Paneth cell extrusion and release of antimicrobial products is directly controlled by immune cell–derived IFN-γ

To test if PC degranulation is induced in direct response to bacterial stimulation, we studied the effect of microbial antigens added to the culture medium of small intestinal organoids. 12 h after addition of peptidoglycan (PGN), poly(I:C), LPS, flagellin, CpG oligodeoxynucleotides (ODNs), MDP, and heat-inactivated or live E. coli bacteria PC granules were visualized using UEA-1 (Ulex europaeus agglutinin 1) lectin staining (Fig. 1 A). We observed no change in organoid morphology and no effect on PC granules under all conditions tested. As a positive control, cholinergic stimulation with Carbamylcholine (Cch) caused complete loss of epithelial granules, demonstrating the physiological competence of cultured PCs similar to their counterparts in vivo (Satoh et al., 1992). The absence of any PC secretory response to bacterial stimuli was confirmed by Western blot analysis of lysozyme content in organoids that were mechanically sheared and extensively washed to separate the pool of intracellular lysozyme from the pool that had been secreted into the crypt lumen (Fig. 1, B and C).

Addition of the various agents to intact organoids results in basolateral exposure to bacterial antigens, similar to disruption of the epithelial barrier in vivo. We next explored the effects of exposure on both the apical and basal sites. For this, we generated organoid fragments by mechanical shearing. These were resuspended in culture medium containing the panel of bacterial agents. Lysozyme content was quantified after 2 h at 37°C—an experimental setting analogous to the stimulation of freshly isolated crypts used previously (Ayabe et al., 2000). No significant decrease of the epithelial lysozyme content was measured for any of the bacterial stimuli, whereas controls using Cch again resulted a complete discharge of PCs (Fig. 1, B and C). We concluded that bacterial stimulation from both basolateral and apical epithelial sides represents no direct trigger for PC degranulation in our system.

By RT-PCR analysis, we found that the expression of microbial pattern receptors (Tlr2,3,7,8,9, MD-2 and Cd14) is maintained in organoids compared with freshly isolated crypts (Fig. 1 D), arguing against nonresponsiveness due to transcriptional down-regulation. We used quantitative RT-PCR analysis to study if microbial stimuli can cause transcriptional activation of NF-κB signaling. Despite expression of cognate receptors (Tlr2, 3, and 9 for PGN, poly(I:C), and CpG ODN, respectively) stimulation in organoids did not induce expression of Icam1 or other NF-κB targets (Fig. 1 E and not depicted). TNF stimulation did cause marked Icam1 induction, suggesting that upstream events in TLR signaling mediate refractoriness to bacterial stimuli. As a positive control, human colon cancer cell lines (Caco-2 and HT-29) strongly responded to all microbial stimuli by induction of IL-8 expression (Fig. 1 F).

We hypothesized that PC degranulation as observed in freshly isolated intestinal crypts (Ayabe et al., 2000) might depend on bacterial sensing by co-isolated leukocytes. Using RT-PCR analysis, we indeed found that Cd3e- and Cd45-expressing immune cells were present in primary crypt preparations, but not in established organoid cultures (Fig. 2 A). We subsequently aimed to identify immune cell mediators that might indirectly cause PC degranulation after bacterial stimulation by testing the activity of several (proinflammatory) cytokines. PCs were morphologically unaffected in the presence of recombinant TNF, IL-1b, IL-6, or IL-22 (all at 20 ng/ml; 2 d; Fig. 2 B). In contrast, treatment with recombinant IFN-γ caused complete loss of PC granules as indicated by morphology, lysozyme immunostaining, and measurement of epithelial lysozyme content (Fig. 2, B–D). The minimal effective concentration of recombinant IFN-γ was 0.3 ng/ml (Fig. 2 E). Histologically, IFN-γ–treated organoids were characterized by a loss of mucus-containing goblet cells (Periodic acid-Schiff [PAS] staining) and mature enterocytes (as observed by alkaline phosphatase [AP] staining; Fig. 2 F), demonstrating a direct effect of this cytokine on other differentiated cell lineages as well.

