In response to antigenic assault, naïve CD4+ T cells differentiate into various subsets of effector T helper (TH) cells with differential cytokine production profiles and distinct functions1,2,3. Inflammatory cytokines direct the differentiation of antigen-specific CD4+ T cells by inducing the expression of subset-specific transcription factors1. For instance, interleukin 12 (IL-12) activates STAT4 and induces T-bet expression, which promotes TH1 differentiation and IFN-γ production4,5,6. IL-6 signaling through STAT3, in concert with TGF-β, induces RORγt expression and initiates the differentiation of TH17 cells7,8,9, which is further enhanced by TNF-α, IL-23, and IL-1β10.

By coordinating both innate and adaptive effector cell activities, CD4+ T cells including TH1, TH2, and TH17 play critical roles in host defense against infectious agents and in the pathogenesis of various autoimmune diseases11. For example, both TH1 and TH17 cells are considered major mediators of autoimmune neuroinflammation in multiple sclerosis (MS) and experimental autoimmune encephalomyelitis (EAE)2,12,13. However, mice deficient in IFN-γ or IL-12 (p35) show exacerbated EAE development, whereas mice deficient in IL-23 (p19) are resistant to EAE14,15,16,17. Loss of RORγt or STAT3, the master regulators of TH17 cells, attenuates the development of EAE8,18. These studies suggest that TH17, rather than TH1 cells, are the main encephalitogenic population in autoimmune neuroinflammation. However, the role of TH17 in MS and EAE is still in debate since none of the TH17-hallmark cytokines, including IL-17, IL-17F, and IL-22, is mandatory for EAE development12,19,20. More recently, granulocyte-macrophage colony-stimulating factor (GM-CSF) secreted by autoreactive T cells was identified as a potential encephalitogenic factor to sustain neuroinflammation21,22,23.

STAT5 transmits IL-2 signal and is crucial for regulatory T (Treg) cell development24, but also negatively regulates TH17 differentiation25,26. However, its function in T cell-mediated autoimmune diseases has not been well documented. In this study, surprisingly, we found that STAT5 was indispensible for the encephalitogenicity of autoreactive CD4+ T cells in EAE. Further investigation showed that IL-7-STAT5 signaling axis induced optimal GM-CSF production in pathogenic CD4+ T cells, which was important for inducing effective neuroinflammation. In vitro studies showed that GM-CSF-producing CD4+ T cells regulated by IL-7-STAT5 signaling axis may represent a new TH subset with a distinct differentiation program and cytokine production profile.


Mice with Stat5 deletion in T cells are resistant to EAE

To examine the role of STAT5 in T cell-mediated pathogenesis, we induced EAE in Cd4-Cre; Stat5f/f (Stat5−/−) mice27, where Stat5a/b loci were specifically deleted in CD4+ and CD8+ T cells, and Stat5f/f (Stat5+/+) mice. We found diminished incidence and severity of EAE disease in Stat5−/− mice compared with Stat5+/+ mice (Figure 1A and Supplementary information, Figure S1A and S1B), which was opposite to our expectation based on an inhibitory role of STAT5 in TH17 generation. Consistent with EAE resistance, we found a remarkable reduction of immune cell infiltration in the CNS of Stat5−/− mice (Figure 1B and Supplementary information, Figure S1C-S1F). However, the frequencies of IL-17+ and IFN-γ+ cells among CD4+ T cells in the central nervous system (CNS) were comparable between Stat5+/+ and Stat5−/− mice (Figure 1C), suggesting that the resistance to EAE in Stat5−/− mice is independent of TH1 and TH17 cells. Furthermore, we detected decreased CD4+CD25+ population and reduced Foxp3 expression in Stat5−/− mice (Figure 1D and Supplementary information, Figure S2A), indicating the resistance to EAE is unlikely due to altered Treg cell development.

Figure 1
figure 1

Stat5-conditional knockout mice are resistant to EAE. (A) Clinical EAE scores (left) and incidence (right, n = eighteen of three experiments pooled) of Stat5+/+ and Stat5−/− mice immunized twice with MOG35-55/CFA. (B) Histology of spinal cord sections obtained from EAE mice on day 9 after 2nd immunization. Scale bars, 200 μm (top), 50 μm (bottom). (C) Flow cytometric analysis of IL-17 and IFN-γ expression by CNS-infiltrating CD4+ T cells at peak of disease. (D) Percentage of CD4+CD25+ T cells in the CNS at peak of disease (n = 3). (E, F) Clinical EAE scores (E, left) and incidence (E, right) of Rag2−/− mice (n = 5 per group) after adoptive transfer of 2 × 106 MOG35-55-reactive Stat5+/+ or Stat5−/− CD4+ T cells, respectively. IL-17 and IFN-γ expression by CNS-infiltrating CD4+ T cells was measured by intracellular cytokine staining at the peak of disease (F). Data represent at least two independent experiments. *P < 0.05; ns, not significant.

Intrinsic defect in encephalitogenicity of STAT5-deficient CD4+ T cells

To examine whether T cell-specific deletion of Stat5 resulted in peripheral lymphopenia, we analyzed T cell populations in spleens of MOG35-55/CFA-immunized mice. Consistent with a previous report28, we detected reduced CD8+ T cell number but similar number of CD4+ T cells in Stat5−/− mice compared with Stat5+/+ mice (Supplementary information, Figure S2B and S2C). Furthermore, we detected increased frequencies of both IL-17+ and IFN-γ+ CD4+ T cells in the spleens of Stat5−/− mice (Supplementary information, Figure S2D). To validate the function of STAT5 in TH1 and TH17 generation, we performed in vitro T-cell differentiation. As reported25,26, STAT5 mediated the suppressive effect of IL-2 on TH17 differentiation (Supplementary information, Figure S3A and S3B). STAT5 deficiency led to slightly decreased TH1-cell generation (Supplementary information, Figure S3C). Therefore, the resistance to EAE in Stat5−/− mice is unlikely due to impaired TH1- and TH17-cell generation in vivo.

