Granulocyte Macrophage Colony-Stimulating Factor-Activated Eosinophils Promote Interleukin-23 Driven Chronic Colitis

1Chronic intestinal inflammation is characterized by dysregulated T helper 1 (Th1) and Th17 cell and innate lymphoid cell responses with excessive production of inflammatory cytokines, leading to increased production of granulocyte-monocyte progenitors (GMPs) and accumulation of inflammatory myeloid cells in the target tissue. Previously we described an interleukin-23 (IL-23)-granulocyte macrophage colony-stimulating factor (GM-CSF) axis as a key driver of dysregulated hematopoiesis in colitis; however, the relative contribution of distinct innate effector cells downstream of this pathway remains unknown. Neutrophils are considered a major culprit in IL-23-Th17-cell-type-mediated tissue damage, while the pathogenic role of eosinophils has primarily been established for Th2 cell-mediated conditions such allergic skin and lung disease.

2Eosinophils, which arise from GMPs through an eosinophil progenitor (EoP) intermediate, are rare in the blood but more abundant in tissues such as the gastrointestinal tract, although their contribution to intestinal homeostasis remains enigmatic. Beyond their role in Th2 cell immunity, eosinophil secrete various inflammatory mediators (e.g., TNF, IL-13, CXCL1) and have been implicated in activation of dendritic cells (DCs) and neutrophils. They can also release anti-microbial compounds toxic for viruses and bacteria and promote the survival of immunoglobulin A (IgA)-secreting plasma cells in the intestine, suggesting a possible anti-microbial function.

3A dysregulated eosinophil response can cause immune pathology, and this is most evident in atopic diseases such as asthma and eczema, Th2 cell-mediated eosinophilic esophagitis, and hypereosinophilic syndrome. However, the molecular signals that drive eosinophils from protective to tissue damaging cells are ill-defined and require further characterization. Similar to neutrophils, eosinophils produce a range of cytotoxic mediators; matrix metalloproteinases and reactive oxygen species, as well as specific proteins such as eosinophil peroxidase (EPO) and eosinophil cationic protein (ECP). These molecules are toxic for invading microorganisms but can also lead to collateral damage to host tissues including the intestinal epithelium. Indeed, intestinal eosinophil accumulation has been implicated in the pathogenesis of a chemically induced model of acute colonic injury and increased eosinophil numbers and activation has been reported in inflammatory bowel disease (IBD). However, despite their abundance in the intestine, the regulation of eosinophils by colitogenic cytokines and their functional role in chronic intestinal inflammation is not known.

4Our previous work identified IL-23-driven GM-CSF as a key mediator of chronic inflammation in T cell transfer colitis. GM-CSF promoted intestinal inflammation at several levels, including skewing of hematopoiesis toward granulo-monocytopoiesis and accumulation of highly proliferative GMPs in the intestine. Using experimental models of chronic colitis, we now show that GM-CSF promoted IL-23-driven intestinal inflammation through local accumulation of activated eosinophils and potentiation of their effector functions. In addition, it also promoted bone marrow (BM) eosinopoiesis in synergy with IL-5. Because IL-23 is a well-known driver of the Th17 cell response, these results provide evidence of a link between the Th17-cell-type response and eosinophils in intestinal inflammation and suggest that targeting the GM-CSF-eosinophil axis might have therapeutic utility in some forms of IBD.

5To investigate the relative contribution of granulocyte subsets to chronic intestinal inflammation, we used a well-characterized T cell transfer model of IL-23 driven colitis. In this model, chronic colitis develops ∼6 weeks after transfer of T cells into Rag1−/− mice and is accompanied by increased granulopoiesis. Standard markers were used to discriminate eosinophils (Siglec-F+Gr1int) from Siglec-F−Gr1hi neutrophils. We also used the level of expression of Siglec-F as a measure of eosinophil activation. Thus, Siglec-Fint cells are immature or resting eosinophils that reside in lymphoid organs and in uninflamed tissues, whereas Siglec-Fhi cells are mature or activated eosinophils mostly found in extralymphoid tissues and increased during inflammation. In chronic colitis, we found both populations among colonic lamina propria leukocytes (cLPL), with high granularity (SSChi) and expression of the eotaxin receptor CCR3 confirming them as eosinophils (Figures 1A and S1A). Percentages of Siglec-Fhi eosinophils were as high as those of neutrophils in the inflamed intestine, both of which were ∼2-fold increased in colitic compared to control Rag1−/− mice. This increase was equivalent to a ∼40-fold increase in absolute numbers (Figures 1A and 1B). The abundance of intestinal eosinophils was confirmed in situ, with a high density of Siglec-F+ cells observed in inflamed colons (Figure S1E). IL-23-deficient Rag1−/− mice, which only develop mild colitis after T cell transfer, had a reduced absolute number and percentage of eosinophils among CD45+ leukocytes compared to colitic IL-23 competent mice, suggesting a link between intestinal eosinophil accumulation and IL-23-driven inflammation (Figure S1B).