Recent experiments have shown that prolonged IFN-γ exposure in vivo during T. gondii infection results in a complex intestinal pathology that is characterized but also caused by the loss of PCs (Raetz et al., 2013). However, the kinetics and directness of the epithelial response to IFN-γ have not yet been studied. Long-term exposure during several organoid passages caused successive and dose-dependent reduction in organoid number (Fig. 3 A). Time course experiments showed a strong apoptotic response starting 16 h after addition of 1 ng/ml IFN-γ as measured by FACS analysis of cleaved Caspase3 staining (Fig. 3, A and B). Significant reduction of the epithelial lysozyme content was detected on Western blot after a similar incubation period (Fig. 3 C). qPCR analysis of marker gene expression in PCs (Lyz1 and Defa6) and stem cells (Lgr5 and Olfm4) showed progressive reduction to <20% of control levels after 24 h of treatment, whereas the IFN-γ target gene Irf1 was stably induced (Fig. 3 D). These results indicate that IFN-γ affects epithelial homeostasis by induction of programmed cell death, which may contribute to PC loss. To test if PCs are maintained after degranulation, we performed in situ hybridization of Defa6 mRNA and anti-CD24 immunostainings (Sato et al., 2011a) to label their cytoplasm and membranes, respectively (Fig. 3, E and F). Reduced presence of PCs was found and FACS quantification showed that ∼25% of CD24⁺ cells were left after 5 d of treatment (Fig. 3 G and Fig. S1 B). Ubiquitous induction of the IFN-γ target MHC2, suggested a direct effect on all cell types, including PCs (Fig. 3 E). 4 d after withdrawal of IFN-γ from the culture medium, the organoid morphology had normalized and granules had recovered (Fig. 3 H), demonstrating that the epithelium can restore either by recharging of granules or by de novo generation of PCs.

To study the mechanism and the consequences of degranulation at a single-cell level, we performed time-lapse imaging. PC granules were labeled with ZP-1, a fluorescent and membrane permeable Zn²⁺ chelator (Giblin et al., 2006). Organoid morphology and marker distribution were followed after IFN-γ addition using confocal microscopy (Fig. 4 A and Videos 1–3). We found that degranulation was not synchronous within a given crypt. Individual PCs extruded their fluorescent label in an all-or-nothing manner, rather than through continuous exocytosis of vesicles. At a cellular level this process was completed within minutes. Between organoids, the onset and completion of degranulation was variable, but was clearly dependent on the IFN-γ concentration (6–12 h at 5 ng/ml and 4–8 h with 20 ng/ml IFN-γ).

Transmission electron microscopy also showed collective extrusion of PC granules (Fig. 4 B). Generation of large apical vacuoles was observed, reminiscent of the ultrastructural changes after muscarinergic stimulation in vivo (Satoh et al., 1992). After secretion into the crypt lumen, the electron-dense granules dissolved and the ZP-1 fluorescent signal rapidly dispersed (Fig. 4, A and B; and Videos 2 and 3). In other cases, granules remained densely packed, indicating the release of large PC fragments (Fig. 4 B). We found no indication of PC mitochondrial damage like that observed by electron microscopy after IFN-γ exposure during T. gondii infection (Raetz et al., 2013). After 12 h in the presence of 20 ng/ml IFN-γ, mature PC granules were absent, but we detected cells containing small electron-dense granules that may correspond to discharged or newly differentiating PCs.