To address whether the resistance to EAE in Stat5−/− mice is caused by STAT5 deficiency in CD4+ T cells, we reconstituted Rag2−/− mice with Stat5+/+ or Stat5−/− CD4+ T cells followed by EAE induction. We found that Rag2−/− mice receiving Stat5−/− CD4+ T cells were resistant to the disease compared with mice receiving Stat5+/+ CD4+ T cells (Supplementary information, Figure S4A and S4B), demonstrating that Stat5−/− CD4+ T cells were impaired in mediating EAE development. The expression of chemokine receptors such as CCR6 and CXCR3, which are critical for TH17 or TH1 cell entry into the CNS29,30, was not impaired in Stat5−/− CD4+ T cells (Supplementary information, Figure S5A), indicating that CD4+ T cells with STAT5 deficiency are likely capable of infiltrating the CNS. Consistent with this, we observed comparable numbers of CD4+ T cells in the CNS of Stat5+/+ and Stat5−/− mice before disease onset (on days 7 and 9) (Supplementary information, Figure S5B). However, during the later phase, significantly more CD4+ T cells accumulated in the CNS of Stat5+/+ mice compared with Stat5−/− mice (Supplementary information, Figure S5C). These results indicate that Stat5−/−CD4+ T cells can infiltrate the CNS but fail to induce effective inflammatory responses. To further exclude the possibility that EAE resistance is due to reduction in Stat5−/− CD4+ T cell number in the CNS, we transferred more Stat5−/− CD4+ T cells than Stat5+/+ cells into Rag2−/− mice followed by MOG35-55/CFA immunization, so that comparable numbers of autoreactive CD4+ T cells were present in the CNS after disease onset. We still observed reduced disease severity in mice receiving Stat5−/− CD4+ T cells (Supplementary information, Figure S6A and S6B). Together, these results suggest that the resistance to EAE disease caused by STAT5 deficiency in CD4+ T cells is unlikely due to impaired CD4+ T cell infiltration or survival in the CNS.

To confirm that EAE resistance caused by STAT5 deficiency is due to intrinsic impairment of autoreactive CD4+ T cells, we isolated CD4+ T cells from MOG35-55/CFA-immunized mice and transferred ex vivo-expanded MOG35-55-reactive Stat5+/+ and Stat5−/−CD4+T cells into Rag2/− mice separately without additional immunization. Mice receiving Stat5+/+ cells developed EAE 1 week after the transfer (Figure 1E). In contrast, mice receiving Stat5−/−CD4+T cells had significantly reduced disease severity and incidence (Figure 1E). Of note, the frequencies of IL-17+ and/or IFN-γ+ cells among CD4+ T cells in the CNS were comparable between the two groups (Figure 1F). CD4+ and CD8+ T cell co-transfer experiments demonstrate that the resistance to EAE observed in Stat5−/− mice was not due to CD8+ T cells (Supplementary information, Figure S7A and S7B). Therefore, Stat5−/− CD4+ T cells are intrinsically defective in encephalitogenicity, independent of TH1- and TH17-cell generation.

STAT5 deficiency in CD4+ T cells causes impaired expression of GM-CSF

Communication between CNS-infiltrating CD4+ T cells and myeloid cells is critical for inducing effective neuroinflammation21,22. GM-CSF production by encephalitogenic CD4+ T cells, not other types of cells, is essential for microglial cell activation, peripheral myeloid cell recruitment and EAE development23. To test whether GM-CSF production was impaired upon Stat5 depletion, we examined GM-CSF expression in MOG35-55-specific CD4+ T cells. We found that GM-CSF production was robustly increased in a dose-dependent manner in Stat5+/+, but not in Stat5−/− cells, upon antigen re-stimulation (Figure 2A). Antigen-specific CD4+ T cells with STAT5 deficiency contained significantly reduced percentages of GM-CSF-producers in both IL-17+ and IL-17 populations (Figure 2B). Of note, the frequency of IL-17-producing CD4+ T cells was increased with STAT5 deficiency (Figure 2B). Together, these results suggest that STAT5 is required for GM-CSF expression by antigen-specific CD4+ T cells.

Figure 2
figure 2

Diminished induction of GM-CSF in Stat5−/− CD4+ T cells. (A, B) Splenocytes were obtained from MOG35-55/CFA-immunized Stat5+/+ and Stat5−/− mice (n = 3 per group) before disease onset and challenged with MOG35-55 at various concentrations for 24 h. GM-CSF secretion was measured by ELISA (A). Golgiplug was added in the last 4 h of MOG35-55(20 μg/ml) challenge and the frequencies of IL-17+ and GM-CSF+ cells among CD4+CD44hi T cells were measured (B). (C) IL-17, IFN-γ and GM-CSF expression by CNS-infiltrating CD4+ T cells of Stat5+/+ and Stat5−/− mice was measured by intracellular cytokine staining at peak of disease. (D) IL-17, IFN-γ, and GM-CSF expression by CNS-infiltrating CD4+ T cells of Rag2−/− recipient mice at peak of EAE induced by adoptive transfer of MOG35-55-reactive CD4+ T cells. (E) CNS tissues were collected from naïve or MOG35-55/CFA-immunized mice for RNA extraction (n = 3 per group at each time point). Time-course analysis of cytokine mRNA expression was performed with RT-PCR. The RT-PCR data were normalized to Rn18S, and expression in naïve mice was set to 1. Data represent two independent experiments. *P < 0.05.

