IL-21 Is a Central Memory T Cell–Associated Cytokine That Inhibits the Generation of Pathogenic Th1/17 Effector Cells

Cytokines regulate the differentiation of uncommitted naive T cells to various types of effector T cells that target different classes of pathogens (1, 2). In particular, it is well established that IL-12 induces Th1 differentiation in humans and mice, because it promotes IFN-γ production (3) and upregulates T-bet, the lineage-defining transcription factor of Th1 cells (4). In the case of Th17 cells, the cytokine requirements are more complex and appear to be partially different in humans and mice. In particular, the combination of TGF-β and IL-6 is sufficient to induce Th17 cells that produce high levels of IL-17 (5) and express the lineage-defining transcription factor ROR-γt (6) in mice, but not in humans (7–9). However, IL-1β and IL-23 have been shown to induce or expand Th17 cells both in humans (10, 11) and mice (12), but they also induce the maturation of Th17 cells and the acquisition of Th1-like properties (11, 12). These “alternative” or “pathogenic” Th17 cells coexpress ROR-γt and T-bet, coproduce IL-17 and IFN-γ, and drive experimental autoimmune diseases in mice (12, 13). They are distinct from “classical” Th17 cells, which are not pathogenic, and produce mainly IL-17 and IL-10 (14, 15). Moreover, the production of the cytokine GM-CSF by Th17 cells is important for their pathogenic function (12, 16, 17). Importantly, IL-17/IFN-γ coproducing T cells are also present in human autoimmune patients (18, 19), and single nucleotide polymorphisms in the IL-23R are strongly associated with inflammatory bowel and other autoimmune diseases (20–23). Interestingly, Th17 cells primed by different pathogens produce preferentially IL-10 or IFN-γ (24). Moreover, high salt concentrations have been shown to promote pathogenic Th17 cell generation both in humans and in mice (25). Nevertheless, the conditions that induce pathogenic Th17 cells in humans are incompletely defined.

IL-21 is a pleiotropic cytokine that has both proinflammatory (26) and anti-inflammatory activities (27). It is mainly produced by CD4⁺ T cells and is critical for CD4-mediated help for B cell and cytotoxic lymphocyte responses (28–31). In particular, IL-21 produced by specialized follicular Th (T_FH) cells in B cell follicles is the most potent cytokine to induce B cell differentiation to Ab-secreting plasma blasts (32). Moreover, IL-21 has a critical role to sustain antiviral cytotoxic T cell responses and to prevent chronic infections (33–35). IL-21 can be induced by IL-6 (36, 37) or IL-12 in humans (38, 39), and it promotes its own expression via STAT3 (37, 40). In the mouse, autocrine IL-21 signaling has been proposed to promote Th17 differentiation (36, 41–43), and consistently, human patients with defective IL-21R signaling lack Th17 cells (44). However, although IL-21 can promote autoimmune diseases such as systemic lupus erythematosus (45, 46) or type 1 diabetes (47) by its effect on B cells (48) and dendritic cells (DCs) (49), respectively, inconsistent results have been published on the role of IL-21 in Th17 cell differentiation and Th17-driven experimental autoimmune encephalomyelitis (41, 50, 51). Interestingly, the addition of exogenous TGF-β and IL-21 was reported to induce Th17 differentiation of human naive T cells (8). However, whether autocrine IL-21 signaling is also important for human Th17 cell differentiation and how IL-21 impacts on the generation of pathogenic human Th17 cells is unclear (15).

In this study, we analyzed the role of IL-21 in human CD4⁺ T cell differentiation. We found that IL-21 is highly produced by central memory T (T_CM) cells, and that it promoted T_CM generation, IL-10 secretion, and differentiation of Th17 cells. However, it inhibited Th1 differentiation and GM-CSF production, and consequently prevented the generation of pathogenic Th1/17 effector cells.

