IL-1β, IL-23, and TGF-β drive plasticity of human ILC2s towards IL-17-producing ILCs in nasal inflammation

Innate lymphoid cells (ILCs) are crucial for the immune surveillance at mucosal sites. ILCs coordinate early eradication of pathogens and contribute to tissue healing and remodeling, features that are dysfunctional in patients with cystic fibrosis (CF). The mechanisms by which ILCs contribute to CF-immunopathology are ill-defined. Here, we show that group 2 ILCs (ILC2s) transdifferentiated into IL-17-secreting cells in the presence of the epithelial-derived cytokines IL-1β, IL-23 and TGF-β. This conversion is abrogated by IL-4 or vitamin D3. IL-17 producing ILC2s induce IL-8 secretion by epithelial cells and their presence in nasal polyps of CF patients is associated with neutrophilia. Our data suggest that ILC2s undergo transdifferentiation in CF nasal polyps in response to local cytokines, which are induced by infectious agents.

C ystic fibrosis (CF) is a life-limiting multisystem genetic disorder caused by mutations in the CF transmembrane conductance regulator (CFTR) gene 1 , a cAMP-regulated epithelial chloride channel 2 . The major consequences of CFTR mutations are abnormally viscous secretions in the upper and lower airways, impaired mucociliary clearance, persistent infection, and chronic inflammation 3 . Consequently, the respiratory tract of most patients with CF becomes infected with opportunistic bacteria, most importantly Staphylococcus aureus (SA) and Pseudomonas aeruginosa (PA) 4 . This eventually results in lung damage and loss of pulmonary function, the major cause of morbidity and mortality in CF disease 5 . Notably, chronic rhinosinusitis with nasal polyps (CRSwNP) is frequent in CF. Up to 40% of CF patients will manifest NP in their lifetime 6 , whereas CRSwNP prevalence in the general population is estimated at 2-5% 7 .
Despite massive neutrophil infiltration and excessive production of pro-inflammatory cytokines, CF patients display persistent and difficult to treat bacterial infections of the upper and lower airways 8 . Recent reports provide evidence that targeting specific inflammatory pathways reduce the exuberant inflammatory response that compromises the host's ability to control infection 9 . IL-1β, which is involved in lung inflammation, promotes PAinduced early neutrophil recruitment in the lung via a synergistic interaction with IL-23 10 . Thus, IL-1β, which is present in sputum of CF patients in high amounts, could be a potent target to ameliorate the pathogenic consequences of the microbial colonization in CF 11 . Furthermore, IL-1β triggers the activation of pathogenic IL-17A-secreting T cells, highly enriched in CF patients 12,13 . Anti-IL-17 therapy significantly suppresses neutrophil recruitment and inflammation in CF mouse models after PA infection 14 . Besides Th17 cells, the type 3 subset of innate lymphoid cells (ILC3) is a major producer of IL-17 15 . This raises the question whether these strategically located cells contribute to the cytokine milieu in CF.
ILCs are involved in early stages of the immune host defense mounted against pathogens and mediate the generation of lymphoid organs and tissue remodeling 16 . ILCs subsets have been identified based on their cytokine profiles and by the expression of key transcription factors that regulate their development and function 17 . ILC1s produce IFN-γ and need the Tbox transcription factor family member T-bet; ILC2s require the GATA-3 transcription factor and produce IL-4, IL-5, and IL-13, and ILC3s produce IL-17, IL-22, GM-SCF, and TNF-α and express RORγt. Strategically located at the body's interface with the environment, ILCs promptly respond to pathogenic threats by secreting effector cytokines and igniting adaptive immune responses in synchrony with myeloid cells 18 . ILCs are activated by cytokines and chemokines secreted by triggered epithelial cells or other cells of the immune system 19 . Several studies in humans and mice have now established that ILCs are highly plastic, adapting their cytokine secretion profiles in response to changes in the microenvironment. Thus, ILC3s and ILC2s can transdifferentiate into IFN-γproducing ILC1s in the presence of IL-12 and ILC1s can shift into ILC3s in response to IL-23, IL-1β, and retinoic acid [20][21][22] . These transdifferentiation processes result in increased IFN-γ producing ILC1-like cells as observed in inflamed tissue of Crohn's patients and in lung tissue from individuals with severe COPD 22,23 .