6Eosinophils in the inflamed intestine showed increased activation based on a number of parameters. First the majority express high amounts of Siglec-F (Figure 1C), a phenotype associated with activation in inflammatory lung disease. Consistent with their increased activation state, Siglec-Fhi eosinophils in the inflamed intestine also expressed higher amounts of other eosinophil activation markers such as CD11b, IL-33R, and Gr1 (Figure 1C). These activated eosinophils in colitis also expressed the degranulation marker CD63. Such cells were 1.7-fold higher in the colon of colitic mice compared to controls with a 60-fold increase in total numbers (Figure 1D). Finally, while resting Siglec-Fint eosinophils in the BM and spleen secreted negligible amounts of TNF, 14% of eosinophils in the inflamed colon were positive for TNF (Figure S1D), a crucial colitogenic cytokine.

7Interestingly, protection from colitis by co-transfer of Foxp3+ regulatory T (Treg) cells with colitogenic CD4+ T cells reduced not only cLPL neutrophil numbers but also eosinophil accumulation and activation, indicating similar inhibitory effects of Treg cells on neutrophils and eosinophils in this model (Figure S1C).

8Taken together, these data show that eosinophils are a major constituent of the IL-23-driven intestinal inflammatory network (∼4 to 12% of total cLPL in colitis, Figure 1B). Although a predominant population of eosinophils in the normal intestinal mucosa had a resting Siglec-Fint phenotype, a shift toward an activated Siglec-Fhi population with signs of degranulation occurred during chronic intestinal inflammation.

9As we observed sustained accumulation of eosinophils in chronic intestinal inflammation, we next sought to investigate the role of eosinopoiesis in this process. Eosinophils can increase their lifespan from a few days to a few weeks within inflamed tissues, therefore we could not exclude that the abundant stock of preformed eosinophils in the BM (Figure 2B) fuelled tissue accumulation without major changes in eosinopoiesis. We first examined this in T cell transfer colitis and found that a striking 3.5-fold expansion of EoPs correlated with a substantial increase of eosinophils in the BM of colitic mice compared to controls (Figures 2A and 2B). There was also a 2-fold increase in the percentage of BM eosinophils positive for the gut-homing α4β7 integrin (Figure 2C). Consistent with these results, ∼7% of CD4+ T cells in the inflamed intestine expressed the eosinopoietin IL-5 (Figure S2A). GM-CSF, which can act in synergy with IL-5 to stimulate eosinopoiesis, was also increased in the inflamed colon compared to controls (Figure S2B) and ∼40% of CD4+ T cells were GM-CSF+ in colitis (Figure S2A).

10Next we investigated eosinopoiesis in a lymphocyte replete model of colitis to ensure our results were not a consequence of altered myelopoiesis in Rag1−/− hosts. For this we used a well-described model of colitis following Helicobacter hepaticus (Hh) infection and concomitant blockade of the IL-10-IL-10R pathway. In this model, there was a similar increase in EoPs and BM eosinophils, as well as accumulation of activated TNF-secreting eosinophils in the colon (Figures 2D, 2E, and S2D).

11Mature granulocytes in peripheral tissue are described as post-mitotic and indeed eosinophils that accumulated in the inflamed colon stained negative for BrdU after a 16 hr pulse-chase assay, whereas almost half of eosinophils developing in the BM had incorporated the dye at this time (Figure 2F). In contrast, BrdU+ eosinophils appeared in the intestine only 2–3 days after initial BrdU pulsing (Figure 2F), suggesting that the increase in colonic eosinophils is supported by sustained BM eosinopoiesis.

