD4⁺ and CD8⁺ T cells were able to mediate organoid suppression (fig. S4, B and C). In addition to murine co-cultures, human T cells suppressed the growth of genetically disparate human duodenal organoids as well (Fig. 3B).

Fig. 3. T-cell-derived IFNγ targets intestinal epithelium leading to reduction of Lgr5⁺ stem cells.

(A) Representative images and number of SI organoids after co-culture of B6 organoid cells with activated by allogeneic B6 dendritic cells or naive allogeneic BALB/c T cells (culture day 7, n = 3–6 wells/group); scale bars = 500μm. (B) Representative images and number of human SI organoids cultured with human allogeneic CD8⁺ T cells (culture day 7, n = 7–13 wells/group); scale bars = 1000μm. (C) Numbers of B6 SI organoids after culture with anti-CD3/CD28-activated B6 syngeneic T cells (culture day 7, n = 3 wells/group). (D) Human LI organoids after culture with autologous human CD4⁺ and CD8⁺ T cells (culture day 7, n = 3 wells/group). (E) Representative images and numbers of B6 SI organoids after culture with anti-CD3/CD28-activated BALB/c T cells and anti-IFNγ neutralizing antibodies (culture day 7, n = 4 wells/group); scale bars = 500μm. (F) Human SI organoids after culture with human allogeneic T cells and anti-IFNγ (culture day 7, n = 9 wells/group). (G) WT or Ifngr^−/− B6 SI organoids cultured with BALB/c T cells (culture day 7, n = 4 wells/group). (H) B6 SI organoids after culture with rmIFNγ (culture day 7, n = 3 wells/group). (I) Representative images after co-culture of BDF1 organoid cells with GFP⁺ allogeneic B6 T cells. Shown are bright field (upper), fluorescent (middle), or overlap (lower) images; scale bars = 50μm. (J) IFNγ ELISA on supernatants from culture of B6 SI organoids with BALB/c T cells (n = 4–8 wells/group; ND, not detected. (K) FACS analysis of Lgr5-GFP^high ISCs in organoids cultured with rmIFNγ for 16 or 72 hours (n = 3–5 wells/group). Data are mean and s.e.m.; comparisons performed with t-tests (two groups) or one-way ANOVA (multiple groups); *P < 0.05, **P < 0.01, ***P < 0.001. Data are representative of at least two independent experiments, or combined from two or three (B, F, J) independent experiments.

As organoid formation and survival were impaired by activated allogeneic T cells but not by naive allogeneic T cells (Fig. 3A), we hypothesized that antigenic disparity was important for T cell activation but was not required for target suppression once the T cells were already activated. Indeed, syngeneic co-cultures with allo-activated T cells or with polyclonally-stimulated T cells both impaired the viability of mouse organoids (Fig. 3C and fig. S4D). Furthermore, upon co-culturing T cells and organoids from the same donors, activated human CD4⁺ and CD8⁺ T cells suppressed the growth of autologous human colon organoids (Fig. 3D). These findings indicated that T cell activation can impair the viability of intestinal epithelium ex vivo, even in the absence of genetic disparity with the epithelial targets.

In order to define the T cell effector pathways mediating organoid toxicity, we performed co-cultures under several conditions with either genetically-deficient T cells or with neutralizing antibodies. Inhibition of perforin, FasL, TRAIL, IL-1β, IL-6, IL-17A, IL-22, and TNFα had no effect on organoid numbers (fig. S5, A to F). In contrast, T cell co-culture with anti-Interferon-γ (IFNγ) neutralizing antibodies restored murine SI organoid growth (Fig. 3E), and IFNγ blockade with neutralizing antibodies also protected human duodenal organoids from human allogeneic T cells (Fig. 3F). Anti-IFNγ antibodies could have protected organoids by preventing paracrine/autocrine IFNγ activity among the T cells or by suppressing IFNγ signaling within the organoids. However, Ifngr^−/− T cells demonstrated intact organoid suppression (fig. S5, G and H), while Ifngr^−/− organoids were significantly resistant to both allogeneic and syngeneic T cells (Fig. 3G and fig. S5I), indicating that IFNγ targeted the epithelium during co-culture.