Next, we examined GM-CSF expression in the CNS during EAE development. Although IL-17 and IFN-γ expression by CNS-infiltrating Stat5−/− CD4+ T cells was not impaired (Figure 1C), we detected a significantly diminished frequency of CD4+GM-CSF+ cells in the CNS of Stat5−/− mice compared with control mice (Figure 2C). Similarly, in passive EAE induction, Rag2−/− mice transferred with STAT5-deficient MOG35-55-reactive CD4+ T cells also showed a reduced frequency of CD4+GM-CSF+ T cells in the CNS compared with mice transferred with wild-type (WT) cells (Figure 2D). Time-course analysis of cytokine induction in the whole CNS tissues showed that GM-CSF mRNA expression in Stat5+/+ mice was markedly increased as early as day 8 after MOG35-55/CFA immunization, whereas GM-CSF induction in Stat5−/− mice was significantly diminished (Figure 2E). Meanwhile, no significant difference in IL-17 or IFN-γ expression was detected between Stat5−/− and Stat5+/+ mice on day 8 post-immunization (Figure 2E). The reduced IL-17 expression in the CNS of Stat5−/− mice at a later stage (day 14, Figure 2E) could be explained by the inability of Stat5−/− CD4+ T cells to induce effective neuroinflammation with a result of decreased inflammatory cell infiltration (Supplementary information, Figures S1 and S5C). We also observed the expression level of IL-23, an important inflammatory cytokine mainly produced by dendritic cells (DCs)16, was reduced in the CNS of Stat5−/− mice compared with Stat5+/+ mice, paralleling decreased IL-17 expression level (Figure 2E). Interestingly, the expression level of IL-23 in the CNS of Stat5+/+ mice was only significantly increased on day 14 after disease induction (Figure 2E), suggesting that IL-23 might not be required for GM-CSF expression and EAE induction at the early stage. Together, these results demonstrate that STAT5 deficiency in CD4+ T cells results in impaired GM-CSF expression, which is associated with EAE resistance.

IL-7-STAT5 signaling induces GM-CSF expression in autoreactive CD4+ T cells

We next investigated the possible cytokine(s) that signal through STAT5 to regulate GM-CSF expression. We stimulated CD4+ T cells with IL-23 and IL-1β, two cytokines that drive GM-CSF expression in TH17 cells21,22. We found that neither IL-23 nor IL-1β was able to induce STAT5 activation (Figure 3A). Furthermore, IL-1R1 expression was not changed, whereas IL-23Rα expression was increased in Stat5−/− CD4+ T cells (Figure 3B), indicating that STAT5-mediated GM-CSF expression is unlikely dependent on IL-23 and IL-1β signaling. In contrast, both IL-2 and IL-7 potently activated STAT5 (Figure 3A). Therefore, we further examined the roles of these two cytokines in GM-CSF induction in CD4+ T cells. Splenocytes were isolated from MOG35-55/CFA-immunized mice before disease onset and challenged with MOG35-55 alone versus in the presence of IL-2 or IL-7 ex vivo. We did not detect an obvious effect of IL-2 on the frequency of GM-CSF-producing cells in CD4+CD44hi population (Supplementary information, Figure S8). In contrast, IL-7 significantly increased the frequency of GM-CSF-producing cells in CD4+CD44hi population and GM-CSF secretion in a STAT5-dependent manner (Figure 3C and 3D).

Figure 3
figure 3

IL-7 promotes GM-CSF expression through STAT5 activation in autoreactive CD4+ T cells. (A) Purified CD4+ T cells were cultured with TGF-β and IL-6 for 3 days, followed by resting for 6 h. Then cells were treated with various cytokines for 30 min, and pSTAT3 and pSTAT5 levels were determined by immunoblotting. STAT3 and STAT5 were further detected after stripping. (B) The mRNA expression of IL-23Rα and IL-1R1 in splenic CD4+ T cells of Stat5+/+ and Stat5−/− EAE mice (n = 3). (C, D) Splenocytes were obtained from MOG35-55/CFA-immunized Stat5+/+ and Stat5−/− mice before disease onset and challenged with MOG35-55 (20μg/ml) in the absence or presence of IL-7 for 48 h. Frequencies of GM-CSF+ and IL-17+ cells among CD4+CD44hi T cells were measured by intracellular cytokine staining and flow cytometry (C). Right panel in C shows overall frequencies of GM-CSF+ and IL-17+ cells in Stat5+/+ group (n = 3). GM-CSF secretion was measured by ELISA (D). Data represent two independent experiments with three mice per group. (E) Splenic CD62LhiCD44lo and CD62LloCD44hi T cells from MOG35-55/CFA-immunized mice were sorted out. Cells were stimulated with anti-CD3 and anti-CD28 in the absence or presence of IL-7 for 4 h and then harvested for the analysis of GM-CSF expression by RT-PCR. *P < 0.05; ns, not significant.

IL-7Rα is expressed by naïve and effector CD4+ T cells, suggesting that IL-7 may directly act on both populations to regulate GM-CSF expression. To address this, sorted CD62LhiCD44lo (naïve) and CD62LloCD44hi (effector) CD4+ T cells from Stat5−/− and Stat5+/+ mice during EAE development were activated by anti-CD3 plus anti-CD28 in the presence or absence of IL-7, followed by GM-CSF expression examination. As shown in Figure 3E, CD62LloCD44hi T cells expressed GM-CSF more robustly than CD62LhiCD44lo cells. IL-7 promoted GM-CSF expression in both cell subsets, which was abrogated by STAT5 deficiency (Figure 3E).