12Thus, the large accumulation of eosinophils in the inflamed intestine was supported by a significant increase in eosinopoiesis, giving rise in the BM to newly formed eosinophils that were preferentially tagged for intestinal migration.

13GM-CSF can directly stimulate eosinopoiesis and eosinophil survival. To test whether the GM-CSF-eosinophil pathway is pathological in colitis in a lymphocyte-replete setting, we turned to the Hh and anti-IL-10R colitis model described above. Lack of a GM-CSF-Rβ signal in Csf2rb−/− mice reduced EoP and eosinophil increases in the BM in this chronic model of colitis (Figure 3A). A change in eosinopoiesis was accompanied by a ∼90% decrease in percentages of intestinal eosinophils in Csf2rb−/− versus WT infected mice, whereas percentages of neutrophils and CD4+ T cells were similar (Figures 3C and S2C). This alteration in the composition of the cellular infiltrate correlated with a significantly reduced colitis score in Csf2rb−/− compared to WT mice (Figure 3B) and a decrease in the ratio of activated to resting eosinophils in the intestine (Figure 3C). It was notable that the decrease in intestinal eosinophils in Csf2rb−/− mice occurred at steady state, whereas the lack of IL-23R signaling had no effect on the accumulation of eosinophils in the normal intestine (Figure S3A). Next we utilized mixed WT and Csf2rb−/− BM chimeras to distinguish cell-intrinsic from non-cell-autonomous secondary effects. While cell-intrinsic GM-CSF-Rβ signaling was not required for accumulation of neutrophils and monocytes in the inflamed intestine, cell autonomous GM-CSF-Rβ signaling was required for eosinophil accumulation in colitis (Figure 3D).

14Together, these data indicate that GM-CSF-Rβ chain signaling promotes eosinophilia and colitis and differentially regulates the accumulation of neutrophils and eosinophils in the inflamed intestine.

15Because the lack of colitis observed in Csf2rb−/− mice correlated with a decrease in the frequency of eosinophils, but not neutrophils, we next investigated the relative contribution of these distinct innate effectors to the pathogenesis of colitis. We employed two strategies to deplete eosinophils, either blockade of the eosinopoietin IL-5 or antibody-mediated depletion of Siglec-F+ cells. In the presence of IL-10R blockade, Hh infected mice treated with anti-IL-5 had a 50% reduction in total cLPL compared to isotype treated controls (data not shown) and an 87% decrease in the percentage of Siglec-Fhi eosinophils (Figure 4A). Most importantly, this was accompanied by a significant reduction in colitis severity compared to isotype treated controls (Figure 4B).

16The IL-5R is constitutively expressed by eosinophils but also by some B cell subsets. Therefore, to further increase the specificity of treatment, we used an anti-Siglec-F depletion approach shown to selectively eliminate eosinophils. Although alveolar macrophages in the lung express Siglec-F, intestinal and peritoneal macrophages do not and are therefore not affected by anti-Siglec-F depletion. Treatment with anti-mouse Siglec-F immune serum reduced colitis severity to the same extent as IL-5 blockade (Figure 4D). This treatment regimen led to an 85% decrease of eosinophils based on reduced CD11b+CCR3+SSChi cells in the colon of anti-Siglec-F versus pre-immune serum treated mice (Figure 4C), whereas there was only a 28% reduction in the percentage of colonic neutrophils (Figure 4E). The small reduction in neutrophils was most likely secondary to reduced overall inflammation, because uninfected mice treated with anti-Siglec-F serum did not display a decrease in neutrophils or any leukocyte populations other than eosinophils (Figure S4A). By contrast with eosinophil-depleting strategies, depletion of neutrophils with an anti-Ly-6G antibody did not have a significant effect on colitis (Figures 4F and S4B).

17Together these results reveal differential roles for eosinophils and neutrophils in chronic colitis. While eosinophils play a non-redundant role in disease, neutrophils are dispensable for the development of chronic intestinal inflammation.