IFNγ alone was sufficient for mediating organoid toxicity, as addition of recombinant murine (rm) IFNγ to cultures without T cells demonstrated concentration-dependent suppression of organoid numbers (Fig. 3H). Tracking T cell kinetics in co-cultures using GFP⁺ T cells, we found that T cell frequencies decreased by day 4 of the culture (Fig. 3I and fig. S4E), suggesting that IFNγ-mediated organoid suppression was initiated within the first few days of the culture. Consistent with this, IFNγ was detected in the culture media on day three of co-culture, and the concentration decreased by day five (Fig. 3J). Moreover, flow cytometry analysis showed a significant reduction of Lgr5-GFP^high ISCs in organoids after 16 hours of incubation with IFNγ, and this ISC depletion progressed substantially by 72 hours after exposure to IFNγ (Fig. 3K).

IFNγ programs stem cell death, and Inhibition of JAK/STAT signaling protects ISCs from IFNγ

We next investigated the signaling pathways involved in IFNγ-mediated organoid and ISC suppression. Ruxolitinib is a JAK 1/2 inhibitor capable of preventing T cell function, suppressing production of inflammatory cytokines by CD4⁺ T cells and promoting increased frequencies of Foxp3⁺ regulatory T cells in the BMT setting (42). Recent work has established ruxolitinib as a therapeutic option in GVHD, particularly for steroid refractory disease, and it has received FDA approval for this indication (43–45). However, the potentially distinct effects of ruxolitinib on T cells and on target tissues have yet to be delineated. We found that ruxolitinib significantly protected intestinal organoids from allogeneic T cells ex vivo (Fig. 4A). Ruxolitinib also protected mouse (Fig. 4B) and human (Fig. 4C) intestinal organoids from IFNγ. Furthermore, ISC frequencies were significantly preserved in organoids cultured with IFNγ in the presence of ruxolitinib, including near-total preservation in the first 16 hours and partial preservation by 72 hours (Fig. 4D).

Fig. 4. JAK/STAT inhibition protects intestinal stem cells from IFNγ.

(A) Representative images and numbers of B6 organoids after culture with BALB/c T cells and ruxolitinib (culture day 7, n = 4 wells/group); scale bars = 500μm. (B) Numbers of B6 SI organoid cells after culture with rmIFNγ and ruxolitinib (culture day 7, n = 4 wells/group). (C) Human SI organoids cultured with rhIFNγ (culture day 7, n = 3 wells/group) and ruxolitinib. (D) FACS analysis of Lgr5–GFP^high ISCs in organoids cultured with rmIFNγ and ruxolitinib for 16 or 72 hours (n = 3–4 wells/group). (E) Jak1-deficient B6 SI organoids from Jak1^fl/flxRosa-cre-ert2 mice cultured with BALB/c T cells or rmIFNγ (culture day 7, n = 4 wells/group). (F) Crypt pSTAT1 western blots after 30 minutes incubation with rmIFNγ +/− ruxolitinib. (G) WT or Stat1^−/− B6 SI organoids cultured with rmIFNγ (culture day 7, n = 4 wells/group). Graphs indicate mean and s.e.m.; t-tests (two groups) or one-way ANOVA (multiple groups); *P < 0.05, **P < 0.01, ***P < 0.001. Data are representative of at least two independent experiments.

Protection of organoids from IFNγ in the absence of T cells suggested that ruxolitinib was acting on the organoids themselves, mediating resistance by suppressing epithelial JAK signaling. Using organoids from Jak1-floxed x Rosa-cre-ert2 mice, we found that passaged organoid cells pretreated with 4-OHT to delete Jak1 were resistant to allogeneic T cells and to IFNγ (Fig. 4E and fig. S6). Furthermore, ruxolitinib prevented phosphorylation of Stat1 by IFNγ in SI crypts (Fig. 4F), and Stat1^−/− organoids were resistant to IFNγ as well (Fig. 4G). We thus concluded that allogeneic T cells and IFNγ targeted the intestinal epithelium via Jak1/STAT1 signaling, and inhibition of epithelial Jak1 could protect intestinal tissue from immune-mediated damage ex vivo.

In addition to the reduction of ISCs identified by flow cytometry (Fig. 3K), qPCR analysis of intestinal organoids showed that gene expression associated with ISCs (Lgr5, Olfm4) decreased quickly in mouse and human cultures treated with IFNγ (fig. S7, A and B). Target genes of Wnt signaling (Axin2) and Notch signaling (Hes1) were also reduced (fig. S7C), and gene expression associated with Paneth cells (Lyz1, Defa1), enterocytes (Alpi), goblet cells (Muc2), enteroendocrine cells (Chga), and tuft cells (Trpm5) were all reduced as well (Fig. S7, D and E). Consistent with these results, qPCR analysis of crypts from GVHD recipients demonstrated reduced gene expression associated with ISCs (Lgr5, Olfm4), Paneth cells (Lyz1, Defa1), goblet cells (Muc2), enteroendocrine cells (Chga), and tuft cells (Trpm5) in comparison with non-GVHD controls, while enterocyte-associated Alpi expression trended down even after BMT without T cells (fig. S7F). Overall, these gene expression patterns suggested that ISCs were not being lost due to increased differentiation of ISCs into their progeny.

