Glial cell-derived neuroregulators control type 3 innate lymphoid cells and gut defence

Abstract

Group 3 innate lymphoid cells (ILC3) are major regulators of inflammation and infection at mucosal barriers¹. ILC3 development has been considered to be programmed¹. Nevertheless, how ILC3 perceive, integrate and respond to local environmental signals remains unclear. Here we show that ILC3 sense their environment and control gut defence as part of a novel glial-ILC3-epithelial cell unit orchestrated by neurotrophic factors. We found that enteric ILC3 express the neuroregulatory receptor RET. ILC3-autonomous Ret ablation led to decreased innate interleukin-22 (IL-22), impaired epithelial reactivity, dysbiosis and increased susceptibility to bowel inflammation and infection. Neurotrophic factors directly controlled innate II22, downstream of p38 MAPK/ERK-AKT cascade and STAT3 activation. Strikingly, ILC3 were adjacent to neurotrophic factor expressing glial cells that exhibited stellate-shaped projections into ILC3 aggregates. Glial cells sensed microenvironmental cues in a MYD88 dependent manner to control neurotrophic factors and innate IL-22. Accordingly, glial-intrinsic Myd88 deletion led to impaired ILC3-derived IL-22 and pronounced propensity to gut inflammation and infection. Our work sheds light into a novel multi-tissue defence unit, revealing glial cells as central hubs of neuron and innate immune regulation via neurotrophic factor signals.

Group 3 innate lymphoid cells (ILC3) produce pro-inflammatory cytokines, regulate mucosal homeostasis and anti-microbial defence¹. In addition to their well-established developmentally regulated program, ILC3 are also controlled by microbial and dietary signals^1–6 raising the hypothesis that ILC3 possess other unexpected environmental sensing strategies. Neurotrophic factors are extra-cellular environmental cues to neurons and include the glial-derived neurotrophic factor (GDNF) family ligands (GFL) that activate the tyrosine kinase receptor RET in the nervous system, kidney and haematopoietic progenitors^7–11.

Analysis of gut lamina propria revealed that ILC3 express high levels of Ret (Fig.1a)^7,12, a finding confirmed at the protein level and by Ret^GFP knock-in mice (Fig.1b-d and Extended Data Fig.1a-d)¹³. ILC3 subsets expressed Ret^GFP and aggregated in Cryptopatches (CP) and Isolated Lymphoid Follicles (ILF), suggesting a role of neuroregulators in ILC3 (Fig.1b-d and Extended Data Fig.1b-j). To explore this hypothesis, we transplanted foetal liver cells from Ret competent (Ret^WT/GFP) or deficient (Ret^GFP/GFP)¹³ animals into alymphoid Rag1^-/-Yc^-/- hosts. Ret deficient chimeras revealed unperturbed ILC3 and CP development (Fig.1e). Strikingly, IL-22 expressing ILC3 were largely reduced despite normal IL-22 producing T cells (Fig.1f,g). In contrast, innate IL-17 was unaffected by Ret ablation (Fig.1f and Extended Data Fig.2a). In agreement, analysis of gain-of-function Ret^MEN2B mice¹⁴ revealed a selective increase of IL-22 producing ILC3 while their IL-17 counterparts were unaffected (Fig.1h and Extended Data Fig.2b). To more specifically evaluate the effects of RET in ILC3, we deleted Ret in RORγt expressing cells by breeding Rorgt-Cre to Ret^fl/fl mice^15,16 (Extended Data Fig.3a,b). Analysis of Rorgt-Cre.Ret^fl/fl (Ret^Δ) mice revealed selective and large reduction of ILC3-derived IL-22, but normal IL-22 producing T cells (Fig.2a and Extended Data Fig.3c,d). IL-22 acts on epithelial cells to induce reactivity and repair genes¹. When compared to their wild-type (WT) littermate controls, the Ret^Δ epithelium revealed normal morphology, proliferation and paracellular permeability, but a profound reduction of epithelial reactivity and repair genes (Fig.2b and Extended Data Fig.3e-h). Accordantly, the Ret^MEN2B epithelium displayed increased levels of these molecules in an IL-22 dependent manner (Fig.2b and Extended data Fig.3i). These results indicate that RET signals selectively control innate IL-22 and shape intestinal epithelial reactivity.

Figure 1. The neurotrophic factor receptor RET drives enteric ILC3-derived IL-22.

a, LTi, NCR^- and NCR⁺ ILC3 subsets, T cells (T), B cells (B), Dendritic cells (Dc), Macrophages (Mø), enteric Neurons (N) and mucosal Glial cells (G). b, Ret^GFP ILC3. c, Left: Ret^GFP gut. White: GFP. Right: ILC3 aggregates. d, Cryptopatches (CP), immature (iILF) and mature (mILF) isolated lymphoid follicles. Green: RET/GFP; Blue: RORγt; Red: B220. e, Ret^GFP chimeras. n=15. f,g, Ret^GFP chimeras. Ret^WT/GFP n=25; Ret^GFP/GFP n=22. h, Ret^MEN2B mice. n=7. Scale bars: 1mm (c left, e); 50µm (c right); 30µm (d). Data are representative of 4 independent experiments. Error bars show s.e.m. *P<0.05; **P<0.01; ns not significant.

Figure 2. ILC3-intrinsic RET signals regulate gut defence.

a, ILC3-derived cytokines. n=11. b, Ret^Δ and Ret^MEN2B mice compared to their WT littermate controls. n=7. c-f, DSS treatment. Ret^fl n=8; Ret^Δ n=8. c, Histopathology. d, Inflammation score and colon length. e, Innate IL-22. f, Bacterial translocation. g-j, DSS treatment. Ret^WT n=8; Ret^MEN2B n=8. g, Histopathology. h, Inflammation score and colon length. i, Innate IL-22. j, Bacterial translocation. k-n, C. rodentium infection. Rag1^-/-.Ret^fl n=15; Rag1^-/-.Ret^Δ n=17. k, Histopathology. l, Inflammation score and colon length. m, Innate IL-22. n, Infection burden. Scale bars: 200µm. Data are representative of 4 independent experiments. Error bars show s.e.m. *P<0.05; **P<0.01; ns not significant.