To analyze PC fate and turnover after IFN-γ stimulation, we generated live cell reporters to trace granules and nuclei simultaneously. Granules were visualized by expression of a Lyz2-RFP fusion protein that was inserted in the Lyz2 genomic region in bacterial artificial chromosome (BAC) transgenic organoids (Schwank et al., 2013). Nuclei were fluorescently labeled by lentiviral expression of Histone 2B-Dendra (H2B-Dendra). We took advantage of the fact that ubiquitous expression driven by the CMV promoter resulted in strongest nuclear fluorescence in post-mitotic PCs that most likely accumulated the label. Strikingly, confocal microscopy showed that IFN-γ treated PCs co-released their nucleus together with their granules into the crypt lumen (Fig. 4 C and Video 4). Tracking of individual PCs (n > 100) showed that the majority of degranulation events resulted in this outcome with only 18% of all PC nuclei remaining after 36 h (1 ng/ml IFN-γ), while control cultures showed only infrequent spontaneous degranulation (Fig. 4 D and Video 5). Our results demonstrate that IFN-γ signaling mobilizes the entire epithelial pool of antimicrobial products by induction of PC extrusion. We conclude that epithelial recovery occurs by de novo differentiation rather than by recharging of PCs.

Next, we studied if endogenous cytokine levels produced by stimulated leukocytes are sufficient to induce PC release. We used the mouse invariant natural killer T (iNKT) hybridoma line DN32.D3, which secretes IFN-γ and other immune mediators upon CD1D-antigen presentation of the lipid antigen α-galactosylceramide (α-GC; Bialecki et al., 2009; Nieuwenhuis et al., 2009). Conditioned cell culture supernatants (SNs) were generated from control iNKT cells and after α-GC stimulation in the presence or absence of CD1D-expressing HeLa cells. These SNs were transferred to organoids (see scheme in Fig. 5 A). iNKT cell stimulation caused pronounced secretion of IFN-γ as observed by ELISA (Fig. 5 B). 2 d after addition to organoids, PC degranulation was analyzed using UEA-1 lectin staining and Western blot quantification of epithelial lysozyme. We found a loss of granules that was dependent on antigen presentation to iNKT cells, whereas direct α-GC stimulation of organoids had no effect (Fig. 5, C–E). Pretreatment with a neutralizing anti–IFN-γ antibody efficiently blocked activity in SNs from activated iNKT cells and also neutralized the effect of recombinant IFN-γ (Fig. 5 F). These results show that IFN-γ (produced at endogenous levels after stimulation) represents the main soluble trigger of PC release from iNKT cells.

Acute T cell stimulation using anti-CD3 antibody injection in vivo causes a severe but transient intestinal pathology that is characterized by epithelial apoptosis and massive villus blunting (Merger et al., 2002; Miura et al., 2005). However, it is unclear which of these effects can be attributed to IFN-γ and whether PC granules are released upon immune activation in vivo. We addressed these questions by i.p. injection of anti-CD3 in C57BL/6 wild-type and IFN-γ knock out (KO) mice. Intestines were collected for lysozyme immunohistochemistry 24 h later. In wild-type mice, anti-CD3 injection caused marked reduction of the number of lysozyme-positive PCs as counted per crypt section (Fig. 6, A and B). This confirmed our in vitro observation that PC granules are rapidly mobilized after immune stimulation. This effect was highly significant and strongest in the proximal intestine (duodenum) where the number of stained PCs was reduced by half (n = 7; P < 10⁻⁶). In homozygous IFN-γ KO mice, however, lysozyme-positive PCs were maintained in all regions of the intestine, demonstrating that IFN-γ is specifically required for granule loss. In situ hybridization for Defa6 mRNA as a cytoplasmic marker of PCs showed a similar response (Fig. 6, C and D), suggesting loss of entire PCs, as observed in vitro. T cell activation stimulated apoptosis broadly in the epithelium as observed by cleaved Caspase-3 immunostaining (Fig. 6, E–G). Importantly, the number of apoptotic cells was unaffected in IFN-γ KO intestines, demonstrating that PC rescue in this genotype is not a general result of an attenuated inflammatory reaction. Villus damage was most apparent in distal regions of the intestine (ileum) and clearly ameliorated in IFN-γ KO mice (Fig. 7 A). Here, PAS staining showed an IFN-γ–dependent release of goblet cell mucus that also massively occured in the colonic epithelium. In colon organoids, extrusion of goblet cell mucus after IFN-γ treatment was recapitulated (Fig. 7 B), demonstrating a more general role of this cytokine as a direct intestinal secretagogue for both Paneth and goblet cells.