Consistent with the facilitating effect of IL-7-dependent T-cell differentiation on EAE, mice treated with an IL-7Rα-specific antibody (clone SB/14) during EAE development showed a significant reduction of disease severity accompanied with reduced CNS inflammation, but without T cell depletion31 (Supplementary information, Figure S9A-S9C). Notably, blocking IL-7 signaling resulted in decreased GM-CSF expression in CNS-infiltrating CD4+ T cells (Supplementary information, Figure S9D-S9F). These findings demonstrate that IL-7 induces STAT5 activation to promote GM-CSF expression in autoreactive CD4+ T cells, which is critical for the development of neuroinflammation.

TH17 or TH1 differentiation condition inhibits GM-CSF expression

To further understand IL-7/STAT5-mediated GM-CSF expression in CD4+ T cells, we stimulated naïve CD4+ T cells with various conditions. We found that anti-CD3 together with anti-CD28 induced the expression of both GM-CSF and IFN-γ (Supplementary information, Figure S10A). Interestingly, both TH1 (IL-12 + anti-IL-4) and TH17 (blocking both IFN-γ and IL-4 in combination with TGF-β + IL-6 or IL-6 + IL-23 + IL-1β) differentiation conditions greatly suppressed the expression of GM-CSF (Figure 4A and 4B). Conversely, neutralization of both IL-12 and IFN-γ promoted the generation of GM-CSF-producing cells, consistent with a previous report21, which was not affected by IL-23 and IL-1β (Figure 4A). In addition to TGF-β-mediated inhibition of GM-CSF expression22, we found that IL-6, an essential cytokine for TH17 differentiation, had a profound inhibitory effect on GM-CSF expression (Figure 4C), indicating that STAT3 could be a negative regulator of GM-CSF expression. We used STAT3-deficient CD4+ T cells to test this hypothesis. As expected, naïve Stat3−/− CD4+ T cells were impaired in TH17 differentiation (Supplementary information, Figure S10B). In WT cells, TH17 differentiation condition (anti-IFN-γ + anti-IL-4 + IL-6 + IL-23 + IL-1β) greatly inhibited GM-CSF expression (Figure 4A). However, deficiency of STAT3 abrogated the inhibitory effect of IL-6 on GM-CSF expression (Supplementary information, Figure S10B). Interestingly, even without exogenous IL-6, STAT3 exhibited a suppressive effect on GM-CSF expression as Stat3−/− cells showed increased GM-CSF expression compared to WT cells (Supplementary information, Figure S10B). In addition, GM-CSF expression in CD4+ T cells is independent of RORγt and T-bet22. Thus, our data support that differentiation of GM-CSF-producing CD4+ T-cell is distinct from TH1 or TH17.

Figure 4
figure 4

Regulation of GM-CSF-producing TH cells in vitro. (A-E) Naïve CD4+ T cells were primed with plate-bound anti-CD3 and soluble anti-CD28 in the presence of a combination of various cytokines and neutralizing antibodies as indicated. GM-CSF, IL-17, and IFN-γ expression was analyzed by intracellular cytokine staining (A, C, D), RT-PCR (B) or ELISA (E). (F, G) Stat5+/+ and Stat5−/− naïve CD4+ T cells were activated with anti-CD3 and anti-CD28 in the presence IL-7 for 3 days. GM-CSF, IL-17, and IFN-γ expression was analyzed by intracellular cytokine staining (F). GM-CSF secretion was measured by ELISA (G). (H) Naïve CD4+ T cells were activated with anti-CD3 and anti-CD28 in the presence of IL-7 or/and anti-IFN-γ as indicated. GM-CSF, IL-17, and IFN-γ expression was analyzed. Data represent more than two independent experiments. *P < 0.05; ns, not significant.

IL-7-STAT5 promotes GM-CSF-producing TH-cell differentiation

Our findings above suggest the possibility of a potential new TH cell subset that is regulated by IL-7-STAT5 signaling. To further test this possibility, we investigated GM-CSF-producing TH cell differentiation in vitro by activating naïve CD4+ T cells with anti-CD3 and anti-CD28 in the presence of different concentrations of IL-7. We found that addition of 0.5 ng/ml IL-7 greatly increased the frequency of GM-CSF-producing cells and the secretion of GM-CSF, which were further increased upon increase in IL-7 concentration (1 ng/ml) (Figure 4D and 4E). Without STAT5, IL-7 was unable to promote the generation of GM-CSF-producing cells (Figure 4F and 4G). Chromatin immunoprecipitation (ChIP) analysis showed that IL-7 activated STAT5 directly bound to promoter regions of the Csf2 gene (Supplementary information, Figure S11A and S11B). We noticed the presence of a small proportion of IFN-γ-producing cells in this condition (Figure 4D). Therefore, we included IFN-γ-blocking antibody in the culture and found that a combination of IL-7 and anti-IFN-γ induced the highest frequency of GM-CSF+ cells, where few IL-17+ or IFN-γ+ cells were detected (Figure 4H). Therefore, the in vitro generation of GM-CSF-producing TH cells requires the transcription factor STAT5, optimal concentration of IL-7, and IFN-γ neutralization in addition to TCR and CD28 signaling.