We next evaluated the role of programmed cell death. Gene expression in mouse SI organoids cultured with IFNγ demonstrated multiple transcriptional changes consistent with induction of apoptosis, as expression of the anti-apoptotic genes Bcl2 and Bcl2l1 (Bcl-x_L) decreased and expression of the pro-apoptotic gene Bak increased (Fig. 5A). Similar transcriptional changes were observed in human duodenal organoids treated with IFNγ (Fig. 5B). No changes were observed in expression of the anti-apoptotic gene Mcl1 or the pro-apoptotic gene Bax (fig. S8A). Moreover, Annexin V analysis showed increased Annexin⁺DAPI⁻ and Annexin⁺DAPI⁺ Lgr5-GFP⁺ cells after IFNγ treatment, consistent with increased early apoptotic and dead ISCs, even though there was already a statistically significant reduction in ISC frequency at that point (Fig. 5, C and D). Increased apoptosis was also identified in human intestinal organoids incubated with IFNγ, as determined by increased caspase-3/7 activity and confirmed by increased detection of cleaved caspase-3 (Fig. 5, E and F). Additionally, ruxolitinib inhibited the pro-apoptotic transcriptional changes observed in intestinal organoids treated with IFNγ (Fig. 5G), thus specifically linking JAK signaling to the apoptotic phenotype. Therefore in total, ex vivo experiments indicated that T cells induced organoid toxicity via IFNγ, which activated JAK1-dependent STAT1 activation within the epithelium, resulting in an apoptotic transcriptional program and loss of ISCs.

Fig. 5. IFNγ programs stem cell death.

(A) Apoptosis-related genes expression in mouse SI organoids cultured with rmIFNγ for 6 hours (n = 6 wells/group); Mann–Whitney U analysis. (B) Apoptosis-related genes expression in human SI organoids cultured with rhIFNγ for 24 hours (n = 9–10 wells/group, data are from 3 different SI donors); Mann–Whitney U analysis. (C and D) FACS plots (C) and quantifications (D) of Lgr5-GFP^high cells and Annexin V analysis from SI organoids cultured with rmIFNγ for 16 hours (n = 4 wells/group). (E) Relative caspase-3/7 activity as evaluated by Caspase-Glo assay; fold increase over baseline after treatment with rhIFNγ for 24 hours (n = 6 wells/group). (F) Human organoid cleaved caspase-3 western blot after 48 hours incubation with rhIFNγ. (G) Apoptosis-related genes expression in mouse SI organoids cultured with rmIFNγ and ruxolitinib for 24 hours (n = 6 wells/group); Kruskal–Wallis analysis. Graphs indicate mean and s.e.m.; comparisons performed with t-tests (two groups) or one-way ANOVA (multiple groups) unless otherwise stated; *P < 0.05, **P < 0.01, ***P < 0.001. Data are representative of two, or combined from two (E) independent experiments.

T-cell-derived IFNγ promotes stem cell apoptosis and intestinal pathology in immune-mediated GI damage in vivo

We next sought to evaluate the role of IFNγ in T-cell-mediated stem cell injury in vivo. Anti-IFNγ antibody treatment after allogeneic BMT significantly protected ISC numbers in Lgr5-LacZ reporter mice (Fig. 6A). Additionally, anti-IFNγ neutralizing antibodies increased ISC frequencies during autoimmunity occurring after Treg depletion (Fig. 6B). Furthermore, ruxolitinib treatment significantly protected ISCs in transplant recipients after allogeneic BMT (Fig. 6C and fig. S1B).

Fig. 6. T-cell-derived-IFNγ decreases ISCs in vivo.