18In order to further understand the colitogenic role of GM-CSF, we investigated whether GM-CSF and IL-5 had differential effects on eosinophil production and activation. Because GM-CSF-R and IL-5R share the same β-receptor subunit, we tested whether GM-CSF blockade would reproduce the decrease in eosinophil activation and accumulation observed in Csf2rb−/− mice. Interestingly, IL-5 and GM-CSF were produced at steady state by ILCs but were not increased in lymphocyte-replete colitis (Figure S5B). By contrast, GM-CSF production by CD4+ T cells was increased in the inflamed colon compared to controls, while percentages of IL-5 producers were unchanged (Figure S5A). Accordingly, colonic GM-CSF, but not IL-5, mRNA and protein levels were augmented in chronic colitis, possibly highlighting a more homeostatic role for IL-5 compared with the more activation-induced functions of GM-CSF (Figure 5A and 5B). Regarding eosinophil chemoattractants, eotaxin-1 and RANTES were increased in early and late phases of colitis, respectively (Figure 5A).

19When treated with anti-GM-CSF, Hh infected and anti-IL-10R-treated mice exhibited significantly reduced colonic infiltrates and colitis score compared to mice treated with isotype control (Figure 5C) and displayed a striking 50% decrease in the frequency and activation status of colonic eosinophils (Figure 5D). While anti-IL-5 treatment inhibited the general accumulation of colonic eosinophils (Siglec-Fint and Siglec-Fhi) (Figures 4A and S5D), GM-CSF blockade only decreased the most activated population (Figures 5D and S5D) suggesting a role for GM-CSF in intestinal eosinophil activation in the inflamed colon.

20Blockade of either GM-CSF or IL-5 led to reductions in the number of eosinophils in the BM; however, only GM-CSF blockade inhibited the accumulation of GMPs and downstream EoPs (Figure 5E). These results indicate that during intestinal inflammation, GM-CSF sustains eosinophilic granulopoiesis, whereas IL-5 mediates a more specific function promoting the terminal differentiation of EoPs into Siglec-F+ cells (Figure 5E). A differential effect of GM-CSF and IL-5 was also evident on the “gut-tagging” of newly produced BM eosinophils, as upregulation of α4β7 integrin in colitis was only inhibited by IL-5 blockade (Figure 5F).

21Overall, these results highlight the synergy between GM-CSF and IL-5 in the regulation of eosinopoiesis and reveal the key role of GM-CSF in driving chronic intestinal inflammation through accumulation of activated eosinophils in the colon.

22We next sought to characterize further the differential regulation of eosinophils in the periphery by GM-CSF and IL-5. Both cytokines can promote eosinophil survival, and we confirmed this observation in colitis. Annexin-V staining on freshly isolated peripheral eosinophils was decreased in colitis and increased in the presence of GM-CSF or IL-5 blockade (Figure 6A). As GM-CSF production was increased during colitis, while IL-5 levels stayed constant (Figure 5B), we hypothesized that GM-CSF would be a key driver of eosinophil effector functions in the inflamed intestine. Indeed, anti-GM-CSF treatment inhibited the increase in CD11b and increase in side scatter (correlating with granularity), while IL-5 blockade did not have a significant effect on these markers of activation (Figure 6B). In addition, cell-sorted eosinophils exhibited morphological changes in vitro in the presence of GM-CSF, notably increased diameter as a sign of activation (Figure 6F). Interestingly CD64 (FcγRI), which is increased on neutrophils in IBD, was induced on eosinophils during colitis in a GM-CSF-dependent but IL-5-independent manner (Figure 6C). Furthermore, the amount of CD64 was higher on Siglec-Fhi than Siglec-Fint eosinophils (Figure 6C), consistent with their more activated status.

23Regarding expression of cytokines involved in epithelial cell dysregulation and damage, intestinal eosinophils expressed higher amounts of Tnf, Il6, and Il13 mRNA in colitis compared to uninflamed controls (Figure 6D). In vitro analysis of cell-sorted intestinal eosinophils showed that GM-CSF stimulated Tnf and Il13 mRNA expression, but had no effect on Il6 (Figure 6E). Altogether, these data demonstrate that GM-CSF and IL-5 promote the survival of peripheral eosinophils, but only GM-CSF promotes their activation and inflammatory cytokine production, revealing one of the key colitogenic effects of GM-CSF during chronic intestinal inflammation.