GM-CSF-producing TH cells represent a potential new subset distinct from TH1 or TH17

To further characterize GM-CSF-producing TH cells, we differentiated TH1, TH17, and GM-CSF-producing TH cells from naïve CD4+ T cells in vitro. The expression of RORγt and T-bet was examined. We found that unlike TH1 or TH17 cells, the expression of T-bet or RORγt was minimal in GM-CSF-producing TH cells (Figure 5A). Next, we performed microarray analysis to examine gene expression profiles of TH1, TH17, and GM-CSF-producing TH cells. We identified a list of 202 genes preferentially expressed in TH1 cells compared with naïve, TH17 and GM-CSF-producing TH cells, among which IFN-γ, Gzmb, and T-bet were on the top of the list (Figure 5B, left panel and Supplementary information, Table S1). Similarly, TH17 feature genes, including IL-17, IL-17F, RORγt, and RORα, were identified in the list of 411 genes specific to TH17 cells (Figure 5B, middle panel, and Supplementary information, Table S1). The GM-CSF-producing TH cell-specific gene list contains 210 genes with genes encoding GM-CSF and IL-3 as the top genes in the list (Figure 5B, right panel and Supplementary information, Table S1).

Figure 5
figure 5

Distinct features of GM-CSF-producing TH cells. (A) The mRNA expression of T-bet and RORγt in naïve, TH1 (IL-12 + anti-IL-4), TH17 (TGF-β + IL-6 + anti-IFN-γ + anti-IL-4) and GM-CSF-producing TH (IL-7 + anti-IFN-γ) cells. The RT-PCR data were normalized to Gapdh, and expression in naïve T cells was set to 1. (B) Naïve CD4+ T cells were differentiated into TH1, TH17 and GM-CSF-producing (GM-CSF+) TH cells in vitro. Microarray analysis was performed to examine their gene expression profiles. Hierarchical clusters of preferentially expressed genes for TH1, TH17, or GM-CSF-producing TH cells were shown (biological duplication). (C) GM-CSF secretion by three TH subsets in vitro. (D) Flow cytometric analysis of cytokine expression (GM-CSF, IL-3, IL-17, and IFN-γ) by three TH subsets in vitro. (E) Frequency of IL-3+ cells generated with or without IL-7. (F) GM-CSF and IL-3 expression by WT or STAT5-deficient GM-CSF-producing TH cells. (G) Clinical EAE scores of Rag2−/− mice (n = 3-6 mice per group) after adoptive transfer of 6 × 105 various MOG35-55-reactive TH subsets. Data represent two independent experiments. *P < 0.05.

Next, we further verified the microarray findings. Cytokine expression analysis showed that GM-CSF was predominately expressed in GM-CSF-producing TH cells compared with TH1 or TH17 cells (Figure 5C and 5D). Interestingly, IL-3, a cytokine that is coregulated with GM-CSF32, was also highly expressed in GM-CSF-producing TH cells, not TH1 or TH17 cells (Figure 5D). Further examination showed that IL-3 expression was also regulated by IL-7-STAT5 signaling (Figure 5E and 5F).

IL-2, which also signals through STAT5, did not promote GM-CSF-producing TH differentiation (Supplementary information, Figure S12A), possibly due to the lack of IL-2Rα expression on naïve CD4+ T cells and thus the unresponsiveness of STAT5 to the IL-2 signal early in differentiation (Supplementary information, Figure S12B-S12D). To further confirm this possibility, we stimulated activated CD4+ T cells with IL-2 or IL-7, and found both cytokines induced STAT5 activation, STAT5 binding to Csf2 promoter, and increase in GM-CSF mRNA levels (Supplementary information, Figure S13A-S13C). Notably, IL-2 induced a prolonged STAT5 activation compared with IL-7 (Supplementary information, Figure S13A).

To test the hypothesis that the GM-CSF-producing TH subset is the primary encephalitogenic effector cells, we performed adoptive transfer of different subsets of MOG35-55-reactive CD4+ T cells into Rag2−/− mice for EAE induction. As shown in Figure 5G, GM-CSF-producing TH cells were preferentially able to induce a more robust EAE compared with TH17 and TH1 subsets.

Together, these data demonstrate that IL-7-STAT5-signaling controls the differentiation of a new T helper cell subset that is distinct from TH1 or TH17 and predominantly expresses GM-CSF and IL-3.

STAT5-deficient CD4+ T cells retain the capacity to induce colitis

The finding from the EAE model prompted us to test whether STAT5-deficient T cells also lacked pathogenicity in other T cell-mediated autoimmune diseases. We assessed the pathogenic potential of STAT5-deficient autoreactive CD4+ T cells in colitis by reconstituting Rag2−/− mice with Stat3−/−, Stat5−/− or WT CD4+CD25CD45RBhi naïve T cells. Consistent with a previous report33, mice reconstituted with Stat3−/− T cells continued to gain weight, whereas mice receiving Stat5−/− or WT T cells lost weight (Figure 6A), and showed enlarged spleens and mesenteric lymph nodes (MLNs) (Figure 6B). Marked colonic inflammation and inflammatory cell infiltration in the colon were observed in mice that received either WT or Stat5−/− T cells at 8 weeks after reconstitution (Figure 6C and 6D). Consistent with the resistance to colitis in mice receiving Stat3−/− cells, an obviously reduced frequency of CD4+ T cells and IL-17-producing CD4+ T cells in their lamina propria lymphocytes (LPLs) was observed compared with mice receiving Stat5−/− or WT cells (Figure 6D and 6E). There was no defect in either IL-17 or IFN-γ production by CD4+ T cells in LPLs of mice reconstituted with Stat5−/− T cells (Figure 6E). However, we detected a significant reduction in GM-CSF-producing T cells in the absence of STAT5 when transferred into Rag2−/− mice (Figure 6E). Therefore, GM-CSF-producing CD4+ T cells might not play a critical role in colitis. This view was further supported by the observation that although there were comparable GM-CSF+ cells in Stat3−/− and WT-transferred T cells, Stat3−/− CD4+ T cells failed to induce colitis.