(A) ISCs 10 days after LP-into-B6 BMT with isotype or anti-IFNγ antibodies. Representative images and frequency of SI Lgr5-LacZ⁺ ISCs (n = 15–27 independent sections/group); scale bars = 500μm (Upper) or 50μm (Lower). (B) SI Lgr5⁺ ISCs in Foxp3-DTR⁺ or Foxp3-DTR⁻ Lgr5-LacZ reporter mice 5 days after DT treatment along with Isotype or anti-IFNγ antibodies (n = 29–31 independent sections/group). (C) Frequency of SI Lgr5-LacZ⁺ ISCs 10 days after LP-into-B6 BMT with vehicle or ruxolitinib (n = 6–10 independent sections/group). (D) Frequency of SI Lgr5-LacZ⁺ ISCs 10 days after B6-into-BDF1 BMT with wild type or Ifng^−/− marrow (n = 5–13 independent sections/group). (E to M) B6-into-BDF1 BMT with wild type or Ifng^−/− T cells. (E) Frequency of SI Lgr5-LacZ⁺ ISCs (n = 5–11 independent sections/group). (F) Intestinal GVHD histopathology score 10 days after BMT (n = 6–12 mice). (G) Crypt numbers (n = 6–21 independent sections/group) and villus blunting histopathology scores (n = 6–12 mice/group) 10 days after BMT. (H) Representative images and quantification of Ki67 IHC in the crypt area 10 days after BMT (n = 47–77 crypts/group); scale bars = 100μm. (I) Apoptosis-related genes expression in mouse SI crypts 10 days after BMT (n = 10 mice; Mann–Whitney U analysis). (J) Images and quantification of crypt cleaved caspase-3 IHC 10 days after BMT. Arrows indicate cleaved caspase-3⁺ apoptotic crypt cells (n = 489–974 crypts/group); scale bars = 50μm. (K) Quantification of crypt TUNEL staining 10 days after BMT (n = 251–491 crypts/group). (L) Double immunofluorescent staining of β-gal (green) and cleaved caspase-3 (red) from Lgr5-LacZ recipient mice 10 days after BMT. Shown are representative images and average frequencies of cleaved caspase-3⁺ apoptotic ISCs per mouse ileum as a percentage of the total Lgr5⁺ ISCs detected; scale bars = 50μm. (M) Day 5 SI organoid numbers per 100 crypts cultured 10 days after BMT (n = 5–7 mice/group). Graphs demonstrate mean and s.e.m.; comparisons performed with t-tests (two groups) or one-way ANOVA (multiple groups) unless otherwise stated; *P<0.05, **P<0.01, ***P<0.001. Data are representative of two, or combined from two (A, B, F, G and I) independent experiments.

To identify in vivo sources of IFNγ promoting ISC reduction post-BMT, we first phenotyped IFNγ⁺ cells in the mucosa of transplant recipients. After mechanically dissociating the villi to enrich for crypt region tissue, lamina propria lymphocytes (LPLs) were isolated and incubated with a golgi inhibitor prior to IFNγ analysis. Substantially more IFNγ⁺ cells were identified in recipient intestinal mucosa after allogeneic BMT than after syngeneic BMT, and the vast majority of the IFNγ⁺ cells identified were indeed donor T cells (fig. S9, A and B). Further analysis indicated T-bet⁺ Th1 helper T cells with an activated phenotype (fig. S9, C and D).

We next evaluated donor-derived IFNγ functionally. While transplantation with Ifng^−/− donor marrow did not impact ISC numbers (Fig. 6D), allogeneic BMT with Ifng^−/− donor T cells resulted in significantly greater ISC recovery (Fig. 6E), confirming donor T cells to be the critical source of IFNγ resulting in ISC reduction. Furthermore, broader histopathologic analysis after BMT with Ifng^−/− donor T cells showed significant reduction in overall GVHD pathology, including reduced SI crypt loss and villus blunting (Fig. 6, F and G, and fig. S1C). In addition, we observed increased epithelial proliferation along with the tissue injury occurring in GVHD, which was also significantly reduced in recipients of IFNγ-deficient T cells (Fig. 6H).

Similar to the gene expression changes induced by IFNγ ex vivo, qPCR analysis of crypts isolated post-BMT indicated increased Bcl2, increased Bcl2l1, and decreased Bak1 in recipients of Ifng^−/− T cells (Fig. 6I). Additionally, anti-cleaved-caspase-3 and TUNEL staining both indicated significantly reduced crypt apoptosis in mice transplanted with Ifng^−/− T cells (Fig. 6, J to L). Furthermore, anti-cleaved-caspase-3 and anti-β-gal double immunofluorescent staining demonstrated apoptotic Lgr5⁺ ISCs in GVHD, and the frequency of cleaved caspase-3⁺ ISCs was significantly reduced in mice receiving Ifng^−/− T cells (Fig. 6L and fig. S10). Finally, SI crypts isolated from recipients of Ifng^−/− T cells demonstrated significantly greater organoid formation than crypts from mice transplanted with wild-type (WT) T cells (Fig. 6M), indicating functional improvement in addition to the improved histology. Overall, in vivo studies thus supported the ex vivo findings of ISC reduction induced by T-cell-derived IFNγ and JAK/STAT signaling.