Previously we reported that ILC2s accumulate in NP of patients suffering from CRS 24,25 . Here we observed that IL-17producing ILCs predominate in NP from CF patients. Unexpectedly, we found that majority of these IL-17-producing ILCs are derived from ILC2s which transdifferentiated to IL-17producing cells by the epithelium-derived cytokines IL-1β, IL-23, and TGF-β. We present evidence that these ILC contribute to the increase of neutrophils which mediate the inflammation in NP of CF patients.

Results
ILC3s accumulate in nasal polyps from CF patients. To evaluate the role of ILCs in the sinonasal inflammation in CF patients, we examined the composition of ILC subsets in NP of these patients as compared with those from CRSwNP patients and healthy nasal tissue (inferior turbinates) (patient characteristics in Supplementary Fig. 1a). Although the total frequency of ILCs in these three tissues were similar (Fig. 1a), the composition differed. A significant increase in the proportion of NKp44 − ILC3s was found in CFwNP as compared with CRSwNP. (Fig. 1b, c). In contrast, ILC2s predominate in CRSwNP patients (Fig. 1b, c) and were almost undetectable in NP from CF patients and inferior turbinates. The CD117 + NKp44 − CRTH2 − ILCs from CFwNP did not express KLRG1, CD25 and had low ICOS expression (Supplementary Fig. 1b). We did observe significant GATA-3 expression in CD117 + NKp44 − CRTH2 − ILCs from CFwNP, but not in CD117+NKp44 − CRTH2 − ILCs from CRSwNP patients ( Supplementary Fig. 1b). Even though the ILC composition in NP from CF patients and healthy turbinate was similar, only in the inflamed tissue inflammatory cytokines could be detected (Fig. 1d). NP from CF patients displayed a significant increase in IL-17A when compared with NP from CRS patients whereas IL-5 levels were similar. Moreover, ILC3s from CFwNP patients produced IL-17A, whereas ILC3s and ILC2s from healthy turbinate tissue or from NP of CRS patients did not (Fig. 1e). Although we cannot exclude that Th17 cells and other cells can also be a source of IL-17 in these patients, this indicates that NKp44 − ILC3s are at least partly responsible for the IL-17 production in NP of CF patients.
Nasal epithelialium derived cytokines drive ILC2 plasticity. Our findings raised the question of why NP from CF patients do not contain ILC2s whereas in morphologically similar NP from CRS patients ILC2s predominate. One possibility is that the cytokine microenvironment in these polyps is different, resulting in altered ILC compositions. If correct, we would predict that the microenvironment in CF is disadvantageous and that of CRS advantageous for ILC2s. Indeed, we published previously that NP from CRS patients contain IL-4 and TSLP that strongly stimulate ILC2s 22 . Given that CF patients suffer from infections with SA or PA, we hypothesized that these pathogens might be responsible for a reduction of ILC2s observed in NP from these patients. Therefore, we cultured peripheral blood ILC2s directly with these pathogens or with NP epithelial cells (from CRS patients) that were exposed to SA or PA. We did not observe a reduction of ILC2s due to cell death, but surprisingly, coculture of SA or PAexposed epithelial cells with ILC2s resulted in an increase of IL-17A, without production of IL-5 by these cells (Fig. 2a, b), suggesting that these bacteria induced transdifferentiation of ILC2s to IL-17A-producing ILCs. ILC2s did not respond to the bacteria in the absence of epithelial cells. Culturing of blood-derived ILC1s or CD117 + CRTH2 − ILCs with SA or PA exposed epithelial cells did not result in IL-17A production ( Supplementary Fig. 2a). We reasoned that cytokines derived from epithelial cells drive IL-17A production by ILC2s. Infection with SA-induced or PA-induced high production of IL-1β, IL-23, and TGF-β in nasal epithelial cells, whereas no IL-33 or TSLP could be detected (Fig. 2c). IL-1β is a general activator of ILC2s and important for allowing ILC2 to ILC1 transdifferentiation 22,26 . Both in NP from CRS and CF patients, high levels of IL-1β were detected. We could not detect IL-23, but TGF-β was significantly increased in NP from CF as compared with those from CRS patients (Fig. 2d). These cytokines are implicated in Th17 polarization 15,27 and indeed, the combination of these three cytokines induced potent IL-17A production by proliferating ILC2s (Fig. 2e, Supplementary  Fig. 2b, c). Additionally, exposure of nasal epithelium to SA or PA induced high production of IL-6 ( Fig. 2c), a cytokine which has also been implicated in Th17 polarization in vitro 28 . However, IL-6 did not contribute to the transdifferentiation of ILC2s into IL-17-producing ILCs, but instead reduced this process (Fig. 2e,  Supplementary Fig. 2f).
