ILC3 GM-CSF production and mobilisation orchestrate acute intestinal inflammation

1Innate lymphoid cells (ILCs) are a recently defined family of evolutionarily ancient cells involved in many facets of host defence. As with conventional Th1, Th2 and Th17 T cells, ILCs can be functionally classified based on the expression of transcription factors and associated signature cytokines. ILC1 are defined by Th1-like and ILC2 by Th2-like cytokine responses, whereas ILC3 express RORγt, and can secrete IL-17 and/or IL-22 upon activation. The ILC3 family includes the prototypic foetal lymphoid tissue inducer (LTi) cells that play a non-redundant role in lymphoid tissue development via lymphotoxin-dependent interactions with stromal cells. This pathway is required not only for lymph node development but also for organised lymphoid structures in the gut such as dendritic cell (DC) and ILC containing cryptopatches (CP), B cell, DC and ILC containing isolated lymphoid follicles (ILFs), and small intestinal Peyer’s Patches (PP) that have a very similar structure to lymph nodes. Postnatally, the LTi population can also make IL-17 and IL-22 and contribute to host defence against pathogens, particularly in the gut.

2The multi-functional roles of ILCs in disease development and pathogenesis, as well as host defence and repair have been the focus of much interest. Polyfunctional ILCs have also been described that do not fall into distinct ILC1 or ILC3 phenotypes but express both ILC1 and ILC3 lineage defining transcription factors Tbet and RORγt and secrete multiple cytokines such IL-17A, IL-22 as well as IFNγ.

3We previously identified a critical role for a phenotypically distinct population of IL-23R+ RORγt+ CD4- NKp46- ILCs in the development of innate intestinal inflammation in Rag-/- mice following Helicobacter hepaticus infection or αCD40 stimulation. Similar ILC populations were enriched in the colonic mucosa of patients with inflammatory bowel disease (IBD), implicating IL-23-responsive, RORγt expressing ILCs in the pathogenesis of inflammatory gut disease in mice and humans. However, it remains unclear how ILCs, that are numerically sparse in vivo can initiate inflammatory processes that lead to colitis.

4Despite advances in understanding of the functions of ILCs, little is known about their location in tissue at different stages of the inflammatory response, and how putative structural and cytokine-mediated functions are co-ordinated. Since its description in 2006, the induction of colitis by injecting agonistic anti-CD40 antibody has become an important tool to assess ILC-driven acute colitis. By contrast with other models, anti-CD40 induced colitis follows discrete phases at well-defined time points following initiation, offering the opportunity to probe the role of leukocytes in the development and amplification of the inflammatory response. Experiments have demonstrated that intestinal inflammation was mediated via Thy1+ ILCs in a rorc dependent manner, making it an ideal system to study how ILCs contribute to pathogenesis. A recent study investigating potential biomarkers for anti-IL-23 therapy described similar changes in the colons of both anti-CD40-treated mice and patients with active Crohn’s disease.

5Many recent publications have focused on the specific functions of ILC subsets within effector sites, and the location of ILCs has been proposed to contribute to their ability to affect systemic cytokine levels. Despite histological and flow cytometry data demonstrating the presence of ILCs within lymphoid structures in the gut, it isn’t clear whether they function as sedentary, cytokine producing cells or play a more active role in cell interactions and organization. In vivo microscopy is a tool that provides an opportunity to look at the behaviour of ILCs within the tissue. By combining anti-CD40 stimulation with intra-vital microscopy we are able to reliably track cellular changes at discrete phases of disease and capture cell movement at key timepoints.

6Our results show two novel mechanisms through which the small number of ILCs found in vivo orchestrate the intestinal inflammatory response. IL-23-driven GM-CSF production by ILC3s is critical for the development of colitis, and ILCs mobilise from cryptopatches after activation in a GM-CSF-dependent manner. Both of these behaviours likely contribute to the ability of ILCs to coordinate the immune response in the gut. Initiation and perpetuation of disease occur in distinct anatomical compartments, indicating both a temporal and spatial switch of ILC function during inflammatory conditions.

7Anti-CD40 induced colitis is dependent on a RORγt/IL-23 axis but key downstream cytokines are less well understood. As IL-17 and IL-22 are major downstream effectors of the IL-23 signalling axis we first investigated their role in anti-CD40 colitis. However, blockade of IL-17A failed to modify anti-CD40-induced systemic or intestinal disease (Figure 1A,B), indicating that IL-17A is dispensable for development of acute colitis in this model. Blocking the closely related molecule IL-17F also failed to modify disease (Figure 1—figure supplement 1).