IFNγ directly targets intestinal stem cells and induces apoptosis

Given the broad expression of the IFNγ receptor (IFNγR) on numerous cell types, T-cell-derived IFNγ could have many targets in vivo, leading to indirect effects on the stem cell compartment. To examine if T-cell-derived IFNγ was targeting the recipient epithelium directly in vivo, we performed allogeneic BMT into Ifngr1^fl/fl x Villin-Cre (Ifngr^ΔIEC) mice. Staining for Olfm4 indicated that targeted deletion of IFNγR from recipient intestinal epithelium significantly protected Olfm4⁺ ISCs from allogeneic T cells (Fig. 7A and fig. S1D). Ifngr^ΔIEC recipients also demonstrated reduced overall GVHD pathology, greater crypt numbers, decreased villus blunting, and significantly less crypt apoptosis (Fig. 7, B to D).

Fig. 7. IFNγ directly targets intestinal stem cells and induces apoptosis.

(A to D) Allogeneic B10.Br-into-B6 (Allo) or syngeneic B6-into-B6 (Syn) BMT using Ifngr^fl/flxVillin-Cre (Ifngr^ΔIEC) or Cre-negative Ifngr^fl/fl (Ifngr^WT) littermate controls. (A) Representative images and frequency of SI (ileum) IHC for Olfm4⁺ ISCs 7 days after BMT (n = 6–17 independent sections/group); scale bars = 100μm. (B) Intestinal GVHD histopathology score 9 days after BMT (n = 3–5 mice/group). (C) Crypt number quantification (n = 9–17 independent sections/group) and villus blunting histopathologic scoring 9 days after BMT (n = 3–5 mice/group). (D) Quantification of crypt cleaved caspase-3 (cCaspase-3) IHC 9 days after BMT (n = 489–974 crypts/group). (E) FACS analysis of CD119 (IFNγR1) expression on ISCs and Paneth cells. (F) Numbers of B6 SI organoids 7 days after culture with BALB/c T cells +/− Wnt3a and Jagged1 (n = 3 wells/group). (G) Paneth-cell-deficient Atoh1^ΔIEC SI organoids cultured in WNT3-supplemented ENR media for 7 days +/− BALB/c T cells or IFNγ (n = 4 wells/group). (H) SI organoids from sort-purified SI Lgr5-GFP^high ISCs and sort-purified Paneth cells cultured for 7 days +/− BALB/c T cells (n = 3–6 wells/group). (I) RNAseq indicating IFNγ-responsive gene expression in sorted Lgr5-GFP^high ISCs incubated with IFNγ for 1.5 hours. (J) Organoids from sorted WT or Ifngr^−/− Lgr5-GFP^high SI ISCs cultured for 6 days +/− BALB/c T cells (n = 7–8 wells/group). (K to Q) ISC colonies cultured in WENR with HDAC and GSK3β inhibition +/− IFNγ. (K) Representative confocal images of cleaved caspase-3 (cCaspase-3) immunofluorescence in ISC colonies cultured +/− IFNγ (culture day 6, arrows indicate apoptotic ISCs in the cellular layer, and arrow heads indicate apoptotic ISCs in the colony lumen), scale bars = 50μm. (L) Frequency of epithelial-layer cCaspase-3⁺ ISCs (n = 40–72 colonies/group). (M) cCaspase-3 staining intensity in the lumen area (n = 76–136 colonies/group). (N) qPCR analysis of apoptosis-related genes in mouse SI ISC colonies cultured with rmIFNγ for 24 hours (n = 6 wells/group; Mann–Whitney U analysis). (O to Q) Representative images and viability quantification of ISC colonies cultured with IFNγ. Images show bright field microscopy (Upper), Hoechst staining (Middle), or propidium iodide (Lower); scale bars = 200 μm. (O) Images of Lgr5-GFP⁺ SI ISC colonies cultured with rmIFNγ +/− caspase inhibitor Q-VD-OPh (culture day 7). (P) Images of SI ISC colonies initiated from sorted WT or Bak/Bax double knockout (DKO) SI ISCs cultured with IFNγ (culture day 7). (Q) Quantification of ISC colony survival after cultured with IFNγ (n = 3 colonies/group); t-tests at each concentration of IFNγ. Graphs indicate mean and s.e.m.; comparisons performed with one-way ANOVA unless otherwise stated; *P<0.05, **P<0.01, ***P<0.001. Data are representative of at least two, or combined from three independent experiments (I).

