Introduction
Since the initial description of T helper type 1 (TH1) and TH2 cells1, cytokines have seemed increasingly important for the induction, regulation and function of distinct T helper subsets. In the TH1-TH2 paradigm, single cytokines such as interleukin 12 (IL-12) or IL-4 induce the differentiation of TH1 or TH2 cells, respectively2. Other parameters, such as the dose of antigen or type of costimulation, are able to modulate TH1 or TH2 responses2.
Another subset of T helper cells that produce IL-17 (TH-17 cells) has been identified as being distinct from TH1 and TH2 cells3, 4. TH-17 cells have specific functions in antimicrobial immunity5, 6 and autoimmune inflammation7, 8, 9. In mice, many cytokines are required and act in a coordinated way to induce TH-17 differentiation, with a critical function for transforming growth factor-
(TGF-; A002271) and IL-6 (refs. 8,10,11). IL-6 induces IL-21 production, which subsequently favors TH-17 differentiation in an autocrine way12, 13, 14. Mouse TH-17 cells produce not only IL-17 but also IL-21 (ref. 15), IL-22 (ref. 16) and, in some cases, IL-10 (ref. 17). It is unclear at present whether TH-17 cells can produce additional T helper cytokines and to what extent the requirements for induction of these TH-17-associated cytokines are similar. Understanding the regulation of the global TH-17 cytokine profile is essential, as each T helper cytokine has specific functions.
Characterizing the factors driving human TH-17 differentiation is of particular interest because of the importance of TH-17 cells in health and disease. Five independent reports have addressed this issue with unexpectedly contradictory results. Three studies showed IL-118, IL-23 (ref. 19) or polyclonal stimulation with antibody to CD3 (anti-CD3) and anti-CD28 (ref. 20) to be sufficient for the generation of human TH-17 cells, in contrast to the many factors required in mice. Two other studies were not able to differentiate naive CD4 T cells into TH-17 cells, even with conditions shown to be efficient in mouse or human systems21, 22. Finally, TGF-, which has been shown to be essential for mouse TH-17 differentiation, has been reported as a negative regulator in humans18, 19. Thus, the requirements for human TH-17 differentiation remain controversial23. Here we show that TGF-, IL-23 and proinflammatory cytokines (IL-1 and IL-6) were essential components of human TH-17 differentiation and expression of IL-17A, IL-17F, the IL-23 receptor (IL-23R) and the transcription factor ROR
t. However, experimental and computational methods showed that each TH-17-promoting cytokine had a specific function in the regulation of the global TH-17 cytokine profile.
Results
Driving IL-17 production
To define the cytokine requirements for the induction of human TH-17 differentiation, we did a standard naive CD4 T cell differentiation assay in the presence of polyclonal stimulation with anti-CD3 and anti-CD28. We systematically tested all cytokines shown to be involved in the polarization of IL-17-producing cells in mouse systems. In the first set of experiments, we independently considered TGF-, IL-23 and proinflammatory cytokines (IL-1, IL-6 and tumor necrosis factor (TNF); Fig. 1a). None of these components individually was sufficient to induce detectable IL-17 (Fig. 1a). A combination of these three components, however, induced high production of IL-17 (Fig. 1a). IL-17 dropped to undetectable amounts in the absence of proinflammatory cytokines and decreased by 70% in the absence of TGF- or IL-23 (Fig. 1a), which indicated that the three components were required for optimal IL-17 production. We obtained similar results with CD3+CD4+CD45RA+ peripheral blood naive T cells obtained by negative or positive selection and with total CD4+ T cells from cord blood (data not shown).
Figure 1: TGF-
, IL-23 and proinflammatory cytokines are required for the differentiation of human CD4+ TH-17 cells.
![Figure 1 : TGF-|[beta]|, IL-23 and proinflammatory cytokines are required for the differentiation of human CD4+ TH-17 cells.](/ni/journal/v9/n6/thumbs/ni.1613-F1.jpg)
(a,b) ELISA of IL-17 in supernatants of naive T cells differentiated for 5 d in the presence of anti-CD3 plus anti-CD28 and various combinations of TGF-, IL-23 and proinflammatory cytokines (Infl; IL-1, IL-6 and TNF), measured after 24 h of restimulation with anti-CD3 plus anti-CD28. <10, concentration below 10 pg/ml. *, P < 0.01; **, P < 0.001 (Wilcoxon test). Data (mean and s.e.m.) are representative of eight independent experiments. (c) Flow cytometry of intracellular IL-17 and IFN- in naive T cells stimulated in presence of anti-CD3 plus anti-CD28 and various cytokines (above plots), assessed after 6 h of restimulation with anti-CD3 plus anti-CD28. TH0, no polarizing cytokines; TH1, IL-12; Infl, IL-1 and IL-6 and TNF. Numbers in quadrants indicate percent IL-17+IFN--
cells (top left), IL-17+IFN-+ cells (top right) or IL-17-
IFN-+ cells (bottom right). Data are from one experiment representative of five independent experiments.
