Research ArticleMultiple Sclerosis

IL-17 and GM-CSF Expression Are Antagonistically Regulated by Human T Helper Cells

See allHide authors and affiliations

Science Translational Medicine  18 Jun 2014:
Vol. 6, Issue 241, pp. 241ra80
DOI: 10.1126/scitranslmed.3008706


Although T helper 17 (TH17) cells have been acknowledged as crucial mediators of autoimmune tissue damage, the effector cytokines responsible for their pathogenicity still remain poorly defined, particularly in humans. In mouse models of autoimmunity, the pathogenicity of TH17 cells has recently been associated with their production of granulocyte-macrophage colony-stimulating factor (GM-CSF). We analyzed the regulation of GM-CSF expression by human TH cell subsets. Surprisingly, the induction of GM-CSF expression by human TH cells is constrained by the interleukin-23 (IL-23)/ROR-γt/TH17 cell axis but promoted by the IL-12/T-bet/TH1 cell axis. IL-2–mediated signal transducer and activator of transcription 5 (STAT5) signaling induced GM-CSF expression in naïve and memory TH cells, whereas STAT3 signaling blocked it. The opposite effect was observed for IL-17 expression. Ex vivo, GM-CSF+ TH cells that coexpress interferon-γ and T-bet could be distinguished by differential chemokine receptor expression from a previously uncharacterized subset of GM-CSF–only–producing TH cells that did not express TH1, TH2, and TH17 signature cytokines or master transcription factors. Our findings demonstrate distinct and counterregulatory pathways for the generation of IL-17– and GM-CSF–producing cells and also suggest a pathogenic role for GM-CSF+ T cells in the inflamed brain of multiple sclerosis (MS) patients. This provides not only a scientific rationale for depleting T cell–derived GM-CSF in MS patients but also multiple new molecular checkpoints for therapeutic GM-CSF suppression, which, unlike in mice, do not associate with the TH17 but instead with the TH1 axis.


Granulocyte-macrophage colony-stimulating factor (GM-CSF) is a hematopoietic growth factor that is produced by many cell types, including T cells. GM-CSF promotes the maturation and activation of monocytes and dendritic cells and exerts proinflammatory functions by increasing their antigen presentation and release of inflammatory cytokines (1). The onset of autoimmune tissue inflammation is contingent on GM-CSF as has been shown in mouse models of neuroinflammation, arthritis, and myocarditis (25).

In recent years, substantial experimental evidence that supports a crucial role for TH17 cells in the pathogenesis of autoimmune diseases has emerged. In mice, it has been demonstrated that IL-23 is essential for induction of pathogenic properties in T helper 17 (TH17) cells (6, 7), whereas in humans, we could demonstrate a similar role for interleukin-1β (IL-1β) (8). Recently, GM-CSF has been identified as a T cell cytokine that is induced by IL-23 and IL-1β and that is crucial for the pathogenesis of experimental autoimmune encephalomyelitis (EAE) (911). Murine TH17 cells were identified as the chief source of GM-CSF, although T cells with other polarization patterns were also able to produce GM-CSF by mechanisms that remain to be explored. Whereas the TH17 signature cytokine IL-17 was insufficient to drive EAE pathology (12), GM-CSF sustained neuroinflammation via myeloid cells that infiltrated the central nervous system (CNS) and induced IL-23 secretion by dendritic cells, thereby perpetuating TH17 cell generation and chronic inflammation (9, 10). Thus, the missing link between IL-23 and murine TH17 cell pathogenicity was identified to be GM-CSF. The role of T cell–derived GM-CSF in human autoimmune diseases, instead, and its regulation still represent critical gaps of knowledge.

Here, we come to the unexpected conclusion that GM-CSF production in humans is linked to TH1 cells and constrained by the TH17 axis. We could furthermore delineate a population of GM-CSF–only–producing TH cells as a distinct subset of TH cells based on cytokine production, transcriptional regulation, priming requirements, and a unique set of surface markers. GM-CSF–producing T cells had CNS-homing properties and were associated with inflammation in multiple sclerosis (MS).


GM-CSF production is associated with human TH1 but not TH17 cell regulation

Pathogenic properties of TH17 cells have recently been associated with GM-CSF production in murine autoimmune diseases (2). To explore whether GM-CSF production by human T cells is also regulated by the TH17 cell axis, we isolated TH1, TH2, and TH17 cell subsets from the peripheral blood of healthy donors according to the differential expression of chemokine receptors as described previously (8) and assessed GM-CSF secretion after short ex vivo restimulation. Surprisingly, TH1 but not TH17 cells were the main producers of GM-CSF (Fig. 1A). This association of GM-CSF production with the TH1 cell subset was also supported by high coexpression of GM-CSF on the single-cell level with T-bet, the TH1-specific transcriptional master regulator, but low coexpression with ROR-γt, GATA-3, or FOXP3, the TH17, TH2, and regulatory T (Treg) cell transcriptional master regulators, respectively (Fig. 1, B and C, and fig. S1). Collectively, these results demonstrate that, although GM-CSF is expressed by all human TH cell subsets, it is preferentially associated with TH1 rather than with TH17 cells.

Fig. 1. GM-CSF expression does not correlate with TH17 but with TH1 regulation—heterogeneity of GM-CSF–producing TH cells.

(A) TH cell subsets were sorted according to the differential expression of chemokine receptor surface markers as TH1 (CXCR3+CCR4CCR6), TH2 (CXCR3CCR4+CCR6), and TH17 cells (CXCR3+CCR4CCR6+) and restimulated with anti-CD3 and phorbol 12,13-dibutyrate (PDBu) for 8 hours. Supernatant was analyzed by enzyme-linked immunosorbent assay (ELISA) (n = 3, mean ± SEM). (B and C) Flow cytometry in ex vivo isolated CD4+ T cells (B) and pooled data (mean ± SEM, n = 6) (C). (D) Expression of IFN-γ and IL-17 in GM-CSF+ and GM-CSF freshly isolated CD4+ T cells and expression of IL-4 in IFN-γ and IL-17 negatively subgated GM-CSF+ and GM-CSF CD4+ T cells shown by intracellular cytokine staining and flow cytometry. One representative experiment is shown. (E) Pooled data (n = 3) are shown as the percentage (pie chart and table) of cytokine- and cytokine combination–positive cells within freshly isolated CD4+GM-CSF+ T cells. (F) Isolation and intracellular cytokine staining of a resting CD4+ T cell clone after 14 days of clonal expansion under nonpolarizing conditions (n = 600, representative of about 2% of all analyzed CD4+ T cell clones). (G) Real-time polymerase chain reaction (PCR) analysis of the expression TBX21, RORC2, and GATA3 transcripts in TH1, TH2, TH17, and GM-CSF–only clones (mean ± SEM; three individual clones of each type). GM-CSF–only clones were defined as CD4+ T cell clones with >70% GM-CSF–, <2% IFN-γ–, <2% IL-4–, and <2% IL-17–producing cells. Resting T cell clones were restimulated for 3 hours with phorbol 12-myristate 13-acetate (PMA)/ionomycin before RNA extraction. A.U., arbitrary units. All P values were determined by the unpaired Student’s t test.