8Like IL-17 blockade, treatment with a neutralising anti-IL-22 mAb had no effect on systemic disease or colitis in the proximal colon (Figure 1A&B), indicating that in our facility, IL-22 is also redundant for disease induction in this model. However, blockade of IL-17A and anti-IL-22 reduced intestinal inflammation but not systemic disease indicating redundant roles for IL-17A and IL-22 in the intestinal inflammatory response (Figure 1—figure supplement 1).

9As we have described an important role for GM-CSF in IL-23 driven T cell dependent colitis we next investigated the role of GM-CSF in innate colitis models. Strikingly, administration of an anti-GM-CSF blocking mAb reduced both weight loss (Figure 1C) and severity of colitis (Figure 1D). Indeed amelioration of colitis with GM-CSF blockade was similar to that observed with anti-IL-23R mAb treatment. Similar effects of GM-CSF blockade were also observed in bacteria-induced innate colitis following Helicobacter hepaticus infection of 129SvEv Rag2-/- mice (Figure1—figure supplement 2) supporting a pivotal role for this cytokine in both T cell dependent and innate colitis. GM-CSF has been shown to promote CNS inflammation through effects on inflammatory monocytes. Consistent with this and in line with a recent report GM-CSF blockade led to a marked reduction in the number of inflammatory monocytes in the colon of anti-CD40 treated mice (Figure 1E). The number of neutrophils and eosinophils also decreased, but this was not significant. These changes were accompanied by an increase in both the percentage and total number of CD11b- CD103+ dendritic cells (Figure 1E).

10We next investigated GM-CSF expression by flow cytometry to identify the cellular source in intestinal inflammation. Only a small proportion of epithelial, stromal or lineage positive (CD11b, CD11c, Gr-1, B220, CD49b) cells expressed GM-CSF (Figure 2A). The majority of these lineage positive GM-CSF producers were NK cells, and they expressed a lower amount of GM-CSF than ILCs (Figure 2—figure supplement 1A & B). However, using Rag and IL-15 receptor double deficient mice that lack NK cells, we found, as previously reported, that NK cells were not required for the development of anti-CD40 mediated colitis (Figure 2—figure supplement 1C–F). By contrast, a significant proportion of ILCs were capable of producing GM-CSF, even in the healthy intestine, and this proportion increased during colitis (Figure 2B). Consistent with an IL-23-dependent ILC3 phenotype, GM-CSF+ ILCs were CCR6-, cKit- and NKR-, and expressed only low amounts of T-bet (Figure 2A).

11We next investigated the cytokine milieu that regulates GM-CSF production by ILCs in vivo. To determine whether IL-23 signalling induces GM-CSF, we cultured total colonic lamina propria cells overnight with IL-12 or IL-23 and measured GM-CSF production in the cultures. IL-12 had no effect on either the proportion of GM-CSF-expressing ILCs or the amount of GM-CSF per cell (Figure 2C). By contrast and contrary to a previous study using sorted ILC3s, culture with IL-23 increased both the proportion of GM-CSF expressing ILCs as well as GM-CSF production per cell (Figure 2C).

12Analysis of Il23a mRNA expression within colonic tissue revealed a peak in expression at day 1 following anti-CD40 treatment that rapidly declined but was followed by increased expression of Csf2, the gene for GM-CSF, at day 3 (Figure 2D). The increase in both Il23a and Csf2 was abrogated in mice treated with a blocking antibody to GM-CSF, indicating that GM-CSF may be required to maintain and amplify expression of Il23a.

13To assess whether GM-CSF production may also be relevant in human IBD, we first analyzed GM-CSF production from blood ILCs. PMA and ionomycin stimulated blood ILC3s produced GM-CSF with very little production from other ILC subsets (Figure 3A&B, gating strategy in Figure 3—figure supplement 1). Indeed, a greater proportion of ILCs produced GM-CSF than did T cells from the same healthy donors (Figure 3C). ILCs secreting GM-CSF are further enriched in the colon (Figure 3D). Critically, the proportion of blood ILCs capable of GM-CSF production along with other disease-associated cytokines such as IFNγ and TNFα was greater in IBD patients than in healthy controls (Figure 3E).

14Analysis of a publically available dataset of colon tissue from healthy controls and IBD patients showed a significant increase in CSF2 gene expression in the colons of patients with either Crohn’s disease or ulcerative colitis compared with controls (Figure 3F). This was confirmed by gene expression analysis of colon biopsies from the Oxford cohort of IBD patients (Figure 3G). CSF2 expression was also higher in uninflamed biopsies from IBD patients compared with controls. There was significantly greater expression again in lesional areas in active disease, suggesting that GM-CSF expression correlates with, and may be a driver of, inflammation in IBD.