In a second set of experiments, we addressed the function of individual proinflammatory cytokines. In the presence of TGF- and IL-23, the removal of TNF had only a small effect on IL-17 production (Fig. 1a). We obtained similar results in the presence or absence of IL-1 and/or IL-6, which confirmed that TNF does not have a substantial effect on IL-17 production (Supplementary Fig. 1 online), contrary to what has been reported for mice11. The removal of IL-1 or IL-6 induced a comparably substantial decrease (over 50%; Fig. 1b). Any combination of one, two or three of the proinflammatory cytokines was not sufficient to induce detectable IL-17 production in the absence of TGF- and IL-23 (Fig. 1b). T cell population expansion on day 5 of culture was similar in all cytokine combinations, which indicated that differences in cytokine production could not be attributed to insufficient expansion (Supplementary Fig. 2a,b online).
Intracellular cytokine staining confirmed that TGF-, IL-23 and proinflammatory cytokines induced a well defined IL-17-producing cell population (Fig. 1c), which dropped by 70% in the absence of TGF- or IL-23. That paralleled the data obtained by enzyme-linked immunosorbent assay (ELISA). The decrease in IL-17 was accompanied by an increased cell population producing interferon- (IFN-), up to threefold in the absence of TGF- (Fig. 1c). IL-17-producing cells did not make IFN-; this distinguished them from TH1 cells, which are generated in the presence of IL-12. We detected no IL-17-producing cells among unpolarized T cells (TH0 cells) or in TH1 conditions (Fig. 1c), which again confirmed the ELISA data (Fig. 1a). Using the frequency of the IL-17-producing population, we calculated an average production of 0.006 pg IL-17 per cell. That is similar to the amount of IL-17 produced by in vitro–differentiated or ex vivo memory TH-17 cells in other human studies (range, 0.001–0.016 pg/cell)19, 22.
Many endogenous factors present in the T cell cultures could have possibly altered the cytokine requirements for IL-17 production, and it was important to clarify their function in our system. First, we addressed the effect of serum TGF-, as we used medium containing 10% fetal calf serum for our experiments. When we added a TGF--blocking monoclonal antibody to the complete TH-17 combination, IL-17 decreased considerably, consistent with the inhibition of exogenous TGF- (Supplementary Fig. 3a online). In the absence of exogenous TGF-, we noted a residual small amount of IL-17, which was not significantly affected by monoclonal antibody to TGF- (P = 0.5; Supplementary Fig. 3a). This indicated that endogenous (serum) TGF- had only a marginal function in our system. We also did TGF- 'titration', which confirmed that TGF- acted positively to regulate IL-17 production in a dose-dependent way (Supplementary Fig. 3b). Second, we addressed the function of serum itself, which may affect TH-17 differentiation independently of TGF-. Although absolute IL-17 production was higher in the absence of serum than in the presence of serum, we found that the cytokine requirements and regulation of IL-17 production were similar in these two types of culture media (Supplementary Fig. 4 online); TGF-, IL-23 and proinflammatory cytokines were all required for IL-17 production. We also addressed the function of endogenous IL-4 and IFN-, two cytokines described in the mouse to inhibit TH-17 differentiation4. We repeated the same type of experiment in the presence of monoclonal antibodies blocking IL-4 and IFN- and found that the cytokine requirements to induce IL-17 production were not affected (Supplementary Fig. 4). In summary, none of endogenous factors tested modified the cytokine requirements for the induction of IL-17 production, and TGF- invariably had a critical function independently of the experimental system.
IL-17 production is associated with typical TH-17 features
To address whether IL-17 production was associated with the acquisition of typical features of TH-17 cells, we first measured mRNA encoding various IL-17 family members. Optimal TH-17 conditions (TGF-, IL-23 and proinflammatory cytokines) induced the most mRNA encoding IL-17A and IL-17F (Fig. 2a), two cytokines associated with TH-17 differentiation24. However, we did not detect measurable amounts of other IL-17 family members, such as IL-17E (IL-25), which is related more to TH2 responses25 (data not shown). TGF- and proinflammatory cytokines were also required for the induction of IL-23R mRNA (Fig. 2a), another important characteristic of TH-17 cells10, 26.
Figure 2: TGF-, IL-23 and proinflammatory cytokines induce typical TH-17 features.