GM-CSF–only–producing T cells are a distinct population of TH cells in humans

Despite the association of GM-CSF expression with TH1 regulation, we also observed a subset of GM-CSF+ cells that was negative for interferon-γ (IFN-γ) and T-bet (Fig. 1, B and D). To assess the heterogeneity of GM-CSF–producing TH cells in more detail, we analyzed the simultaneous coexpression of GM-CSF with IFN-γ, IL-17, and IL-4 in freshly ex vivo isolated resting CD4+ T cells (Fig. 1, D and E). 21 ± 2% (mean ± SEM) of CD4+ TH cells expressed GM-CSF ex vivo. IFN-γ expression was enriched in GM-CSF+ as compared to GM-CSF TH cells (sevenfold). We also observed a small population of IL-17–producing cells that coexpressed GM-CSF, although a third of them were also simultaneously IFN-γ coproducers. A smaller proportion of CD4+ T cells coexpressed GM-CSF with IL-4 or simultaneously with multiple cytokines (Fig. 1, D and E). 23 ± 3% (mean ± SEM) of freshly ex vivo isolated CD4+GM-CSF+ cells did not coexpress any TH cell signature cytokine, which indicated the existence of a separate GM-CSF+IFN-γIL-17IL-4 TH cell subset (Fig. 1, D and E). These GM-CSF–only–producing cells constituted 2.2 ± 0.4% (mean ± SEM) of the entire CD4+ TH cell population in healthy donors. In addition, we performed further functional profiling of GM-CSF–producing TH cells and could exclude coexpression with other lineage-defining cytokines such as IL-9 (TH9) as well as IL-13, IL-10, and IL-21 for the majority of GM-CSF–producing TH cells (fig. S2). The majority of IL-22+ cells within ex vivo isolated CD4+ cells coexpressed GM-CSF, although we could also detect a population of GM-CSFIL-22+ and GM-CSF+IL-22 cells (fig. S3A). TH22 cells have recently been characterized as CCR10+CCR6+CCR4+CXCR3 cells (13, 14). They contained a high fraction of GM-CSF+ cells (42 ± 3; mean ± SEM) (fig. S3B). To further demonstrate the existence of IL-22+ and IL-22 GM-CSF–producing TH cells, we cloned CCR10+ T cells and could distinguish these two T cell populations on the basis of TH cell clones (fig. S3C). Addition of the aryl hydrocarbon receptor ligand FICZ to naïve T cell cultures augmented IL-22 in accordance with previous studies (13) but not GM-CSF, which indicated differential priming requirements for both cytokines (fig. S3D).

Together, these data demonstrate the existence of GM-CSF–only–producing TH cells, which are in addition to all lineage-defining cytokines also independent of IL-22 expression and independent of multiple other TH cell cytokines.

To further confirm the existence of a distinct GM-CSF–only–producing TH cell population, we cloned in vivo primed memory CD4+ TH cells by limiting dilution and analyzed the cytokine profile by flow cytometry on the single-cell level after a 2-week cloning culture period. About 2% of all analyzed CD4+ T cell clones (n > 600 T cell clones) had high expression levels of GM-CSF but did not express IFN-γ, IL-4, IL-17, or IL-22 (Fig. 1F). On the basis of the ex vivo analysis of the polyclonal T cell population and on the isolated T cell clones, we conclude that the GM-CSF–only–producing T cell phenotype is a stable and distinct property of a human memory T cell subset.

Because TH cell subsets can traditionally be distinguished on the basis of their master transcription factors, we isolated in vivo primed TH1, TH2, and TH17 clones and also TH clones that produced GM-CSF but were negative for IFN-γ, IL-17, and IL-4 expression, and analyzed transcript levels of T-bet, GATA-3, and ROR-γt (Fig. 1G). As expected, T-bet, GATA-3, and ROR-γt transcripts were selectively expressed by TH1, TH2, and TH17 clones, respectively. GM-CSF+IFN-γIL-17IL-4 T cell clones, however, had low or undetectable expression of all three master transcription factors, demonstrating that their transcriptional regulation was different from that of TH1, TH2, and TH17 cells (Fig. 1G). Collectively, these data support the existence of a distinct population of GM-CSF–only–producing TH cells in humans.

GM-CSF–producing TH cells have specific migration properties

Effector functions and migratory capacity are co-regulated during TH cell differentiation. We therefore analyzed whether GM-CSF–producing TH cells have a distinct chemokine receptor profile that would control tissue-specific migration and that would provide surface markers for their detection and ex vivo isolation. We first compared GM-CSF expression in central versus effector memory T cells (15), which can be distinguished on the basis of the lymph node–homing marker CCR7 (Fig. 2, A and B). GM-CSF–producing cells were absent in naïve TH cells but were significantly enriched in CCR7 effector memory compared to CCR7+ central memory T cells. We then freshly isolated CD45RA memory TH cells that were positive or negative for the expression of specific chemokine receptors and analyzed ex vivo cytokine expression on the single-cell level (Fig. 2, C to E). We first focused on the chemokine receptors CCR6, CXCR3, CCR4, and CCR10 that we and others have previously used to isolate TH1, TH2, TH17, and TH22 cell subsets (8, 13, 16), respectively, and that direct migration to the CNS (1719) and skin (20) (Fig. 2, C to E). Significantly more cells expressed GM-CSF within the CXCR3+, CCR6+, and CCR10+ compartments than within their respective chemokine receptor–negative counterparts, whereas CCR4 expression did not distinguish between GM-CSF+ and GM-CSF T cells (Fig. 2, C and D). CCR4 did, however, as well as CXCR3 and CCR10, serve to distinguish between IFN-γ+ and IFN-γ subsets within GM-CSF–producing TH cells (Fig. 2, C and E, and fig. S4).