24Because TNF, IL-6, and IL-13 are expressed by various leukocytes, we decided to investigate whether eosinophil-specific products could also drive chronic intestinal inflammation. For this purpose, we tested whether EPO, which is produced exclusively by eosinophils and can be tissue-toxic, contributed to chronic colitis. EPO levels and activity in the intestine were greatly increased during chronic inflammation, confirming substantial eosinophil degranulation. In addition, a reduction in eosinophil numbers during anti-GM-CSF or anti-IL-5 treatment was accompanied by a significant decrease in EPO (Figure 7A). Resorcinol is a potent inhibitor of EPO leading to decreased anti-bacterial activity of eosinophils. In Hh-induced colitis, daily treatment of WT mice with resorcinol led to significantly reduced EPO activity and decreased colitis (Figures 7B and 7C). This was accompanied by decreased markers of colonic inflammation compared to PBS treated mice, including reduced leukocyte infiltration, lower neutrophil percentages, and a trend toward reduced IFN-γ+ CD4+ T cells (Figures 7B and 7C). EPO inhibition, however, did not affect the frequency of eosinophils among cLPL (Figure 7C), consistent with previous in vivo observations.

25Overall, the pathogenic effect of uncontrolled accumulation of activated eosinophils in chronic colitis could be attenuated by inhibition of EPO, an enzyme well known to mediate oxidative tissue damage in eosinophil-dependent inflammatory diseases.

26Our study newly identifies a GM-CSF-eosinophil axis as a crucial component of IL-23-driven chronic colitis. Our previous work described GM-CSF as a pivotal downstream effector of IL-23 in the inflammatory cascade that drives aberrant responses to commensal microbiota through increases in myelopoiesis in T cell transfer colitis. Neutrophils are widely accepted as tissue-toxic cells in IL-23-mediated colitis. However, our results challenge this view and indicate a more prominent and unexpected role for eosinophils in this response. We show a marked accumulation of activated eosinophils in the colon of colitic mice, supported by increased eosinopoiesis, and a direct colitogenic role through production of eosinophil peroxidase and inflammatory cytokines. Although eosinophils are abundant in the intestine, their role in chronic intestinal inflammation is rarely considered. Here we identified GM-CSF as a key molecular switch diverting eosinophils from a tissue-protective to a tissue-toxic state of activation. These results extend the paradigm of eosinophil-mediated immune pathology beyond Th2 cell-type responses to effectors of IL-23-GM-CSF-driven dysregulated tissue immunity.

27GM-CSF is emerging as a central cytokine at the crossroads of various types of effector T cell responses and can be produced by Th1 and Th2 cells to stimulate increased myeloid cell activity. More recently it was shown that IL-23 stimulated Th17 cells to produce GM-CSF, which was pathogenic in EAE although its functional role was not established. In the inflamed intestine, IL-23 stimulated polyfunctional IFNγ+IL-17A+Th cells to produce GM-CSF, which triggered extramedullary hematopoiesis. In this report, we show that GM-CSF increased eosinopoiesis and numbers of highly activated eosinophils in the inflamed intestine. GM-CSF promoted increases in GMPs and downstream EoPs, which both express GM-CSF-R-α and β chains. Increased EoPs have been observed in Th2 cell-mediated asthma and anti-helminth responses, however, our study constitutes the first report of chronic EoP accumulation during IL-23-Th17 cell type-mediated immune disease, extending our previous observation of dysregulated hematopoiesis in colitis to the eosinophilic lineage.

28IL-5 or GM-CSF blockade resulted in a substantial decrease of eosinophils in the inflamed intestine, however there were marked differences in their action. Although both cytokines promoted eosinopoiesis during colitis, IL-5 specifically increased the differentiation of EoPs into Siglec-F+ eosinophils and promoted imprinting of α4β7 integrin expression. However by contrast with GM-CSF it had no effect on the upstream GMP.

