IL-21-mediated non-canonical pathway for IL-1β production in conventional dendritic cells

The canonical pathway for IL-1β production requires TLR-mediated NF-κB-dependent Il1b gene induction, followed by caspase-containing inflammasome-mediated processing of pro-IL-1β. Here we show that IL-21 unexpectedly induces IL-1β production in conventional dendritic cells (cDCs) via a STAT3-dependent but NF-κB-independent pathway. IL-21 does not induce Il1b expression in CD4+ T cells, with differential histone marks present in these cells versus cDCs. IL-21-induced IL-1β processing in cDCs does not require caspase-1 or caspase-8 but depends on IL-21-mediated death and activation of serine protease(s). Moreover, STAT3-dependent IL-1β expression in cDCs at least partially explains the IL-21-mediated pathologic response occurring during infection with pneumonia virus of mice. These results demonstrate lineage-restricted IL-21-induced IL-1β via a non-canonical pathway and provide evidence for its importance in vivo.

I nterleukin-1b (IL-1b) is a cytokine important for host defence to pathogens 1,2 . It triggers inflammatory responses by various mechanisms, including the induction of cyclooxygenase-2 and inducible nitric oxide synthase, which control the release of prostaglandin E2 and nitric oxide and recruitment of inflammatory cells 2,3 . IL-1b also promotes differentiation of CD4 þ T cells into Th17 cells 4 , and the production of IL-17 and IL-22 from these cells further recruits neutrophils to the inflammatory sites and stimulates them to release antimicrobial peptides 1 . Local inflammatory responses during pathogen invasion are often beneficial to the host 5 , but they can also be deleterious if not properly controlled. Moreover, dysregulation of IL-1b expression leads to the development of autoimmune diseases, such as rheumatoid arthritis, type 2 diabetes and gout, as well as autoinflammatory diseases, such as familial Mediterranean fever and cryopyrin-associated periodic syndromes 3,6,7 .
Production of biologically active IL-1b requires immune cells to receive two distinct signals. First, Il1b mRNA expression is induced by the toll-like receptor-or IL-1 receptor (IL-1R)mediated signalling pathways, which activate transcription factor NF-kB, followed by translating the mRNA to the biologically inactive pro-IL-1b (ref. 8). The cells then activate the caspase-1containing 8 or the more recently identified caspase-8-containing 9 inflammasomes when receiving second signals such as bacterial metabolites, with caspase-mediated cleavage of pro-IL-1b into mature IL-1b (ref. 8). However, signals for triggering IL-1b production under conditions of sterile inflammation remain largely unknown, and the mechanism(s) for releasing IL-1b to the extracellular environment remain unclear.
IL-21 is a pleiotropic type 1 cytokine produced by CD4 þ T cells and natural killer T cells. It promotes immunoglobulin class switch 10 , augments antibody production 10 and drives terminal differentiation of B cells 11 . IL-21 is critical for the development of multiple forms of autoimmunity in animal models 12 , and antibodies to IL-21 are being tested in phase 1 clinical trials for rheumatoid arthritis (NCT01208506 and EudraCT-2011-005376-42). Moreover, IL-21 exhibits antitumor activity 13,14 , with IL-21 in phase 2 clinical trials for various cancers 12 . Although the roles of IL-21 in T-and B-cell function have been extensively studied, there is limited knowledge on the actions of IL-21 on dendritic cells (DCs). IL-21 has been reported to inhibit the lipopolysaccharide (LPS)-stimulated production of several proinflammatory cytokines, including IL-1b, by bone marrow-derived DCs (BMDCs) 15 , and we recently demonstrated that IL-21 is proapoptotic for conventional DCs (cDCs), at least in part by its induction of BIM 16 . Here we show an unanticipated role of IL-21 in inducing IL-1b expression in cDCs. IL-21-induced expression of Il1b mRNA is independent of NF-kB activation but requires STAT3. Moreover, we found that this IL-21-STAT3 pathway for regulating IL-1b expression is DC-specific, which correlates with the presence of cell typespecific enhancer landscapes. Importantly, IL-21-mediated production of mature IL-1b does not require a second signal for the activation of inflammasomes and in fact is independent of canonical caspase-containing inflammasomes, but it depends on IL-21-mediated cell death, which also requires STAT3 (ref. 16). Finally, we demonstrate that this IL-21-STAT3-dependent IL-1b expression can at least in part explain the pathologic immune response mediated by IL-21 during infection with Pneumonia Virus of Mice (PVM), a mouse model for human respiratory syncytial virus.