Fig. 2. A distinct chemokine receptor expression profile characterizes GM-CSF–producing TH cells.

(A and B) Intracellular cytokine staining and flow cytometry of freshly isolated naïve (CCR7+CD45RA+), central (CCR7+CD45RA), and effector memory (CCR7CD45RA) TH cells. (A) Data are representative of three individual experiments and donors. (B) Pooled data (mean ± SEM, n = 3). (C and D) Comparison of freshly isolated peripheral blood CD4+CD45RA cells sorted positive or negative for the indicated chemokine receptors by intracellular cytokine staining and flow cytometry. (C) Data are representative of six individual experiments and donors. (D) Pooled data (mean ± SEM, n = 6). Numbers above bars indicate P values (Student’s t test). (E) Comparison of the percentage of IFN-γ+ TH cells within the GM-CSF–producing CD4+CD45RA cell population by intracellular cytokine staining and flow cytometry. Shown are pooled data (mean ± SEM, n = 3). Numbers above bars indicate P values (Student’s t test). (F) Intracellular cytokine staining and flow cytometry of freshly isolated CXCR3CCR4+CCR6CCR10+ T cells. Data are representative of four experiments. (G) Cytometric bead array (CBA) and ELISA of cytokine production by freshly isolated TH cell subsets isolated according to the differential expression of chemokine receptor surface markers after 36 hours of stimulation with CD3 and CD28 monoclonal antibodies (mAbs) followed by 3-hour PMA/ionomycin treatment (mean ± SEM, n = 3).

We next searched for a combination of chemokine receptor surface markers that would identify GM-CSF–only–producing cells ex vivo. We therefore isolated CD4+ TH cells according to the combinatorial expression of chemokine receptor surface markers and analyzed ex vivo cytokine expression upon PMA and ionomycin restimulation on the single-cell level and by ELISA. CCR10+CCR4+CXCR3CCR6 cell surface expression stringently identified GM-CSF–only–producing TH cells (Fig. 2, F and G). Analysis of cytokine secretion confirmed that CCR10+CCR4+CXCR3CCR6 cells produced GM-CSF but not IL-17, IFN-γ, and IL-22 and low levels of IL-4 (Fig. 2G). TH1, TH2, and TH17 cells, which could be identified according to the differential expression of CXCR3, CCR4, and CCR6, not only produced their proprietary cytokines as expected from previous reports (8) but also coproduced GM-CSF with highest levels observed within the CXCR3+CCR4CCR6 subset that enriched for TH1 cells but not within the CXCR3CCR4+CCR6+ subset that enriched for TH17 cells. To further confirm that GM-CSF–only–producing cells could indeed be characterized as CCR10+CCR4+CXCR3CCR6 cells and distinguished from their GM-CSF+IFN-γ+ counterpart, which was contained within the CXCR3+ and CCR4 population (Fig. 2, C and E), we analyzed T-bet expression on the transcriptional and protein level. T-bet expression levels of ex vivo isolated CCR10+CCR4+CXCR3CCR6 cells were similar to naïve T cells, whereas CCR4CXCR3+ cells had high T-bet expression (fig. S5, A and B).

Collectively, these data not only show that the production of GM-CSF is associated with specific migration properties through the preferential expression of chemokine receptors such as CXCR3, CCR6, CCR7, and CCR10. They also support the association of GM-CSF with the TH1 axis and, in addition, identified a subset of GM-CSF–only–producing cells that could be characterized and phenotypically distinguished from other TH cell subsets by a unique combination of chemokine receptor surface markers.

GM-CSF–only TH cells represent a stable TH cell subset with the ability to modulate cytokine expression in distinct extracellular cytokine milieus

We next determined the influence of polarizing cytokine milieus on the stability of GM-CSF expression by freshly isolated TH cells (Fig. 3). Memory TH cells down-regulated GM-CSF production upon restimulation in the presence of the TH17-polarizing cytokines IL-6 and IL-23 and, in particular, by the combination of IL-1β and IL-23, which instead enhanced IL-17 production (Fig. 3A). In contrast, IL-12, which regulates TH1 and antagonizes TH17 cell differentiation, significantly enhanced GM-CSF expression by human memory TH cells, whereas it has been reported to suppress GM-CSF induction in murine T cells (9). IL-27 suppressed GM-CSF production by memory CD4+ T cells, which is in accordance with previously published work (21) (Fig. 3A). These data demonstrate preferential association of GM-CSF expression with TH1 but not TH17 regulation.

Fig. 3. Stability and plasticity of GM-CSF–producing TH cells.

(A) CD4+CD45RA memory T cells were cultured in the presence or absence of cytokines, counted on day 5, and restimulated at equal cell numbers with anti-CD3 (2 μg/ml) and PDBu for 8 hours (mean ± SEM, n = 3). (B). TH cell clones were generated during a 14-day culture period with irradiated allogeneic feeder cells and phytohemagglutinin. GM-CSF–only–producing TH cell clones (n = 6), which were defined by less than 2% IFN-γ, IL-17, or IL-4 expression, were restimulated repeatedly with CD3 and CD28 mAb on days 0 and 10. Intracellular cytokine staining and flow cytometry were performed every 5 days. Data are means ± SEM. (C) GM-CSF–only–producing T cell clones (<2% IFN-γ, IL-17, or IL-4 expression) were restimulated in the presence or absence of the indicated TH1, TH2, and TH17 lineage–skewing cytokines for 5 days with plate-bound anti-CD3 and anti-CD28 (48 hours). Intracellular cytokine staining and flow cytometry. Data are representative of seven experiments and donors.