15To date attention has focussed on the capacity of ILCs in tissue to produce immune modulatory cytokines with little emphasis on how ILC positioning within a tissue may impact on functional outcomes. Under non-inflammatory conditions, ILCs are found within lymphoid structures in the colon, and in Rag-deficient hosts these structures are limited to cryptopatches as there are no B cells to form ILFs. Under homeostatic conditions ILCs primarily reside within lymphoid aggregate cryptopatch (CP) structures (Figure 4A) and are found only rarely within the non-inflamed lamina propria. Within CP, RORγt+ and IL-7Rα+ ILC3s are present at high density surrounded by CD11c and MHCII expressing cells. Some of these ILCs within CP also express CD4, suggesting that some but not all are classical LTi cells that are known to be involved in lymphoid organogenesis (Figure 4A). In addition to RORγt expression, ILCs are known to express IL-23R. FACS analysis of Rag-deficient mice that express GFP under control of the Il23r demonstrates that these mice can be used to study ILC behaviour in vivo (Figure 4B). Two-photon imaging of Il23rgfp/+ Rag-/- mice shows ILCs (green) present within a cryptopatch (Figure 4C, and Video 1). Quantification of ILC localization in 3D indicates that greater than 90% of ILCs are present within cryptopatches under steady-state conditions (Figure 4D). This suggests that IL-23R+ ILCs are perfectly positioned in close proximity to a high density of CD11c+ DCs that can be activated by colitogenic stimuli. Indeed Il23r mRNA increased in the colon 6 hr following anti-CD40 injection indicating that the first ILC changes occur in the hours following CD40 stimulation (Figure 4—figure supplement 1A).

16We further examined the changes in IL-23R expression following anti-CD40 stimulation, and found a decrease in IL-23R+ ILCs at day 3 (Figure 4—figure supplement 1B&C), which may indicate a change in ILC phenotype. In support of this hypothesis, analysis of RORγt protein revealed a transient decrease in the proportion of ILCs expressing it 3 days following anti-CD40 stimulation (Figure 4—figure supplement 1D&E), similar to a previous report. Whilst these data suggested that ILCs are reacting to changes in the environment, it remained unclear how this leads to the downstream inflammatory cascade.

17It is possible that ILCs that are resident within lymphoid structures produce cytokines locally to carry out their effects. Alternatively, ILCs may be able to mobilize into the tissue to coordinate the immune response. To understand which mechanism ILCs utilize, we adopted the strategy of McDole et al used to study leukocytes in the gut. In the steady-state, ILCs were much more motile than expected. While little motility could be observed within the cryptopatch, ILCs can be observed to enter and exit the cluster near the basal collagen layer (Figure 5A and Videos 2 and 3). Interestingly, the cryptopatch did not have an obvious point of entry or exit, implying that trafficking into and out of the structure may be governed by mechanisms different to that of the lymph node. By gating on motile cells outside the cluster, we could see that the average velocity of motile ILCs (approximately 9 μm/min) is grossly similar to that of other lymphocytes and while cell velocity decreases after treatment with anti-CD40 (Figure 5B), this difference does not appear to affect overall cell trafficking. To determine whether ILC trafficking changed in colitis, we plotted the displacements of ILCs in steady-state and after anti-CD40 treatment (Figure 5C, Video 4, and Video 5). While no gross changes appeared in track displacements, the overall movement of cells shifted. In steady-state, roughly equal numbers of cells move toward and away from the cluster; however, just 4–6 hr after anti-CD40 treatment, a greater proportion of cells exit the tissue and a smaller proportion enter (Figure 5D). The ratio of entering to exiting cells indicates a significant skewing toward egress from cryptopatches (Figure 5E). To assess whether GM-CSF may be playing a role in ILC mobilization, we treated mice with anti-GM-CSF 24 hr before imaging. The shift toward exit that is normally observed after anti-CD40 treatment was strikingly reversed when mice were treated with a GM-CSF blocking antibody 24 hr before imaging (Figure 5D and E, Video 6). It is possible that many of these cells move through blood vessels, but by labelling the blood vessels with texas-red dextran, ILCs appeared to be adjacent to blood vessels, not within them (Figure 5F). This indicates that ILCs may be mobilizing into the adjacent tissue from cryptopatches.

18The initial characterisation of the anti-CD40 model identified an influx of myeloid cells, particularly CD40+ MHC-II+ CD11chi DCs, in specific inflammatory foci within the tip of the villi of the proximal colon during colitis. We hypothesised that development of colitis is driven by initiation of the immune response in the cryptopatch, followed by accumulation of immune cells within anatomically distinct villus inflammatory foci.