To interrogate whether neurotrophic factors regulate intestinal defence we tested how varying degrees of RET signals control enteric aggressions. While Ret^Δ mice treated with Dextran Sodium Sulfate (DSS) had increased weight loss and inflammation, reduced IL-22 producing ILC3, decreased epithelial reactivity/repair genes and pronounced bacterial translocation from the gut, Ret^MEN2B mutants were highly protected over their WT littermate controls (Fig.2c-j and Extended Data Fig.4). Since DSS mostly causes epithelial injury, we tested whether ILC3-autonomous RET signals are required to control infection. To this end, Ret^Δ mice were bred to Rag1^-/- mice to formally exclude adaptive T cell effects. Rag1^-/-.Ret^Δ mice were infected with the attaching and effacing bacteria Citrobacter rodentium. When compared to their littermate controls, Rag1^-/-.Ret^Δ mice had marked gut inflammation, reduced IL-22 producing ILC3, increased C. rodentium infection and translocation, reduced epithelial reactivity genes, increased weight loss and reduced survival (Fig.2k-n and Extended Data Fig.5). Altogether, these data indicate that ILC3-intrinsic neurotrophic factor cues regulate gut defence and homeostasis.

Formal definition that IL-22 is the molecular link between RET-dependent ILC3 activation and epithelial reactivity was provided by a multi-tissue organoid system. Addition of GFL to ILC3/epithelial organoids strongly induced epithelial reactivity genes in an IL-22 and RET dependent manner (Fig.3a,b and Extended Data Fig.6a). To further examine how RET signals control innate IL-22 we investigated a gene signature associated with ILC identity¹. While most of those genes were unperturbed, notably the master ILC transcription factors Runx1, Id2, Gata3, Rora, Rorgt, Ahr and Stat3, II22 was significantly reduced in Ret^Δ ILC3 (Fig.3c and Extended Data Fig.6b). In agreement, activation of ILC3 with all or distinct GFL/GFRα pairs in trans efficiently increased Il22 despite normal expression of other ILC3-related genes (Fig.3d and Extended Data Fig.6c). Activation of RET by GFL leads to p38 MAPK/ERK-AKT cascade activation in neurons, while phosphorylation of STAT3 shapes Il22 expression^7,17. Analysis of Ret^Δ ILC3 revealed hypo-phosphorylated ERK1/2, AKT, p38/MAP kinase and STAT3 (Fig.3e and Extended Data Fig.6d). Accordantly, GFL-induced RET activation in ILC3 led to rapid ERK1/2, AKT, p38/MAP kinase and STAT3 phosphorylation and increased Il22 transcription (Fig.3d,f and Extended Data Fig.6e,f). In agreement, inhibition of ERK, AKT or p38/MAP kinase upon GFL activation led to impaired STAT3 activation and Il22 expression (Fig.3g,h). Finally, inhibition of STAT3 upon GFL-induced RET activation led to decreased Il22 (Fig.3h). To examine whether GFL directly regulate Il22 we performed chromatin immunoprecipitation (ChIP) (Fig.3i,j)¹⁸. Stimulation of ILC3 with GFL resulted in increased binding of pSTAT3 in the Il22 promoter and increased trimethyl-H3K36 at the 3’ end of Il22, indicating active Il22 transcribed regions (Fig.3d,j)¹⁹. Thus, cell-autonomous RET signals control ILC3 function and gut defence via direct regulation of Il22 downstream of STAT3 activation.

Figure 3. ILC3-autonomous RET signals directly control Il22 downstream of pSTAT3.

a,b, Epithelial/ILC3 organoids. n=9. c, Ret^Δ ILC3 compared to their WT controls. n=4. d, ILC3 activation by GFL. n=4. e, Ret^Δ ILC3. pERK n=8; pAKT n=12; phosphorylated p38/MAP kinase n=6; pSTAT3 n=14. f, ILC3 activation by GFL. pERK n=10; pAKT n=16; phosphorylated p38/MAP kinase n=3; pSTAT3 n=15. g, pSTAT3 in ILC3 cultured with medium (n=7), GFL (n=11) or GFL and inhibitors for: p38 MAPK/ERK-AKT (LY) (n=7); ERK (PD) (n=7); AKT (VIII) (n=8); and p38 MAPK (SB) (n=6). h, Il22 in ILC3 cultured with GFL (n=17) or GFL and the inhibitors LY (n=18); PD (n=16); VIII (n=15); SB (n=15); and the STAT3 inhibitor (S3I) (n=8). i, Il22 locus. j. ChIP analysis of ILC3 stimulated with GFL. n=10. Data are representative of 3 independent experiments. Error bars show s.e.m. *P<0.05; **P<0.01; ns not significant.

Propensity to inflammation and dysregulation of intestinal homeostasis have been associated to dysbiosis^20,21. When compared to their WT littermates, Ret^Δ mice have altered microbial communities as evidenced by quantitative analysis, weighted UniFrac analysis and significantly altered levels of Sutterella, unclassified Clostridiales and Bacteroides (Fig.4a and Extended Data Fig.7). Discrete microbial communities may have transmissible colitogenic potential^20,21. Nevertheless, germ-free mice colonised with the microbiota of Ret^Δ or their control littermates revealed similar susceptibility to DSS-induced colitis and identical innate IL-22 (Fig.4b-d). In agreement, co-housed Ret^Δ and WT littermates had differential propensity to intestinal inflammation (Fig.2c,d). Together, these data indicate that dysbiosis per se is insufficient to cause altered innate IL-22 and susceptibility to gut inflammation as observed in Ret^Δ mice (Fig.2c-f). Thus, we hypothesised that GFL producing cells integrate commensal and environmental signals to control innate IL-22. Accordingly, antibiotic treatment of Ret^Δ and their WT littermate controls resulted in similar ILC3-derived IL-22 (Fig.4e)²².

Figure 4. Glial cells set GFL expression and innate IL-22, via MYD88-dependent sensing of the microenvironment.

a, Weighted Unifrac PCoA analysis and genus-level comparisons from co-housed Ret^fl (white circles) and Ret^Δ (black circles) littermates. n=5. Purple: Unclassified S24-7; Red: Bacteroides; Green: Sutterella; Blue: Unclassified Clostridiales; Grey: Other. b-d, DSS treatment of colonised germ-free (GF) mice. n=5. b, Histopathology. c, Inflammation score. d, Innate IL-22. e, Innate IL-22 after antibiotic treatment. n=8. f, Ret^GFP.Gfap-Cre.Rosa26^RFP mice. Green: RET/GFP; Red: GFAP/RFP. g,h, Glial cell activation with TLR2, TLR4, IL-1β receptor and IL-33 receptor ligands. n=6. i, TLR ligands, IL-1β and IL-33 activation of co-cultured ILC3 with WT (white bars) or Myd88^-/- glial cells (black bars). n=6. j-m, DSS treatment of Gfap-Cre.Myd88^Δ mice. n=12. j, Histopathology. k, Inflammation score and colon length. l, Innate IL-22. m, Body weight. Scale bars: 200µm (b, j); 10µm (f). Data are representative of 3-4 independent experiments. Error bars show s.e.m. *P<0.05; **P<0.01; ns not significant.