ILC2s give rise to IL-17 producing cells at a clonal level. Next, we analyzed the progeny of single ILC2s derived from NP from CRS patients or blood in the presence of IL-2, IL-7, IL-1β, IL-23, and TGF-β (Fig. 3a, Supplementary Fig. 3a). With a cloning efficiency of 5-22%, we were able to grow clones within 14 days of culture on OP9 stromal cells. Almost all clones gave rise to IL-13-producing progeny, while a substantial part of each clone was producing IL-17A, similar to the inflammatory ILC2s (iILC2s) that have been described in mice 30 (Fig. 3a, Supplementary  Fig. 3a). A minor part still produced IL-5. There was a lot of variation in the fraction of IL-17A-producing cells that was obtained from each clone, with some clones exclusively producing IL-17A, whereas other clones gave rise to IL-5, IL-13, and IL-17A producing cells. (Fig. 3b, c), suggesting that within the progeny of one clone cells with different cytokine secretion profiles developed. PB ILC2 clones were more potent in generating IL-17producing cells as compared with NP ILC2s. Comparison of IL-17A-producing with non-IL-17A-producing clones revealed that the former ones had a higher RORC expression (Fig. 3d). When the progeny of IL-17A-producing clones were cultured in the presence of IL-33 and TSLP, these cells regained the ability to produce IL-5 to a similar extent as the progeny from non-IL-17Aproducing clones (Fig. 3e, f, Supplementary Fig. 3b). At the same time, these cells retained their IL-17A production following stimulation with TSLP and IL-33. In summary, these data indicate that ILC2s are a heterogeneous population and that a fraction of ILC2s can transdifferentiate into-IL-17A-secreting cells, whereas part of these cells remains resilient to conversion. The capacity of ILC2s to produce IL-17A is, at least in some cases, a stable feature.