19To understand what impact the early mobilization of ILCs has on the tissue, we looked at several chemokines that could be influencing the development of inflammatory foci. 24 hr after anti-CD40 treatment the chemokines Ccl19 and Ccl2 were increased in the tissue. The receptor for Ccl19, Ccr7, was also increased (Figure 6A). Blocking GM-CSF prevented the increase of Ccr7 and Ccl2, supporting the idea that there is cross-talk between the GM-CSF and migration axes. Staining for CCR7 and CCR2 on lineage+ and ILC subsets suggested that it is the lineage positive cells that are capable of responding to the increased tissue chemokines (Figure 6B). Histological analysis of the proximal colon 7 days after anti-CD40 treatment showed immune foci at the tips of villi (Figure 6C). Immunofluorescence staining for CD11c and RORγt showed colocalization of ILCs with the CD11c positive immune infiltrate. The latter may reflect the inflammatory monocytes identified by flow cytometry from colon digests (Figure 6D and 1E). Quantification of cryptopatch cellularity from H&E after anti-CD40 treatment showed a decrease in cellularity after 24 hr (Figure 6E). The net egress of cells from cryptopatches over a short imaging window (Figure 5D) corresponds to the loss of cells from the tissue over 24 hr. This decrease is followed by an increase in inflammatory infiltrates starting one day later (Figure 6F). Blocking GM-CSF resulted in maintenance of cryptopatch structure as would be predicted from the imaging data (Figure 6G). Taken together, these data indicate that ILC mobilization from cryptopatches is associated with a decrease in cryptopatch cellularity, increase in tissue chemokines, recruitment of inflammatory cells, and development of inflammatory foci associated with active disease.

20The results of this study reveal a new IL-23-driven GM-CSF axis that promotes ILC3-mediated acute colitis. Our data indicate a spatiotemporal model for innate colitis that involves cellular activation, GM-CSF dependent mobilization of IL-23R+ ILCs from cryptopatches, and subsequent perpetuation of the response through separate ILC-containing inflammatory foci and GM-CSF dependent accumulation of inflammatory monocytes.

21Although IL-23 dependent, we found that anti-CD40 induced colitis was IL-17A and IL-17F independent in keeping with earlier reports (Figure 1 and ). Two recent reports indicated a functional role for IL-22, possibly through neutrophil recruitment. However our data show no impact of IL-22 blockade or neutrophil depletion (data not shown) on the incidence or severity of disease in this model. By contrast, double blockade of IL-17A and IL-22 reduced colitis severity, indicating that these cytokines may play a redundant role in driving intestinal inflammation possibly through shared effects on intestinal epithelial cells. The pathogenic role of IL-22 in CD40-triggered colitis may depend on the presence of microbial factors that also influence IL-17A production in the intestine. Consistent with this, IL-22 alone was required for the maintenance of bacteria driven innate colitis-associated cancer. Together these results suggest that context-dependent effector functions mediate the colitogenic potential of IL-23.

22Recent studies have focussed on the role of GM-CSF as an important mediator of IL-23 driven tissue inflammation. We found that GM-CSF played a non-redundant role in both systemic and intestinal inflammation in the anti-CD40 model and was required for bacteria-induced innate colitis. The role of GM-CSF in anti-CD40 mediated systemic disease has been recently reported but effects on colitis were not assessed. Our results indicate that IL-23-responsive ILCs can act as the main source of GM-CSF in an innate model. Importantly we also find that ILCs are potent sources of GM-CSF in humans, consistent with a previous report describing cytokine-driven GM-CSF production from an NKp44- ILC population. Our study goes further to show that GM-CSF is produced by ILC3s in the blood and colon of patients. The percent of GM-CSF+ ILCs is larger than the percent of GM-CSF+ T cells. This GM-CSF+ ILC3 population is increased in the blood and colonic tissue of patients with IBD, indicating commonalities between mouse models and human disease pathogenesis.

23IL-23 and GM-CSF may act in an autocrine loop early in the early inflammatory response as we found that anti-CD40-induced early colonic IL-23 expression was GM-CSF dependent and that in turn, IL-23 was required for sustained GM-CSF production by ILCs. A similar positive feedback loop has been described for T cell derived GM-CSF in CNS inflammation. Such a feedback loop would promote more GM-CSF production from ILCs as IL-23 production increases in disease. The role of GM-CSF may be to recruit myeloid cells to sites of inflammation, and impact their ability to coordinate inflammatory responses, as has recently been shown in EAE. Indeed, we found blocking GM-CSF prevented the accumulation of Ly6C+ inflammatory monocytes following anti-CD40 stimulation, in line with recent evidence for a key role for these cells in driving pathogenesis during anti-CD40 induced colitis. However, GM-CSF expression also has wider consequences since it affects the behaviour of ILCs within the cryptopatch.