![Figure 2 : TGF-|[beta]|, IL-23 and proinflammatory cytokines induce typical TH-17 features.](/ni/journal/v9/n6/thumbs/ni.1613-F2.jpg)
(a) RT-PCR analysis of the expression of RORc, IL-17A, IL-17F and IL-23R mRNA in naive CD4+ T cells differentiated in vitro for 5 d in presence of anti-CD3 plus anti-CD28 and various cytokine combinations (below graphs), followed by 24 h of restimulation with anti-CD3 plus anti-CD28. Cycling threshold values are normalized to those of mRNA encoding ribosomal protein L34, and data are normalized to the maximum value obtained for each donor. Data are the mean and s.e.m. of three independent experiments. (b) IL-17A transcript and IL-17 protein from cells cultured in the presence of TH-17-inducing cytokines (IL-1, IL-6, TNF, TGF- and IL-23) or in the absence of individual components of that group, correlated to RORc transcripts with the Pearson correlation (R, correlation coefficient). Data are representative of six experiments. (c) RT-PCR of IL-17A and IL-17F mRNA and ELISA of IL-17 protein in naive T cells cultured in serum-free medium and infected with various dilutions (below graphs) of supernatants containing shRNA-expressing lentiviral vector specific for RORc (shRORc) or empty vector (Mock; negative control) during the first day of TH-17 differentiation, then washed extensively and cultured for additional 5 d in TH0 or TH-17 conditions and analyzed after 24 h of restimulation with anti-CD3 plus anti-CD28. TH0, no polarizing cytokines; TH-17, IL-1, IL-6, TNF, TGF- and IL-23. Data are the mean and s.e.m. of three independent experiments.
The transcription factor RORt has been shown to be critical for mouse TH-17 differentiation26. To assess the relationship between human RORt expression and IL-17 production, we measured the transcription of RORC, which encodes the human ortholog of mouse RORt. We quantified RORc mRNA in many optimal and suboptimal TH-17-polarizing conditions, similar to those used above (Fig. 1a,b). We noted a strong correlation between the amount of RORc transcript and IL-17 transcript or protein (Fig. 2b), which indicated that RORc expression was both a sensitive and specific marker of human TH-17 cells and suggested that RORt could be involved in regulating the production of human IL-17. IL-17F expression was less associated with RORc than was IL-17A expression (Supplementary Fig. 5 online), which confirmed a published result obtained with mice27. To directly address the function of RORt in controlling IL-17 production, we used short-hairpin RNA (shRNA) to 'knock down' RORc expression. RORc-specific shRNA but not control shRNA induced a decrease of about 50% in RORc mRNA expression (Supplementary Fig. 6 online); this decrease was sufficient to inhibit IL-17A mRNA and protein at all concentrations of shRNA tested (Fig. 2c). The effect on IL-17F expression was weaker but was dose dependent (Fig. 2c). Expression of the 'housekeeping' genes RPL34 (encoding ribosomal protein L34) and HPRT (encoding hypoxanthine guanine phosphoribosyl transferase) was not affected by RORc-specific or control shRNA (data not shown); expression of other genes not reported before to depend on RORc, such as those encoding IFN-, TNF and IL-23R, was also not affected by shRNA treatment, which indicated that the inhibition of IL-17 was specific (Supplementary Fig. 6). Overall, our data show that a combination of TGF-, IL-23 and proinflammatory cytokines was both necessary and sufficient to induce typical features of TH-17 differentiation.
Cytokine profiles of TH1, TH2 and TH-17 cells
We then defined the detailed TH-17 cytokine profile relative to that of standard TH0, TH1 and TH2 subsets. TH1 differentiation was driven by IL-12, TH2 differentiation was driven by IL-4, and TH0 differentiation was driven by polyclonal stimulation (anti-CD3 plus anti-CD28) in the absence of any polarizing cytokines (Fig. 3). We did not use blocking monoclonal antibodies in any of these conditions to avoid interfering with potential autocrine loops. We also assessed the expression of transcription factors associated with each of the T helper cell subsets. As expected, TH1 and TH2 conditions induced the highest expression of the transcription factors T-bet and GATA-3, respectively, and RORc was highly specific for TH-17 cells (Fig. 3a), which confirmed data obtained with mice26. We measured the expression of ten T helper cell–associated cytokines for the TH0, TH1, TH2 and TH-17 subsets (Fig. 3b). IL-17 and IL-6 were the most specific for TH-17 and were either absent or produced in very small amounts in TH1 and TH2 conditions; a second set of cytokines, IL-21, IL-22, TNF and IFN-, could be detected in TH-17 and TH1 conditions and, notably, IL-21 and IL-22 had similar or higher expression in TH1 conditions versus TH-17 conditions; and a third set of cytokines, IL-10, IL-4, IL-5 and IL-13, was produced mainly in TH2 conditions. Thus, TH-17 conditions induced the production of many cytokines in addition to IL-17.