Gut mucosal immunity is controlled by complex interactions among epithelium, microbiota, and the immune system, which complicates the dissection of signaling mechanisms. Here, we have used purely epithelial cultures to show that release of antimicrobial peptides and PC turnover are directly controlled by immune signals. IFN-γ was identified as the main mediator of PC degranulation after CD1D-dependent stimulation of iNKT cells (Nieuwenhuis et al., 2009) and after anti-CD3 stimulation of T cells in vivo. We propose that the proinflammatory cytokine may represent a danger signal to induce PC release and to react to microbial challenges. The exact cellular source of IFN-γ − most likely − depends on the physiological situation. Besides systemic IFN-γ release such as upon α-GC stimulation of iNKT cells (Nieuwenhuis et al., 2002), activation of intra-epithelial lymphocytes could also represent a paracrine source of IFN-γ (Guy-Grand et al., 1998; Cheroutre et al., 2011).

Organoid cultures have allowed us to visualize the rapid and complete discharging of PCs by IFN-γ or by muscarinergic stimulation, but cultured PCs were insensitive to a broad spectrum of microbe-associated molecular patterns, which contrasts with previous data obtained using freshly isolated crypts (Ayabe et al., 2000). While evidence suggests that microbial patterns can directly trigger PC function in vivo (Rumio et al., 2012; Brandl et al., 2007; Vaishnava et al., 2008), lack of Tlr4 expression in PCs (Tanabe et al., 2005) indicates an indirect nature of the effect seen after stimulation of primary crypts LPS. Also, the fact that degranulation after bacterial inoculation of germ-free mice can be inhibited by atropine favors an indirect mechanism, such as involvement of acetylcholine-secreting enteric neurons (Satoh, 1988). Interestingly, bacterial colonization is required for PC loss after T. gondii infection (Raetz et al., 2013). Here, however, bacterial sensing (and Myd88 expression) by T cells was required, demonstrating that continuous immune stimuli act on PCs in this model. Future research will benefit from the use of organoid cultures to identify mechanisms that regulate endotoxin tolerance of intestinal epithelial cells, which has been noted previously (Lotz et al., 2006). Given the broad distribution of bacterial antigens within the gut lumen, nonresponsiveness to microbial patterns (as observed by us) could be beneficial to avoid continuous stimulation by commensals.

Our study shows that organoids are a useful model to address PC function because analysis of pure epithelium prevents indirect effects of co-isolated (intraepithelial) leukocytes that are unavoidable in primary material, and organoids, even ex-Matrigel, are less fragile than fresh crypts—the latter degenerate instantly after isolation and allow short stimulations only. Thus, organoids offer a more sensitive system because full PC discharging can be achieved, whereas maximal secretion using fresh tissue never exceeds >20–30% under the conditions used (Ayabe et al., 2000). In addition, live imaging offers the unique possibility to follow cellular mechanisms and consequences upon degranulation.