Figure 6
figure 6

STAT5 is not required in T cell-dependent colitis. (A) Rag2−/− mice were reconstituted with CD4+CD25CD45RBhi naïve T cells derived from WT, Stat3−/− or Stat5−/− mice (3-5 mice per group). Body weight loss was monitored and calculated over 7 weeks. Data are representative of two independent experiments. (B) Systemic inflammation was assessed by comparing size of spleen and MLNs 8 weeks after transfer. (C) Colonic inflammation was assessed histologically. Images shown are representative of three mice per group. (D) Percentages of CD4+ T cells and CD11b+ cells in MLNs and LPLs were analyzed by flow cytometry. Data are representative of at least three mice per group. (E) IL-17, IFN-γ, and GM-CSF production by CD4+ T cells in LPLs was measured. *P < 0.05, **P < 0.005.


Effector TH cells, differentiated from naïve T cells after TCR-mediated antigen recognition with the influence of costimulation and the instruction from specific cytokines, are classified by their specific cytokine expression and immune-modulatory functions1,3,34. Here we have demonstrated that IL-7, signaling through STAT5, induces the development of a potential new TH subset that predominantly expresses GM-CSF (GM-CSF-producing TH cells). In addition to GM-CSF, we identified that this TH subset highly expresses IL-3, a cytokine important in regulating the function of myeloid-derived immune cells. The development of GM-CSF-producing TH cells is independent of the mechanisms required for TH1 or TH17 development. In fact, conditions for TH1 or TH17 differentiation suppressed the development of GM-CSF-producing TH cells. It is known that differentiated effector TH cells could produce specific cytokines to create a cytokine environment that favors the differentiation of their own while suppressing the differentiation of other TH subsets. For instance, IL-12 as well as IFN-γ induces TH1 differentiation. The differentiated TH1 cells produce large amounts of IFN-γ to amplify TH1 differentiation, whereas they suppress TH17 or GM-CSF-producing TH differentiation. It is possible that the GM-CSF-producing TH cells produce certain cytokines to promote the differentiation of their own, but suppress the differentiation of TH1 and/or TH17 cells. It is also possible that this new subset of TH cells produce cytokines such as GM-CSF to enhance TH1 or TH17 responses via inducing the production of inflammatory cytokines from myeloid cells. Further studies are needed to explore such possibilities.

While this manuscript was under submission, a report was published, showing that IL-17 and GM-CSF expression in human TH cells are antagonistically regulated35. Similar to our finding in mice, Noster et al.35 found that GM-CSF-producing TH cells represent a substantial population in the cerebrospinal fluid of MS patients, suggesting a pathogenic role of these cells in neuroinflammation. Our work utilizing the mouse model for a thorough functional study has provided clear genetic evidence showing the essential role of STAT5 in GM-CSF and IL-3 production in this novel TH cell subset. This subset is essential, even in the context of normal TH1 or TH17 activity, for the pathogenesis of EAE. We propose to tentatively refer to this T helper cell subset as TH-GM, which is critically regulated by STAT5 and predominantly produces GM-CSF and IL-3.

GM-CSF governs the activities of myeloid-derived cell populations and is implicated in various inflammatory and autoimmune diseases such as rheumatoid arthritis (RA) and MS. Therapeutic interventions targeting GM-CSF, such as Mavrilimumab (human anti-GM-CSFRα Ab) and MOR103 (human anti-GM-CSF mAb), are under phase 2 clinical trial in RA and RA/MS, respectively ( IL-3 is a cytokine also involved in several autoimmune diseases36,37,38,39. Therefore, directly targeting this new subset of TH cells, the dominant source of such pathogenic mediators, in various autoimmune diseases could lead to better outcomes than targeting a single factor, such as GM-CSF or IL-3.

STAT5 transmits IL-2 signals. A previous widely accepted model is that T-cell survival and proliferation require IL-2, based on studies using in vitro systems. There is now much evidence arguing that IL-2 is dispensable for the induction of T cell-dependent immunity in vivo40. In our EAE model, we indeed observed Stat5-conditional knockout mice had fewer CD4+ T cells in the CNS after disease onset as STAT5-deficient CD4+ T cells had inability to induce effective inflammatory responses. However, in EAE induced by CD4+ T cell transfer, STAT5-deficient CD4+ T cells with a number comparable to that of WT cells in the CNS still failed to induce the disease (Supplementary information, Figure S6A and S6B). These results suggest that the role of STAT5 in CD4+ T cell survival and proliferation is unlikely the major causal factor of EAE resistance in Stat5-conditional knockout mice.

IL-7, signaling through the common γc and IL-7Rα, is required for T cell homeostasis41. A previous report showed that IL-7 could acutely stimulate GM-CSF production from memory T cells42. Here, our work demonstrates that IL-7 through STAT5 activation induces the generation of pathogenic TH-GM cells for GM-CSF production in mediating neuroinflammation. Our finding is in line with the association of IL-7Rα variants with MS in patient GWAS and the beneficial effect of IL-7Rα-neutralizing antibody in EAE treatment43,44,45. A recent report indicates IL-7 promotes IFN-γ production by TH1 cells and contributes to a TH1-driven subtype of MS45. However, mice deficient in IFN-γ show exacerbated EAE development15. Thus, the requirement for IL-7 in EAE development is unlikely due to its role in promoting IFN-γ production.