Figure 3: The TH-17 cytokine profile has specific features but also features that overlap with those of other T helper cell–polarizing conditions.
![Figure 3 : The TH-17 cytokine profile has specific features but also features that overlap with those of other T helper cell|[ndash]|polarizing conditions.](/ni/journal/v9/n6/thumbs/ni.1613-F3.jpg)
(a) RT-PCR analysis of the expression of T-bet, RORc and GATA-3 mRNA in naive T cells differentiated with anti-CD3 plus anti-CD28 in TH0, TH1, TH2 or TH-17 conditions and then restimulated for 24 h with anti-CD3 plus anti-CD28. Cycling threshold values are normalized to those of L34. Data are the mean and s.e.m. of three independent experiments. (b) Cytometric bead assay or ELISA of IL-17, IL-21, IL-22, IL-4, IL-5, IL-6, IL-10, IL-13, IFN- and TNF in culture supernatants of naive T cells differentiated in TH0, TH1, TH2 or TH-17 conditions and then restimulated for 24 h with anti-CD3 plus anti-CD28. TH0, no polarizing cytokines; TH1, IL-12; TH2, IL-4; TH-17, IL-1, IL-6, TNF, TGF- and IL-23. Data are the mean and s.e.m. of eight independent experiments.
Differential regulation of individual TH-17 cell–derived cytokines
How IL-17-promoting cytokines regulate the production of diverse TH-17-associated cytokines is not known. To address that issue, we measured T helper cell cytokines in the presence or absence of individual TH-17-promoting cytokines. As shown before (Fig. 1a,b), removal of any of the five TH-17-promoting cytokines, except for TNF, decreased the production of IL-17 by over 50% (Fig. 4). Notably, each TH-17-associated cytokine was regulated in a specific way: IL-22 was generally stable, even in the absence of critical TH-17-inducing cytokines such as IL-1, IL-6 and IL-23, and the removal of TGF- induced a significant increase in IL-22 production, which indicated that IL-17 and IL-22 were differentially regulated by TGF- (Fig. 4). IL-21, shown before to be an important autocrine factor in the induction of mouse TH-17 differentiation12, 13, was specifically dependent on IL-23 (Fig. 4 and Supplementary Fig. 7 online). Notably, although TH-17 conditions induced only low production of IL-10, this production was higher in the absence of IL-1 (Fig. 4). We obtained an opposite result with IFN-, which indicated that IL-1 differentially regulated IL-10 and IFN-. Finally, IL-6 production was mostly dependent on TGF- and IL-23, a regulation that is more closely related to that of IL-17. Overall, each TH-17-associated cytokine was regulated in a specific way. This suggested that individual TH-17-promoting cytokines might not only control the amount of IL-17 produced but also modulate quantitatively and qualitatively the global T helper cytokine profile, potentially inducing a shift in the type of T cell response.
Figure 4: TH-17 cell–derived cytokines are differentially regulated by TH-17-promoting cytokines.
![Figure 4 : TH-17 cell|[ndash]|derived cytokines are differentially regulated by TH-17-promoting cytokines.](/ni/journal/v9/n6/thumbs/ni.1613-F4.jpg)
ELISA and/or cytometric bead assay of IL-17, IL-6, IL-21, IL-22, IL-10 and IFN- in supernatants of naive T cells differentiated with anti-CD3 plus anti-CD28 in the presence of IL-1, IL-6, TNF, TGF- and IL-23 (TH-17 conditions) or in the absence of individual components of that group (below graphs) and then restimulated for 24 h with anti-CD3 plus anti-CD28. *, P < 0.05; **, P < 0.01 (Wilcoxon test). Data are the mean and s.e.m. of seven independent experiments.
TH-17 profile modulated by cytokines
To test the hypothesis that TH-17-promoting cytokines might drive or modulate the global T helper cytokine profile, we measured all ten T helper cytokines in control T helper cell conditions (TH0, TH1 and TH2), in 'optimal' TH-17 conditions and after the removal of individual TH-17-promoting cytokines (Fig. 5a). We obtained complete data sets (ten cytokines in nine polarizing conditions) from six independent experiments. To allow for comparison among profiles, we normalized values obtained for each cytokine to the maximum value obtained for that cytokine across the whole data set for each donor (Fig. 5a). Three cytokine sets characterized each profile: IL-10, IL-4, IL-5 and IL-13 were highest in TH2 conditions; IFN- and IL-22 were highest in TH1 conditions; and IL-17, IL-6 and TNF were highest in TH-17 conditions. IL-21 was similarly high in TH1 and TH-17 conditions (Fig. 5a). Overall, the TH-17 profile was distinct from those of TH1 and TH2. The analysis of cytokine profiles generated in the absence of each TH-17-promoting factor identified a notable diversity. It became apparent that each of these cytokines not only controlled IL-17 production but induced substantial changes in the global TH-17 cytokine profile. For example, removing IL-1 decreased IL-17 and IFN- and increased IL-10; removing TGF- induced a decrease in the entire TH-17 'cluster' and an increase in the TH1 'cluster'; and removing IL-23 decreased the TH-17 sets without any substantial change in the other cytokines (Fig. 5a).