The tissue cytokine balance modulates ILC2 plasticity. The observation that the ILC composition in NP of CF or CRS patients differed, prompted us to evaluate the presence of distinct inflammatory cytokines in the local nasal tissues that can modulate the process of plasticity. Whereas TGF-β protein was increased in CF as compared with CRS ( Fig. 2d), IL-4 gene was only found in CRSwNP tissue (Fig. 4a). Since we have previously shown that IL-4 is able to revert the IL-12-mediated transdifferentiation of ILC2s into ILC1s 22 , we investigated whether the absence of IL-4 alters the ILC composition in NP of CF patients. We isolated CD117 + CRTH2 − NKp44 − ILCs from CFwNP patients and activated these cells in the presence of IL-4 and IL-1β. Indeed, under these conditions, CD117 + CRTH2 − NKp44 − ILCs from CF patients acquired the ability to secrete IL-5 (Fig. 4b, c). Furthermore, IL-4 alone upregulated CRTH2 in part of the CD117 + CRTH2 − NKp44 − ILCs (Fig. 4d, e). In contrast, CD117 + CRTH2 − NKp44 − ILCs isolated from CRSwNP patients did not become ILC2s in the presence of IL-4. We assessed whether IL-4 also affected the ILC2 into ILC3 conversion induced by IL-1β, IL-23, and TGF-β. Culturing of ILC2s with IL-1β, IL-23, and TGF-β in the presence of IL-4 inhibited ILC2 transdifferentiation into IL-17A-producing cells (Fig. 4f, g). Transdifferentiation of ILC2s was prevented as well when IL-4 pre-stimulated ILC2s were cocultured with SA or PA infected epithelial cells (Fig. 4h). We next investigated the mechanism underlying the plasticity process. IL-1β is known to induce activation and cytokine responsiveness of ILC2s 31 and indeed pre-stimulation with IL-1β made the cells responsive to IL-23 by STAT3 phosphorylation (Fig. 4i). Similar STAT3 phosphorylation was observed in the presence or absence of TGF-β, but TGF-β addition induced additional phosphorylation of SMAD2/3, indicating that additional activation of this pathway is probably necessary to induce optimal expression of RORγt and IL-17 32 . IL-4 inhibition of the transdifferentiation is accompanied by downregulating the TGFBR1 and IL1R1 gene expression (Fig. 4j) and inhibition of STAT3 phosphorylation (Fig. 4k) indicating that IL-4 makes ILC2s unresponsive to IL-1β and TGF-β.
To further study the functional and phenotypic consequences of ILC2 plasticity, we compared the transcriptome of ILC2s in response to cytokine stimulations that result in differentiation into IL-5 or IL-17-producing cells. Both stimulations resulted in many genes being differentially expressed as compared with freshly isolated ILC2s (Fig. 5d, Supplementary Dataset 1) and all three groups formed distinct clusters ( Supplementary Fig. 4d). Further analysis was done to identify specific expression patterns for IL-5 and IL-17-producing ILC2s. We identified two distinct clusters of genes that were specifically expressed upon stimulation with either IL-33 plus TSLP, or IL-1β plus IL-23 and TGF-β (Fig. 5e, Supplementary Fig 5). Cluster 1, consisting of 497 genes (p values < 0.01; a 2log change in expression of over 1.4 fold) that are enriched upon stimulation with IL-1β, IL-23, and TGF-β, among which ILC3 and Th17 related genes such as RORC and CCR6 ( Supplementary Fig. 5). CCR6 was also among the top 25 of most significantly expressed genes as compared with cells stimulated with IL-33 and TSLP (Fig. 5f). Cluster 2 contained 186 genes that are enriched upon stimulation with IL-33 and TSLP, including ILC2 and Th2 related genes such as IL1RL1, IL5, or IL13 ( Supplementary Fig. 5). The most significantly differentially expressed genes included known ILC2 receptors such as KLRG1 and CRLF2 (which encodes the TSLP receptor) (Fig. 5f,  Supplementary Fig. 4f).
Many genes encoding markers that have been described to be specific for ILC2s are affected by stimulation (Fig. 6a). Blood ILC2s are naïve cells as the expression of most ILC2 markers, such as KLRB1, IL1RL1, CRLF2, PTGDR2, IL17RB, and GATA3 was increased upon stimulation with IL-33 and TSLP. Only elevated expression of IL1RL1 and CRLF2 was specific for IL-5/ IL-13-producing cells, whereas expression of the other markers was also elevated upon induction of IL-17 production, suggesting that, similar to mouse iILC2s, human ILC2s that are induced by IL-1β, IL-6, IL-23, or TGF-β are still responsive to IL-25 30 . Expression of RORC was in concordance to what we observed after intranuclear staining with a RORγt-specific antibody (Figs. 5c, 6b), whereas AHR was highly expressed in freshly isolated ILC2 and remained high upon stimulation. IL-17producing ILCs also displayed higher expression of the ILC3 related transcription factors RUNX3, EGR2, BATF3 and SMOX 33,34 (Fig. 6b). Neither type of stimulation induced TBX21 nor EOMES.