Neurotrophic factors of the GDNF family were shown to be produced by enteric glial cells, which are neuron-satellites expressing the glial fibrillary acidic protein (GFAP)^7,23. Strikingly, double reporter mice for ILC3 (Ret^GFP) and glial cells (Gfap-Cre.Rosa26^RFP) revealed that stellate-shaped projections of glial cells are adjacent (4.35μm±1.42) to RORγt⁺ ILC3 within CP (Fig.4f and Extended Data Fig.8a). These data suggest a paracrine glial-ILC3 crosstalk orchestrated by neurotrophic factors. In agreement, lamina propria glial cells were main producers of GFL (Extended Data Fig.8b). Recent studies have shown that glial cells express pattern recognition receptors, notably Toll-like receptors (TLRs)^24,25. Activation of neurosphere-derived glial cells revealed they specifically respond to TLR2, TLR4, and the alarmins IL-1β and IL-33, which efficiently controlled GFL expression and induced robust innate Il22 in a MYD88 dependent manner (Fig.4g-i and Extended Data Fig.8c-g). To formally demonstrate the physiological importance of MYD88-dependent glial cell sensing on innate IL-22, we deleted Myd88 in GFAP expressing glial cells by breeding Gfap-Cre to Myd88^fl/fl mice^26,27. Remarkably, glial-intrinsic deletion of Myd88 resulted in decreased intestinal GFL, increased gut inflammation, impaired ILC3-derived IL-22, and increased weight loss (Fig.4j-m; Extended Data Fig.9a-d). In agreement, Gfap-Cre.Myd88^Δ mice had increased susceptible to C.rodentium infection (Extended Data Fig.9e-h). Thus, mucosal glial cells orchestrate innate IL-22 via neurotrophic factors, downstream of MYD88-dependent sensing of commensal products and alarmins.

Defining the mechanisms by which ILC3 integrate environmental cues is critical to understand mucosal homeostasis. Our work sheds light on the relationships between ILC3 and their microenvironment, notably through decoding a novel glial-ILC3-epithelial cell unit orchestrated by neurotrophic factors (Extended Data Fig.10). Glial-derived neurotrophic factors operate in an ILC3-intrinsic manner by activating the tyrosine kinase RET, which directly regulates innate IL-22 downstream of p38 MAPK/ERK-AKT and STAT3 phosphorylation (Extended Data Fig.10). Future studies will elucidate further the mechanisms inducing RET expression in ILC3.

Our data demonstrate that in addition to their well-established capacity to integrate dendritic cell-derived cytokines¹, ILC3 perceive distinct multi-tissue regulatory signals leading to STAT3 activity and IL-22 expression, notably via integration of glial cell-derived neuroregulators. Thus, rather than providing hard-wired signals for ILC3-immunity, we propose that RET signals critically fine-tune innate IL-22 leading to efficient gut homeostasis and defence.

Previous studies demonstrated that neurons can indirectly shape foetal lymphoid tissue inducer cells and that ablation of glial cells leads to gut inflammation^28,29; here we reveal glial cells as central hubs of neuronal and innate immune regulation. Notably, neurotrophic factors are the molecular link between glial cell sensing, innate IL-22 and intestinal epithelial defence. Thus, it is tempting to speculate that glial/immune cell units might be also critical to the homeostasis of other barriers, notably in the skin, lung and brain³⁰. From an evolutionary perspective, coordination of innate immunity and neuronal function may ensure efficient mucosal homeostasis and a co-regulated neuro-immune response to various environmental challenges, including xenobiotics, intestinal infection, dietary aggressions and cancer.

Methods

Mice

C57BL/6J mice were purchased from Charles River. Ret^{GFP 13}, Rag1^-/-γc^{-/- 31,32}, Ret^{MEN2B 14}, Rosa26^{YFP 33}, Rosa26^{RFP 34}, Ret^{fl/fl 16}, Rorgt-Cre ¹⁵, Il1b^{-/- 35} and Myd88^{-/- 36} were in a full C57BL/6J background. Gfap-Cre²⁶ bred to Myd88^{fl/fl 27} were in F8-F9 to a C57Bl/6J background. All lines were bred and maintained at IMM Lisboa animal facility. Mice were systematically compared with co-housed littermate controls. Both males and females were used in this study. Randomization and blinding were not used unless stated otherwise. All animal experiments were approved by national and institutional ethical committees, respectively Direção Geral de Veterinária and iMM Lisboa ethical committee. Germ-free mice were housed at Instituto Gulbenkian de Ciência, Portugal, and Institut Pasteur, France, in accordance to institutional guidelines for animal care. Power analysis was performed to estimate the number of experimental mice.

Generation of foetal liver chimeras

For reconstitution experiments, 5x10⁶ foetal liver cells were isolated from E14.5 Ret^WT/GFP or Ret^GFP/GFP mice and injected intravenously into non-lethally irradiated (200rad) alymphoid Rag1^-/-γc^-/- hosts. Mice were analysed 8 weeks post-transplantation.

Dextran Sodium Sulphate-induced colitis

Dextran Sodium Sulphate (DSS) (molecular mass 36,000-50,000 Da; MP Biomedicals) was added into drinking water 3% (w/v) for 5 days followed by 2 days of regular water. Mice were analysed at day 7. Body weight, presence of blood and stool consistency was assessed daily.

Citrobacter rodentium infection

Infection with Citrobacter rodentium ICC180 (derived from DBS100 strain)³⁷ was performed by gavage inoculation of 10⁹ colony forming units^37,38. Acquisition and quantification of luciferase signal was performed in an IVIS system (Caliper Life Sciences). Throughout infection, weight loss, diarrhoea and bloody stools were monitored daily.

Antibiotic treatment

Pregnant females or new born mice were treated with streptomycin 5g/L, ampicillin 1g/L and colistin 1g/L (Sigma-Aldrich) into drinking water with 3% sucrose. Control mice were given 3% sucrose in drinking water as previously described²².