24Functional analyses of ILC populations have mainly been restricted to cytokine production, and recent evidence that they do not recirculate suggested a primarily sedentary role for these cells. However, that study did not assess ILC movement within tissues as a mechanism through which ILCs may modulate the immune and inflammatory response. To investigate this, we utilized intra vital microscopy to image ILC movement within cryptopatches under homeostatic conditions and in anti-CD40 induced intestinal inflammation. Cryptopatches may provide a platform for the rapid amplification of the immune response and may provide a potential mechanism by which a very small number of ILCs could initiate an immune cascade that culminates in colonic inflammation. Because of the large overlap of cell surface molecule and transcription factor expression between ILCs and T cells, imaging Rag-/- mice provided the cleanest system to study ILC motility. The anti-CD40 model provided temporal synchronization, which is a feature unique to this model of colitis. To our surprise, a subset of ILCs associated with the cryptopatch was migratory under steady-state conditions. After anti-CD40 treatment, ILC motility shifted toward egress from the cryptopatch. In agreement with data that suggests they do not recirculate through the blood, ILCs were not observed migrating into the peripheral blood but appeared to crawl directly through the tissue, possibly following a local chemokine gradient. Future studies will be required to determine what factors govern entrance and exit from lymphoid structures within the tissue, which appears very different from the lymph node. Further understanding of this response may identify pathways that can be targeted to break the proinflammatory cycle within the tissue. As tools become available, studying ILC behaviour in lymphorepleate hosts will further expand these findings and allow increased interrogation of cellular interactions between ILCs and other immune cells as well as spatio-temporal control points in the inflammatory response.

25Consistent with chemokine-induced migration of ILCs, we found an early peak in mRNA expression of the chemokines Ccl2, Ccl19 and its receptor Ccr7 in the tissue, suggesting an early role for CCR7 and CCR2. Despite data showing that migration of LTi ILC3s from the gut to the mesenteric lymph nodes is dependent on CCR7 expression, expression of CCR7 on ILCs was not assessed in that study and we were unable to identify CCR7 or CCR2 on intestinal ILC3s in anti-CD40 induced colitis. While this argues against CCR7 driven chemotaxis of the ILCs, it does not preclude an organizational function for ILCs in the tissue driven by other mediators. Indeed, our data indicate a role for GM-CSF in the net egress of ILCs from cryptopatches in anti-CD40 induced colitis. Similarly, increased amounts of ccl2 and ccr7 mRNA in the colon were also GM-CSF-dependent, as was accumulation of inflammatory monocytes which also express CCR7 and CCR2. Together, the data are compatible with a model in which GM-CSF-dependent activation of ILCs leads to exit from the cryptopatch to the villus tip. This coincides with the recruitment of CCR7 and CCR2 expressing inflammatory monocytes through a CCL19 and CCL2 gradient, culminating in tissue damage. Sustained production of GM-CSF by ILCs in inflammatory foci could recruit inflammatory progenitor cells and contribute to differentiation of myeloid effector cells. GM-CSF has been shown to control the differentiation of inflammatory monocytes, and this population was markedly reduced following anti-GM-CSF treatment. GM-CSF may also impact on this pathway through altered haematopoiesis and accumulation of granulocyte macrophage progenitors as described in T cell dependent colitis. Given the multiple activities of GM-CSF on myeloid cell development and function, further studies will be required to delineate the key downstream pathways.

26IL-23 has been suggested as a therapeutic target in IBD. However, our studies indicate that neutralisation of GM-CSF is as efficacious as blockade of IL-23 in mouse models. In addition, increases in GM-CSF secreting ILCs are a feature of IBD. However, targeting GM-CSF may not be straightforward in IBD as there are clear host protective functions of GM-CSF in the intestine. Indeed neutralising anti-GM-CSF autoantibodies are increased in IBD, particularly in patients with ileal and stricturing disease. Furthermore, administration of recombinant GM-CSF (sargramostim) has been used to treat Crohn’s disease although further trials did not show any substantial benefit of GM-CSF treatment. Despite these issues, in light of recent developments testing anti-GM-CSF in clinical trials for the treatment of other inflammatory conditions such as multiple sclerosis it will be interesting to see if subgroups of patients with hyperinflammatory innate immune activation-driven IBD could benefit from GM-CSF blockade.

27IBD and colorectal cancer (CRC) patients were recruited through the Oxford IBD cohort study, and samples were obtained from Oxford GI Biobank in collaboration with Oxford Radcliffe Biobank. Colon specimens were collected from patients undergoing surgery for severe disease. Macroscopically healthy sections from CRC patients were used as controls. For gene expression data, intestinal mucosal pinch biopsies were taken from patients undergoing routine endoscopy. From 20 patients with evidence of active intestinal inflammation, matched biopsies were collected from both inflamed tissue and regions of the intestine with no apparent inflammation. IBD blood samples were collected from patients attending the outpatient clinic and in some cases from patients undergoing surgery. Control blood samples were obtained from healthy donors. Ethical approval was obtained from the Oxfordshire Research Ethics Committee (Reference numbers: 11/YH/0020 and 09/H0606/5) and informed written consent was given by all study patients.