As expected, IL5 and to a lesser extend IL13 were specifically induced by IL-33 and TSLP, whereas stimulation with IL-1β, IL-23, and TGF-β increased expression of IL17F, but also of IL9, CCL20, and TNFSF13B (encoding BAFF) (Fig. 6c). Some IL-9 production was induced by IL-33 and TSLP, however IL-1β, IL-23, and TGF-β induced significantly higher IL-9 production ( Supplementary Fig. 4a). Stimulation with IL-33 and TSLP Cytokine (pg/mL) resulted in expression of IL10 and IL19, whereas both type 2 and type 3 stimulations induced IL26 expression (Fig. 6c). In contrast to IL17F, we did not observe an increase of IL17A transcripts in ILC2s stimulated for 6 days with IL-1β plus IL-23, and TGF-β compared with unstimulated ILC2 (Fig. 6c), even though transcripts of RORC, that controls both IL-17A and IL-17F expression, were increased and secretion of IL-17A protein was specifically induced (Fig. 5a, b). This may be due to mRNA instability or to the kinetics of induction of IL-17A transcripts. KEGG pathway analysis revealed that many differentially expressed genes that were selectively upregulated upon stimulation with IL-1β plus IL-23, and TGF-β belong to the Th17 pathway (Fig. 6e). A number of surface markers were selectively upregulated upon stimulation with either IL-1β, IL-23, and TGF-β; or IL-33 and TSLP (Fig. 6d). IL-17-producing ILC2s expressed chemokine receptors that were reported to be expressed on Th17 cells such as CCR6, CCR4, and CXCR6 35,36 suggesting that IL-17-producing ILC2s are migratory cells as also described for mouse iILC2 37 . We observed increased expression of the co-inhibitory SLAM family member CD244, which is also expressed on NK and CD8 + T cells and the mannose receptor MRC2, which can be induced on and TGF-β also led to increased protein expression as was confirmed by staining with specific antibodies (Fig. 6f, g). We observed that whereas a subset of blood ILC2s expressed CCR6, this expression was retained on IL-17-producing ILC2s, but lost upon stimulation with IL-33 and TSLP or IL-1β and IL-12 (Fig. 6g). CD117 + CRTH2 − NKp44 − ILCs from blood, which include ILC precursors, did contain CCR6 + cells, but these cells did not co-express CCR4 nor CXCR6 ( Supplementary Fig. 6a). The CCR6 + IL-17-producing ILCs resemble the MHCII expressing ILC3 subpopulation found in human tonsils ( Supplementary  Fig. 6b) 39 . In mice, these CCR6 + MHCII + ILC3s regulate CD4 + T cells in the intestine 40 . Stimulation with IL-1β, IL-23, and TGFβ also resulted in decreased expression of a cluster of 15 genes, including PTGER2 (encoding the prostaglandin E2 receptor), which ligand was shown to be a suppressor of ILC2 activation ( Supplementary Fig. 6c) 41 . KLRG1, which is expressed on blood ILC2s and maintained by IL-33 and TSLP, was also downregulated by IL-1β, IL-23, and TGF-β (Fig. 6h).