Buffy-coated blood from healthy donors and cord blood, as well as intestinal and tonsil specimens from patients, were obtained from the IRCCS Policlinico Ospedale Maggiore, Milan, Italy, and the Charité Hospital and the Deutsches Rotes Kreuz in Berlin, Germany. The ethical committee approved the use of intestinal and tonsillar specimens for research purposes (permission 2476 and EA1/107/10, respectively), and informed consent was obtained from patients. Human mononuclear cells from peripheral blood in the intestinal lamina propria and tonsils were isolated by Ficoll-Hypaque gradient (Sigma-Aldrich). Mononuclear cells from peripheral blood in the intestinal lamina propria were purified as previously described (37). CD4⁺ T cells were enriched with magnetic beads on an AutoMacs Pro cell separator (Miltenyi), and T cell subsets were sorted on a FACSAriaII (BD). CD45RA⁺ naive T cells were sorted as CD4⁺CD25⁻CD45RA⁺CCR6⁻ cells, whereas CD31⁺ naive T cells were sorted as CD4⁺CD25⁻CD45RA⁺CCR6⁻CXCR5⁻CD31⁺ cells. T_CM and effector memory T (T_EM) subsets were sorted as CD4⁺CD25⁻CD45RA⁻ according to CCR7 and CXCR5 expression, whereas tonsillar T_FH and non-T_FH cells were sorted as CD4⁺CD25⁻CD45RA⁻ cells according to CXCR5 and ICOS expression. CCR6⁺ memory cells were sorted as CD4⁺CD25⁻CD45RA⁻CCR6⁺CXCR3⁻ cells. Anti-CD4 Ab was purchased from Miltenyi; anti-CCR6, anti-CD31, and anti-CXCR3 Abs from BD; anti-CD25 and anti-ICOS Abs from eBioscience; anti-CXCR5 and anti-CCR7 Abs from R&D; and anti-CD45RA Ab was produced in-house. Purity of T cell populations was >97%.

CD1c⁺ myeloid DCs (mDCs) were isolated from peripheral blood mononuclear cells by magnetic enrichment followed by cell sorting as described previously (52). mDC1 purity was >95%.

In vitro stimulation of sorted T cells was performed with plate-bound anti-CD3 and anti-CD28 Abs (BD). Neutralizing Abs against IL-12, IFN-γ, and IL-4 were used at 2 μg/ml (BD). All recombinant cytokines (R&D) were added at 10 ng/ml, with the exception of TGF-β isoforms that were used at 1 ng/ml. Soluble IL-21R (sIL-21R; R&D) was added at 5 μg/ml, and neutralizing anti–IL-10 Abs (Miltenyi) at 10 μg/ml. Under neutral conditions, cells were stimulated in the presence of anti–IL-12, anti–IFN-γ, and anti–IL-4. Under Th1 conditions, cells were cultured for 6 d with IL-12 plus anti–IL-4, and under Th2 conditions with IL-4 and anti–IL-12 and anti–IFN-γ. Under Th17 conditions, rTGF-β1 or rTGF-β3, IL-1β, IL-6, and IL-23 were added in addition to the neutralizing Abs. Because IL-17 was induced late (data not shown), intracellular cytokine production under Th17 conditions was analyzed after 14 d. For DC priming, CD31⁺ naive T cells were incubated with matured allogenic DCs at a 1:5 ratio for 7 d.

For data acquisition, a FACS Canto II (BD) with DIVA software was used, whereas the analysis of the data was performed using the FlowJo software (Tree Star). The transcription factors T bet and Foxp3 were stained after 4 d of stimulation with PE- and allophycocyanin-conjugated Abs purchased from BD and eBioscience, respectively, after fixation and permeabilization with a fixation/permeabilization kit (eBioscience). For intracellular cytokine staining ex vivo or after in vitro activation, cells were cultured for 24–48 h in uncoated wells and were then stimulated for 5 h with phorbol-12, 13-dibutyrate, Ionomycin in the presence of brefeldin A (Sigma-Aldrich). Cells were fixed with 2% paraformaldehyde and permeabilized with 0.5% Saponin. Abs specific for cytokines were purchased from eBioscience (IL-17, GM-CSF) or BD (IL-21), whereas anti–IFN-γ was produced in-house. Chemokine receptors were stained after 4 d of stimulation with PE-conjugated anti-CCR6 (BD) or FITC-conjugated anti-CCR7 Abs (R&D). To assess cellular proliferation, we stained cells with cell tracer or CFSE, and fluorescence intensity was performed after 4 d of stimulation.