Results
IL-21-mediated Il1b gene expression in cDCs requires STAT3. We previously performed microarray analysis to identify genes that are regulated by IL-21 in cDCs 16 . Surprisingly, although IL-21 inhibits the LPS-induced IL-1b expression in BMDCs 15 , we observed a significant induction of Il1b mRNA by IL-21 in cDCs. Transcription of Il1b mRNA is induced in response to IL-1 itself or microbial products acting via toll-like receptors 2 , but Il1b induction by IL-21 was not anticipated. Therefore, we validated this result by reverse transcription-PCR and confirmed that IL-21 indeed rapidly (within 30 min) induced Il1b mRNA (Fig. 1a) and that this induction did not occur in Il21r À / À cells (Fig. 1b), excluding the possibility that it resulted from contaminating endotoxin. We compared the effect of IL-21 with the classical Il1b-inducing ligand LPS and found that IL-21 more robustly induced Il1b mRNA expression in cDCs than did LPS (Fig. 1c). Because NF-kB is known to be critical for the induction of Il1b mRNA expression by LPS 8 , we treated cDCs with an IkB kinase b inhibitor, MLN120B 17 and this strongly inhibited LPS-induced, but not IL-21-induced, Il1b mRNA expression (Fig. 1d), indicating that the effect of IL-21 was independent of NF-kB. Correspondingly, LPS treatment of cDCs induced phosphorylation of IkBa, which promotes its ubiquitination and degradation, and allows nuclear translocation of NF-kB, whereas IL-21 had no effect ( Fig. 1e and Supplementary Fig. 1). Because IL-21 can activate STAT3 in cDCs 16 , we next crossed Stat3-floxed mice to CD11c-Cre transgenic mice to delete Stat3 in cDCs (herein denoted as Stat3 À / À mice), and this markedly decreased IL-21-induced ( Fig. 1f) but not LPS-induced ( Fig. 1g) Il1b mRNA expression. Importantly, when we tested human cDCs isolated from peripheral blood, IL-21 also induced tyrosine phosphorylation of STAT3 (Fig. 1h) and augmented IL1B mRNA expression (Fig. 1i) in human cDCs. Thus, IL-21 uses a distinctive, STAT3-dependent pathway for the induction of Il1b mRNA expression in cDCs.
Pro-IL-1b induction by IL-21 in cDCs is cell type-specific.
To determine if the effect of IL-21 extended to other cytokines that activate STAT3, we stimulated cDCs with IL-6 and IL-10 and confirmed that these cytokines also induced tyrosine phosphorylation of STAT3 in cDCs (Fig. 2a, upper panel) and correspondingly, these cytokines also induced pro-IL-1b production (Fig. 2a,lower panel,and Fig. 2b) in these cells, with the degree of induction approximately correlating with the magnitude of STAT3 activation. In contrast, FLT3 ligand, a cytokine that promotes cDC survival 18 , did not induce phosphorylation of STAT3 nor induce pro-IL-1b (Fig. 2a,b). IL-21 has been shown to suppress LPS-stimulated IL-1b production in BMDCs 15 , and IL-10 is an anti-inflammatory cytokine and suppressor of cytokine production 19 . We therefore pre-treated cDCs with IL-21 or IL-10 and then determined the effect of LPS on Il1b mRNA expression. Unlike the situation for BMDCs, IL-21 did not inhibit LPS-induced Il1b mRNA expression in cDCs. Corresponding to the effect of IL-10 on pro-IL-1b expression, IL-10 induced Il1b mRNA expression but did not suppress the LPS-induced Il1b mRNA expression (Fig. 2c). Indeed, neither IL-21 nor IL-10 suppressed LPS-induced Il6 (Fig. 2d) or Tnf (Fig. 2e) mRNA expression, suggesting that the effects of IL-21 and IL-10 on cDCs are different from those on BMDCs.