Transdifferentiated ILC2s enhance neutrophil recruitment. Whereas NP of CRS patients are characterized by an eosinophilic inflammation, the granulocyte population within CFwNP patients consist mainly of neutrophils (Fig. 7a) 42 . Analysis of ILCs and granulocytes in NP from CRS and CF patients revealed a positive correlation between ILC2s and the eosinophil to neutrophil ratio in NP, whereas the proportions of CD117 + NKp44 − ILC3s negatively correlated with this ratio (Fig. 7b) suggesting that ILCs are involved in the recruitment of eosinophils and neutrophils. Consistent with this notion is our previous observation that ILC2s in CRSwNP activate eosinophils 22 . Here we found that IL-1β, IL-23, and TGF-β induced secretion of IL-17A and TNF-α (Fig. 7c, d), which are potent activators of epithelial cells 43 . We observed that stimulation of epithelial cells with supernatant from ILC2s activated with IL-1β, IL-23, and TGF-β induced strong IL-8 production by epithelial cells (Fig. 7e). This was not due to the remaining IL-2, IL-1β, IL-23, and TGF-β in these supernatants as these cytokines did not directly activate epithelial cells (Supplementary Fig. 7a). The IL-8 production by these epithelial cells could be partially blocked by adding anti-IL-17A antibody to the conditioned medium (Fig. 7f), indicating that IL-17A secreted by ILC2s is partly responsible for IL-8 secretion by epithelial cells. IL-1β, IL-23 and TGF-β also induced production of IL-8 and in addition GM-CSF which can contribute to neutrophil recruitment ( Supplementary Fig. 7b). Since stimulation with IL-1β, IL-23 and TGF-β inhibits IL-5 production by ILC2s it is likely that transdifferentiated ILC2s are responsible for the low eosinophil to neutrophil ratio in CFwNP.

Discussion
CRSwNP is a disease of the upper airways characterized by tissue eosinophilia, high IL-5 and IgE, Th2-polarized responses, and an enrichment of ILC2s. We and others have previously shown that ILC2s are enriched in NP from CRS patients 24,22 and that their frequencies correlate with the allergic state 48 , and eosinophilia 49 . In the present study, we observed that the ILC composition in NP of CF patients is very different, despite morphological and clinical similarities with NP of CRS subjects. In NP of CF patients, the frequency of ILC2s was diminished, whereas that of NKp44 − ILC3s was increased, as compared with NP of CRS patients. These ILC3s produce IL-17 and together with Th17 cells, which are also enriched in NP of CF patients, may be responsible for the increased levels of IL-17 these inflamed nasal tissues 42 . Whether the same holds true in the lungs of CF patients remains to be confirmed. ILC3s can be divided into different subsets based on the expression of NCRs (NKp44 and NKp46). Earlier findings indicate that in a number of tissues NKp44 − ILC3s produce IL-17A whereas NKp44 + ILC3s produce IL-22 20,50,51 . Also in CF NP, NKp44 − ILC3s were found to produce IL-17A, but our data strongly suggest that these cells are derived from ILC2s that do not produce IL-22 or IFN-γ. Next to transdifferentiated ILC2s, ILC precursors that specifically develop into IL-17-producing ILC3s can also contribute to the observed IL-17 production.
VD3 deficiency is correlated with more severe inflammation of the nasal mucosa in CRSwNP patients 53 and associated with the presence of NP in CF 54 . Recently, it was shown that VD receptordeficient mice accumulate IL-22-producing ILCs in their intestines and that these mice are more susceptible to colitis 55 . In the present study, we show that VD3 selectively blocks IL-17A, IL-9 production, and RORC expression by ILC2s exposed to IL-1β, IL-23, and TGF-β. VD3 has been shown to regulate Th17 responses as VD receptor knockout mice have elevated Th17 cells and IL-17 production 56 . Our findings are consistent with observations that downregulation of the IL-23 receptor upon exposure to VD3 plays a critical role in modifying the IL-17 pathway in mucosal ILC3s 47 . Further identification of modulators of ILC function may be of interest for the design of therapeutic interventions, especially those targeting the downstream or upstream pathways that regulate ILC activation and recruitment 57 . Current treatment options for CF aim to improve the airway function and prevent microbial infection by physical removal of mucus combined with anti-microbial and anti-inflammatory agents, however new therapeutic strategies include CFTR correctors that are designed to improve defective protein processing, trafficking