The function of IL-23, a cytokine required for TH17 terminal differentiation46, in EAE suggests the involvement of TH17 cells in this disease16. However, the major cytokines produced by TH17 cells, IL-17A, IL-17F, and IL-22, were found to be dispensable for the development of EAE19,20. However, IL-23 was found to induce the expression of GM-CSF in TH17 cells, a factor essential for encephalitogenicity of T cells21,22, which seems to strengthen the link between TH17 cells and EAE. Paradoxically, a recent study reported IL-23/RORγt axis-suppressed GM-CSF expression in human TH cells35. Moreover, anti-IL-23 receptor antibodies that worked well for treating peripheral inflammation such as psoriasis failed in treating MS47, suggesting that the function of IL-23 in MS could be compensated by other factors. In this study, we found that TH-GM cells produced greater amounts of GM-CSF than TH1 or TH17 cells (Figure 5), suggesting that TH-GM cells are the major source of GM-CSF in T cell-mediated neuronal inflammation. Interestingly, we found that the expression of IL-23p19 in CNS during EAE development only occurred after the onset of the disease and Stat5 gene deletion in T cells abolished its expression (Figure 2E), suggesting that IL-23 is not required for the initiation of EAE and signaling transduced by STAT5 is required for IL-23 expression in CNS in neuronal inflammation. It is possible that GM-CSF-producing TH-GM cells provide GM-CSF to induce the expression of IL-23 from DCs, macrophages, and other CNS-residential cells to sustain the inflammation. Therefore, our study does not exclude a scenario where GM-CSF-producing TH-GM cells cooperate with TH1 and/or TH17 to mediate the development of EAE. Further study on TH-GM cells including their physiological functions is warranted for targeting these cells for the development of therapeutic interventions for human inflammatory diseases such as MS.

It is interesting to see that STAT5-regulated TH-GM cells are not required in a T-cell transfer model of colitis (Figure 6). EAE represents sterile CNS inflammation, in which autoreactive CD4+ T cells are the major producers of GM-CSF, a factor crucial for EAE pathogenesis23. However, numerous inflammatory cytokines, such as TH1-related IFN-γ and TH17-related IL-17F, are implicated in the pathogenesis of colitis48. STAT5-deficient CD4+ T cells are not impaired in TH1 or TH17 generation. In addition, GM-CSF could be induced in other types of cells in the intestine, such as Paneth cells49. Thus, STAT5-mediated GM-CSF production in T cells may not be mandatory for colitis.

In summary, we found that a distinct subset of T helper cells (TH-GM), which is regulated by IL-7-STAT5 signaling axis and predominantly produces GM-CSF and IL-3, is critical for autoimmune neuroinflammation.

Materials and Methods


Stat5f/f mice were provided by L Hennighausen (National Institute of Diabetes and Digestive and Kidney Diseases)27. Stat3f/f mice were generated as described50. Cd4-Cre transgenic mice were purchased from Taconic Farms. Rag2/− mice were obtained from Jean-Pierre Abastado (Singapore Immunology Network). All mice are on a C57BL/6 genetic background and housed under specific-pathogen-free conditions at the National University of Singapore. All experiments were performed with 6-8-week-old mice and approved by the Institutional Animal Care and Use Committee of NUS.

In vitro T-cell differentiation

CD4+ T cells were obtained from spleens and lymph nodes by positive selection and magnetic separation (Miltenyi Biotec), followed by purification of naïve CD4+ T cell population (CD4+CD25CD62LhiCD44lo) sorted with FACS Aria. Naïve CD4+ T cells were stimulated with plate-bound anti-CD3 (3 μg/ml; BD Pharmingen) and anti-CD28 (1 μg/ml; BD Pharmingen) in the presence of different combinations of neutralizing antibodies and cytokines for 3-4 days: for neutral conditions, no addition of any cytokine or neutralizing antibody; for TH1 conditions, IL-12 (10 ng/ml), and anti-IL-4 (10 μg/ml, BD Pharmingen); for TH17 conditions, hTGF-β (3 ng/ml), IL-6 (20 ng/ml), anti-IFN-γ (10 μg/ml, eBioscience), and anti-IL-4 (10 μg/ml); for an alternative TH17 conditions, IL-6 (20 ng/ml), IL-23 (10 ng/ml), IL-1β (10 ng/ml), anti-IFN-γ (10 μg/ml), and anti-IL-4 (10 μg/ml). For GM-CSF-producing cell differentiation, naïve CD4+ T cells were stimulated with plate-bound anti-CD3 (3 μg/ml) and soluble anti-CD28 (1 μg/ml) with the addition of IL-7 (2 ng/ml) and/or anti-IFN-γ (10 μg/ml) as indicated. All cytokines were obtained from R&D Systems. All cells were cultured in RPMI 1640 supplemented with 10% FBS, 100 units/ml penicillin, 0.1 mg/ml streptomycin, 1 mM sodium pyruvate, 0.1 mM nonessential amino acid and 5 μM beta-mercaptoethanol. After polarization for 3-4 days, cells were washed and restimulated with phorbol 12-myristate 13-acetate (PMA) and ionomycin in presence of Golgiplug for 4-5 h, followed by fixation and intracellular staining with a Cytofix/Cytoperm kit from BD Pharmingen. Foxp3 staining was done with a kit from eBioscience. Cells were acquired on the LSR II (BD Biosciences) and analyzed with FlowJo software (Tree Star).