To further compare the effects of IL-21 and IL-10 on cDCs, including on LPS-treated cells, we performed RNA-Seq analysis. As expected, the number of genes regulated by LPS was greater than those regulated by IL-21 or IL-10 ( Supplementary Fig. 2a and Supplementary Data 1a-c). Both IL-21 and IL-10 strongly activate STAT3 in cDCs, and they shared overlapping but also have distinctive gene regulation patterns ( Supplementary Fig. 2b,c and Supplementary Data 1d-f). Among 2,278 genes that were regulated by LPS, only 58 were affected by IL-21, and only 15 out of these 58 genes were repressed by IL-21 ( Supplementary Fig. 2d and Supplementary Data 2a), indicating that IL-21 did not globally suppress LPS-induced gene expression in cDCs. IL-10 had a stronger effect than IL-21, with 189 out of 2,278 genes regulated by LPS being affected by IL-10, and among these genes, 90 were repressed (Supplementary Fig. 2e and Supplementary Data 2b). We also compared the effects of IL-21 and IL-10 in regulating LPS-mediated gene expression and found that 89 of 105 genes that were differentially expressed by LPS versus LPS þ IL-21 stimulation were also regulated by IL-10 ( Supplementary  Fig. 2f,g and Supplementary Data 2c-e), and overall, there were more IL-10 regulated genes, suggesting that IL-10 had a more potent suppressive effect than IL-21 on LPS-mediated gene regulation. These data indicate that IL-21 and IL-10 have both overlapping and distinctive effects on cDCs.
Divergent STAT3 binding patterns in cDCs versus CD4 þ T cells. Above, we noted differential effects of IL-21 in cDCs and CD4 þ T cells, even though IL-21 strongly activates STAT3 in both cell types. We therefore compared the genome-wide STAT3 binding induced by IL-21 in cDCs versus CD4 þ T cells using ChIP-Seq analysis. Remarkably, STAT3 binding sites in IL-21-stimulated cDCs and CD4 þ T cells were mostly non-overlapping ( Fig. 3a and Supplementary Data 3a,b), with low STAT3 binding in CD4 þ T cells at sites where IL-21-activated STAT3 binds in cDCs (Fig. 3b, green versus blue lines) and vice versa (Fig. 3c, blue versus green lines). Nevertheless, motif analysis showed that IL-21-activated STAT3 in cDCs and CD4 þ T cells bound primarily to GAS (gamma-activated site, TTCnnnGAA) motifs    Supplementary Fig. 3). We therefore determined whether the differential STAT3 binding correlated with the presence of active enhancer landscapes. Increased histone H3 lysine-4 monomethylation (H3K4me1) and H3K27 acetylation (H3K27ac) indicate poised or active enhancer landscapes that promote transcription factor binding 20 , so we compared IL-21-induced STAT3 binding (Fig. 3a-c) with H3K4me1 (Fig. 3d,e) and H3K27ac (Fig. 3f,g) patterns. The 'dips' at the center of the histone marks represent open chromatin corresponding to nucleosome depletion that occurs at active promoters and enhancers and promotes transcription factor binding 21 . Sites binding STAT3 in cDCs but not CD4 þ T cells (Fig. 3b, blue line) had less H3K4me1 (Fig. 3d, blue and gold lines) and H3K27ac (Fig. 3f, blue and gold lines) at the STAT3 peaks but increased H3K27ac adjacent to the STAT3 peaks (Fig. 3f, blue and gold lines), indicating an active enhancer in cDCs but not CD4 þ T cells. Analogously, sites binding STAT3 in CD4 þ T cells (Fig. 3c, green line) had less H3K4me1 (Fig. 3e, red and green lines) and H3K27ac (Fig. 3g, red and green lines) at the STAT3 peaks but increased H3K27ac adjacent to the STAT3 peaks Differential epigenetic marks at the Il1b and Il21 loci. The ChIP-Seq results prompted us to determine whether differential STAT3 binding could explain the cell type-specific functions of IL-21 in cDCs and CD4 þ T cells. Therefore, we compared IL-21-induced gene regulation in these cells, and we found that these STAT3 binding differences correlated with differential IL-21-induced gene expression in cDCs versus CD4 þ T cells ( Fig. 4a) (for example, IL-21 could induce Il1b in cDCs but not CD4 þ T cells, whereas it induced Il21 in CD4 þ T cells but not in cDCs ( Fig. 4a,b). Interestingly, analysis revealed IL-21-induced STAT3 binding in cDCs at the Il1b promoter and 5 0 upstream region (Fig. 4c) to GAS-like motifs (Supplementary Table 1), with enriched H3K4me1 and H3K27ac marks at those sites (Fig. 4c), whereas STAT3 did not bind to these sites in CD4 þ T cells and active enhancer marks were correspondingly absent (Fig. 4d).