Culture supernatants for the analysis of cytokine concentrations of cells stimulated with anti-CD3 and anti-CD28 Abs were collected at day 4 for IL-10 and on day 12 for IL-17. ELISAs were performed following the manufacturers’ protocols (IL-10 ELISA [BD], IL-17A ELISA [eBioscience]).

Because we observed unspecific staining by commercially available anti–IL-23R Abs (data not shown), we assessed the expression of functional IL-23R by IL-23–induced STAT3 phosphorylation: after 4 d of stimulation with anti-CD3 and anti-CD28 Abs, cells were transferred to uncoated wells for additional 48 h. Cells were washed with fresh culture medium without cytokines, and the washing step was repeated twice every 2 h. Cells were then incubated in culture medium with or without 10 ng/ml IL-23 for 30 min. For staining of p-STAT3, cells were fixed with Cytofix Fixation Buffer (BD) followed by permeabilization with Permbuffer III (BD). Thereafter, p-STAT3 was stained with allophycocyanin-conjugated anti–p-STAT3 (pY705; BD). p-STAT3 in the absence of IL-23 was undetectable under these conditions and used as the negative control. To measure IL-23R or RORC mRNA levels, we isolated total RNA using the mirVana Kit (Ambion). mRNA expression levels were assessed by TaqMan Gene expression assay (Applied Biosystems) and normalized on 18s rRNA.

Lentiviral particles were produced according to the standard protocol indicated in the SBI User Manual instructions. Naive CD4⁺ T cells were activated as described earlier and simultaneously transduced with either pLV-ctrl vector, pLV-T-bet, or pLV-RORC2. Cells were detached on day 4 and cultured in uncoated wells in IL-2 and analyzed for cytokine production at days 10–14. Cytokine production was analyzed on gated RORC2⁺ and T-bet⁺ cells, and compared with Mock-transduced cells.

Statistical analysis was performed with Prism 5 software (Graph Pad), and Student t tests for paired and unpaired samples were used to evaluate differences between groups of variables. Statistical significance was set at *p < 0.05, **p < 0.005, and ***p < 0.0005. Asterisk on top of columns indicates significance as compared with the neutral condition.

IL-21 is a helper cytokine that is mainly produced by Ag-experienced CD4⁺ T cell subsets. In human peripheral blood, CCR7⁺CD45RA⁻ T_CM cells recirculate through secondary lymphoid organs and possess low levels of effector functions, whereas CCR7⁻CD45RA⁻ T_EM cells are recruited preferentially to nonlymphoid tissues and produce higher levels of effector cytokines (53). Notably, however, T_CM cells produced significantly higher amounts of IL-21 than T_EM cells (Fig. 1A), whereas T_EM cells produced higher levels of IFN-γ as expected. In some donors, CXCR5⁺ T_CM secreted higher levels of IL-21 than CXCR5⁻ T_CM, but this difference was not consistent and did not reach statistical significance. CXCR5 is characteristic for T_FH cells that are generated in inflamed secondary lymphoid organs. In human tonsils, CD4⁺ T cells secreted indeed higher levels of IL-21 as compared with CD4⁺ T cells from peripheral blood or the inflamed gut of IBD patients (Fig. 1B), but both CXCR5⁺ICOS⁺ T_FH cells and CD45RA⁻CXCR5⁻ICOS⁻ non-T_FH cells secreted high levels of IL-21 (data not shown).

Although IL-21 was highly produced by T_CM cells, it was nevertheless coproduced with IFN-γ, IL-4, and in particular with IL-17 in CD4⁺ T cells from peripheral blood and tonsils (Fig. 1C), consistent with the view that IL-21 production is not limited to a particular T cell differentiation lineage. Because the transcription factors T-bet and RORC2 are required to induce, respectively, Th1 and Th17 effector cell differentiation, we analyzed IL-21 production in CD4⁺ T cells that were forced to express one of these lineage-defining transcription factors. Strikingly, both T-bet and RORC2 inhibited IL-21 production (T-bet: 52 ± 6%, RORC2: 64 ± 5% inhibition), whereas they promoted as expected IFN-γ and IL-17, respectively (Fig. 1D). These findings show that IL-21 is produced by various CD4⁺ T cell lineages, but is most prominent in T_CM cells and downregulated upon effector cell differentiation.