Figure 5: IL-23 and proinflammatory cytokines induce a TH1-like profile that 'converts' to a TH-17 profile after the addition of TGF-.
![Figure 5 : IL-23 and proinflammatory cytokines induce a TH1-like profile that 'converts' to a TH-17 profile after the addition of TGF-|[beta]|.](/ni/journal/v9/n6/thumbs/ni.1613-F5.jpg)
(a) ELISA and/or cytometric bead assay of cytokine production by differentiated T cells in TH1, TH2, TH-17 and suboptimal TH-17 conditions (absence of individual TH-17-promoting cytokines). Cytokine amounts are normalized to the maximum value obtained for that cytokine across the entire data set for each donor. Open bars, highest expression in TH2 conditions; filled bars, highest expression in TH1 conditions; light gray bars, highest expression in TH-17 conditions; dark gray bars, similarly high expression in TH1 and TH-17 conditions. Data are the mean and s.e.m. of six independent experiments. (b) Cluster analysis of the data in a by Pearson correlation distance. Culture conditions are separated into clusters by comparison of their linkage distance. Agglomerative coefficient, 0.85 (reflects data structure; values near 1 indicate well separated clusters); resampling similarity index, 0.093 (values near 0 indicate a robust cluster). (c) Principal component analysis of the data in a. T helper conditions (ovals) are positioned in a space defined by the principal components 1 and 2, which are the two 'best' axes representing the entire data set. Black ovals, TH0, TH1, TH2 and TH-17 profiles; gray ovals, removal of cytokines from the TH-17 conditions. Red arrows indicate T helper cell–derived cytokines that contribute to the differences among culture conditions. The direction and length of such vectors indicate the importance of each T helper cytokine in discriminating the T helper profiles (Supplementary Methods online).
Full size image (84 KB)We sought to determine whether the profiles generated in suboptimal TH-17 priming conditions (through the removal of individual TH-17-promoting cytokines) represented new T helper cell profiles or whether they were related to any of the standard TH1, TH2 or TH-17 profiles. Computational methods were necessary because of the complexity of the data. We used cluster analysis as an exploratory tool to identify similarities among all the profiles (Fig. 5b). The agglomerative coefficient of 0.85 obtained indicated that the clustering allowed the separation of samples into clusters of conditions. The resampling similarity index of 0.093 reflected highly robust clusters and suggested high statistical significance. The standard T helper conditions showed that the TH1, TH2 and TH-17 conditions segregated into different clusters, which confirmed their distinct cytokine profiles. The TH0 condition segregated with the TH2 condition, probably because of the baseline production of small amounts of TH2 cytokines in the absence of TH1 and TH-17 cytokines (Fig. 3b). Among the suboptimal TH-17 conditions, the removal of TNF induced only a minor change in the profile, which clustered with the optimal TH-17 condition. Notably, the removal of TGF- induced a shift in the profile, which clustered with the TH1 condition; the removal of IL-1, IL-6 or IL-23 induced profiles with no distinct similarity to the TH1, TH2 or TH-17 conditions (Fig. 5b).
We used principal component analysis to complement the cluster analysis and applied this to the average cytokine values obtained for the six donors for each T helper condition (Fig. 5c). This analysis can be viewed as a simplification of the data projected along two axes (the two principal components) that best represented the entire data set and preserved maximum dispersion of the data28. We projected the T helper cell profiles generated in each polarizing condition onto this two-dimensional space (Fig. 5c). Each polarizing condition presented in this way was thus a simplified representation of the cytokine profiles shown for the same polarizing conditions described above (Fig. 5a). The two principal components enabled good discrimination among the TH0, TH1, TH2 and TH-17 profiles, which confirmed that they represented distinct entities (Fig. 5c). The removal of TNF induced the smallest deviation from the TH-17 profile; in the absence of TGF-, the remaining IL-23 and proinflammatory cytokines induced a profile that was more closely related to that of TH1, which was 'converted' to a TH-17 profile in the presence of TGF-. The cytokine vectors indicated the importance of the individual T helper cell–derived cytokines in the discrimination of the different T helper profiles according to the length and direction of the vector (Fig. 5c). For example, the IL-4 and IFN- vectors pointed in opposite directions, which indicated that they were inversely correlated. Accordingly, the TH2 profile was determined not only by IL-4 but also by the lack of IFN- (opposite vector). The TH1-like profile induced in the absence of TGF- 'segregated away' from TH-17 along the second principal component. According to the cytokine vectors, the presence of IFN- and IL-22, along with the decreased IL-17 and IL-6, explained the separation of these two profiles (Fig. 5c). Similarly, IL-10, along with smaller amounts of IL-17 and IL-6, explained the separation between the TH-17 profile and the TH-17 profile without IL-1. Our computational analysis of the global T helper profiles thus demonstrated that the removal of individual TH-17-promoting cytokines generated a diversity of distinct T helper cytokine profiles. Although TNF had a small effect on the global TH-17 profile, the removal of TGF- induced substantial changes and a shift toward a TH1-like profile.