The modulation of GM-CSF expression in total memory TH cells could be biased by a preferential response of specific TH cell subsets to the respective cytokines and their different relative proportions within the memory T cell pool. We therefore tested the effects of TH1- versus TH17-promoting cytokines on GM-CSF expression by individual freshly isolated TH cell subsets, which could be distinguished on the basis of the differential expression of chemokine receptor surface markers (fig. S6). All tested TH cell subsets including TH17 cells significantly up-regulated GM-CSF expression in response to restimulation with IL-12, although TH2 cells were less responsive to IL-12 restimulation in accordance with lower expression levels of IL-12Rβ2 and thus reduced signal transducer and activator of transcription 4 (STAT4) signaling (fig. S7) (22). The combination of IL-1β + IL-6 + TGF-β (transforming growth factor–β) instead significantly reduced GM-CSF expression in all human TH cell subsets, corroborating antagonistic regulation of GM-CSF by TH1 and TH17 cytokines (fig. S6).

To further assess the stability and plasticity of GM-CSF–only TH cells and to avoid that preferential outgrowth of undefined T cells contributed to the observed effects, we cloned in vivo primed memory TH cells long term (14 days) under nonpolarizing conditions and restimulated GM-CSF–producing TH cell clones that were negative for IFN-γ, IL-4, and IL-17 (GM-CSF–only clones) again in the absence or presence of polarizing cytokines. GM-CSF–only–producing cells did not change their cytokine profile under nonpolarizing conditions during the cloning procedure (first stimulation for 14 days) as well as upon repetitive restimulations with anti-CD3 and anti-CD28 over a total culture period of more than a month (Fig. 3B). They up-regulated neither IL-4 nor IL-17 under TH2- or various TH17-polarizing conditions, respectively, but acquired the ability to coexpress IFN-γ in the presence of IL-12 as shown on the single-cell level and by cytokine secretion (Fig. 3C and fig. S8).

Collectively, the data demonstrate that GM-CSF–producing memory T cells remained stable upon polyclonal repetitive restimulations and expansion under TH2 and TH17 cell conditions but had the plasticity to acquire IFN-γ expression as memory cells under TH1 conditions.

GM-CSF– and IL-17–producing TH cells have differential priming requirements

The identification of GM-CSF–only–producing T cells as a distinct TH cell subset raised the question, which cytokines might be required for the induction of GM-CSF in naïve TH cells? In mice, it has been demonstrated that IL-23 and IL-1β induce GM-CSF, whereas TH1 cytokines inhibit the polarization of GM-CSF–producing cells (9, 10). To analyze the effect of these cytokines on GM-CSF polarization in human naïve TH cells, we isolated naïve TH cells from adult peripheral blood and stimulated them in the presence or absence of IL-12 and TH17 cell–inducing cytokines or cytokine combinations (23) (Fig. 4, A and B). In the absence of exogenous cytokines, a high proportion of cells up-regulated GM-CSF upon activation and proliferation (29 ± 1%, mean ± SEM). IL-12, which induced TH1 and inhibited TH17 cell generation in accordance with published reports (Fig. 4B) (24), significantly increased GM-CSF production in a STAT4-dependent manner (fig. S9). The combination of IL-1β and IL-6, which promotes the generation of human TH17 cells (8, 24), suppressed GM-CSF production instead (Fig. 4A). Recently, proinflammatory (IL-1β + IL-6 + IL-23) and anti-inflammatory (TGF-β + IL-6) TH17 cell–priming conditions have been demonstrated in mice (6). We showed that both reduced GM-CSF induction compared to the control condition during human naïve T cell polarization (Fig. 4B). This was primarily due to the GM-CSF–inhibitory effect of IL-6 because IL-1β and TGF-β did not alter GM-CSF expression levels significantly. Likewise, tumor necrosis factor–α (TNF-α) and IL-6, which enhance TH22 generation (14), suppressed the development of GM-CSF–producing cells. IL-27, which is known to support the generation of Tr1 cells (25), suppressed GM-CSF up-regulation in naïve TH cells (Fig. 4B). Although IL-23 was reported to induce GM-CSF production in mice (9, 10), we could show that it did not in human naïve T cells and even had an inhibitory effect (Fig. 4B). IL-6 and IL-1β modulated the generation of GM-CSF–single versus GM-CSF/IFN-γ–coproducing TH cells (Fig. 4C) that we were also able to detect ex vivo (fig. S4). IL-6 had a strong down-regulatory effect on IFN-γ up-regulation and thus favored the generation of GM-CSF–only cells but reduced overall GM-CSF expression, whereas IL-1β favored the generation of GM-CSF/IFN-γ–coproducing TH cells from naïve T cell precursors (Fig. 4C).

Fig. 4. Antagonistic regulation of GM-CSF and IL-17 induction in naïve TH cells.

(A to C) Naïve carboxyfluorescein diacetate succinimidyl ester (CFSE)–labeled TH cells were freshly sorted ex vivo from peripheral blood and stimulated for 5 days in the presence or absence of the indicated cytokines with plate-bound anti-CD3 and anti-CD28 (48 hours) as well as IL-2 (50 IU/ml). Intracellular cytokine staining and flow cytometry were done on day 5. (A) One representative experiment is shown (n = 4). (B) Pooled data from four to eight experiments for GM-CSF induction, from four experiments for IL-17 induction, and from six experiments for IFN-γ induction; n.d., not detectable; Student’s t test, compared to control. (C) Data are representative of three experiments. (D) Intracellular cytokine staining and flow cytometry were done on day 4 (anti-CD3 and anti-CD28 stimulation for 48 hours) after stable lentiviral transduction of naïve TH cells with an empty lentiviral vector (pCCL) or with a vector containing human RORC2 (pCCL.RORC2) in IL-2–containing complete medium (50 IU/ml). Gated on CD271+ cells. Data are representative of three experiments. (E) TH cells were restimulated with plate-bound anti-CD3 and PDBu for 8 hours for collection of supernatant after transduction with a lentiviral vector containing pCCL or pCCL.RORC2. Cells were sorted CD271+ after 48 hours of anti-CD3 and anti-CD28 stimulation in the presence of the lentiviral vectors. Cytokines were quantified by CBA (n = 3, mean ± SEM).

The inhibitory effects of TH17 cell–priming cytokine combinations on the polarization of GM-CSF–producing TH cells prompted us to assess whether ROR-γt is functionally involved in the polarization of GM-CSF–producing TH cells. Lentiviral overexpression of RORC2 (pCCL.RORC2) in naïve TH cells significantly reduced the frequency of GM-CSF+ cells in transduced cells compared to overexpression with the control lentivirus (pCCL) (Fig. 4D). As expected, RORC2 overexpression also reduced IFN-γ and increased IL-17 induction in accordance with published reports (26) (Fig. 4E). Collectively, these data demonstrate a negative regulatory effect of ROR-γt on GM-CSF production in differentiating cells and thus corroborate counterregulatory mechanisms for the development of IL-17– and GM-CSF–producing cells.