and cell surface expression and potentiators that increase the activity of mutant CFTR at the cell surface 58 . Specific targeting of IL-17 combined with anti-microbial treatment might result in a reduced neutrophilia and severity of inflammation in the lower and upper Correlation of ILC2 and NKp44 -ILC3 frequencies to eosinophil/neutrophil ratio in CRSwNP (n = 13). c, d Representative flow cytometry analysis (c) and quantification (d) of intracellular expression of IL-17A and TNF-α in blood ILC2s exposed to IL-2, IL-33, and TSLP or IL-2, IL-1β, IL-23, and TGF-β for 5-7 days after PMA/ionomycin (n = 9-10). e Quantification of IL-8 production by NCI-H292 epithelial cells after 48 hr exposure to conditioned medium of ILC2s that were stimulated with IL-2, IL-33, and TSLP; or IL-2, IL-1β, IL-23, and TGF-β for 5-7 days (n = 3-4). f Quantification of IL-8, IL-6, and GM-CSF production by NCI-H292 epithelial cells in the presence or absence of IL-17-blocking antibody upon culture as described in e (n = 3). Data are presented as individual values with mean of two to four independent experiments (d-f). *p < 0.05, **p < 0.01, ***p < 0.001 as determined by one way ANOVA or Student's t test airways. The increase in airway IL-17 in CF is most likely due to a dysregulated inflammatory response to infection which subsequently results in a massive influx of neutrophils. In a cftr −/− mouse model the importance of IL-17 in the pathology of PAinduced airways inflammation was demonstrated 14 . The neutralization of IL-17 prior to infection significantly improved the outcomes in the CF mice, suggesting IL-17 as a therapeutic target. This is supported by the reduced airway tissue damage and burden of infection in an Il17a −/− model of PA infection 59 . Notably, in the ctfr −/− , il1r1 −/− d/d mouse model of PA infection, neutralization of IL-1β attenuated the infection and improved bacterial clearance 10 . This suggests that IL-1β could be another interesting target for controlling the CF pathology.
ILC2 transdifferentiation into IL-17A-producing ILC2s required TGF-β to block IL-5 production and induce RORγt and IL-17 production in a SMAD2 dependent manner. Upon stimulation with IL-1β, IL-23, and TGF-β we generated cells that coproduce IL-13 and IL-17, similar to the iILC2s that have been described in mice 30,37 . iILC2s are recruited from the intestines in response to IL-25 or to an infection with N. brasiliensis. In mice, iILC2s can be identified by the expression of KLRG1, CCR9, and IL-25R, while lacking the IL-33R. Here we show that in humans, both IL-5-producing ILC2s and IL-17-producing iILC2 express the IL-25R, with the latter lacking the IL-33R. In contrast to mouse iILC2s, human IL-17-producing ILC2s lack KLRG1 and CCR9. In mice, iILC2s are derived from an intestinal precursor and are able to differentiate into natural IL-5 and IL-13producing ILC2s in vivo 60 . The nature of the intestinal ILC2 precursor in mice is unknown. Our findings suggest that IL-17producing iILC2s can be derived from mature nILC2s, although we cannot exclude the existence of a common ILC2 precursor for IL-17-producing ILC2s and nILC2s in humans as well. The IL-17-producing cells we describe here are very different from the recently described murine ILC2s that produce IL-5, IL-13, and IL-17 in an AHR-dependent, but RORyt-independent manner in response to IL-33 61 . These cells retain their ILC2 characteristics and are also very different from the iILC2s as they express similar levels of ST2 as nILC2s and do not depend on IL-25. An interesting fundamental difference between iILC2s and nILC2s in mice is that iILC2s are migratory cells whereas nILC2s are tissue resident 37,62 . Whether or not the IL-17-producing human ILC2s are migratory cells is unknown, but it is noteworthy that they express chemokine receptors, including CCR4, that are not expressed on nILC2s.
In summary, our findings demonstrate that the functional activities of ILC2s are strongly influenced by the microenvironment. Plasticity permits tissue resident ILC2s to adapt to changing conditions and may also change their migratory properties. Consequently, ILC2 transdifferentiation will facilitate the eradication of invading pathogens that require different types of immune responses or pathogens that colonize multiples niches along their life cycle. Henceforward, upon disease remission, ILCs might undergo cell-reprograming to avoid excessive mucosal damage, restore the mucosal tissue, and re-establish a beneficial interplay with commensal flora.