Conversely, for the Il21 gene, in cDCs, where the gene is not expressed, neither STAT3 nor H3K4me1/H3K27ac marks were found at the promoter region after IL-21 stimulation (Fig. 4e), but IL-21-activated STAT3 bound to the Il21 promoter region in CD4 þ T cells, which express the gene, with enrichment of H3K4me1 and H3K27ac marks (Fig. 4f). Thus, gene expression correlated with the active enhancer landscapes and IL-21-induced STAT3 binding in cDCs versus CD4 þ T cells. Moreover, we found IL-21-induced H3K4me3 at the Il1b promoter (indicating an open chromatin structure 20 ) in cDCs where Il1b is expressed ( Fig. 4c) but not in CD4 þ T cells where it is not (Fig. 4d); conversely H3K4me3 was strongly present at the Il21 promoter in CD4 þ T cells where Il21 is expressed (Fig. 4f), but not in cDCs where it is not (Fig. 4e). Interestingly, Il1b is one of only 24 genes in which we observed much higher H3K4me3 in cDCs than in CD4 þ T cells ( Supplementary Fig. 4). The presence of H3K27me3 repressive marks (indicative of inactive genes 20 ) at the Il1b locus in CD4 þ T cells and Il21 locus in cDCs corresponded to the non-expression of these genes in these cell types. Interestingly, the Il21 locus in CD4 þ T cells exhibited both H3K27me3 and H3K4me3 marks (Fig. 4f), indicative of a bivalent gene poised for induction or repression 22 .
A non-canonical pathway for IL-21-mediated IL-1b production. cDCs develop from precursor cells distinct from those that develop into monocyte-differentiated DCs and macrophages 23 . Whereas IL-1b production has been extensively studied in these latter cells 1 , the molecular machinery for processing pro-IL-1b into mature IL-1b in cDCs has not been reported. We treated cDCs with IL-21 or LPS and determined the expression of pro-IL-1b at different time points. Pro-IL-1b was detected at 2 and 6 h of IL-21 stimulation but then rapidly declined ( Fig. 5a and Supplementary Fig. 5). Interestingly, this decline correlated with an increase of secreted IL-1b (from 16-to 24-hour time points, Fig. 5b). The antibodies recognizing secreted IL-1b did not cross-react with pro-IL-1b ( Supplementary Fig. 6a,b), and correspondingly, pro-IL-1b was not detected in the cell culture supernatant after IL-21 stimulation ( Supplementary Fig. 7), indicating that the protein detected was indeed the mature IL-1b. Importantly, the secreted IL-1b was bioactive and induced IL-2Ra expression in Th17 cells (Fig. 5c). Moreover, this induction was blocked by anti-IL-1b and did not occur in Il1r À / À Th17 cells (Fig. 5c), confirming that IL-21 induces production of functional IL-1b.
In macrophages and BMDCs, IL-1b production in response to LPS þ ATP involves an inflammasome containing caspase-1, NLRP3 and ASC (apoptosis-associated speck-like protein containing a CARD, encoded by the Pycard gene) as critical components. We thus examined IL-1b production in cDCs after stimulation with LPS þ ATP, and found that it was greatly diminished in Casp1-, Nlrp3-or Pycard-deficient cDCs (Fig. 5d), indicating that the canonical caspase-1-containing inflammasome is functional in cDCs. Strikingly, however, IL-21-induced production of IL-1b was independent of caspase-1, NLRP3 and ASC (Fig. 5e), indicating a distinctive mechanism. Recently, a caspase-1-independent but caspase-8-dependent non-canonical inflammasome was reported 9 . We therefore, next examined the H3K4me1 H3K4me1 ARTICLE role of caspase-8 in IL-21-induced IL-1b production using cDCs from Ripk3 and Casp8 double-deficient mice, in which RIPK3-dependent necrosis resulting from caspase-8 deficiency is prevented 24 , but we found normal IL-21-induced IL-1b in caspase-8-deficient cDCs (Fig. 5f), indicating that caspase-8 is not essential for IL-21-induced IL-1b production in these cells.