28A previously published study with a publically available dataset (Arijs I., Van Lommel L., Van Steen K., De Hertogh G., Geboes K., Schuit F. and Rutgeerts P. 2009. Mucosal expression profiling in patients with inflammatory bowel disease before and after first infliximab treatment. Accession number GSE16879, Gene Expression Omnibus repository) was used to confirm mRNA observations of the Oxford cohort, using colonic mucosal biopses obtained at endoscopy from healthy controls or patients refractory to corticosteroids and/or immunosuppression. Gene expression data were generated using the Affymetrix Human Genome U133 Plus 2.0 array as described in.

29All mice were bred and maintained under specific pathogen-free conditions in accredited animal facilities. C57BL/6 Rag1-/-, 129SvEvS6.Rag2-/- and C57BL/6 Rag1-/-Il23rgfp/+ (from Daniel Cua, Merck Research Laboratories, Palo Alto, USA) strains used in this study were bred and maintained at the University of Oxford. C57BL/6 Rag1-/-Il15ra-/- and C57BL/6 Rag1-/- were bred and maintained at the National Institute for Medical Research. Experiments were conducted in accordance with local animal care committees (UK Scientific Procedures Act of 1986). Mice were routinely screened for the absence of pathogens, and were kept in individually ventilated cages with environmental enrichment.

30For cLP isolation, colon tissue was cut into 1cm pieces, and epithelial cells were removed by incubation (2x) in RPMI containing 5% heat-inactivated FCS and 5mM EDTA. Remaining tissue was incubated in RPMI containing 5% FCS, 15mM HEPES and 300U/ml of Collagenase VIII (Sigma-Aldrich, St Louis, MO) to digest the remaining tissue. Cell populations were purified by 75%:40%:30% Percoll (GE Healthcare, Little Chalfont, U.K.) gradient centrifugation (600 x g, 20 min). Lymphocytes were isolated from the 40%/30% interface.

31Peripheral blood was diluted in an equal volume of PBS and layered onto Ficoll-Hypaque (GE-Healthcare). Following centrifugation, PBMCs were collected from the Ficoll-plasma interface and washed 3x in PBS prior to culture or flow cytometry.

32Specimens were incubated in 1 mM DTT solution, followed by three washes in 0.75 mM EDTA solution. Tissues were then digested overnight in 0.1 mg/ml collagenase A solution (Roche Diagnostics, Basel, Switzerland). Cell populations were purified by 100%:60%:40%:30% Percoll (GE Healthcare) gradient centrifugation (900 x g, 30 min). Lymphocytes were isolated from the 60%/40% interface.

33For human biopsy samples, tissue was immediately placed in RNAlater solution (ThermoFisher Scientific, Waltham, MA) and stored at -20oC until processed. Human and murine cLP cells were lysed in RLT buffer (Qiagen, Hilden, Germany) containing ß-mercaptoethanol. RNA was isolated using the RNEasy kit (Qiagen) including a DNAse I digestion. Content and purity of RNA was measured using a Nanodrop spectrophotometer (Thermo Fisher Scientific). cDNA was synthesised using the Superscript III reverse transcription kit (Life Technologies, Paisley, U.K.) or the High Capacity cDNA kit (ThermoFisher Scientific, Waltham, MA). Quantitative real-time PCR for the candidate genes was performed using the Taqman system. cDNA samples were analysed in triplicate (technical replicates) using the CFX96 detection system (Bio-Rad Laboratories, Hercules, CA) or the Viia7 system (Applied Biosystems, Foster City, CA). Values were normalised to Hprt expression, and analysed using the d-Ct method. Taqman Gene Expression Assays (Life Technologies) for mouse Ccl19, Ccl20 Ccr7, Ccl2, Il23r and Csf2, and human RPLPO and CSF2 were used.

34To induce acute innate colitis, 25-–50 mg of αCD40 IgG2a mAb (clone FGK45, BioXCell, West Lebanon, NH) was administered via i.p. injection to mice on a C57BL/6 Rag-/- background. Unless otherwise indicated, mice were sacrificed 7 days post αCD40 injection, and weight loss was monitored throughout the course of the experiment.

35To induce colitis using Helicobacter hepaticus, Hh NCI-Frederick isolate 1A (strain 51449) was grown as previously described. Mice on a 129SvEv Rag2-/- background were fed 3x on consecutive days with Hh 1A (1.0x108 CFU) by oral gavage. Animals were analysed at week 6 following infection.