We found that IFN-γ acts as a potent intestinal secretagogue both on Paneth and goblet cells. Goblet cell mucus serves as matrix to retain and concentrate antimicrobial PC products (Meyer-Hoffert et al., 2008). Given that T cell activation can cause severe damage to the intestine, such as apoptosis and disruption of the intestinal barrier (Merger et al., 2002; Miura et al., 2005), this massive release of antimicrobial factors might help to limit bacterial entry. Extrusion of PCs has been noted previously upon induction of ER stress or perturbation of zinc ion homeostasis (Sawada et al., 1991; Zhao et al., 2010; Grootjans et al., 2011; Adolph et al., 2013). Co-extruded genomic DNA might serve as a clot to seal crypts or act as a scaffold to trap bacteria, analogous to extracellular DNA webs that are released by cell death from activated neutrophils (Brinkmann et al., 2004). It should be noted that human Defensin 6 (HD6) itself can form nanonets to trap bacteria (Chu et al., 2012). Although mice lack a functional HD6 homologue, it has been reported that defensins have DNA-binding activity (Tewary et al., 2013) and might associate with released PC DNA. After elimination of PCs, new PCs can rapidly be generated de novo from adjacent stem cells as early as 12–24 h after challenge (Sawada et al., 1991). Continuous IFN-γ stimulation did not cause desensitization but a progressive loss of PCs. This occurred at concentrations comparable to IFN-γ serum levels reported in common mouse inflammatory bowel disease (IBD) models (Alex et al., 2009), arguing for a direct effect on PCs in chronic pathologies such as persistent murine norovirus infection in Atg16L1 mice or toxoplasmosis (Cadwell et al., 2010; Raetz et al., 2013). Future studies should address if PC dysfunction and associated changes of the gut flora are causally involved in other IFN-γ–driven IBD models (Kontoyiannis et al., 2002; Saha et al., 2010). Our observations might explain the reduction of PC numbers observed in other IFN-γ–dependent pathologies such as graft versus host disease and celiac disease (Nilsen et al., 1998; Di Sabatino et al., 2008; Eriguchi et al., 2012; Levine et al., 2013). Possibly the PC status might serve as sensitive reporter for IFN-γ signaling in clinical biopsies.

Given that organoid cultures can be established from human biopsies (Sato et al., 2011b), the phenotypic characterization of PC function in vitro may provide an important diagnostic tool to dissect disease mechanisms in these patients. The pharmacologic modulation of IFN-γ signaling to control PC function and survival may allow a new strategy to modify host defense.

Small intestinal organoids were established from wild-type C57BL/6 mice, cultured as described previously (Sato et al., 2009; Farin et al., 2012) and passaged weekly at a 1:6 split ratio. Organoids were stimulated for the indicated times by addition of recombinant murine IFN-γ (1 to 20 ng/ml; R&D Systems), mTNF-α (PeproTech), mIL-1b (Sigma-Aldrich), mIL-6 (eBioscience), mIL-22 (R&D Systems; all at 20 ng/ml) or Cch (25 µM; Sigma-Aldrich). Microbial stimuli were PGN (10 µg/ml; InvivoGen from S. aureus), p(I:C) (100 µg/ml; InvivoGen), LPS (1 µg/ml; Sigma-Aldrich from E. coli 026:B7), flagellin (0.1 µg/ml; InvivoGen from S. typhimurium), CpG ODN 1668 (1 µg/ml; Enzo), and MDP (100 µg/ml; InvivoGen). Log phase E. coli bacteria (strain K-12) were washed and added to the culture medium directly (in absence of penicillin/streptomycin), or after heat inactivation for 10 min at 95°C; per epithelial cell 100 bacteria were added. Mouse colon organoids were cultured as described previously for human colon cultures (Sato et al., 2011b). Colon organoids were differentiated toward goblet cells by addition of γ-secretase inhibitor DAPT (10 µM; Sigma-Aldrich) for 5 d before IFN-γ treatment. For stimulation of fragmented epithelia, organoids were collected in cold medium and mechanically sheared using glass Pasteur pipettes. Dispersal of crypts was confirmed by light microscopy and 100 crypts each were stimulated at 37°C, 5% CO₂.

BAC transgenesis of organoids has been described recently (Schwank et al., 2013), in brief a tagRFP-PGK::neo-cassette was inserted before the STOP codon of the mouse Lys2 locus (ENSMUSG00000069516) using BAC clone RP11-1105J23. BAC DNA was transfected in organoid cells using Lipofectamine 2000 reagent (Life Technologies) and G418 resistant, RFP⁺ clones were expanded. Viral transduction was done as previously described (Koo et al., 2012) using a pLV lentivirus containing a CMV-promoter-H2B-Dendra2-IRES-puro cassette and maintained under continuous puromycin selection to assure transgene expression.