Discussion
Here we have shown that TGF-, IL-23 and proinflammatory cytokines were essential in driving and regulating four key aspects of human TH-17 differentiation: IL-17 production; the acquisition of TH-17-specific features; individual TH-17-derived cytokines; and the global TH-17 cytokine profile. The TH-17 pathway has been linked to the pathogenesis of several autoimmune diseases, including psoriasis29, experimental allergic encephalomyelitis9, arthritis7 and colitis30. It is also crucial in immunity to mycobacteria31 and Candida albicans6. In mice, several studies have shown that TGF- and IL-6 are essential in driving TH-17 differentiation8, 10, 11, with regulatory T cells as a potential source of TGF-11. This indicates that proinflammatory cytokines in the absence of TGF- are not sufficient to induce a TH-17 response, which gives a central function to TGF- in the generation of both regulatory T cells and TH-17 cells. Two studies in human systems have questioned the importance of TGF-, showing that IL-118 or IL-23 (ref. 19) is sufficient to induce TH-17 differentiation and that TGF- negatively regulates this response18, 19. Those studies18, 19 used much longer T cell assays than the mouse studies8, 10, 11 and detected substantial IL-17 production in control medium18 or IL-2 alone19, which suggests that T cells might have less stringent requirements for IL-17 production. In our study, although we obtained a lower yield in IL-17-producing cells, we did not find measurable IL-17 in the absence of polarizing cytokines (TH0) either by ELISA or by intracellular flow cytometry. This suggests that the standard 5-day T cell assay is less sensitive but more specific than systems with longer culture duration. In these conditions, we found that TGF- was required for optimal human TH-17 differentiation and the acquisition of typical TH-17-associated features, such as expression of IL-17A, IL-17F, IL-23R and RORc.
Other factors that might explain the discrepancies among human TH-17 studies18, 19, 20, 21, 22 include the following: the cytokine combinations used, which do not always overlap18, 19, 22; the use of exogenous IL-2, which produces population expansion of differentiated TH-17 cells18, 32; and the use of monoclonal antibody blocking IL-4 and/or IFN-18, 33, two cytokines that inhibit mouse TH-17 differentiation4. The differences might also be due to the serum added to the culture medium, which usually contains TGF- and may also affect TH-17 differentiation in a TGF--independent way. Because of the many experimental parameters that could potentially affect TH-17 differentiation, we confirmed our basic findings in six different systems: CD3+CD4+CD45RA+ peripheral blood naive T cells purified by positive or negative selection; total cord blood CD4 T cells; serum-containing medium; serum-containing medium and monoclonal antibody blocking TGF-; serum-free medium; and serum-free medium and monoclonal antibody blocking IL-4 and IFN-. Although the absolute amount of IL-17 varied, the cytokine requirements to induce optimal IL-17 production in each of these experimental systems were similar.
Another important issue that remains controversial is the function of IL-23, a cytokine of the IL-12 family associated with TH-17 responses34. In mice, IL-23 acts on IL-23R-expressing differentiated TH-17 cells to induce their population expansion in vitro11 and in vivo35 but does not influence TH-17 differentiation10, 11. In humans, IL-23 is sufficient for TH-17 differentiation19. Our data have confirmed an important function for IL-23 in human TH-17 differentiation but only in synergy with TGF- and proinflammatory cytokines, which were essential for inducing IL-23R expression.
Studies have shown that human and mouse TH-17 cells can produce IL-22 (refs. 16,18). However, IL-22 is also produced by polyclonally stimulated naive CD4+ cells36 and TH1 cells37 and is inhibited by TH-17-inducing conditions in memory CD4+ T cells20. Our study has provided further evidence that IL-22 is not specific for TH-17 cells and could have even higher production by TH1 cells. This could explain the different functions of IL-17 and IL-22 in inflammatory responses38 and autoimmune diseases39.