Previously, we showed that IL-21-induced apoptosis of cDCs could be rescued by granulocyte-macrophage colony-stimulating factor (GM-CSF) 16 , so we investigated the effect of GM-CSF and whether apoptosis is required for IL-1b production. Interestingly, GM-CSF did not inhibit Il1b mRNA expression, and in fact GM-CSF slightly induced its expression, with GM-CSF þ IL-21 inducing slightly more Il1b mRNA than IL-21 alone (Fig. 6a). However, adding GM-CSF enhanced the accumulation of unprocessed pro-IL-1b (Fig. 6b, upper panel), but GM-CSF significantly decreased IL-21-induced production of mature IL-1b  protein (Fig. 6c). We hypothesized that cell death might be required for the release of IL-1b, and indeed inhibition of death by GM-CSF (Fig. 6b, lower panel) correlated with diminished release of mature IL-1b (Fig. 6c). Furthermore, corresponding to GM-CSF inhibiting IL-21-mediated cDC death by inhibiting BIM expression 16 , we observed markedly decreased IL-21induced apoptosis in the absence of BIM, which is encoded by the Bcl2l11 gene (ref. 16 and Fig. 6d), and whereas IL-21-induced Il1b expression was not significantly altered in BIM-deficient cDCs (Fig. 6e), IL-21-induced IL-1b production was essentially abolished (Fig. 6f). In contrast, LPS þ ATP-mediated apoptosis (Fig. 6d) and IL-1b production (Fig. 6f) were not dependent on BIM. These data indicate that IL-21 uses a distinctive, noncanonical pathway for IL-1b production.
Apoptosis is critical for IL-21-mediated IL-1b production. We next compared the effect of IL-21 to those of IL-6 and IL-10, which also activated STAT3 and could induce pro-IL1b (Fig. 2a,b). IL-21 appeared somewhat distinctive, as IL-1b production and cDC death were only slightly induced by IL-6 and IL-10 ( Fig. 7a,b), even though these cytokines were more similar to IL-21 in their induction of pro-IL-1b (Fig. 2a,b). These results underscore the distinctive ability of IL-21 to potently induce apoptosis and the production of mature IL-1b in cDCs. Despite its apoptotic effect on cDCs, IL-21 does not diminish the viability of BMDCs 16 , and although IL-21-induced pro-IL-1b expression in BMDCs (Fig. 7c), it did not induce the production of mature IL-1b, in contrast to its potent induction by LPS þ ATP (Fig. 7d). cDC death may result in the induction of serine protease activity, (c) CD4 þ T cells from WT or Il1r À / À mice were cultured in Th17 cell differentiation conditions for 2 days, then supernatant from a 24 h, IL-21-treated cDC culture was added to the Th17 cells and incubated for 2 days, with or without addition of 10 mg ml À 1 of anti-IL-1b. Expression of IL-2Ra (MFI) was determined by flow cytometry. The amount of biologically active IL-1b was determined using a standard curve constructed by assaying recombinant IL-1b. Shown are the combined results of two independent experiments with total of six samples. (d,e) WT, Casp1 À / À , Nlrp3 À / À and Pycard À / À cDCs were rested 1 h. In d, cDCs were then treated with 100 ng ml À 1 LPS for 20-24 h with 5 mM ATP added in the final 1 h, and IL-1b assessed. Data are from two experiments; error bars are means ± s.e.m. In e, cDCs were then treated with IL-21 for 20-24 h and IL-1b protein determined. Data are from five experiments. P values of IL-21-treated WT samples as compared with Casp1 À / À , Nlrp3 À / À and Pycard À / À samples are 0.99, 0.22 and 0.96, respectively; error bars are means ± s.e.m. (f) cDCs from Ripk3 À / À , Ripk3 þ / À Casp8 þ / À and Ripk3 À / À Casp8 À / À mice were treated as in e, and IL-1b assessed. Data shown are from three experiments. P values of IL-21-treated Ripk3 À / À sample compared with Ripk3 þ / À Casp8 þ / À and Ripk3 À / À Casp8 À / À samples are 0.57 and 0.93, respectively. In b and d-f, IL-1b production in the culture supernatant was determined by ELISA. Pro-IL-1b induced by IL-21 in the culture supernatant was minimal, based on a pro-IL-1b-specific ELISA. Statistical analysis was performed by Student's t-test.