GM-CSF–producing TH cells are induced by STAT5 but inhibited by STAT3 signaling

The high proportion of cells that spontaneously differentiated into GM-CSF–producing TH cells in the absence of polarizing cytokines (Fig. 4) suggested that autocrine cytokine signaling might be responsible for induction of GM-CSF expression. Autocrine GM-CSF signaling has been excluded previously by demonstrating the absence of GM-CSF receptor expression on CD3+ T cells (10, 27). IL-2 is known to be produced by activated T cells and to regulate differentiation into TH cell lineages by modulation of key cytokine receptor expression (28) and by regulating the relative activation of STAT3 and STAT5 (29, 30). We therefore tested the effect of IL-2 on GM-CSF induction during TH cell differentiation (Fig. 5). To this end, we stimulated naïve TH cells in the absence (anti–IL-2) or presence of titrated concentrations of IL-2. We observed higher GM-CSF expression with increasing concentrations of IL-2 (Fig. 5, A and B). To exclude that differences in GM-CSF production were due to effects on proliferation or more advanced terminal differentiation, we tracked T cell proliferation by labeling the cells with CFSE. The absence of IL-2 signaling by addition of anti–IL-2 prevented proliferation and GM-CSF induction (Fig. 5, A and B). Endogenous IL-2 production (medium control) was sufficient to induce three rounds of division (Fig. 5B). This was further amplified by exogenous IL-2, which allowed us to track up to six divisions (Fig. 5, A and B). We compared the percentage of GM-CSF–producing cells between each division and compared the effect of titrated IL-2 concentrations on the percentage of GM-CSF–producing cells within each division. GM-CSF expression increased with cell division. In addition, we observed an increase in the percentage of GM-CSF–producing cells per cell cycle with increasing concentrations of IL-2 (Fig. 5, A and B). These data demonstrate that IL-2 directly induced GM-CSF production in differentiating cells even in an activation- and proliferation-independent way.

Fig. 5. Antagonistic regulation of GM-CSF induction by STAT5 and STAT3 signaling.

(A and B) Intracellular cytokine staining and flow cytometry of naïve CFSE-labeled TH cells on day 5 after 48-hour activation with plate-bound anti-CD3 and anti-CD28 in the presence or absence (anti–IL-2) of titrated concentrations of IL-2 during the 5-day culture period. (A) Numbers in the upper right corners indicate the percentage of total GM-CSF+ cells. Vertical numbers indicate percentage of GM-CSF+ cells out of all cells within each indicated mitosis. Vertical lines separate cells of individual mitoses (M). (B) Pooled data of three individual experiments and donors (n = 2 for condition 50 IU IL-2). (C and D) Intracellular cytokine staining and flow cytometry of naïve CFSE-labeled TH cells on day 5 after 48 hours of activation with plate-bound anti-CD3 and anti-CD28 in the presence or absence of cytokines, which were added at the start of the culture. (C) STAT5 inhibitor (STAT5i) was added after 24 hours of culture. (D) STAT3 inhibitor (STAT3i) was added after 4 hours and anti–IL-2 after 24 hours of culture. Data are representative of three individual experiments and donors. Numbers in quadrants indicate percentage of GM-CSF+ cells.

To further corroborate this finding, we added a pharmacological STAT5 inhibitor to IL-2–stimulated naïve TH cells. It was added after 24 hours of stimulation with anti-CD3 and anti-CD28 to allow the naïve CD4+ T cells to enter mitosis. GM-CSF production was blocked after STAT5 inhibition (Fig. 5C), again supporting IL-2–mediated but activation- and proliferation-independent GM-CSF up-regulation.

IL-2 signaling via STAT5 has been shown to constrain TH17 cell generation by competitive inhibition of STAT3 binding to the Il17 locus (30). These counterregulatory functions of STAT5 and STAT3 and our finding that IL-2–mediated STAT5 signaling enhanced GM-CSF production prompted us to test whether STAT3 signaling might instead suppress GM-CSF production. We therefore stimulated naïve TH cells in the presence or absence of STAT3 activating cytokines as well as anti–IL-2 and added a pharmacological STAT3 inhibitor after 4 hours of activation. Addition of IL-6, IL-23, and IL-21 strongly suppressed GM-CSF expression. Upon addition of the STAT3 inhibitor, TH cells regained GM-CSF expression (Fig. 5D), which demonstrated that GM-CSF up-regulation in differentiating TH cells was constrained by STAT3 signaling. Collectively, these data show antagonistic regulation of GM-CSF and IL-17 on the level of STAT transcription factors. This is supported by evolutionary conserved binding domains for STAT5 and STAT3 in the Csf2 promoter (fig. S10).

GM-CSF–producing TH cells are associated with MS

The chemokine receptors CXCR3 and CCR6 have been reported to arm TH cells with the ability to migrate to the inflamed CNS (19). We could demonstrate here that both chemokine receptors enriched for GM-CSF– and IFN-γ–producing TH cells (Fig. 2, C to E). To test whether GM-CSF–producing human TH cells were also associated with inflammation in MS, we isolated TH cells from the cerebrospinal fluid as well as blood of patients with active MS flares and of control patients with noninflammatory neurological pathologies. TH cells from the cerebrospinal fluid of MS patients had sixfold higher expression levels of GM-CSF ex vivo compared to control patients (Fig. 6, A and B). We also observed elevated frequencies of IFN-γ+ but not IL-17+ T cells. Cytokine expression in TH cells isolated ex vivo from the peripheral blood did not significantly differ between MS patients, control patients, and healthy donors (Fig. 6C).

Fig. 6. GM-CSF–producing TH cells in MS.

(A and B) Intracellular cytokine staining and flow cytometry of CD3+CD4+ T cells isolated from the cerebrospinal fluid (CSF) of MS (n = 7) and control patients (non-MS, n = 7). Shown is one representative experiment. (B) Pooled data from four to seven experiments (Student’s t test). (C) Intracellular cytokine staining and flow cytometry of CD4+ T cells freshly isolated from the peripheral blood [n = 6 to 8; P = 0.71, one-way analysis of variance (ANOVA)]. Numbers in quadrants indicate percentage of cells. Data are representative of six experiments.