In mice, IL-21 is dependent on IL-6 (refs. 12,23) and is produced in TH-17 conditions but not in TH1 or TH2 conditions13, 15. In contrast, we have shown that human IL-21 was produced in TH1-polarizing conditions as well as TH-17-polarizing conditions and that its production depended on IL-23. Given those results and our data on IL-23R expression, we can infer a sequence wherein TGF- and proinflammatory cytokines induce IL-23R, which enables IL-23 to induce IL-21 production in a second step. The production of IL-21 during TH-17 responses might enhance B cell immunity40, which is involved in the physiopathology of autoimmune diseases such as lupus erythematosus and multiple sclerosis.
IL-10 is produced by mouse TH-17 cells driven by TGF- and IL-6 and confers regulatory functions on them17. Our results suggest that in humans, the presence of IL-1 in a TH-17 environment could inhibit IL-10 production. During the resolution of inflammation, a decrease in or lack of IL-1 may simultaneously decrease the production of IL-17 and enhance the production of IL-10, which would further favor immune contraction through its anti-inflammatory properties41.
Studies of TH1 and TH2 cells have shown that T helper subsets can produce a broader array of cytokines than initially described and that some T cell cytokines have limited specificity for a given T helper subset but are potentially associated with important functional properties, such as proinflammatory, for TNF42, or regulatory, for IL-10 (refs. 43,44,45). Thus, focusing on a single T helper cell–derived cytokine, such as IFN- for TH1 cells or IL-4 for TH2 cells, gives only a partial view of a complex T helper cell response. Here we have shown that optimal TH-17-polarizing conditions also drove the production of an array of cytokines, including IL-21, IL-22, IL-6, TNF and IFN-. Although each of these cytokines has different functions, they could collectively affect the global outcome of a TH-17 response. Most notably, we have also shown that TH-17 cell–derived cytokines were regulated in a specific way. We speculate that in vivo, priming of naive T cells might occur in optimal but also suboptimal polarizing conditions, depending on spatiotemporal factors. Each polarizing environment might induce a different T helper cytokine profile, contributing to the diversity and regulation of an immune response. In our in vitro model, proinflammatory cytokines and IL-23 polarized CD4 T cells toward a TH1-like profile, so they produced mainly IFN-, IL-21 and IL-22. The addition of TGF- in such an environment, which could mimic the onset of TGF--producing regulatory T cells, induced a switch toward a typical TH-17 profile. It has been reported that TGF- 'antagonizes' TH1 responses by inhibiting expression of IFN- as well as of T-bet46, 47. We have shown that TGF- might act in a more global way and was able to convert a TH1-like profile into a TH-17 response.
To our knowledge, this is the first study to analyze global T helper cell cytokine profiles with computational approaches. This could open new perspectives for the pharmacological modulation of T helper responses and could help elucidate and/or allow the prediction of the outcome of a specific therapeutic intervention. Applied to the TH-17 cytokine profile, our data could help identify and target pathogenic components while preserving or enhancing protective aspects in the same T helper response.
Methods
Purification of naive CD4+ T lymphocytes from adult blood.
Peripheral blood mononuclear cells were separated by Ficoll-Hypaque centrifugation (Amersham Biosciences) from buffy coats obtained from samples from healthy blood donors (Saint Antoine-Crozatier Blood Bank, Paris). CD4+ T Lymphocytes were purified by immunomagnetic depletion with the human CD4+ T Cell Isolation Kit II (Miltenyi Biotec), with the addition of biotinylated anti-CD45RO (C2400-67; USBiological). Naive CD4+ T cells (CD3+CD4+CD45RA+CD45RO- ) had a purity of over 96%, as shown by flow cytometry (Supplementary Fig. 8 online). For some experiments, peripheral blood naive CD4+ T cells were isolated with the CD4+ T Cell Isolation Kit II (Miltenyi Biotec), followed by staining with allophycocyanin–anti-CD4 (VIT4; Miltenyi Biotec) and phycoerythrin–anti–CD45RA (PNIM1834; Immunotech) and cell sorting of double-positive cells (purity, over 99%) with a FACSAria (BD Bioscience). Human cord blood was obtained by an ethically approved convention (Necker Hospital, Paris), and total CD4+ T cells were purified with the CD4+ T Cell Isolation Kit II (Miltenyi Biotec).
T helper cell differentiation assay.