Because Th1 and Th17 cells produced IL-21, we analyzed the effects of Th1- and Th17-promoting cytokines on IL-21 production in TCR-stimulated naive T cells. In CD45RA⁺ naive T cells from cord blood, both Th1 and Th17 conditions induced IL-21 production as expected (Fig. 2A). However, IL-21 induction under Th17 conditions was not statistically significant in CD4⁺CD45RA⁺ T cells from adults (data not shown), because in many donors, TCR-activated CD45RA⁺ T cells produced IL-21 in the absence of exogenous cytokines (Fig. 2B). The spontaneous IL-21 production in CD45RA⁺ T cells from adults could be explained by the presence of CD45RA⁺ memory T cells (54, 55); in particular, a small subset of CXCR5⁺CD45RA⁺ T cells that were largely CD31⁻ (data not shown) secreted IL-21 ex vivo (Fig. 2C). However, when naive T cells from adults were purified as CXCR5⁻CD45RA⁺CD31⁺ cells (referred to as CD31⁺ for the rest of this report), spontaneous IL-21 production was strongly reduced, and Th17-promoting cytokines induced significant amounts of IL-21 (Fig. 2D). Among Th17-promoting cytokines, IL-6, IL-21 itself, and to a lesser extent IL-1β had a positive effect, whereas TGF-β was inhibitory (Fig. 2A, 2D, data not shown). Interestingly, IL-12 induced higher levels of IL-21 in CD31⁺ naive T cells from adults as compared with neonates (Fig. 2A, 2D). Moreover, CD1c⁺ mDCs, which require both LPS and R848 stimulation to produce bioactive IL-12 (52), also induced high levels of IL-21 upon T cell priming in a partially IL-12–dependent manner (Fig. 2E). In summary, IL-12 and IL-6 induce IL-21 in naive T cells under Th1- and Th17-promoting conditions, respectively, but the presence of IL-21–producing CD45RA⁺ memory T cells in adults masks IL-21 production induced under Th17 conditions.

To assess the effect of IL-21 on human T cell differentiation, we first added exogenous IL-21 to TCR-stimulated naive T cells, and analyzed proliferation and differentiation in the absence of other exogenous cytokines. IL-21 strongly inhibited the downregulation of CCR7 on proliferating T cells (Fig. 3A), and it also inhibited T-bet expression (Fig. 3B). Because IL-12 induced high levels of IL-21, we then analyzed the effects of autocrine IL-21 signaling on IL-12–induced Th1 differentiation by blocking IL-21 with sIL-21R. Unexpectedly, IL-21 neutralization inhibited neither anti-CD3/28 nor mDC-induced T cell proliferation (Fig. 3C). However, it significantly enhanced IL-12–induced IFN-γ production (Fig. 3D), suggesting that IL-21 inhibits Th1 differentiation. Consistently, although IL-21 was efficiently induced under Th1 conditions, addition of rIL-21 had an additional weak inhibitory effect on IFN-γ production (data not shown). Moreover, IL-21 neutralization promoted GM-CSF production under Th1 conditions, whereas exogenous IL-21 inhibited GM-CSF production under neutral conditions (Fig. 3E). Notably, IL-21 neutralization promoted IFN-γ and GM-CSF production not only upon naive T cell priming with anti-CD3 and anti-CD28 Abs, but also with mDCs (Fig. 3D, 3E). In contrast with IFN-γ and GM-CSF, production of IL-10 was promoted by IL-21 and inhibited by IL-21 neutralization under Th1 conditions (Fig. 3F). Altogether, these results show that IL-21 is induced by IL-12, but inhibits the priming of Th1 effector cells in a negative feedback loop and promotes IL-10 production.