Methods
Tissue collection. All tissues were collected after subjects provided informed consent in accordance with approved protocols by the Medical Ethical Committee of the Academic Medical Center in Amsterdam, the Netherlands. Uninflamed nasal inferior turbinates were obtained from patients who underwent corrective surgery for hypertrophy with or without septoplasty. Nasal polyps were obtained from CRSwNP patients with or without cystic fibrosis undergoing endoscopic sinus surgery. Buffy coats were provided by the blood bank at Sanquin, Amsterdam.
For some bulk and all cloning experiments, cells were cultures on the murine stromal cell line OP9-DL1 (3000 cells were pre-seeded in 96-well round bottom plates one night before coculture). For bulk cultures (500-1000), ILC2s were stimulated for 8-10 days with IL-2 (20 U/mL), IL-7 (20 ng/mL), and various combinations of IL-33, TSLP, IL-1β, TGF-β, IL-23, and IL-4 (all at a concentration of 20 ng/mL). For bulk culture, fresh cytokines were added after 5 days. For cloning experiments, cytokines and medium were replenished once per week and cells were analyzed after 2-3 weeks.
Nasal epithelial cells were grown in 75 mL culture flasks in Bronchial Epithelial Cell Growth Medium (Lonza) until confluency or in PneumaCult-ALI medium (Stemcell Technologies) in the air-liquid interface (ALI) model. For ALI, epithelial cells were seeded onto permeable membrane of culture inserts. The basal epithelial cell surface was in contact with liquid culture medium, whereas the apical surface was exposed to air. Upon differentiation, cells were exposed to PA or SA (500,000 of bacteria cells/mL) for 5-7 days. At the same time, ILC2s (3000-5000) were added to the culture medium.
Quantitative real-time PCR. Total RNA was extracted with a NucleoSpin RNA XS kit (Macherey-Nagel) according to the manufacturer's instructions. cDNA was synthesized with a High-Capacity cDNA Reverse Transcription kit (ThermoFisher Scientific). PCRs were performed in BioRad iCycler (BioRad) with IQ SYBR Green Supermix (Bio-Rad,) using the primers described in Supplementary Table 1. Bio-Rad CFX Manager 3.1 software was used for quantification of expression. All samples were normalized to the expression of control genes encoding GAPDH and β-actin and results are presented in arbitrary units. The sequences of primers used in this study are collected in the Supplementary Table 1.
Microarray analysis. To isolate total RNA, sorted cells were flash frozen in PBS immediately after sorting and stored at −80°C prior to RNA extraction. QIAzol Lysis Reagent (Qiagen) was added to the cells, and RNA was isolated and purified using the RNeasy kit (Qiagen). The concentration was measured on a NanoDrop ND-2000 (Thermo Scientific) and RNA integrity was examined using the 2200 TapeStation System with Agilent RNA ScreenTapes (Agilent Technologies). Total RNA was amplified using the GeneChip WT Pico Kit (Thermo Fisher Scientific) generating biotinylated sense-strand DNA targets. The labeled samples were hybridized to human Clariom D arrays (Thermo Fisher Scientific). Washing and staining was performed by the GeneChip Fluidics Station 450 and the scanning was performed using the GeneChip Scanner 3000 (both Thermo Fisher Scientific). All cell populations were generated in triplicate. All data analysis was performed in RStudio. Raw data were normalized using the RMA algorithm implemented in the limma R-package. Adjusted p-values were calculated using the Benjamini-Hochberg method. Data were visualized using ggplot2 and pheatmap R-packages.
Statistical analysis. Statistical significance was determined with ANOVA, Student's t-test, or Mann-Whitney U-test. GraphPad Prism software was used.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. The source data underlying Figs. 1a, c-e, 2b-i, 3a, c-e, 4a, c, e, g, h, j, k, 5b, c, 6f, 7a, b, d-f