Collectively, these data suggest an association of GM-CSF–producing TH cells with MS and most importantly corroborate the dissociation of IL-17 and GM-CSF production within human TH cells on the single-cell level in a relevant autoimmune disease such as MS in vivo.


Studies in the EAE mouse model have associated GM-CSF production with the murine TH17 cell lineage and have established a causal link between this cytokine and the role of TH17 cells in murine autoimmune pathology (2). Here, we demonstrate that GM-CSF production in humans is instead constrained by the TH17 cell regulation axis and linked to the TH1 cell subset. We furthermore identified a population of GM-CSF–only–producing TH cells as a distinct subset of TH cells, on the basis of cytokine production, surface markers, transcriptional regulation, and priming requirements. GM-CSF–producing TH cells had CNS-homing properties and were associated with autoimmune inflammation in MS.

TH17 cells have an arsenal of proinflammatory cytokines at their disposal. Although GM-CSF production has been closely linked to TH17 cell–mediated tissue inflammation in mouse models of autoimmunity (9, 10), our study shows that GM-CSF is not a specific cytokine of human TH17 cells given its disconnection from IL-17 expression, its production by ROR-γt TH cells, and its down-regulation by TH17 cell–priming cytokines. We also observed suppression of GM-CSF production by ectopic ROR-γt expression in differentiating TH cells. This is in line with direct regulation of Csf2 transcription by ROR-γt via an evolutionary conserved ROR response element that has previously been reported (9). In mice, however, some discord about the ability of ROR-γt–deficient T cells to produce GM-CSF remains (9, 10).

Although previous observations in mice have attributed a GM-CSF down-regulatory function to the combination of the TH1 cytokines IL-12 and IFN-γ (9), we observed co-regulation of GM-CSF and IFN-γ in humans. GM-CSF–producing TH cells were heterogeneous with respect to cytokine coexpression but preferentially coproduced IFN-γ. The majority of GM-CSF+ TH cells also coexpressed T-bet and was contained within the CXCR3+ but CCR4 TH cell population, which is known to enrich for IFN-γ–producing cells (8). In addition, IL-12 directly up-regulated GM-CSF expression by a STAT4-dependent mechanism in naïve differentiating as well as freshly isolated memory TH cells. The reason for this discrepancy between the preferential association of GM-CSF production with TH1 versus TH17 cells in mouse and man is currently not clear. Species differences together with the inherent experimental differences that mouse and human studies entail (31) could account for this observation. Despite the disconnection of human GM-CSF from the TH17 cell axis and its preferential association with TH1 cells, it is still coproduced to varying extent by all classical T cell subsets, including TH1, TH2, TH17, and TH22 cells. This is in accordance with previous findings in the EAE mouse model, which demonstrated spontaneous in vivo acquisition of GM-CSF production by in vitro polarized GM-CSF non-TH17 cells and thus GM-CSF production beyond the limitations of the TH17 cell lineage (9). It is also in accordance with the IL-2–mediated mechanism that we provide for GM-CSF up-regulation in TH17 and other effector TH cells and which could potentially also reconcile the phenomenon that mouse TH1 cells can coproduce GM-CSF despite antagonistic regulation of GM-CSF by the TH1 axis (IL-12, T-bet) (9, 10). Human TH17 cells, however, have a relative impairment to produce IL-2 in response to CD3 and CD28 stimulation as well as a reduced ability to respond to IL-2 (32). This is in line with our finding that human TH17 cells have reduced production of GM-CSF as compared to TH1 cells, which readily produce and respond to IL-2 and whose differentiation is even promoted by STAT5 signaling (33).

We demonstrated that naïve human TH cells acquired the ability to produce GM-CSF in the absence of polarizing cytokines in antigen-presenting cell–free cultures. A similar observation was made when TH22 cells were identified as a distinct TH cell subset (13). Our study revealed that GM-CSF expression was promoted by autocrine IL-2–mediated STAT5 signaling. STAT3 activating cytokines, including IL-6, IL-21, and IL-23, had a down-modulatory effect on GM-CSF production instead. This is in accordance with a recent study in mice that showed STAT3-driven suppression of GM-CSF production in total CD4+ T cells via IL-21 signaling (34), although it remains to be explained how IL-23 and IL-6, which also signal via STAT3, at the same time promote GM-CSF in mice. Together, these data suggest that the STAT5/STAT3 ratio regulates GM-CSF production. This is supported by binding sites for STAT5 and STAT3 on the GM-CSF promoter. Opposing actions of STAT5 and STAT3 have previously also been reported for IL-17 expression, but with STAT3 promoting and STAT5 inhibiting the production of IL-17 (29, 30). This confirms reciprocal regulation of GM-CSF and IL-17 also on the transcriptional level of STATs.

TH cell subsets are traditionally defined by their cytokine profile, priming requirements, and transcriptional master regulators. In the case of a small subset of GM-CSF–producing TH cells, the GM-CSF–only T cells, we could exclude the association with the classical TH1, TH2, TH17, and TH22 cell subsets because of a lack of IFN-γ, IL-4, IL-17, and IL-22 coproduction. The identification of TH22 and TH9 cells has demonstrated that signature cytokines can be shared by different TH cell subsets. IL-9 is also produced by TH17 and Treg cells (35, 36), whereas IL-22 is shared by TH17 and TH1 cells (37, 38). Likewise, we found GM-CSF production beyond TH cell lineage boundaries. We were able to isolate GM-CSF–only–producing cells ex vivo based on the differential expression of chemokine receptor surface markers, which characterized and distinguished this T cell population from other TH cell subsets. GM-CSF–only memory T cells did not acquire IL-4– or IL-17–producing properties in TH2 or TH17 cytokine milieus. However, they up-regulated IFN-γ after restimulation with IL-12, which demonstrates their selective plasticity to acquire a TH1 phenotype and thus provides new insights into precursor-product relationships of polyfunctional TH cell subsets.