DNase-treated RNA was used for RT-PCR analysis as previously described (Farin et al., 2012). For a list of primers used see Table S1. Caco-2 and HT-29 cells were grown in adherent culture in DMEM supplemented with 10% fetal bovine serum, penicillin/streptomycin, and 1X Glutamax (Life Technologies) and stimulated at 70% confluence.

Organoids were collected in cold medium and sheared by pipetting to generate epithelial fragments that were subsequently washed twice in ice-cold medium to separate cells from debris and lysozyme was secreted into the crypt lumen. Cells were lysed using RIPA buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.1% SDS, 0.5% Na-Deoxycholate, and 1% NP-40) containing Complete protease and phosphatase inhibitors (Roche). Protein content was quantified using standard Bradford assay. 10 µg of protein was run on a 14% SDS page and eventually transferred to a PVDF membrane and probed with rabbit anti-lysozyme (Dako; 1:10,000 in TBS-T 5% milk) and for normalization mouse anti–α-tubulin (Sigma-Aldrich; 1:3,000). After washes, membranes were incubated with secondary HRP-conjugated antibodies (goat anti–rabbit HRP 1:20,000 and rabbit anti–mouse 1:6,000; GE healthcare) and exposed to x-ray films using ECL reagents (GE Healthcare) and signals were quantified using ImageJ software (National Institutes of Health).

Whole-mount immunofluorescence and immunohistochemistry, histological stainings and in situ hybridization on formalin fixed paraffin sections organoids were performed as previously described (Farin et al., 2012). Antibodies used were rabbit anti-lysozyme (Dako; 1:500) and rabbit anti-cleaved Caspase-3 (Cell Signaling Technology; 1:400). For lectin staining, organoids were fixed in Matrigel (2% PFA; o/n at 4°C). After permeabilization (0.2% Triton X-100) organoids were stained by addition of FITC-coupled Ulex europaeus lectin (UAE-1-FITC; Sigma-Aldrich; 5 µg/ml, 1 h at room temperature), washed, and documented on an EVOS fl inverse fluorescence microscope (Advanced Microscopy Group). For membrane stainings, organoids were collected in cold medium and incubated (unfixed) with PE-conjugated monoclonal antibodies anti-CD24 (clone M1/69) and FITC-conjugated anti-MHC class II (clone M5/114.15.2; both from eBioscience at 1 µg/ml final concentration) for 1 h on ice, after washes, and documentation on an EVOS fl microscope.

For live imaging, organoids were seeded on glass-bottom plates in drops of 50% Matrigel/50% medium to facilitate dye penetration. After 3 d of culture, PC granules were labeled by addition 10 µM ZP-1 (Santa Cruz Biotechnology) to the culture medium for 16 h before the cultures were washed twice with CO₂-equilibrated warm medium and transferred to the culture chamber (37°C; 5% CO₂; 20% O₂) of a Leica SP5 confocal microscope. Images were acquired at 15-min intervals using low laser power. BAC transgenic organoids were imaged (as above) in 30-min intervals using 10-µm z-steps. Individual PCs were tracked using LAS software (Leica) and scored as degranulated if the red signal was <50% of the starting level. For each condition, >100 random PCs were tracked that were clearly positive for both reporters.

Transmission electron microscopy was done as described previously (Sato et al., 2009).