Microscopy

Intestines from Ret^GFP and Ret^GFP chimeras were imaged in a Zeiss Lumar V12 fluorescence stereo microscope with a NeoLumar S 0.8x objective using the GFP filter. Whole-mount analysis was performed as previously described^2,9. Briefly, adult intestines were flushed with cold PBS (Gibco) and opened longitudinally. Mucus and epithelium was removed and intestines were fixed in 4% PFA (Sigma-Aldrich) at room temperature for 10 minutes and incubated in blocking/permeabilising buffer solution (PBS containing 2% BSA, 2% goat serum, 0.6% Triton X-100). To visualise three-dimensional structures of the small intestine, samples were cleared with benzyl alcohol-benzyl benzoate (Sigma-Aldrich) prior dehydration in methanol^2,9. For analysis of thick gut sections intestines were fixed with 4% PFA at 4°C overnight and were then included in 4% low-melting temperature agarose (Invitrogen). Sections of 100μm were obtained with a Leica VT1200/VT1200 S vibratome and embedded in Mowiol (Calbiochem)². Slides or whole-mount samples were incubated overnight or for 1–2 days respectively at 4°C using the following antibodies: rat monoclonal anti-B220 (RA3-6B2) (eBioscience), mouse monoclonal anti-RORγt (Q31-378) (BD Pharmigen), mouse monoclonal anti-GFAP (GA-5) (Sigma-Aldrich), mouse monoclonal anti-GFAP Cy3 (GA-5) (Abcam), anti-GDNF antibody (Abcam), DAPI (4',6-Diamidino-2-Phenylindole, Dihydrochloride) (Invitrogen). A647 goat anti-rat, A568 goat anti-rat, A647 goat anti-mouse, A488 rabbit anti-GFP, and A488 goat anti-rabbit secondary antibodies were purchased from Invitrogen. Neurospheres and cultured glial cells were fixed in PFA 4% 10 minutes at room temperature and permeabilised in PBS-Triton 0.1% during 30 seconds. After several washing steps with PBS cells were incubated with antibodies during 3h at room temperature and then mounted in Mowiol³⁹. Samples were acquired on a Zeiss LSM710 confocal microscope using EC Plan-Neofluar 10x/0.30 M27, Plan Apochromat 20x/0.8 M27 and EC Plan-Neofluar 40x/1.30 objectives. Three-dimensional reconstruction of images was achieved using Imaris software and snapshot pictures were obtained from the three-dimensional images. For analysis of confocal images, cells were counted using in-house software, written in MATLAB (Mathworks, Natick, MA). Briefly, single-cell ILC3 nuclei were identified via RORγt by thresholding and particle analysis. Regions of interest (ROIs) (Extended Data Fig.1i; Bottom panels) were defined from each nucleus for analysis in the GFP channel, where staining was considered positive if a minimum number of pixels (usually 20) were above a given threshold. The software allows for batch processing of multiple images and generates individual report images for user verification of cell-counting results and co-expression analysis: (https://imm.medicina.ulisboa.pt/en/servicos-e-recursos/technical-facilities/bioimaging).

Histopathology analysis

Colon samples were fixed in 10% neutral buffered formalin. The colon was prepared in multiple cross-sections or “Swiss roll” technique⁴⁰, routine-processed for paraffin embedding and 3-4μm sections were stained with haematoxylin and eosin. Enteric lesions were scored by a pathologist blinded to experimental groups, according to previously published criteria^41–43. Briefly, lesions were individually scored (0-4 increasing severity) for the following criteria: 1-mucosal loss; 2-mucosal epithelial hyperplasia, 3-degree of inflammation, 4-extent of the section affected in any manner and 5-extent of the section affected in the most severe manner as previously described⁴³. Final scores were derived by summing the individual lesion and the extent scores. The internal diameter of the crypts was measured in at least five fields (10x magnification), corresponding to the hotspots in which the most severe changes in crypt architecture were seen. Measurements were performed in an average of 35 crypts per sample/mouse, from proximal to distal colon. Intestinal villus height was measured in the jejunum. Measurements were performed in slides scanned using a Hamamatsu Nanozoomer SQ digital slide scanner running NDP Scan software.

Enteric glial cell isolation

Enteric glial cells isolation was adapted from previously described protocols^44,45. Briefly, the muscularis layer was separated from the submucosa with surgical forceps under a dissection microscope (SteREO Lumar.V12, Zeiss). The lamina propria was scraped mechanically from the underlying submucosa using 1.5mm cover-slips (Thermo Scientific). Isolated tissues were collected and digested with Liberase TM (7,5 µg/mL; Roche) and DNase I (0.1mg/ mL; Roche) in RPMI supplemented with 1% hepes, sodium pyruvate, glutamine, streptomycin and penicillin and 0.1% β-mercaptoethanol (Gibco) for approximately 40min at 37°C. Single-cell suspensions were passed through a 100μm cell strainer (BD Biosciences) to eliminate clumps and debris.

Flow cytometry and cell sorting

Lamina propria cells were isolated as previously described⁴⁶. Briefly, intestines were digested with collagenase D (0.5mg/mL; Roche) and DNase I (0.1 mg/ mL; Roche) in RPMI supplemented with 10% FBS, 1% hepes, sodium pyruvate, glutamine, streptomycin and penicillin and 0.1% β-mercaptoethanol (Gibco) for approximately 30min at 37°C under gentle agitation. For cytokine analysis, cell suspensions were incubated 4h in PMA/Ionomycin (Sigma-Aldrich) and Brefeldin A (eBioscience) at 37ºC. Intracellular staining was performed using IC fixation/permeabilisation kit (eBioscience). Cells were stained using PBS, 1% FBS, 1% hepes and 0.6% EDTA (Gibco). Flow cytometry analysis and cell sorting were performed using FORTESSA and FACSAria flow cytometers (BD Biosciences). Data analysis was done using FlowJo software (Tristar). Sorted populations were >95% pure. Cell suspensions were stained with anti-CD45 (30-F11), anti-TER119 (TER-119), TCRβ (H57-597), anti-CD3ε (eBio500A2), anti-CD19 (eBio1D3), anti-NK1.1 (PK136), anti-CD11c (N418), anti-Gr1 (RB6-8C5), anti-CD11b (Mi/70), anti-CCR6 (29-2L17), anti-CD127 (IL-7Rα; A7R34), anti-Thy1.2 (53-2.1), anti-CD49b (DX5), anti-TCRδ (GL3), anti-NKp46 (29A1.4), anti-IL-17 (eBio17B7), anti-IL-22 (1H8PWSR), Rat IgG1 isotype control (eBRG1) antibodies, 7AAD viability dye, anti-Mouse CD16/CD32 (Fc block), anti-RORγt (AFKJS-9); Rat IgG2a_k Isotype Control (eBR2a) and streptavidin fluorochrome conjugates all from eBioscience; anti-CD4 (GK1.5), anti-CD31 (390), anti-CD8α (53-6.7), anti-CD24 (M1/69), anti-Epcam (G8.8) antibodies were purchased from Biolegend. Anti-RET (IC718A) antibody was purchased from R&D Systems. LIVE/DEAD Fixable Aqua Dead Cell Stain Kit was purchased from Invitrogen. Cell populations were defined as: ILC3 - CD45⁺Lin^-Thy1.2^hiIL7Rα⁺RORγt⁺; For ILC3 subsets additional markers were employed: LTi - CCR6⁺Nkp46^-; ILC3 NCR^- - CCR6^-Nkp46^-; ILC3 NCR⁺ - CCR6^-Nkp46⁺; Lineage was composed by CD3ε, CD8α, TCRβ, TCRγδ, CD19, Gr1, CD11c and TER119; Glial cells - CD45^-CD31^-TER119^-CD49b^{+ 47}; T cells - CD45⁺CD3ε⁺; γδ T cells - CD45⁺CD3ε⁺γδTCR⁺; B cells - CD45⁺CD19⁺B220⁺; Macrophages - CD45⁺CD11b⁺F4/80⁺; Dendritic cells - CD45⁺CD19^-CD3ε^-MHCII⁺CD11c⁺; enteric neurons - CD45^-RET/GFP^{+ 13}, Epithelial cells - CD45^-CD24⁺Epcam⁺.