Exogenous IL-21 promotes Th17 differentiation, but the relevance of autocrine IL-21 signaling in human Th17 cell differentiation has not been addressed. We analyzed TCR-stimulated CD31⁺ naive T cells in the presence of the Th17-promoting cytokines TGF-β1, IL-6, IL-1β, and IL-23 for the upregulation of key molecules of Th17 cells in the absence or presence of sIL-21R. Under these Th17 conditions, IL-21 neutralization significantly reduced the surface expression of CCR6, the characteristic chemokine receptor of Th17 cells (Fig. 4A). We also assessed IL-23–induced STAT3 phosphorylation to assess the expression of functional IL-23Rs at the single-cell level (Fig. 4B). The majority of CD31⁺ naive T cells acquired IL-23 responsiveness under Th17, but not under neutral conditions, and consistently IL-23R mRNA was detected under Th17, but not under neutral conditions (data not shown). Notably, blocking autocrine IL-21 signaling reduced IL-23–induced STAT3 phosphorylation (Fig. 4B). Moreover, the mRNA levels of RORC, the lineage-defining transcription factor of Th17 cells, were significantly reduced when IL-21 was neutralized under Th17 conditions (Fig. 4C). In addition, the Treg-specific transcription factor Foxp3, which inhibits Th17 differentiation (56), was promoted by TGF-β and inhibited by proinflammatory cytokines as expected, and this inhibitory effect was largely mediated by IL-21 (Fig. 4D). Importantly, also IL-17 secretion, but not proliferation, was reduced upon IL-21 neutralization under Th17 conditions (Fig. 4D), and IL-6 in the Th17 cytokine mixture could be replaced by exogenous IL-21 (data not shown). Finally, in CD31⁺ naive T cells, we could detect low levels of IL-17 by ELISA following sustained culture with a combination of TGF-β plus IL-21 or IL-6 (Fig. 4E). Altogether, these results indicate that IL-21 is one of the critical cytokines that promotes human Th17 differentiation. However, they also show that Th17-promoting cytokines are partially redundant and that Th17 cells can be consequently generated in the absence of IL-21 signaling.

Given the opposing effects of IL-21 on Th1 and Th17 differentiation, we wondered whether IL-21 would regulate the acquisition of Th1 characteristics by Th17 cells, which are typical for pathogenic effector cells. Indeed, IL-21 neutralization promoted IFN-γ (Fig. 5A) and GM-CSF production (Fig. 5B) under Th17 conditions, but inhibited IL-10 secretion (Fig. 5C). IL-10 is known to inhibit Th1 responses, but the inhibitory effects of IL-21 on GM-CSF and IFN-γ production were not IL-10 mediated, because IL-10 neutralization had no effect on the production of GM-CSF (Fig. 5D) or IFN-γ (data not shown).

We then compared the generation of IL-17 and IFN-γ coproducing cells from CD31⁺ naive T cells under different conditions that were reported to promote the generation of pathogenic Th17 cells, that is, in the absence of TGF-β, with TGF-β3 or under high salt conditions (25, 57, 58). Under standard Th17 conditions with TGF-β1 or TGF-β3, few IL-17/IFN-γ double-producing cells could be detected (Fig. 5E). In the absence of TGF-β, IL-17⁺IFN-γ⁺ cells were selectively increased, but they were still rare. Conversely, high-salt Th17 conditions (25) induced overall high levels of IL-17–producing cells including a distinct minor population of IL-17⁺IFN-γ⁺ cells (Fig. 5E). Strikingly, although IL-17–producing T cells generated under standard Th17 conditions had only low IFN-γ– and GM-CSF–producing capacities, relevant fractions of IL-17⁺ Th17 cells coproduced IFN-γ or GM-CSF when IL-21 was neutralized, in particular, under high-salt Th17 conditions (Fig. 5F). Notably, polyfunctional IL-17⁺IFN-γ⁺GM-CSF⁺ effector cells were quite selectively generated when IL-21 signaling was blocked under both standard and high-salt Th17 conditions. In summary, IL-21 promotes the differentiation of Th17 cells that produce IL-10, but inhibits the generation of Th1/17 effector cells that coproduce IFN-γ and GM-CSF.