These results indicated that IFNγ could act directly on the intestinal epithelium, but it remained possible that effects on ISCs were secondary to targeting some other cell population in the intestinal epithelium. Crypt base ISCs repopulate other intestinal epithelial cells including Paneth cells, which produce several supportive factors such as Wnt3, EGF, and Notch ligands (4). As Paneth cell frequencies and gene expression were reduced in our experiments (Fig. 1, C and E, fig. S2B, and fig. S7, D and F), and IFNγ can induce apoptosis and loss of Paneth cells (30, 31), ISC reduction and organoid elimination could have been due to indirect effects resulting from damage to the Paneth cell niche. FACS analysis confirmed expression of IFNγR1 (CD119) on both Paneth cells and ISCs (Fig. 7E and fig. S2A). Given that ISC-restricted cre-driven gene deletion is not possible in vivo because genetic manipulation of ISCs is rapidly transmitted to their progeny (with faster kinetics than GVHD pathophysiology and at times even faster kinetics than the genetic manipulation can manifest protein-level changes within the ISCs), we therefore examined ex vivo if T-cell-mediated injury was due to targeting of the ISCs or Paneth cells. Arguing against an essential role for Paneth cell targeting, supplementation of the culture media with the Paneth-cell-derived factors Wnt3 and Jagged1 did not protect intestinal organoids from T cells (Fig. 7F). Similarly, prevention of Paneth cell targeting through use of Paneth-cell-deficient Atoh1^ΔIEC organoids did not protect the organoids from T cells or IFNγ either (Fig. 7G). We next co-cultured purified ISCs and purified Paneth cells such that organoid formation from ISCs was dependent on support provided by the Paneth cells. However, cultures with WT and Ifngr^−/− Paneth cells both remained sensitive to T cells (Fig. 7H). Additionally, RNA sequencing of sorted Lgr5^high cells 1.5 hours after exposure to IFNγ showed upregulation of several IFNγ-related genes confirming direct activity of IFNγ in ISCs (Fig. 7I). Moreover, organoid growth from WT ISCs was significantly reduced by co-culture with allogeneic T cells, but Ifngr^−/− ISCs demonstrated intact organoid-forming capacity and were thus resistant to T-cell-mediated suppression (Fig. 7J).

To exclude the possibility that cultures of WT ISCs were sensitive to IFNγ because of damage to their immediate progeny and to further evaluate the direct effects of IFNγ on ISCs, we utilized a niche-independent high-purity ISC culture system composed nearly entirely of Lgr5⁺ cells (46). Combination of GSK3β and histone deacetylase inhibitors enables culture of homogenous symmetrically dividing Lgr5⁺ ISC colonies from purified ISCs, and despite the potent Wnt and Notch pathway activation maintaining a high Lgr5⁺ frequency (fig. S8B), IFNγ directly induced apoptosis within the ISC colonies (Fig. 7K). As evidenced by staining for cleaved caspase-3, apoptotic ISCs identified within the cellular layer of the colonies peaked after eight hours (Fig. 7, K and L) and subsequently accumulated in the colony lumen (Fig. 7, K and M). Gene expression analyses revealed the same apoptotic program identified in organoids exposed to IFNγ, with downregulation of Bcl2 and Bcl2l1, upregulation of Bak, and no significant change in expression of Mcl1 or Bax (Fig. 7N and fig. S8C). IFNγ-induced ISC apoptosis also led to substantial colony death confirmed by propidium iodide uptake (Fig. 7, O and P). Furthermore, addition of the pan-caspase inhibitor Q-VD-Oph to WT ISC colonies (Fig. 7, O and Q) or genetic deletion of Bak and Bax using ISC colonies derived from double-deficient (Bak^−/−/Bax-floxed x Rosa-cre-ert2) mice (Fig. 7, P and Q) both maintained ISC colony viability despite exposure to IFNγ. In conclusion, T-cell-derived IFNγ directly targeted the intestinal epithelium leading to ISC reduction and intestinal pathology, and IFNγ induced Bak/Bax-dependent ISC apoptosis by directly acting on the stem cells themselves.