Quantitative RT-PCR

Total RNA was extracted using RNeasy micro kit (Qiagen) or Trizol (Invitrogen) according to the manufacturer’s protocol. RNA concentration was determined using Nanodrop Spectrophotometer (Nanodrop Technologies). Quantitative real-time RT–PCR was performed as previously described^2,8,9. Hprt and Gapdh were used as housekeeping genes. For TaqMan assays (Applied Biosystems) RNA was retro-transcribed using a High Capacity RNA-to-cDNA Kit (Applied Biosystems), followed by a pre-amplification PCR using TaqMan PreAmp Master Mix (Applied Biosystems). TaqMan Gene Expression Master Mix (Applied Biosystems) was used in real-time PCR. TaqMan Gene Expression Assays (Applied Biosystems) were the following: Gapdh Mm99999915_g1; Hprt Mm00446968_m1; Artn Mm00507845_m1; Nrtn Mm03024002_m1; Gdnf Mm00599849_m1; Gfra1 Mm00439086_m1; Gfra2 Mm00433584_m1; Gfra3 Mm00494589_m1; Ret Mm00436304_m1; Il22 Mm01226722_g1; Il17a Mm00439618_m1; Il23r Mm00519943_m1; Rorgt Mm01261022_m1; Il7ra Mm00434295_m1; Ahr Mm00478932_m1; Stat3 Mm01219775_m1; Cxcr6 Mm02620517_s1; Nfkbiz Mm_00600522_m1; RegIIIa Mm01181787_m1; RegIIIb Mm00440616_g1; RegIIIg Mm00441127_m1; Defa1 Mm02524428_g1; Defa-rs1 Mm00655850_m1; Defa5 Mm00651548_g1; Defa21 Mm04206099_gH; Muc1 Mm00449599_m1; Muc3 Mm01207064_m1; Muc13 Mm00495397_m1; Gfap Mm01253033_m1; Ascl2 Mm01268891_g; Tff3 Mm00495590_m1; Relm-b Mm00445845_m1; Pla2g2a Mm00448160_m1; Pla2g5 Mm00448162_m1; Wnt3 Mm00437336_m1; Ctnnb1 Mm00483039_ m1; Axin2 Mm00443610_m1; Dll1b Mm01279269_m1; Il18 Mm00434225_m1; Tnfa Mm00443260_g1; Lyz1 Mm00657323_m1; Lrg5 Mm00438890_m1; Tbx21 Mm00450960_m1; Id2 Mm00711781_m1; Runx1 Mm01213404_m1; Notch1 Mm00435249_m1; Notch2 Mm00803077_m1; Gata3 Mm00484683_m1; Bcl2 Mm00477631_m1; Bcl2l1 Mm00437783_m1; Arntl Mm00500226_m1; Glpr2 Mm01329475_m1; Gja1 Mm01179639_s1; Ednrb Mm00432989; S100b Mm00485897_m1; Sox10 Mm00569909_m1. Real-time PCR analysis was performed using ABI Prism 7900HT Sequence Detection System or StepOne Real-Time PCR system (Applied Biosystems).

ILC3 activation and cell signalling

Sorted intestinal ILC3 cells were starved for 3 hours in RPMI at 37°C in order to ensure ILC3 viability. Ret^fl or Ret^Δ were analysed directly ex vivo. To test ERK, AKT, p38-MAPK (Cell Signaling Technology) and STAT3 (BD Pharmigen) upon GFL stimulation WT ILC3 were activated with 500ng/mL (each GFL) and co-receptors (rrGFR-α1, rmGFR-α2, rrGFR-α3 and rrGNDF from R&D Systems; rhNRTN and rhARTN from PeproTech) for 10 and 30min. When referring to the use of ‘GFL’, we have employed GDNF, NRTN, ARTN and their specific co-receptors in combination. For inhibition experiments cells were incubated 1h at 37°C before GFL stimulation, to test ERK, AKT, p38/MAPK and STAT3 phosphorylation, or during overnight stimulation with GFLs, to determine Il22 expression levels. Inhibitors were purchased from Sigma-Aldrich: p38 MAPK/ERK-AKT - LY294002 (LY); ERK - PD98059 (PD); AKT - AKT Inhibitor VIII (VIII); p38 MAPK - SB 202190 (SB); and pSTAT3 – S3I-201 (S3I).