Naive CD4+ T cells were cultured in 48-well plates (Falcon) at a density of 8 
104 cells per well in Yssel's medium (a gift from H. Yssel) containing 10% (vol/vol) FCS (Hyclone) or X-VIVO 15 serum-free medium (Lonza) in presence of Dynabeads CD3/CD28 T Cell Expander (one bead per cell; Invitrogen) and the following cytokines: IL-1 (10 ng/ml), IL-6 (20 ng/ml), TNF (10 ng/ml), TGF- (1 ng/ml), IL-23 (100 ng/ml), IL-4 (25 ng/ml) and/or IL-12 (10 ng/ml; R&D Systems). For some experiments, anti-TGF- (human LAP; 27235; R&D Systems), anti-IFN- (B27; BD Biosciences) and/or anti-IL-4 (34019; R&D Systems) were added to the cultures at a concentration of 10 
g/ml. For interference with RORc function, in some experiments lentiviral vectors were used that contained a plasmid encoding shRNA selected for its ability to suppress RORc mRNA expression (33658; Open Biosystems) or empty pLKO.1 vector (Open Biosystems), generated as described48. After 5–6 d, cells were collected and washed extensively and their viability was determined by trypan blue exclusion. Cells (1 106 cells/ml) were restimulated for 6 h (for flow cytometry intracellular staining) or for 24 h (for ELISA and RT-PCR) with Dynabeads CD3/CD28 T Cell Expander (one bead per cell). For shRNA experiments, naive T cells were cultured in serum-free medium at a density of 1 105 cells per well in 96-well round-bottomed plates and were infected with various concentrations of lentiviral vector expressing shRNA or with empty vector (negative control) for the first day of TH-17 differentiation. Cells were washed extensively and were cultured for an additional 5 d in TH0 or TH-17 conditions. IL-17A and IL-17F transcripts and IL-17 protein were analyzed after 24 h of restimulation with anti-CD3 and anti-CD28.
Analysis of cytokine production.
Cytokines in culture supernatants were measured by IL-17 ELISA (R&D Systems), IL-21 ELISA (eBioscience) or IL-22 ELISA (Antigenix) or with IL-4, IL-5, IL-6, IL-10, IL-13, IFN- or TNF cytometric bead assay Flex Sets (BD Bioscience) according to the manufacturer's instructions. Cells producing IFN- and IL-17 were analyzed by intracellular cytokine staining after the addition of brefeldin (10 g/ml) during the final 3 h of restimulation. Cells were made permeable with Cytofix/Cytoperm reagents (BD Biosciences). Cells were stained with fluorescein isothiocyanate–conjugated anti-IFN- (4S.B3; BD Pharmingen) and phycoerythrin-conjugated anti-IL-17 (eBio 64DEC17; eBioscience) and washed and then were analyzed by flow cytometry (FACScan; Becton Dickinson).
Real-time quantitative RT-PCR.
Total RNA was extracted with an RNeasy Micro kit (Qiagen). A mixture containing random hexamers, oligo(dT)15 (Promega) and SuperScript II Reverse Transcriptase (Invitrogen) was used for cDNA synthesis. Transcripts were quantified by real-time quantitative PCR on an ABI PRISM 7900 sequence detector (Applied Biosystems) with Applied Biosystems predesigned TaqMan Gene Expression Assays and Absolute QPCR ROX mix (Thermo Fisher Scientific). The following probes were used (Applied Biosystems assay identification numbers in parentheses): IL-17A (Hs00174383_m1), IL-17F (Hs00369400_m1), RORc (Hs01076112_m1), IL-23R (Hs00332759_m1), T-bet (Hs00203436_m1), GATA-3 (Hs00231122), TNF (Hs 00174128_m1) and IFN- (Hs00174143_m1). For each sample, mRNA abundance was normalized to the amount of ribosomal protein L34 (Hs00241560_m1).
Statistical analysis.
A nonparametric two-tailed Wilcoxon test was used for pairwise comparisons of cytokines. P values of 0.05 or less were considered statistically significant. The Pearson correlation coefficient was used to assess the significance of correlation among IL-17A, IL-17F mRNA or IL-17 protein and RORc. Data for the clustering and principal component analysis (Supplementary Methods online) were corrected for the 'donor effect' through the application of a linear model. For information summaries, replicates were aggregated in each condition to their barycentric value for each cytokine and the principal component analysis was computed with these variables. The Pearson correlation distance and the Ward's criteria as an agglomerative method were used for hierarchical clustering analysis.
Accession code.
UCSD-Nature Signaling Gateway (http://www.signaling-gateway.org): A002271.
Note: Supplementary information is available on the Nature Immunology website.
Author contributions
E.V. did experiments and drafted the manuscript; N.S. did computational and statistical analysis; R.Z. did quantitative RT-PCR analysis and helped with the computational data analysis; S.I.B. did some experiments; P.H. did computational and statistical analysis; E.B. supervised the computational and statistical analysis; and V.S. designed and supervised the study and wrote the manuscript.