36To block specific cytokine or cellular activity in vivo, aCD40-treated mice received 0.375 mg IL-17A blocking mAb (UCB Celltech, Slough, U.K.), 0.15 mg IL-22 blocking mAb (Genentech, San Francisco, CA, clone 8E11), 0.8 mg IL-23R blocking mAb (D. Cua, Merck, Kenilworth, NJ), 0.125 mg GM-CSF blocking mAb (CSL Ltd, Parkville, Australia), 100 mg IL-17F blocking antibody (eBioscience, San Diego, CA) or isotype controls (GP120 10E7.1D2, Genentech or 22E9.11, CSL Ltd). Treatment started on day -1 and antibodies were injected i.p. at day -1 and day 3, except blocking mAb to IL-22 which was administered at days -1, 1, 3 and 5. Blocking mAb to GM-CSF during Helicobacter hepaticus infection was administered i.p. 2x per week starting at day 0.

37The following antibodies were purchased from eBioscience: fixable viability dye, GM-CSF-PE (MP1-22e9), CD127-FITC/PE/PECy5 (A7R34), Thy1.2-PECy7 (53.2.1), T-bet-eFluor660 (eBio4B10), RORγt-PE (AFKJS-9), B220-PerCP-Cy5.5 (RA3-6B2), CD11b-PerCP-Cy5.5 (M1/70), CD11c-PerCP-Cy5.5/APC (N418), Gr-1-PerCP-Cy5.5 (RB6-8C5), CD64-PE (X54-5/7.1), F4/80-PerCP-Cy5.5 (BM8), MHCII-AlexaFluor700 (M5/114.15.2), CCR7-PE (4B12), perforin-APC (eBioOMAK-D) and cKit-Alexa-Fluor700 (ACK2). The following antibodies were purchased from Biolegend (San Diego, CA): NKp46-FITC (29A1.4), CD49b-PerCP-Cy5.5 (DX5), Ly6C-PECy7 (HK1.4), Ly6G-BV570 (1A8), CCR6-BV421 (29-2L17), CD4-BV605 (RM4-5), CCR2-AF647 (SA203G11) and NK1.1-BV650 (PK136). The following antibodies were purchased from BD Horizon (Franklin Lakes, NJ): NKp46-V450 (29A1.4), CD11b-V500 (M1/70), CD45-V500/BV650 (30-F11), SiglecF-BV421 (E5-2440) and NKG2D-PE CF594 (CX5). Lineage staining was performed using PerCP-Cy5.5-conjugated antibodies against B220, CD11b, CD11c, Gr-1, F4/80 and CD49b.

38The following antibodies were purchased from eBioscience: CD1a-FITC (HIH9), CD3-BV650 (OKT3), CD8-FITC (RPA-T8), CD19-FITC (HIB19), IL-17A-PE (eBio64DEC17), IL-22-PECy7 (22URT1). The following antibodies were purchased from Biolegend: CD3-FITC (UCHT1), CD4-FITC (OKT4), CD11c-FITC (B-LY63.9), CD14-FITC (MSE2), CD16-FITC (3G8), CD34-FITC (581), CD123-FITC (6H6), TCRαβ-FITC (IP26), TCRγδ-FITC (B1), CD45-AlexaFluor700 (HI30), CD56-PerCP-Cy5.5/BV510 (HCD56), CD127-BV421 (A019D5), GM-CSF-PerCP-Cy5.5 (BVD2-21C11), IL-5-APC (TRFK5) and IFNγ-BV605 (4S.B3). Lineage staining was performed using FITC-conjugated antibodies against CD1a, CD8, CD19, CD3, CD4, CD14, CD16, CD34, CD123, TCRαβ and TCRγδ.

39Single cell suspensions were prepared from mouse spleen or colonic LP or human LPMC or PBMC cells as described above. For intracellular cytokine staining mouse or human were directly cultivated for 3 or 4 hr in the presence of phorbol 12-myristate 13-acetate (PMA, Sigma-Aldrich) (100ng/ml), ionomycin (Sigma-Aldrich) (1μg/ml), monensin (BD Biosciences) and brefeldin A (eBioscience). Where indicated, mouse and human cLP cells were stimulated overnight with 10 ng/ml of recombinant IL-12 (Peprotech, Rocky Hill, NJ) or IL-23 (mouse - R&D Systems (Minneapolis, MN), human - Peprotech), and then for the last 4 hr PMA, ionomycin and brefeldin A were added.

40Cells were incubated with CD16/32 (eBioscience) to block Fcγ receptors, prior to incubation with fixable viability dye (eFluor780, eBioscience), and mAb conjugates in 100 μl PBS containing 2% BSA and 1 mM EDTA (PBS-E) for 30 min. Cells were subsequently washed twice in PBS-E. For cytokine staining, cells were stained with antibodies for 1 hr using the Cytofix/Cytoperm kit (BD Biosciences). For transcription factor staining, cells were stained with transcription factor antibodies for 1 hr using the Foxp3 staining kit (eBioscience). Cells were washed twice in PBS-E. All labelled cells were then acquired on a LSRII SORP (BD Biosciences) and analysed using Flowjo software (Tree star, Ashland, OR).