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8988 ARTICLE and previous studies suggested that pro-IL-1b can also be processed by serine proteases 25 . We thus investigated the role of serine proteases in IL-21-mediated IL-1b expression by treating cDCs with a general serine protease inhibitor, 3,4 dichloroisocoumarin 26 before IL-21. This inhibitor substantially inhibited IL-21-induced IL-1b production (Fig. 7e) but did not affect IL-21-induced apoptosis (Fig. 7f). Thus, activation of serine protease(s) is not required for IL-21-mediated cell death but rather occurs in association with or following cell death. These data demonstrated that IL-21-STAT3 axis induces a noncanonical, cell death-and serine protease-dependent pathway for IL-1b production in cDCs.
IL-21-mediated IL-1b in the pathologic response to PVM. IL-21 promotes the pathological immune response of PVM but the mechanism is unclear. We found less Il1b mRNA in the lungs of Il21r À / À mice than in wild-type (WT) mice after infection with PVM 27 . Correspondingly, cDCs from Il21r À / À mice had lower expression of pro-IL-1b after PVM infection (Fig. 8a,b). Consistent with the role of STAT3 in IL-21-mediated IL-1b expression, Il1b mRNA was also lower in lungs from Stat3 À / À than from WT mice after PVM infection (Fig. 8c). The absence of STAT3 significantly lowered pro-IL-1b expression in cDCs (Fig. 8d,e), whereas pro-IL-1b production tended to be lower in macrophages, although the decrease was not statistically significant (Fig. 8d,f), analogous to the in vitro data (Fig. 2f). Moreover, Il1r-deficient mice had less perivascular oedema and fewer immune cells (Fig. 8g, right versus left panel), and particularly fewer neutrophils (Fig. 8h) in the lung. These results indicate that IL-1b is a key mediator of inflammation, and collectively indicate the importance of the IL-21-STAT3-IL-b axis in the pathological immune response to PVM infection.

Discussion
In this study, we have identified an unanticipated role of IL-21 in the regulation of IL-1b expression in cDCs and provide an example by which a type 1 cytokine regulates IL-1b. cDCs develop from precursor cells that are distinct from those giving rise to monocyte-derived DCs 28 . Whereas monocyte-derived DCs are important for inflammatory responses, cDCs play critical roles in triggering immune responses as well as maintaining immune self-tolerance through their interaction with T cells and the production of cytokines and chemokines 29 . The molecular mechanisms controlling IL-1b in cDCs previously have not been studied, and we found that in addition to using a canonical pathway (for example, for LPS) that is dependent on NF-kB for Il1b gene expression and the caspase-1-ASC-NLRP3 inflammasome for pro-IL-1b processing, cDCs can also use an alternative STAT3-dependent pathway for IL-21-induced Il1b gene expression. STAT3 bound at the Il1b promoter and 5 0 upstream region, suggesting that it directly regulates Il1b mRNA expression. Interestingly, this IL-21-STAT3 pathway for regulating IL-1b seems to be cDC-specific, as IL-21 did not induce Il1b mRNA in CD4 þ T cells even though it strongly activated STAT3 in these cells. Indeed, IL-21-activated STAT3 bound mostly to non-overlapping sites in cDCs and CD4 þ T cells, and these different binding patterns likely account for the cell type-specific gene regulation by IL-21 (for example, induction of Il1b in cDCs versus Il21 in CD4 þ T cells). We further show that this distinctive STAT3 binding pattern correlates with the differential presence of marks for active enhancer landscapes in the different cell types, suggesting that epigenetic regulation is involved in the control of cell type-specific actions of IL-21.