Discussion

T-cell-mediated tissue damage, particularly in the BMT setting, is the culmination of a systemic process involving cellular activation, migration, and effector function. Given this complexity and the involvement of numerous cell types in various tissues at specific time points, it is challenging to comprehensively and accurately elucidate the specific interactions occurring between T cells and individual subsets of intestinal epithelial cells. To overcome these limitations, we established an ex vivo co–culture system of intestinal organoids and T cells. Using this system to model T-cell-induced ISC damage, we identified a direct role for T-cell-derived IFNγ and subsequent JAK/STAT signaling in ISC apoptosis occurring during immune-mediated GI damage. Consistent with this cytokine-mediated pathology, whole-mount 3-D microscopy after allogeneic BMT demonstrated that donor T cells invading the SI post-transplant primarily infiltrated the lamina propria of the crypt compartment. Donor T cells thus invaded the intestinal mucosa near the stem cells, but were mostly not located precisely within the targeted epithelium. Further analysis of these T cells confirmed that they were the principle source of IFNγ. Surprisingly, few T cells were present within the crypt epithelium at baseline, as most intraepithelial T cells were found within the villi. The lamina propria showed roughly similar T cell densities within the crypt and villus regions at baseline, so the preferential infiltration of the crypt region after allogeneic BMT represented a substantial redistribution of T cell localization within the mucosa of the ileum. Interestingly, a recent study showed that depletion of donor CD4⁺ T cells immediately after BMT resulted in increased serum IFNγ and reduced intestinal GVHD pathology with fewer donor CD8⁺ T cells in the colon (47), suggesting that IFNγ from donor T cells in proximity to the crypt compartment, rather than the IFNγ present in circulation, may be essential for its direct epithelial toxicity.

While naive T cells had no impact on epithelial growth in ex vivo cultures, activated T cells induced substantial toxicity and reduction of ISCs. This did not require genetic disparity between the T cells and their epithelial targets once the T cells were activated, leading to suppression of both allogeneic and syngeneic mouse intestinal organoids. This was also true in human models, with human T cells eliminating allogeneic human organoids and even eliminating autologous organoids as well. This was driven by T-cell-derived IFNγ, which also induced the ISC reduction observed ex vivo as well as in vivo. Antigen specificity in this ex vivo model was dependent on the initial T cell activation, which led to substantial production of IFNγ. These results are consistent with experiments indicating that inflammatory cytokines can mediate tissue damage induced by allogeneic T cells irrespective of antigen presentation by epithelial cells (48), although antigen presentation by epithelial cells including ISCs may be critical for certain other immune responses (49–51).

Effects of IFNγ on the GI tract have been studied in various experimental models, and it has been reported to induce epithelial toxicity through both cell-autonomous and non-autonomous negative regulatory feedback loops (33, 52, 53). IFNγ has also been found to induce loss of Paneth cells in models of infection and autoimmunity (30, 31). Studies in transplant models have identified both pathologic and protective roles for IFNγ in GVHD, as discussed below. However, there has been little exploration of the direct interactions between ISCs and IFNγ. The exclusive use here of primary cells, containing the full diversity of lineages present in normal epithelial tissue, allowed for identification of ISCs as direct targets of T-cell-mediated cytokine-dependent GI damage. This would not have been possible with typical cell lines lacking a stem cell compartment. We found that reduction of ISCs preceded any reduction of Paneth cells in immune-mediated GI damage. The reduction of ISCs was validated using two distinct functional approaches: culturing organoids from crypts isolated after in vivo challenge and lineage tracing for stem cell-derived progeny. Both functional approaches were consistent with the kinetics indicated by ISC quantifications. Furthermore, these kinetics suggested that ISCs were the primary target, though a decrease in ISCs could also have been due to niche dysfunction, rather than loss of the niche. While Paneth cells are not the only component of the stem cell niche, which includes stromal and immunologic members as well (8, 9, 14, 54, 55), and consideration should also be paid to progenitors and mature epithelial cells for a comprehensive understanding of intestinal immunopathology (56–59), ex vivo modeling revealed that IFNγ-receptor-deficient ISCs were resistant to T cells and IFNγ, while niche-dependent cultures of ISCs with IFNγ-receptor-deficient Paneth cells as well as cultures of niche-independent stem cell colonies were not. These findings thus indicated that ISCs were indeed direct targets of T cells and IFNγ.

Further investigation of the stem cell compartment indicated that IFNγ could directly program ISC death. BCL-2 and BCL-X_L are anti-apoptotic BCL-2 family members, and their downregulation can result in activation of pro-apoptotic effectors BAX and BAK (60). Activated BAX and BAK can then form oligomers which permeabilize the mitochondrial outer membrane and release cytochrome c to activate caspases and further propagate the apoptotic cascade (61). The downregulation of Bcl2 and Bcl2l1, and the upregulation of Bak, along with stable expression of Bax clearly implicate initiation of apoptosis as a major direct effect of IFNγ in ISCs. Furthermore, we observed increased caspase-3 activity and cleaved caspase-3 protein in human organoids treated with IFNγ. In vivo IFNγ signaling blockade with neutralizing antibodies, IFNγ deletion from donor T cells, IFNγR deletion from the intestinal epithelium, and ruxolitinib all protected ISCs from T cells in vivo. While the specific roles of IFNγ may be distinct between models of GVHD and autoimmunity, these results suggest that IFNγ is a potent mediator of T-cell-induced ISC impairment and that a secreted cytokine can kill stem cells via activation of JAK/STAT signaling and driving programmed cell death.