Organoids were collected in in TrypLE solution (Life Technologies) and single-cell suspensions were obtained after 2 cycles of mechanical break-up (using narrow-tipped pulled glass Pasteur pipettes) and incubation at 37°C (5 min). Cells were washed twice in PBS 1% BSA, and resuspended at room temperature in 500 µl PBS with 1% BSA and 2,000 U/ml DNaseI (Roche). Cells were fixed by addition of an equal volume of 4% paraformaldehyde (in PBS) for 15 min at room temperature, before washing, permeabilization using 0.1% Triton X-100, and antibody incubation (all in PBS 1% BSA) using rabbit anti-cleaved Caspase-3 (as in the previous section; 1:20 dilution; 1 h at room temperature). After washes, cells were incubated with Alexa Fluor 488–coupled secondary antibody (Molecular Probes; 1:200 dilution; 30 min at room temperature) before washes, filtering, and measurement on a FACSCalibur (BD). Single cells were gated using FSC and SSC parameters (Fig. S1 A). For anti-CD24 FACS (Fig. S1 B), dissociated, unfixed cells were incubated with FITC-conjugated anti-CD24 antibody (clone M1/69; final concentration 0.2 µg/ml; 1 h on ice). Dead cells were excluded after 7-AAD staining.

HeLa cells with or without stable overexpression of hCD1D were grown in DMEM supplemented with 10% heat-inactivated bovine calf serum, 1X Glutamax, 1X penicillin/streptomycin (all from Life Technologies) and incubated with 100 ng/ml α-GC for 24 h. Subsequently, iNKT DN32.D3 hybridoma cells (Bendelac et al., 1995) were incubated with HeLa and HeLa-CD1D cells for another 24 h in serum-free advanced DMEM/F12 medium supplemented with 1X l-Glutamine, 1X penicillin/streptomycin (Life Technologies), and rhIL-2 (10 U/ml; Proleukin; Novartis). SNs were centrifuged for cell depletion and included into the regular organoid medium as 40% (vol/vol). IFN-γ levels were measured with a commercially available Elisa kit from BD. A standard curve was recorded using recombinant mouse IFN-γ protein. In some experiments, SNs were treated with functional grade anti–mouse IFN-γ monoclonal (clone XMG1.2) or rat IgG1 K isotype control (clone eBRG1; both from eBioscience; final concentrations 0.5 µg/ml) 30 min before addition to the organoids.

Experiments were performed according to guidelines and reviewed by the Dier Experimenten Commissie (DEC) of the KNAW. IFN-γ knockout (Dalton et al., 1993) and background/sex/age matched wild-type C57BL/6 mice were obtained from The Jackson Laboratory. 8-wk-old wild-type and knock-out mice were injected i.p. with functional grade anti–mouse CD3e antibody (50 µg per animal; clone 145-2C11; eBioscience), or carrier alone (PBS). Mice were sacrificed 24 h after injection and the intestine was processed for histology as described (van Es et al., 2005). On 4-µm sections of lysozyme, Defa6 and anti-cleaved Caspase-3⁺ cells were counted. In each segment of the intestine, 3 × 15 adjacent crypts were examined (45 crypts/animal/segment).

Fig. 1 shows FACS sorting of organoid cells using intracellular anti-cleaved Caspase-3 staining and anti-CD24 membrane staining. Videos 1-3 show time-lapse analysis of PC degranulation after IFN-g exposure using Zinpyr-1 staining. Videos 4 and 5 show time-lapse visualization of PC extrusion after IFN-g exposure using transgenic reporter organoid lines. Table S1 shows primer sequences used for RT-PCR analysis.

We want to thank Maaike van den Born, Harry Begthel, Jeroen Korving, Stieneke van den Brink, and Karien Hamer for technical assistance and Bon-Kyoung Koo for critical discussions.

H.F. Farin was supported by an EMBO long-term fellowship ALTF 706-2009. W.R. Karthaus was supported by EU/TORNADO-KBBE-222720. G. Schwank was supported by the SNF fellowship for advanced researchers PA00P3 139732 and the Human Frontiers in Science Program long-term fellowship LT000422/2012.

H. Clevers is an inventor of several patents involving the organoid culture system. The remaining authors disclose no competing financial conflicts.