EAE induction

EAE induction procedures were modified from a previous report51. For active EAE induction, mice were immunized in two sites on the hind flanks with 300 μg MOG35-55 in 100 μl CFA containing 5 mg/ml heat-killed M. tuberculosis strain H37Ra (Difco) on day 0 and day 7. Pertussis toxin (List Bio Lab) was administrated intraperitoneally at the dosage of 500 ng per mouse on day 1 and day 8. For single MOG35-55/CFA immunization, the similar procedure was performed on day 0 and day 1 only. In an alternative active EAE induction, LPS (600 μg/ml in IFA, O111:B4 from Sigma) was used as adjuvant. For active EAE induction in Rag2−/− mice, CD4+ T cells derived from Stat5f/f or Cd4-Cre; Stat5f/f mice were transferred, followed by MOG35-55/CFA immunization as described above. Clinical symptoms were scored as follows: 0, no clinical sign; 1, loss of tail tone; 2, wobbly gait; 3, hind limb paralysis; 4, hind and fore limb paralysis; 5, death. IL-7Rα neutralizing antibody (SB/14, BD Pharmingen) and isotype control was administrated intraperitoneally at 200 μg per mouse every other day. For analysis of CNS-infiltrating cells, both spinal cord and brain were collected and minced from perfused mice, and mononuclear cells were isolated by gradient centrifuge with Percoll (GE Healthcare).

For passive EAE induction with Stat5+/+ or Stat5−/− CD4+ T cells, splenocytes and LNs were harvested 10-14 days post-immunization and passed through a 70 μm cell strainer (BD Falcon). Cells were cultured in vitro for 3 days with MOG35-55 (20 μg/ml) in the presence of IL-23 (5 ng/ml) and IL-1β (2 ng/ml). After harvesting, CD4+ T cells were purified by positive selection to a purity > 90%. CD4+ T cells (2 × 106 in sterile PBS) were injected intraperitoneally into Rag2−/− mice, followed by Pertussis toxin administration on the following day. Mice were observed daily for the signs of EAE as described above. For EAE induction by transferring various TH subsets, similar procedures were performed as described above. Different subsets skewing conditions were as follows: Non-skewed, MOG35-55 only; TH1: MOG35-55 plus IL-12 (10 ng/ml) and anti-IL-4 (5 μg/ml); TH17: MOG35-55 plus TGF-β (3 ng/ml), IL-6 (10 ng/ml), anti-IFN-γ (5 μg/ml) and anti-IL-4 (5 μg/ml); GM-CSF-producing TH: MOG35-55 plus IL-7 (5μg/ml), and anti-IFN-γ (5μg/ml). 6 × 105 CD4+ T cells were transferred per recipient mouse.

T-cell transfer model of colitis

CD4+ T cells were isolated from spleens and lymph nodes of WT, Cd4-Cre; Stat3f/f and Cd4-Cre; Stat5f/f mice by positive selection and magnetic separation (Miltenyi Biotech). Naïve CD4+ T (CD4+CD25CD45RBhi) cells were sorted with BD FACS Aria (purify > 98%). Rag2−/− mice were reconstituted with 4 × 105 naïve CD4+ T cells via intraperitoneal injection. Intestine inflammation was monitored up to 8 weeks. Cell suspensions were prepared from spleen, MLNs, and colon lamina propria by methods modified from previous report52.

Histological analysis

For paraffin-embedded tissues, spinal cords, or colons were fixed in 4% PFA. Sections (5 μm) were stained with hematoxylin and eosin (H&E) to assess immune cell infiltration and inflammation. For frozen tissues, spinal cords were embedded in OCT (Tissue-Tek) and snap frozen on dry ice. Sections (10 μm) were fixed in ice-cold acetone and stained with primary anti-CD4 (Biolegend) and anti-CD11b (eBioscience), followed by incubation with fluorescence-conjugated secondary antibodies (Invitrogen).

Real-time PCR

Total RNA was extracted from cells with RNeasy kit (Qiagen) according to the manufacturer's instruction. Complementary DNA (cDNA) was synthesized with Superscript reverse transcriptase (Invitrogen). Gene expressions were measured by 7500 real-time PCR system (Applied Biosystems) with SYBR qPCR kit (KAPA). Actinb, Gapdh, or Rn18S was used as internal control. The primer sequences are available upon request.

Microarray assay

For microarray analysis, RNA from naïve T cells, TH1, TH17 and GM-CSF-producing TH cells was purified with RNeasy kit (Qiagen). Hybridization targets were amplified and labeled using Applause WT-Amp ST System according to the manufacturer's protocol (NuGEN). Labeled cDNA was hybridized to Affymetrix GeneChip Mouse Gene 1.0 ST according to the manufacturer's instructions. All microarray raw data (CEL files) were analyzed together using the Robust Multichip Average method to obtain the gene expression intensities. Normalization was then performed across all samples based on the cross correlation method53. Normalized data were further log2-tranformed and were used for identification of differentially expressed or TH-cell specific genes. The cutoff fold change threshold of 1.5 was used for differential expression.


GM-CSF level was assayed by Ready-SET-Go ELISA kit (eBioscience) according to the manufactures' instructions.

Chromatin immunoprecipitation assays

CD4+ T cells isolated from Stat5f/f or Cd4-Cre; Stat5f/f mice were activated with plate-bound anti-CD3 and anti-CD28 for 3 days. Cells were stimulated with IL-7 (20 ng/ml) or IL-2 (25 ng/ml) for 45 min. Crosslink was performed by addition of formaldehyde at final concentration of 1% for 10 min followed by quenching with glycine. Cell lysates were fragmented by sonication and precleared with protein G Dynabeads, and subsequently precipitated with anti-STAT5 antibody (Santa Cruz) or normal rabbit IgG (Santa Cruz) overnight at 4 °C. After washing and elution, crosslink reversal was done by incubating at 65 °C for 8 h. The eluted DNA was purified and analyzed by RT-PCR with primers specific to Csf2 promoter as described previously54.


Statistical significance was determined by Student's t-test using GraphPad Prism 6.01. P < 0.05 was considered significant. The P values of clinical scores were determined by one-way multiple-range analysis of variance (ANOVA) for multiple comparisons. Unless otherwise specified, data were presented as mean ± SEM.