Despite previous reports about the regulation of GM-CSF by the T cell receptor (TCR)–responsive transcription factors NFκB (nuclear factor κB) and NFAT (nuclear factor of activated T cells) (39), a transcriptional master regulator for GM-CSF–only–producing cells has not yet been identified so far. Therefore, with respect to the classical definition of TH cell subsets, we suggest that GM-CSF–only–producing TH cells should not be termed TH–GM-CSF cells at this stage. Still, regulation by T-bet, ROR-γt, and GATA-3 could be excluded, which demonstrated independence of TH1, TH2, and TH17 cell subsets, respectively. In light of the recent change in the paradigm that master transcriptional regulators do not orchestrate cell identity but have restricted actions on signature cytokines, it will be interesting to identify the global enhancer landscape for GM-CSF–only–producing cells and its impact on the specification of these cells as a distinct TH cell subset (40, 41).

Our findings also have implications for new therapeutic strategies for autoimmune diseases. Not only various mouse models of autoimmunity but also human rheumatoid arthritis and asthma have demonstrated a therapeutic benefit of GM-CSF depletion and have thus highlighted the pathogenic role of GM-CSF (42, 43). However, in Crohn’s disease, GM-CSF administration improved gut inflammation and mucosal repair (44). We found that effector T cells isolated from the cerebrospinal fluid of patients with MS had several-fold higher GM-CSF expression levels compared to cells from patients without inflammatory CNS disease. This suggests a pathogenic role of GM-CSF–producing TH cells in autoimmune CNS inflammation, which will have to be corroborated by future clinical trials.

Our data also identified upstream factors for targeting GM-CSF including cytokine requirements for GM-CSF polarization as well as STAT transcriptional regulators. Therefore, direct GM-CSF depletion could now be complemented by upstream targeting strategies, which might even increase the specificity of treatment by limiting the number of targeted cell types. This is reminiscent of current therapeutic approaches for TH17 cell–mediated diseases such as psoriasis, where upstream targeting with ustekinumab (anti-p40), as well as downstream IL-17–targeting strategies (ixekizumab and brodalumab), is now followed with great clinical efficacy (45).

IL-2 treatment, which is currently being tested in clinical trials with the rationale to induce Treg cells (46), might likewise increase GM-CSF expression in effector T cells or even Treg cells, which are known to have high IL-2 receptor expression. It might therefore represent an attractive upstream target of GM-CSF production. Anti-CD25 treatment (daclizumab) was efficacious in MS (47). Janus kinase inhibitors, which inhibit not only IL-2 signaling but also other cytokines, were successful in various autoimmune diseases as well (48). The downstream effects of these treatments on T cell–derived GM-CSF have not been investigated yet but might account for the benefit or failure of the respective treatments. Genetic variants in the gene encoding IL-2Rα have been highly associated with MS and therefore link IL-2/STAT5 signaling with disease susceptibility (49). Whether these polymorphisms are associated with downstream alterations in GM-CSF secretion by TH cells has never been determined. Our observation that TH cells in the cerebrospinal fluid do not express IL-17 suggests a minor role for IL-17 in MS pathogenesis, which is in line with the data in the EAE mouse model (12) and with insufficient response rates to anti–IL-17 treatment in clinical trials as compared to other autoimmune diseases such as psoriasis. A phase 1b clinical trial with anti–GM-CSF in MS is currently being conducted (NCT01517282). On the basis of our immunological dissection of GM-CSF regulation by human TH cells, we expect that further clinical studies with GM-CSF–blocking antibodies will elegantly link IL-2Rα polymorphisms and IL-2/STAT5–mediated GM-CSF secretion with MS and thus serve as a proof of concept for our findings.

A question that emerged from our study and still remains to be resolved relates to the differential contribution of GM-CSF–only–producing TH cells to health and disease as compared to classical TH cell subsets, which can coproduce GM-CSF. Is GM-CSF per se a proinflammatory cytokine irrespective of its T cell source or is its pathogenicity determined in the context of a particular cytokine microenvironment? From studying TH17 cells, we have recently learned about the existence of two distinct types of TH17 cells, which differed in their pro- versus anti-inflammatory functionalities based on either IFN-γ or IL-10 coexpression (8). A similar scenario could be envisioned for different types of GM-CSF–producing TH cells. In addition, we have shown that GM-CSF–only and classical TH cell subsets, which coproduce GM-CSF, differ in their chemokine receptor profile and thus migration properties and tissue location. This topographic specialization could also translate into different biological functions with implications for the pathogenesis and treatment of tissue-restricted chronic inflammatory diseases.

Here, we have dissected the pathways of GM-CSF regulation by human TH cells and revealed notable differences to previous fundamental findings in mice (fig. S11). Further studies are required to understand the precise transcriptional regulation and to determine the biological significance of GM-CSF–only–producing T cells as compared to GM-CSF–producing classical TH cell subsets in humans.


Study design

This is an experimental laboratory study performed with human blood and tissue (CSF) samples. The objective was to study the regulation of GM-CSF expression in human TH cells with the translational perspective to gain insights into this cytokine’s molecular regulation for therapeutic targeting purposes. Ethics approval was obtained from the Institutional Review Board of the Charité–Universitätsmedizin Berlin, Germany (EA1/221/11, EA1/293/12). All work was carried out in accordance with the Declaration of Helsinki for experiments involving humans. Study components were not predefined. The number of replicates is indicated for each experiment in the respective figure legends. Experiments on T cells from the cerebrospinal fluid from randomly selected hospitalized patients were performed in the laboratory before knowledge about the exact diagnosis (MS versus non-MS) on the day of cerebrospinal fluid acquisition (blinded) and reanalyzed after full clinical information on the diagnosis had been acquired and only if the diagnosis could firmly be established by the clinicians. Mechanistic studies on cells from healthy blood donors were performed with in vitro assays without blinding or randomization. Healthy leukapheresis blood and buffy coats were randomly obtained from the Charité Blood Bank and Deutsches Rotes Kreuz (DRK).