29Resident populations of intestinal leukocytes contribute to the maintenance of basal eosinophil numbers as ILCs constitutively produce IL-5 in the normal intestine and resident macrophages express eotaxins. However, intestinal IL-5 does not appear to be controlled by IL-23 because IL-5 expression was not increased in colitis or reduced in IL-23 deficiency. These results are in contrast to the findings that IL-23 promoted IL-5 and Th2 responses in asthma models suggesting differences in IL-5 regulation in distinct tissue sites. In contrast with IL-5, IL-23 increased GM-CSF expression by CD4+ T cells and ILC in colitis ( and data not shown), pinpointing an IL-23-GM-CSF-eosinophil axis in colitis that can boost basal IL-5 dependent eosinophilia. A recent study showed a role for eosinophils in maintaining intestinal integrity toward the gut microbiota through stimulating IgA+ plasma cells and Foxp3+ Treg cells. Importantly, here we show that the unchecked production of GM-CSF during chronic colitis is a key driver of the eosinophil switch from a resident and homeostatic phenotype (Siglec-Fint) to an over-activated and tissue toxic phenotype (Siglec-Fhi). Siglec-Fhi eosinophils were also increased in lung inflammation and were more resistant to apoptosis than Siglec-Fint eosinophils. Eosinophil activation and cytokine secretion that accompanied colitis was inhibited by GM-CSF but not IL-5 blockade. Furthermore in vitro, GM-CSF acted directly on eosinophils to induce production of colitogenic cytokines TNF and IL-13. Together the data suggest that IL-5 plays a homeostatic role maintaining basal levels of eosinophils in the intestine, whereas GM-CSF promotes their activation and deleterious effector functions in chronic colitis.

30It is worth noting that GM-CSF-Rβ deficiency did not affect the percentage of neutrophils in the intestine in colitis and BM chimera experiments, despite inducing a severe decrease in eosinophil percentages. Thus, GM-CSF is not absolutely required for the neutrophil increase probably owing to the compensatory role of G-CSF, which is a potent inducer of neutrophilia and is increased in T cell transfer colitis. Unexpectedly, while eosinophil depletion dampened colitis, no such effect was provoked by depletion of neutrophils, highlighting a dichotomy in the role of these granulocyte populations in chronic colitis.

31Amelioration of chronic colitis by pharmacological inhibition of EPO, which is implicated in cytotoxic oxidant generation, pinpointed one of the molecular mechanisms by which eosinophils specifically mediate intestinal damage. This pathway has also been implicated in the DSS model of acute colonic injury, suggesting broad relevance in intestinal damage. Interestingly, regulation of eosinophils in acute versus chronic intestinal inflammation is not identical as we found that IL-5 depletion inhibited chronic colitis, whereas IL-5 deficiency had no significant effect in the DSS model contrary to the protective effect of eotaxin deficiency. This suggests that in an acute damage model, mobilization of mature eosinophils from the BM to intestine is sufficient, whereas sustained chronic colitis requires IL-5-dependent eosinopoiesis.

32In human, treatment of eosinophils with GM-CSF in vitro led to increased release of EPO and ECP, providing further evidence that GM-CSF can directly increase the cytotoxic functions of eosinophils. Conversely, IL-10 inhibited LPS-induced TNF release and increased survival of human eosinophils in vitro. Treg cells play an important role in intestinal homeostasis and suppress colitis in part via IL-10. We found that Treg cell-mediated control of colitis correlated with a reduction in eosinophil accumulation and activation. Based on those results, it is tempting to speculate that under homeostatic conditions eosinophils in the intestine are hyporesponsive to TLR activation as a consequence of the IL-10 rich environment, which might be over-ridden by sustained increases in GM-CSF production in chronic inflammation.

33There are several reports of increased GM-CSF in Crohn’s disease (CD) and ulcerative colitis (UC) patients, and concomitant Th17 and IL-5 and IL-13 T cell responses have been observed in ileal CD suggesting a more polyfunctional T cell response in certain patients. However GM-CSF activity is a double-edged sword and when produced in a controlled fashion plays an important role in the steady state accumulation of mononuclear phagocytes and Foxp3+ Treg cells and in promoting host protective immunity in the gut. Consistent with this, intestinal injury in a DSS model was exacerbated in Csf2rb−/− mice further illustrating differences in the mechanisms of acute and chronic intestinal damage. Increased levels of anti-GM-CSF autoantibodies have been observed in pediatric and some forms of adult CD leading to the idea that GM-CSF is protective in IBD. However three clinical trials of recombinant GM-CSF administration failed to show demonstrable protective effects. It is highly likely that GM-CSF will play both protective and pathological roles in IBD and that the context in which it is produced, such as where and for how long, might determine its ultimate functional role.