Chromatin immunoprecipitation (ChIP) assay

Enteric ILC3 from adult C57BL/6J mice were isolated by flow cytometry. Cells were starved for 3h with RPMI supplemented with 1% hepes, sodium pyruvate, glutamine, streptomycin and penicillin and 0.1% β-mercaptoethanol (Gibco) at 37°C. Cells were stimulated with GFLs (500ng/mL each)⁸, lysed, cross-linked and chromosomal DNA-protein complex sonicated to generate DNA fragments ranging from 100-300 base pairs. DNA/protein complexes were immunoprecipitated, using LowCell# ChIP kit (Diagenode)¹⁸, with 3µg of rabbit polyclonal antibody against anti-pSTAT3 (Cell Signalling Technology), rabbit control IgG (Abcam) or H3K36me3 (07-030; Millipore). Immunoprecipitates were uncross-linked and analysed by quantitative PCR using primer pairs (5’-3’) flanking putative sites on Il22. Vehicle (BSA) stimulated ILC3s were used as controls. Il22 primer sequences were previously described^48–50, briefly: a, F-TGCAATCAATCCCAGTATTTTG and R-CTTGTGCAAGCATAAGTCTCAA; b, F-GAAGTTGGTGGGAAAATGAGTCCGTGA and R-GCCATGGCTTTGCCGTAGTAGATTCTG; c, F-ACGGGAGATCAAAGGCTGCTCT and R-GCCAACAAGGTGCTTTTGC; d, F-CTCACCGTGACGTTTTAGGG and R-GTGAATGATATGACATCAGAC; e, F-CGACGAACATGCTCCCCTGATGTTTTT and R-AAACTCATAGATTTCTGCAGGACAGCC; f, F-AGCTGCATCTCTTTCTCTCCA and R-TATCCTGAAGGCCAAAATAGGA; g, F-ACGACCAGAACATCCAGAAGA and R-GCAGAGAAAGAAATCCCCGC; h, F-AGGGGGACTTGCTTTGCCATTT and R-AACACCCCTTCTTTCCTCCTCCAT; i, F-CTGCTCCTTCCTGCCTTCTA and R-CTGAGCCAGGTTTCATGTGA. Primer positions are shown in Fig.3i relative to the transcription start codon of Il22.

Colony forming units and paracellular permeability

Organs were harvested, weighed, and brought into suspension. Bacterial colony forming units (CFU) were determined per gram of tissue and total organ. CFU were determined via serial dilutions on Luria Broth (LB) agar and MacConkey agar (Sigma-Aldrich). Colonies were counted after 2 days of culture at 37ºC. To address intestinal paracellular permeability 16 mg per mouse of Dextran-Fitc (Sigma Aldrich) were administrated by gavage after overnight starvation. Plasma was analysed after 4 hours of Dextran-Fitc administration using a Microplate Reader TECAN Infinity F500.

BrdU administration and Ki-67 labeling

BrdU was administrated by i.p. injection (1.25 mg/mouse). For flow cytometric analysis of epithelial cell proliferation anti-BrdU (Staining Kit for flow Cytometry- eBioscience) and anti-mouse Ki-67 antibody (BioLegend) were employed.

Quantitative PCR analysis of bacteria in stool at the Phylum level

DNA from faecal pellet samples was isolated with ZR Fecal DNA MicroPrep^™ (Zymo Research). Quantification of bacteria were determined from standard curves established by qPCR. qPCR were performed with Power SYBR^® Green PCR Master Mix (Applied Biosystems) and different primer sets using a StepOne Plus (Applied Biosystems) thermocycler. Samples were normalized to 16S rDNA and reported according to the 2^-ΔΔCT method. Primer sequences were: 16S rDNA, F- ACTCCTACGGGAGGCAGCAGT and R- ATTACCGCGGCTGCTGGC; Firmicutes, F- ACTCCTACGGGAGGCAGC and R-GCTTCTTAGTCAGGTACCGTCAT; Bacteroidetes, F- GGTTCTGAGAGGAGGTCCC and R-GCTGGCTCCCGTAGGAGT; Proteobacteria, F- GGTTCTGAGAGGAGGTCCC and R-GCTGGCTCCCGTAGGAGT.

16S rRNA quantification and gene sequencing

Faeces were isolated from co-housed Ret^fl or Ret^Δ littermates. Sequencing of the 16S rRNA gene was performed as previously described⁵¹. Briefly, barcoded primers were used to amplify the V4 region of the 16S rRNA gene, and the amplicons were sequenced on a MiSeq instrument (Illumina, San Diego, USA) using 150 bp, paired-end chemistry at the University of Pennsylvania Next Generation Sequencing Core. The paired ends were assembled and quality filtered, selecting for reads with a quality score ≥30. Reads with >10 bp homopolymers and reads shorter than 248 bp or longer than 255 bp were removed from the analysis. 16S rRNA sequence data were processed using mothur v 1.25.0⁵² and QIIME v 1.8⁵³. Chimeric sequences were removed with ChimeraSlayer⁵⁴. Operational taxonomic units (OTUs) were defined with CD-HIT⁵⁵ using 97% sequence similarity as a cut-off. Only OTUs containing ≥2 sequences were retained; OTUs assigned to Cyanobacteria or which were not assigned to any phylum were removed from the analysis. Taxonomy was assigned using the Ribosomal Database Project (RDP) classifier v 2.2⁵⁶, multiple sequence assignment was performed with PyNAST (v 1.2.2)⁵⁷, and FastTree⁵⁸ was used to build the phylogeny. Samples were rarified to 22,000 sequences per sample for alpha- and beta-diversity analyses. Taxonomic relative abundances are reported as the median with standard deviation. P values were calculated using the Wilcoxon rank-sum test. Statistical tests were conducted in R v. 3.2.0. To determine which factors were associated with microbial community composition, statistical tests were performed using the non-parametric analysis of similarities (ANOSIM) with weighted UniFrac distance metrics⁵⁹.

Data accession

The sequencing data generated in this study have been submitted to the NCBI Sequence Read Archive under BioProject PRJNA314493 (SRA: http://www.ncbi.nlm.nih.gov/sra/?term=PRJNA314493).

Intestinal organoids

IntestiCult™ Organoid Growth Medium and Gentle Cell Dissociation Reagent were purchased from StemCell. Intestinal crypts were isolated from C57BL/6J mice according to the manufacturer’s instructions and were added to previously thawed, ice-cold Matrigel at a 1:1 ratio and at a final concentration of 5,000-7,000 crypts/mL. 15µL of this mix was plated per well of a 96 well round-bottom plate. After Matrigel solidification 100µL of growth medium (100U/mL penicillin/streptomycin) was added and replaced every 3 days. Organoids were grown at 37ºC with 5% CO2 and passaged according to the manufacturer’s instructions. Freshly sorted intestinal ILC3 were added to 5-8 days old epithelial organoids after plating for 24 hours with or without anti-mouse IL-22 antibody (R&D Systems).

IL-22 agonist administration in vivo

150 μg of anti-IL-22 antibody (8E11; gift from Genentech, South San Francisco, CA) or mouse IgG1 isotype control (MOPC-21; Bio X Cell) was administered i.p. to Ret^MEN2B mice every 2 days. Animals were analysed 2 weeks after the first administration.