The results so far were obtained with uncommitted CD31⁺ naive T cells that lack IL-17–producing capacities. We therefore asked whether IL-21 had similar effects on precommitted Th17 cells that might be particularly relevant in established autoimmune disorders. For this aim, we isolated CCR6⁺ memory T cells, excluding IFN-γ–producing CXCR3⁺ Th1 memory cells, stimulated them with IL-1β and IL-23 to induce IL-17⁺IFN-γ⁺ cells (Fig. 6A), and assessed the effects of IL-21 neutralization. We also stimulated CCR6⁺ memory T cells under standard Th17 conditions that enhanced IL-17, but not IFN-γ production (Fig. 6B, 6C). In contrast with naive T cells, the inhibition of autocrine IL-21 signaling had no effect on IL-17 and IFN-γ production by CCR6⁺ memory T cells. Surprisingly, both IL-1β and IL-23 and Th17 conditions inhibited GM-CSF expression, but blocking IL-21 significantly enhanced GM-CSF production (Fig. 6D). Interestingly, IL-21 neutralization also significantly enhanced T-bet upregulation induced by IL-1β and IL-23 (Fig. 6E). Thus, IL-21 is not required for IL-17 or IFN-γ upregulation in precommitted Th17 cells, but it significantly inhibits GM-CSF production and T-bet upregulation.

IL-21 is a pleiotropic cytokine produced by Th cells and is a crucial regulator of lymphocyte differentiation. We showed in this study that IL-21 is associated with CD4⁺ T_CM cells and promoted the differentiation of classical Th17 cells, but inhibited the generation of pathogenic Th1/17 effector cells.

We report in this article the unexpected finding that IL-21 is produced at very high levels by CD4⁺ T_CM cells, whereas we could not detect an association with CXCR5 expression as previously proposed. Because conventional effector cytokines like IFN-γ, IL-4, IL-5, and IL-17 are produced at higher levels by T_EM cells (59, 60), IL-21 is, to our knowledge, the first effector cytokine that is specifically associated with T_CM cells. Importantly, IL-21 inhibited TCR-induced differentiation to CCR7⁻ Th1 effector cells, suggesting that it arrests CD4⁺ T cells at the immature T_CM stage of differentiation. A negative effect of IL-21 on Th1 cell priming has also been observed in mice (61–63), whereas apparently conflicting results of other reports (64, 65) might be explained by IL-21–dependent expansion of pre-existing Th1 cells. IL-21 was also shown to inhibit the differentiation of murine CD8⁺ T cells to effector cells (48, 66), and the inhibition of excessive CD8⁺ T cell differentiation leading to functional exhaustion is important for the protective role of IL-21 in chronic viral infections (33, 34). Altogether, these findings suggest that CD4⁺ T_CM-derived IL-21 preserves the T_CM stage of CD4⁺ and CD8⁺ T cells, and inhibits excessive T cell differentiation. Consistent with the view that IL-21 is not a conventional Th1/17 effector cytokine, it was inhibited by the transcription factors T-bet and RORC2 that are critical transcriptional regulators of Th1 (4) and Th17 effector cells (6), respectively. Nevertheless, we confirmed and extended previous data that IL-6 (36, 37, 41) and IL-12 (38, 39) induce IL-21 under Th1 and Th17 priming conditions, respectively. These apparently contradictory findings could be explained by a modified progressive differentiation model (53), where IL-21 is acquired early by developing Th1 and Th17 cells by cytokines, and maintained by T_CM cells to inhibit excessive effector cell differentiation. Upon progressive differentiation of pre-Th1 and pre-Th17 cells to T-bet or ROR-expressing effector cells, IL-21 is downregulated, and Th1 effector cytokines like IFN-γ and GM-CSF are consequently upregulated.