Previous studies have identified cell type-specific enhancer landscapes in Th1 and Th2 cells 30 , but our results reveal that differential enhancer landscapes in cDCs and CD4 þ T cells determine the cell type-specific effect of IL-21. Strikingly, we found that IL-21 not only induced Il1b mRNA expression in cDCs but also triggered pro-IL-1b processing and release of mature IL-1b, and that a second, inflammasomeactivating signal was not required. Indeed, IL-21 acted in the absence of either caspase-1 or caspase-8, but IL-21-induced IL-1b production was strictly dependent on IL-21-mediated cell death, which requires STAT3-mediated induction of BIM 16 . Interestingly, the actions of IL-21 are distinctive because IL-6 and IL-10 minimally induce cDC death even though these cytokines also activate STAT3.
In this study, we have found that IL-21-mediated mature IL-1b production and release from cDCs are strictly dependent on cell death. Our data indicate that this process requires the activation of serine protease(s), which likely occurs in association with or following IL-21-induced cell death. Interestingly, previous studies have shown that serine proteases such as cathepsin G, elastase and proteinase 3 can cleave pro-IL-1b into active IL-1b in neutrophils 25 , which is responsible for caspase-1-independent IL-1b production during acute arthritis induction in mice 31 . Additional work is required to determine the in vivo importance of the novel IL-21-mediated, serine protease-dependent IL-1b production pathway we have identified as well as the importance of the induction of IL1B mRNA by human DCs. Our studies establish a novel crosstalk between IL-21 and IL-1b, and this crosstalk may contribute to the uncontrolled inflammatory responses during PVM-induced lung disease. Additional work is required to determine the importance of this mechanism versus that of the canonical major NF-kB-inflammasome-dependent pathway in lung DCs and macrophages for the production of IL-1b during PVM infection. Conceivably, the pathway we describe might also contribute to autoimmune diseases where high levels of IL-21 and IL-1b can be detected 7,12 . These studies significantly expand our understanding of the actions of IL-21 in cDCs, with IL-21 both promoting immune tolerance through the induction of apoptosis of cDCs and also promoting inflammatory responses via its induction of IL-1b.

Methods
Mice and reagents. C57BL/6, 129X1/SvJ, Bcl2l11 À / À , Il1r À / À and CD11c-Cre mice were from The Jackson Laboratory. Stat3-deficient DCs were generated by crossing Stat3-floxed mice with CD11c-Cre mice 16 . Il21r À / À mice have been described 10 . Ripk3 and Casp8 double-deficient mice have been described 24 . Casp1 À / À mice were from The Jackson Laboratory or from Dr Richard Flavell, Yale University 32 . Nlrp3 À / À and Pycard À / À mice were from Dr Vishva Dixit, Genentech 33,34 . Both male and female mice from 7 to 12 weeks old were used. All protocols were approved by the NHLBI Animal Care and Use Committee and followed NIH guidelines for using animals in intramural research. MLN120B was provided by Dr Ulrich Siebenlist, NIH.
Isolation of splenic DCs. Mouse spleens were injected with 1 ml of 1 mg ml À 1 collagenase D and 20 mg ml À 1 DNase I (Roche), cut and incubated in collagenase solution at 37°C for 20 min. After passage through a cell strainer, red blood cells were lysed using ACK lysis buffer and the remaining cells were incubated with Fc block (BD Biosciences) at 4°C for 10 min. CD11c þ cells were positively selected with CD11c microbeads (Miltenyi Biotec). Purity of CD11c þ DCs was 93-95%.
Isolation of human peripheral blood CD1c þ cDCs. Buffy coats from healthy volunteers were separated by gradient centrifugation using lymphocyte separation medium. Mononuclear cells were negatively selected using CD19 microbeads to remove the CD1c þ B cells, followed by positive selection with CD1c microbeads (Miltenyi Biotech) to enrich for CD1c þ cDCs (purity was 490%).
Generation of BMDCs and BMMs. Bone marrow cells from mouse femurs and tibias were cultured for 8 days in RPMI-1640 medium containing 10% FBS, 200 mM L-glutamine, 10 IU ml À 1 penicillin, 100 mg ml À 1 streptomycin, 55 mM b-mercaptoethanol and 20 ng ml À 1 GM-CSF (for BMDCs) or macrophage colonystimulating factor (for BMMs) (Peprotech), with medium changed every 3 days. Cells were 490% pure. RNA analysis. Total RNA was isolated using the RNeasy Plus Mini Kit (Qiagen). First-strand complementary DNAs were synthesized using the Omniscript reverse transcription kit (Qiagen) and oligo(dT). Quantitative reverse transcription-PCR was performed on a 7900H sequence detection system (Applied Biosystems). Real-time primers and TaqMan probes were from Applied Biosystems. Expression was normalized to mouse Rpl7 or human ACTB.