41To assess the severity of colitis, samples of the proximal colon were taken and immediately fixed in buffered 10% formalin (3.6% w/v formaldehyde). 4–5 μm paraffin-embedded sections were cut and stained with haematoxylin and eosin (H&E). Inflammation was assessed using previously published criteria. Each sample was graded semiquantitatively from 0 to 4 in each of the 4 following features: epithelial hyperplasia and goblet cell depletion, cellular infiltration of the lamina propria, percentage of the section affected, and markers of severe inflammation (submucosal inflammation, crypt abscesses). Three separate colon sections for each sample were examined. Scores for each criterion were added to give an overall score for each sample of 0–12. Samples were scored in a blinded fashion by two individuals.

42Gut imaging was performed as described in McDole et al.. Briefly, mice were anesthetized with ketamine/xylazine, the gut was exposed and immobilized with an imaging window, and imaging proceeded using isoflurane to maintain anesthesia. Images were collected using a 20x water dipping lens and the spectral detector of a Zeiss 780 upright microscope (Carl Zeiss, Oberkochen, Germany). Images were linearly unmixed using the Zen software (Carl Zeiss) to separate autofluorescence, collagen, EGFP, and Texas red-dextran based on single color controls.

43Mice were anesthetized for imaging and injected intravenously with 50 mg of 70 kDa texas-red labelled dextran (Life Technologies). Imaging proceeded immediately as described above.

44Intravital microscopy was analyzed using Imaris 8 (Bitplane, Belfast, U.K.) after linear unmixing using Zen (Carl Zeiss). Images were drift corrected based on mucus or collagen signal. ILCs were tracked using autoregressive motion algorithm of the spot-tracking module. Motile ILCs were gated based on distance less than 75 μm from a cryptopatch, track duration greater than 5 min, and displacement greater than 14 μm (approximately two cell lengths). Images were smoothed using a Gaussian filter for display.

45For immunofluorescence, samples of the proximal colon were taken and immediately embedded in OCT compound (Tissue-Tek, Sakura, Alphen aan den Rijn, The Netherlands) and frozen in a bath of isopentane on dry ice prior to storage at -80oC. 6 μm frozen sections were cut and collected on frosted glass slides. For staining, slides were fixed in 2% formalin. For RORγt staining, slides were subsequently also treated with Fix-Perm buffer (eBioscience). Endogenous peroxidase activity was blocked with 1% H2O2 (Sigma-Aldrich) and 2% sodium azide and non-specific binding was blocked with 10% donkey serum. Sections were incubated with the following antibodies CD4 (RM4-5, BD Biosciences) IL-7Rα (A7R34, eBioscience) in 10% donkey serum or RORγt (AFKJS-9, eBioscience) in Perm buffer (eBioscience). Sections were then incubated with donkey anti-rat HRP secondary antibody (Jackson ImmunoResearch Laboratories, West Grove, PA). Tyramide signal amplification was the performed (Perkin Elmer, Waltham, MA). Then sections were stained with CD11c (N418, eBioscience) in 10% goat serum followed by goat anti-hamster Cy3, or with MHCII-FITC or AF647 (both M5/114.15.2, Biolegend).

46Sections were mounted with Vectashield containing DAPI. Images were collected using a 710 microscope (Carl Zeiss), and analysed using ImageJ open access software.

47The colons from C57BL/6 Rag1-/-Il23rgfp/+ mice were freshly isolated and maintained in room temperature PBS before imaging. Opened sections of colon were immobilized using Vetbond tissue adhesive (3M, Saint Paul, MN) in dishes containing PBS. CPs were imaged from the muscularis side of the colon using a 20x water dipping objective and spectral detector on a Zeiss 780 upright multiphoton microscope. Images were analysed using the linear unmixing module of Zen (Carl Zeiss) and Imaris (Bitplane).

48Statistical analysis and graphical representations were performed using Prism 5.0a software (GraphPad, La Jolla, CA). For comparison of qPCR results and histopathological analysis, the nonparametric Mann-Whitney U test was used, and for multiple samples, one-way ANOVA with Bonferroni’s post-test was used unless otherwise stated. Technical replicates are from the same sample, analysed on the same day. Biological replicates are from independent experiments.

49Sample size was computed using an estimate of colitis score between 6–8, to detect a score reduction by 50% in a treated group with a standard deviation of 30%, false positive of 0.05, and power of 0.80. Sample size was therefore calculated to be between 6–7. Where smaller changes were expected, sample sizes were increased accordingly.