The cytokine requirements of human Th17 cell differentiation have been the subject of intense research and debates (67). Although there are still some controversies about the contribution of individual cytokines, it is now accepted by many groups that IL-6, IL-1β, IL-23, and low amounts of TGF-β contribute to human Th17 differentiation (68). A previous report showed that IL-21 and TGF-β were sufficient to induce IL-17 in human naive T cells (8). Consistently, by ELISA, we were able to detect low IL-17 production by TGF-β and IL-21 (8) or IL-6 (5) in CD31⁺ naive T cells. However, optimal IL-17 production required the combination of TGF-β, IL-1β, IL-6, IL-23, and autocrine IL-21 signaling. These findings clearly show that IL-21 contributes to human Th17 cell differentiation, consistent with earlier mouse studies (36, 41–43) and the paucity of Th17 cells in IL-21R–deficient patients (44). Notably, however, IL-21 did not affect naive T cell proliferation and was not absolutely required for Th17 differentiation, possibly explaining inconsistent reports on the role of IL-21 in Th17 differentiation in mice (41, 42, 50, 51). Moreover, in Th17 memory cells, IL-17 upregulation was completely IL-21 independent, indicating that the cytokine requirements of uncommitted and precommitted Th17 cells are partially different.

The generation of pathogenic Th17 cells, which are characterized by IFN-γ, GM-CSF, and T-bet expression and low IL-10 production, is a highly relevant issue in autoimmune research, but it is poorly understood in humans. In mice, IL-1β and IL-23 have a critical role (13, 69, 70), and absence of TGF-β1 and/or presence of TGF-β3, as well as high salt concentrations, have been shown to induce pathogenic Th17 cells (25, 57, 58). In humans, high salt conditions have also been proposed to induce pathogenic T cells (25), and IL-1β and IL-23 induce IL-17/IFN-γ coproducing T cells from precommitted precursors (11). However, the role of IL-21 in the induction of classical versus pathogenic Th17 cells (15) has not been addressed in humans. We found that IL-21 promoted the differentiation of Th17 cells and IL-10 production, but also enhanced IL-23R responsiveness that is important for the induction of pathogenic Th17 cells. Nevertheless, despite enhanced IL-23 signaling, IL-21 inhibited the priming of polyfunctional Th1/17 effector cells that coproduced IL-17, IFN-γ, and GM-CSF. Notably, although IL-10 could inhibit GM-CSF production, neutralization experiments indicated that the inhibitory effect of IL-21 on GM-CSF production was not IL-10 mediated under the studied conditions. These findings obviously do not rule out that IL-21–induced IL-10 could contribute to limit GM-CSF production by Th17 cells in vivo. Because IL-21 also promotes IL-10 (27) and can inhibit IFN-γ (62) and GM-CSF (63) in mice, our findings might explain why mice develop experimental autoimmune encephalomyelitis in the absence of IL-21 signaling (50, 51). However, there are also some relevant differences between humans and mice in the cytokine requirements for the generation of pathogenic Th17 cells, because in humans, TGF-β3 was ineffective (58) and IL-1β plus IL-23 enhanced IFN-γ and T-bet expression in Th17 cells, but inhibited GM-CSF production (12, 16). Notably, IL-21 is a pleiotropic cytokine and has been reported to have both proinflammatory and anti-inflammatory effects in autoimmune diseases (48, 71). In precommitted human Th17 cells, which are presumably highly relevant in autoimmune diseases, IL-21 had no effect on IL-17 and IFN-γ upregulation, but it inhibited GM-CSF and T-bet, two key features of pathogenic Th17 cells (12, 16, 58). In the light of these findings, it seems uncertain whether autoimmune patients with established, pathogenic Th17 cell responses will benefit from anti–IL-21 therapy, unless pathogenesis is largely driven by autoreactive B cells and Abs.

In summary, IL-21 is a T_CM-associated cytokine that arrests T cells at the T_CM differentiation stage. Moreover, IL-21 signaling promotes IL-10, inhibits generation of Th1 effector cells, and consequently inhibits the generation of potentially pathogenic human Th1/17 cells. These findings have relevant implications for the treatment of autoimmune patients with anti–IL-21 therapy.

We thank Monika Killig and Merlin Lüttke-Eversloh of the Deutsches Rheumaforschungszentrum Berlin for providing tonsil samples. We thank Dr. Jan Siedentopf of the Department of Gynecology, Charité University Medical School Berlin for help collecting human cord blood samples. We thank Dr. Toralf Kaiser and Jennifer Kirsch for cell sorting.

The authors have no financial conflicts of interest.