34Several studies have reported increased eosinophil numbers and activation in both UC and CD. ECP was also increased in the faeces of IBD patients suggesting eosinophil degranulation. Our results in model systems taken together with an emerging picture in humans suggest that blockade of the GM-CSF/eosinophil axis might be a therapeutic target in particular patient subsets. The fact that sustained depletion of eosinophils in patients with hyper-eosinophilic syndrome treated for up to 6 years with anti-IL-5 did not lead to adverse effects is encouraging for considering this approach in IBD.

35B6.SJL-Cd45.1 mice or C57BL/6 mice: wild-type (WT), Csf2rb−/−, Il23r−/−, Rag1−/−, or Rag1−/−Il23p19−/− were bred and maintained under specific pathogen–free conditions in accredited animal facilities at the University of Oxford. All procedures involving animals were conducted according to the requirements and with the approval of the UK Home Office Animals (Scientific Procedures) Acts, 1986. Mice were negative for Helicobacter spp. and other known intestinal pathogens and were used when 7–12 weeks old.

36Naive CD4+CD25−CD45RBhi T cells and regulatory CD4+CD25+CD45RBlo T cells were sorted by flow cytometry from enriched CD4+ single-cell spleen suspensions (Dynal) to a purity of >99%. For the induction of colitis, 4 × 105 naive T cells were injected intraperitoneally (i.p.) into C57BL/6.Rag1−/− recipients. Where indicated, 2 × 105 protective T reg cells were co-injected i.p.

37Colitis was induced in WT C57BL/6 mice by infecting with Hh (oral gavage) on 2 consecutive days with 5 × 107–2 × 108 CFU Hh and i.p. injection of 1 mg 1B1.2 (anti-IL10R) mAb on days 0 and 7 after Hh infection. Mice were killed 7 days after the last anti-IL10R mAb treatment (weeks 2–3). Where indicated, mice were i.p. injected two times per week with 0.4 mg of anti-GM-CSF (MP1-22E9; CSL Ltd) or isotype control (GL117, rat IgG2a) or 0.5 mg of anti-IL-5 (TRFK5; BioXCell), or three times per week with 0.25 mg anti-Ly6G (1A8; BioXCell) or isotype control (2A3, rat IgG2a) starting from the first day of Hh infection. Where indicated, mice were i.p. injected two times per week with sheep preimmune serum or sheep anti-Siglec-F serum. Where indicated, mice were i.p. injected daily with Resorcinol (1.25mg/kg) or PBS.

38Proximal, mid-, and distal colon samples were fixed in buffered 10% formalin solution. 5 μm paraffin embedded sections were cut and stained with hematoxylin and eosin and inflammation was scored in a blinded fashion. In brief, 4 parameters of inflammation were assessed (scored 0–3): epithelial hyperplasia and goblet cell depletion, leukocyte infiltration in the lamina propria, area of tissue affected, and markers of severe inflammation such as submucosal inflammation. Aggregate scores were taken for each section, to give a total inflammation of 0–12. Colon inflammation scores represent the average score of the three sections taken.

39Single cell suspensions were prepared from spleen, MLN, and cLPL as previously described. In brief, colons were longitudinally opened, cut into 1 cm pieces, and incubated in RPMI 1640 with 10% FCS and 5 mM EDTA at 37°C to remove epithelial cells. Tissue was then digested with 100 U/ml of type VIII collagenase (Sigma) in complete RPMI medium containing 15cmM HEPES during 1 hr at 37°C. The isolated cells were layered on a 30/40/75% Percoll gradient, which was centrifuged for 20 min at 600 g, and the 40/75% interface, containing mostly leukocytes, was recovered. BM cell suspensions were prepared by flushing the marrow out of femur and tibia and were resuspended in PBS with 2% BSA.

40Flow cytometry and cell sorting, quantitation of gene expression using real-time PCR, in vitro stimulation assays, EPO Elisa and EPO colorimetric assay, in vivo BrdU labeling and immunofluorescence were performed as described in Supplemental Experimental Procedures.

41Statistical analysis was performed with Prism 6.0 (GraphPad Software). The nonparametric Mann-Whitney test was used for all statistical comparisons. Differences were considered statistically significant when p < 0.05.

42T.G. and I.C.A planned and performed experiments and wrote the paper. C.P., T.K., C.S., F.F., and J.S. performed particular experiments. F.P. wrote the paper and supervised the study. F.P. and T.G. designed the study. B.S.M. and P.R.C. provided essential materials and were involved in data discussions.