Apoptosis studies. Splenic DCs were rested 1 h, stimulated with different cytokines for 20-24 h, and apoptosis was analysed by Annexin V and 7-AAD staining (BD Biosciences). For some experiments, the cells were incubated with 20 mM of serine protease inhibitor 3,4 dichloroisocoumarin for 1 h before cytokine stimulation.
Bioactivity assay of mature IL-1b. Naïve CD4 þ T cells were isolated from WT and Il1r À / À mice and differentiated into Th17 cells with 5 mg ml À 1 plate-bound anti-CD3, 2 mg ml À 1 soluble anti-CD28, 10 ng ml À 1 IL-6, 2 ng ml À 1 TGF-b, 10 mg ml À 1 anti-IFN-g, 10 mg ml À 1 anti-IL-4 for 2 days. cDC supernatants from un-stimulated or IL-21-stimulated cultures were added to the cells and incubated for 2 days, without or with the addition of 10 mg ml À 1 anti-IL-1b. Expression of IL-2Ra on Th17 cells was determined by flow cytometry. Amounts of mature IL-1b in the supernatants were determined by a standard curve (0-10 ng ml À 1 recombinant IL-1b) PVM infection and lung cell preparation. Virus stock (PVM strain J3666) was prepared as previously described 35 . Mice were anaesthetised briefly by ketamine/ xylazine and inoculated intranasally with 60 plaque-forming unit PVM in 50 ml PBS. For flow cytometry analysis, lung tissue was minced into small pieces using a razor blade and digested in a solution containing 1 mg ml À 1 collagenase D and 200 mg ml À 1 DNase I (Roche) for 30 min at 37°C. Digested tissue was then pushed through a cell strainer with a syringe. Cells were centrifuged, and RBCs were lysed with ACK, followed by two washes with complete RPMI-1640 medium. For RNA analysis, lung tissue was homogenized in TRIzol (Invitrogen) followed by RNA cleanup with the RNeasy kit (Qiagen).
PCR products were barcoded (indexed) and sequenced on an Illumina HiSeq 2000 platform. Sequenced reads (50 bp, single end) were obtained with the Illumina CASAVA1.8 pipeline and mapped to the mouse genome (NCBI37/mm9, July 2007) using Bowtie 0.12.9 36 . Only uniquely mapped reads were retained. To eliminate bias caused by PCR amplification, only non-redundant reads were retained and analysed. The mapped outputs were converted to browser-extensible data files, which were then converted to binary tiled data files (TDFs) using IGVTools 2.3.32 for viewing on the IGV browser (http://www.broadinstitute.org/ igv/home). TDFs represent the average alignment or feature density for a specified window size across the genome. We mapped reads into non-overlapping 200 bp windows (for transcription factors STAT3) or 20 bp windows (for histone modifications such as H3K4me1 or H3K27ac) and the reads were shifted 100 bp from their 5 0 starts to represent the center of the DNA fragment associated with the reads.
See Supplementary Table 2 for the summary of ChIP-Seq libraries.
reads were analysed for peak calling. A transcription factor (such as STAT3) was considered as 'bound' to genes if peaks were within 5 kb upstream of the transcription start site and anywhere across the gene body.
Binding intensity distribution. To compute the binding distribution of transcription factors or histones across selected genomic regions (such as STAT3 peaks), we chose±3 kb regions around either transcription start site or peak summits and divided the regions into bins of 20-bp window size. Reads (or tags) that fell into each bin were counted and normalized by library size. The normalized read density was plotted using Gnuplot 4.6. Heatmap of binding intensity was generated based on K-means clustering and plotted using seqMiner 1.3.3e 38 .
Statistical analysis. Statistical comparison between samples was performed by unpaired Student's t-test, Po0.05 is considered as statistically significant. NS, not statistically significant.