Due to its pleiotropic effects, IFNγ has demonstrated strikingly distinct impacts in various BMT models. Deficiency of donor-derived IFNγ resulted in increased GVHD mortality and limited GVL in a CD8⁺ T-cell-mediated experimental transplant model (28). A subsequent study showed opposing effects of IFNγ in distinct tissues, with IFNγ playing a protective role in the lungs, but mediating GI toxicity in mice with GVHD (18), and another study indicated that IFNγ can reduce intestinal GVHD pathology after depletion of CD4⁺ T cells (47). These studies illustrate the complex role of IFNγ after BMT, and undesirable complications could thus result from targeting IFNγ in clinical BMT. IFNγ signaling is transduced by the JAK/STAT pathway (62), which could represent another approach for interfering with IFNγ-mediated GI damage and loss of ISCs. Indeed, ruxolitinib treatment protected ISCs from T-cell-mediated damage ex vivo and in vivo. While JAK inhibition can suppress T cell activity (42), we found that it also protected intestinal epithelium from T-cell-mediated injury by suppressing the tissue’s response to the T cells. JAK inhibition has been investigated clinically in GVHD (43–45), and it has recently been approved for steroid refractory GVHD. JAK inhibitors thus provide a promising approach for protecting the ISC compartment from pathologic immune responses. Additionally, these findings suggest that the efficacy of JAK inhibition in GVHD, particularly in settings where other immunosuppressive agents have failed, could be due to suppression of pathologic cytokine signaling within the GVHD target organs.

In summary, we found that damage to the ISC compartment was a shared feature of GVHD and autoimmunity, and T-cell-derived IFNγ was a key mediator of ISC reduction in immune-mediated GI damage. Intestinal organoid cultures were used to assay epithelial function during immune-mediated damage in vivo and to interrogate specific interactions between T cells and epithelial targets ex vivo. T cell localization within the intestinal mucosa differed substantially between homeostasis and the post-transplant setting, with donor T cells primarily localizing to the lamina propria of the crypt region where they were the dominant producers of IFNγ. IFNγ directly targeted ISCs, inducing a gene expression program resulting in stem cell apoptosis, and JAK inhibition protected ISCs from T cells by suppressing their response to IFNγ. IFNγ thus played a central role in the T-cell-mediated stem cell damage, and JAK inhibitors may provide clinically efficacious immunosuppression in part by suppressing tissue responses to pathologic signals from the immune system.

Materials and Methods

Study design

The purpose of the study was to investigate mechanisms of T cell effects on ISCs during immune-mediated GI damage. We used two types of in vivo animal models: allogeneic BMT and Treg-depletion-induced autoimmunity. To perform detailed evaluation of direct interactions between T cells and ISCs, we established a method of co-culturing T cells with intestinal organoids. Analyses of experiments were performed with histologic staining, 3-D imaging, flow cytometry, qPCR, and western blotting. Statistical issues are described below. There were no pre-defined study end points. Experiments were generally performed a minimum of two times, and statistical methods are described in the figure legends.

For further details, please see the Supplementary Materials and Methods.

Supplementary Material

Table S1

Table S1. Raw data used to generate all graphs that have n <25.

supplementary main

Fig. S1. Representative images post-BMT.

Fig. S2. FACS plots demonstrating the gating strategy for Lgr5-GFP^high ISCs and for Paneth cells.

Fig. S3. Allogeneic BMT decreases lineage tracing from intestinal stem cells.

Fig. S4. Co-culture of activated T cells and intestinal organoids.

Fig. S5. Investigation of T cell effector pathways responsible for T-cell-mediated intestinal organoid toxicity.

Fig. S6. Tamoxifen treatment does not impair organoid numbers.

Fig. S7. Epithelial lineage markers ex vivo and in vivo.

Fig. S8. Supplemental analyses of IFN-induced apoptosis in intestinal epithelium.

Fig. S9. Phenotyping of IFNγ⁺ cells in recipient intestinal mucosa after BMT.

Fig. S10. IFNγ induces ISC apoptosis in vivo.

Table S2. List of antibodies used.