Neurosphere-derived glial cells

Neurosphere-derived glial cells were obtained as previously described⁶⁰. Briefly, total intestines from E14.5 C57BL/6J and Myd88^-/- mice were digested with collagenase D (0.5mg/mL; Roche) and DNase I (0.1mg/ mL; Roche) in DMEM/F-12, GlutaMAX, supplemented with 1% hepes, streptomycin/penicillin and 0.1% β-mercaptoethanol (Gibco) for approximately 30 minutes at 37°C under gentle agitation. Cells were cultured during 1 week in a CO2 incubator at 37 °C in DMEM/F-12, GlutaMAX™, streptomycin and penicillin and 0.1% β-mercaptoethanol (Gibco) supplemented with B27 (Gibco), EGF (Gibco) and FGF2 (Gibco) 20ng/mL. After 1 week of culture cells were treated with 0.05% trypsin (Gibco), transferred into PDL (Sigma-Aldrich) coated plates and culture in DMEM supplemented with 10% FBS, 1% hepes, glutamine, streptomycin and penicillin and 0.1% β-mercaptoethanol (Gibco) until confluence. Glial cells were activated with TLR2 (5μg/ml) (Pam3CSK4), TLR3 (100μg/ml) (PolyI:C), TLR4 (50ng/ml) (LPS), TLR9 (50μg/ml) (DsDNA-EC) ligands from Invivogen and IL-1β (10ng/mL) (401ML005), IL-18 (50ng/mL) (B002-5), IL-33 (0.1 ng/mL) (3626ML) recombinant proteins from R&D Systems. Cells were also co-cultured with purified ILC3 from WT and Il1b deficient mice. IL-22 expression in glial-ILC3 co-cultures upon TLR4 activation was also performed using GDNF (2μg/mL) (AB-212–NA), NRTN (2μg/mL) (AF-387sp) and ARTN (0.3μg/mL) (AF-1085-sp) blocking antibodies. Cells were analysed after 24 hours of co-culture.

Statistics

Results are shown as mean ± SEM. Statistical analysis used Microsoft Excel. Variance was analysed using F-test. Student’s t-test was performed on homocedastic populations, and Student’s t-test with Welch correction was applied on samples with different variances. Analysis of survival curves was performed using a MAntel-Cox test. Results were considered significant at *p ≤ 0.05; **p ≤ 0.01. Statistical treatment of metagenomics analysis is described in the methods section: 16S rRNA gene sequencing and analysis.

Extended Data

Extended Data Figure 1. ILC3 selectively express the neurotrophic factor receptor RET.

a, Expression of RET protein in gut CD45⁺Lin^-Thy1.2^hiIL7Rα⁺RORγt⁺ ILC3. b, Analysis of gut ILC3 from Ret^GFP mice. Embryonic day 14.5 (E14.5). c,d Analysis of enteric ILC3 subsets from Ret^GFP mice. e, Analysis of cytokine producing ILC3 from Ret^GFP mice. f, Pregnant Ret^GFP mice were provided with antibiotic cocktails that were maintained after birth until analysis at 6 weeks of age. Left: RET/GFP (white). Right: flow cytometry analysis of RET/GFP expression in ILC3. Thin line: Ab treated; Bold line: SPF. g, Ret expression in enteric ILC3 from Germ-Free (GF) mice and Specific Pathogen Free (SPF) controls. n=4. h, Analysis of lamina propria populations from Ret^GFP mice. i, Enteric ILC3 clusters. Green: RET/GFP; Blue: RORγt; Red: B220. Bottom: quantification analysis for RET/GFP and RORγt co-expression (79,97 ±4,72%). j, Rare RET expressing ILC3 in intestinal villi. Green: RET/GFP; Blue: RORγt; Red: CD3ε. Scale bars: 10µm. Data are representative of 4 independent experiments. Error bars show s.e.m. ns not significant.

Extended Data Figure 2. T cell-derived IL-22 and IL-17 in Ret^GFP chimeras and Ret^MEN2B mice.

a, T cell derived IL-17 in Ret^GFP chimeras. Ret^WT/GFP n=25; Ret^GFP/GFP n=22. b, T cell derived IL-22 and IL17 in the intestine of Ret^MEN2B mice and their WT littermate controls. Ret^WT n=7; Ret^MEN2B n=7. Data are representative of 4 independent experiments. Error bars show s.e.m. ns not significant.

Extended Data Figure 3. Enteric homeostasis in steady-state Ret^Δ mice.

a, Rorgt-Cre mice were bread to Rosa26^YFP. Analysis of Rosa26/YFP expression in gut ILC3 from Rorgt-Cre.Rosa26^YFP mice. b, Number of Peyer’s patches (PP). Ret^fl n=10; Ret^Δ n=10. c, T cell derived IL-22 in Ret^Δ mice and their WT littermate controls. Ret^fl n=11; Ret^Δ n=11. d, γδ T cell derived IL-22 in Ret^Δ mice and their WT littermate controls. Ret^fl n=4; Ret^Δ n=4. e, Intestinal villus and crypt morphology. Ret^fl n=6; Ret^Δ n=6. f, Epithelial cell proliferation. Ret^fl n=5; Ret^Δ n=5. g, Intestinal paracellular permeability measured by Dextran-Fitc in the plasma. Ret^fl n=5; Ret^Δ n=5. h, Tissue repair genes in Ret^Δ intestinal epithelium in comparison to their WT littermate controls. n=8. i, Reactivity genes in Ret^MEN2B mice treated with anti-IL-22 blocking antibodies in comparison to Ret^MEN2B intestinal epithelium. Ret^MEN2B n=4; Ret^MEN2B + anti-IL-22 n=4. Data are representative of 3 independent experiments. Error bars show s.e.m. ns not significant.

Extended Data Figure 4. Enteric inflammation in mice with altered RET signals.

Mice were treated with DSS in the drinking water. a, Weight loss of DSS treated Ret^Δ mice and their littermate controls. Ret^fl n=8; Ret^Δ n=8. b, T cell derived IL-22 in Ret^Δ mice and their WT littermate controls after DSS treatment. Ret^fl n=8; Ret^Δ n=8. c, Weight loss of DSS treated Ret^MEN2B mice and their WT littermate controls. Ret^WT n=8; Ret^MEN2B n=8. d, T cell derived IL-22 in Ret^MEN2B mice and their WT littermate controls. Ret^WT n=8; Ret^MEN2B n=8. e, Intestinal villi and crypt morphology. Ret^fl n=6; Ret^Δ n=6. f, Epithelial reactivity gene expression in DSS treated Ret^Δ mice in comparison to their WT littermate controls. n=8. g, Tissue repair gene expression in DSS treated Ret^Δ mice in comparison to their WT littermate controls. n=4. Data are representative of 3-4 independent experiments. Error bars show s.e.m. ns not significant. Error bars show s.e.m. *P<0.05; **P<0.01; ns not significant.

Extended Data Figure 5. Citrobacter rodentium infection in Ret^Δ mice.

a, C. rodentium translocation to the liver of Rag1^-/-.Ret^Δ and their Rag1^-/-.Ret^fl littermate controls at day 6 post-infection. n=15. b, MacConkey plates of liver cell suspensions from Rag1^-/-.Ret^Δ and their Rag1^-/-.Ret^fl littermate controls at day 6 after C. rodentium infection. c, Whole-body imaging of Rag1^-/-.Ret^Δ and their Rag1^-/-.Ret^fl littermate controls at day 6 after luciferase-expressing C. rodentium infection. d, Epithelial reactivity gene expression in C. rodentium infected Rag1^-/-.Ret^Δ mice and their Rag1^-/-.Ret^fl littermate controls. Rag1^-/-.Ret^fl n=15; Rag1^-/-.Ret^Δ n=17. e, Weight loss in C. rodentium infected Rag1^-/-.Ret^Δ mice and their Rag1^-/-.Ret^fl littermate controls. Rag1^-/-.Ret^fl n=8; Rag1^-/-.Ret^Δ n=8. f, Survival curves in C. rodentium infected Rag1^-/-.Ret^Δ mice and their Rag1^-/-.Ret^fl littermate controls. Rag1^-/-.Ret^fl n=8; Rag1^-/-.Ret^Δ n=8. g, C. rodentium translocation to the liver of Ret^Δ and their Ret^fl littermate controls at day 6 post-infection. n=6. h, MacConkey plates of liver cell suspensions from Ret^Δ and their Ret^fl littermate controls at day 6 after C. rodentium infection. i, Whole-body imaging of Ret^Δ and their Ret^fl littermate controls at day 6 after luciferase-expressing C. rodentium infection. j, C. rodentium infection burden. Ret^fl n=8; Ret^Δ n=8. k, Innate IL-22 in in C. rodentium infected Ret^Δ mice and their Ret^fl littermate controls. Ret^fl n=8; Ret^Δ n=8. Data are representative of 3-4 independent experiments. Error bars show s.e.m. ns not significant. Error bars show s.e.m. *P<0.05; **P<0.01; ns not significant.

Extended Data Figure 6. Glial-derived neurotrophic factor family ligand (GFL) signals in ILC3.

a, Multi-tissue intestinal organoid system. Scale bar: 20µm. Black arrows: ILC3. b, Expression of ILC-related genes in ILC3 from Ret^Δ mice in comparison to their littermate controls. n=4. c, ILC3 activation with all GFL/GFRα pairs (GFL); single GDNF family ligand (GDNF, ARTN or NRTN); or single GFL/GFRα pairs (GDNF/GFRα1, ARTN/GFRα3 or NRTN/GFRα2) compared to vehicle BSA. n=5. d, ILC3 from Ret^Δ mice (open black) and their littermate controls (open dash). Isotype (closed grey). e, 30 minutes activation of ILC3 by GFL (open black) compared to vehicle BSA (open dash). Isotype (closed grey). f, 10 minutes activation of ILC3 by GFL. pERK n=8; pAKT n=8; phosphorylated p38/MAP kinase n=8; pSTAT3 n=8. Similar results were obtained in at least 3-4 independent experiments. Error bars show s.e.m. *P<0.05; **P<0.01; ns not significant.

Extended Data Figure 7. Alterations in the diversity of intestinal commensal bacteria of Ret^Δ mice.

a, Quantitative PCR analysis at the Phylum level in stool bacterial from co-housed Ret^fl and Ret^Δ littermates in steady state. n=5. b, Metagenomic Phylum level comparisons in stool bacterial from co-housed Ret^fl and Ret^Δ littermates in steady state (left) and after DSS treatment (right). n=5. c, Genus level comparisons in stool bacterial from co-housed Ret^fl and Ret^Δ littermates in steady state (left) and after DSS treatment (right). n=5. Error bars show s.e.m. *P<0.05; **P<0.01; ns not significant.

Extended Data Figure 8. GFL expressing glial cells anatomically co-localise with ILC3.

a, Intestine of Ret^GFP mice. Green: RET/GFP; Red: GFAP; Blue: RORγt. Similar results were obtained in three independent experiments. b, Purified lamina propria LTi, NCR^- and NCR⁺ ILC3 subsets, T cells (T), B cells (B), Dendritic cells (Dc), Macrophages (Mø), enteric Neurons (N) and mucosal Glial cells (G). c, Neurosphere-derived glial cells. d, M: medium. Activation of neurosphere-derived glial cells with TLR2 (Pam3CSK4), TLR3 (Poli I:C), TLR4 (LPS) and TLR9 (DsDNA-EC) ligands, as well as IL-1β, IL-18 and IL-33. n=6. e, Il22 in co-cultures of glial and ILC3 using single or combined GFL antagonists. n=6. f, Il22 in co-cultures of ILC3 and glial cells from Il1b^-/- or their WT controls. n=3. g, Gdnf, Artn and Nrtn expression in glial cells and ILC3 upon TLR2 stimulation. n=3. Scale bar: 30µm.Similar results were obtained in at least 4 independent experiments.

Extended Data Figure 9. Glial cell sensing via MYD88 signals.

a-c, Intestinal glial cells were purified by flow cytometry. a, Germ-free (GF) and their respective Specific Pathogen Free (SPF) controls. n=3. b, Myd88^-/- and their respective WT littermate controls. n=3. c, Gfap-Cre.Myd88^Δ and their littermate controls (Myd88^fl). n=3. d, Total lamina propria cells of Gfap-Cre.Myd88^Δ and their littermate controls (Myd88^fl). n=6. e-h, Citrobacter rodentium infection of Gfap-Cre.Myd88^Δ mice and their littermate controls (Myd88^fl). n=6. e, Innate IL-22. f, Citrobacter rodentium translocation. g, Infection burden. h, Weight loss. Data are representative of 3 independent experiments. Error bars show s.e.m. *P<0.05; **P<0.01; ns not significant.

Extended Data Figure 10. A novel glial-ILC3-epithelial cell unit orchestrated by neurotrophic factors.

Lamina propria glial cells sense microenvironmental products, that control neurotrophic factor expression. Glial-derived neurotrophic factors operate in an ILC3-intrinsic manner by activating the tyrosine kinase RET, which directly regulates innate IL-22 downstream of a p38 MAPK/ERK-AKT cascade and STAT3 phosphorylation. GFL induced innate IL-22 acts on epithelial cells to induce reactivity gene expression (CBP: Commensal bacterial products; AMP: antimicrobial peptides; Muc: mucins). Thus, neurotrophic factors are the molecular link between glial cell sensing, innate IL-22 production and intestinal epithelial barrier defence.