Cell purification and sorting

Peripheral blood mononuclear cells (PBMCs) were isolated by density gradient sedimentation using Ficoll-Paque Plus (GE Healthcare). CD4+ T cells were isolated from PBMCs by positive selection with CD4-specific microbeads (Miltenyi Biotec). TH cell subsets were sorted to at least 97% purity as follows: TH1 subset, CXCR3+CCR4CCR6CD45RACD25CD8; TH2 subset, CXCR3CCR4+CCR6CD45RACD25CD8; TH17 subset, CXCR3CCR4+CCR6+CD45RACD25CD8; GM-CSF–only–producing T cells, CXCR3CCR4+CCR6CCR10+CD45RACD25CD8; and naïve T cells isolated as CD45RA+CD45ROCCR7+CD25CD8 to a purity of more than 99%. The antibodies used for fluorescence-activated cell sorting were identical to those we have described previously (8). Cells were sorted with FACSAria (BD Biosciences). Healthy leukapheresis blood and buffy coats were obtained from the Charité Blood Bank and DRK. Cerebrospinal fluid (4 to 5 ml) was obtained by nontraumatic lumbar puncture from patients diagnosed with flares of MS (n = 8) and from control patients who suffered from non-demyelinating neurological pathologies including lumbar spinal stenosis (n = 2), normal pressure hydrocephalus (n = 2), pseudotumor cerebri (n = 1), headache (n = 3, MS was excluded), and dementia (n = 1). Patient samples were obtained from the Department of Neurology, Charité–Universitätsmedizin Berlin, and processed immediately. None of the patients was under systemic treatment for their underlying neurological disease at the time of or at least 8 weeks before sample collection.

Cell culture

Cells were cultured in RPMI 1640 medium supplemented with 2 mM glutamine, 1% (v/v) nonessential amino acids, 1% (v/v) sodium pyruvate, penicillin (50 U/ml), streptomycin (50 μg/ml; all from Invitrogen), and 5% (v/v) human serum (Charité Blood Bank, pooled from more than 100 donors). In some experiments, T cell cultures were performed in the presence of recombinant cytokines (all from R&D Systems at 10 ng/ml; IL-6 and IL-27 both at 50 ng/ml), with neutralizing antibodies [anti–IL-2 (5 μg/ml), BD Biosciences] or with the STAT3 inhibitor 6-nitrobenzo[b]thiophene-1,1-dioxide (Stattic, Sigma-Aldrich) or the STAT5 inhibitor N′-[(4-oxo-4H-chrome-3yl)methylene]nicotinohydrazide (573108, Calbiochem). T cells were labeled with CFSE according to standard protocols. T cell clones were generated by limiting dilution under nonpolarizing conditions containing IL-2 (500 IU/ml) as described (50). Restimulations were performed with plate-bound anti-CD3 (0.5 μg/ml, clone TR66) and anti-CD28 (1 μg/ml, CD28.2; BD Biosciences).

Cytokine and transcription factor analyses

For intracellular cytokine staining, cells were restimulated for 5 hours with PMA and ionomycin in the presence of brefeldin A (all Sigma-Aldrich) for the final 2.5 hours of culture. Cells were fixed and permeabilized with Cytofix/Cytoperm (BD Biosciences) according to the manufacturer’s instructions. For transcription factor staining in conjunction with intracellular GM-CSF staining, cells were first fixed, permeabilized (Cytofix/Cytoperm, BD Biosciences), stained for cytokines, and subsequently fixed and permeabilized for staining of transcription factors (Cytofix/Cytoperm, eBioscience). Cells were stained with phycoerythrin- or allophycocyanin-conjugated GM-CSF (6804, R&D Systems; BVD2-21C11, eBioscience) and antibodies identical to those described in (8) and were analyzed with FACSCanto (BD Biosciences). Flow cytometry data were analyzed with FlowJo software (Tree Star) using fluorescently unlabeled matched cell samples as gating controls. Cytokines in culture supernatants were measured by ELISA (R&D Systems), CBA (BD Biosciences), or ProcartaPlex (eBioscience) according to standard protocols after restimulation of cultured T cells with PDBu (50 nM, Sigma-Aldrich) and plate-bound anti-CD3 (1 μg/ml, TR66) for 8 hours or as indicated in the respective figure legends.

Plasmids and lentiviral transduction

The vectors pCCL.RORC2 [containing the complementary DNA (cDNA) encoding for the human RORC variant 2] (51) and pCCL (empty) were provided by M. Levings (University of British Columbia). Both vectors also expressed the ΔNGFR reporter gene (CD271) as a marker for transduction. Lentiviral particles were produced by calcium phosphate cotransfection of human embryonic kidney (HEK) 293T cells with the pCCL or pCCL.RORC2 transfer vectors together with the packaging vectors pRSV-Rev, pMDLg-pRRE (third-generation Pkg) (Addgene plasmid 12260), and pMDG2 (Addgene plasmid 12259) (52). Briefly, HEK293T cells were transfected with a mixture of transfer vector, pMDLg-pRRE, pMDG2.G, and pRSV-Rev at a ratio of 5:2:1.5:1. Viral particles were harvested 24 and 36 hours after transfection and concentrated by ultracentrifugation at 114,000g. Intracellular cytokine staining was performed on day 4 in cultures with 40 to 70% transduction efficiency as determined by CD271 expression (BD Biosciences) on the same day.

Gene expression analysis

Total RNA was extracted using TRIzol reagent (Invitrogen) according to the manufacturer’s instructions. Random hexamer primers and an MMLV reverse transcriptase kit (Stratagene) were used for cDNA synthesis. Transcripts were quantified by real-time PCR with predesigned TaqMan Gene Expression Assays (TBX21, Hs00203436_m1; GATA3, Hs00231122_m1; RORC2, Hs01076112_m1) and reagents (Applied Biosystems). For each sample, mRNA abundance was normalized to the amount of 18S ribosomal RNA (Applied Biosystems) and is expressed as arbitrary units.


Student’s two-tailed unpaired t test was used for statistical comparisons if not indicated otherwise. All statistical tests have been indicated in the figure legends. P values of 0.05 or less were considered significant.


Fig. S1. GM-CSF+ TH cells coexpress less FOXP3 and GATA-3 compared to GM-CSF TH cells.

Fig. S2. Coexpression of GM-CSF with multiple other effector cytokines on the single-cell level in human TH cells ex vivo.

Fig. S3. Existence of GM-CSF–producing TH cells, which are independent of IL-22 production.

Fig. S4. The differential expression of CXCR3 and CCR4 distinguishes IFN-γ+ and IFN-γ GM-CSF–producing TH cells.

Fig. S5. Distinction of T-bet+ and T-bet GM-CSF–producing TH cell subsets by surface markers.

Fig. S6. Modulation of GM-CSF expression in distinct TH cell subsets.

Fig. S7. TH2 cells are less susceptible to IL-12 stimulation compared to TH1 and TH17 cells as determined by STAT4 phosphorylation.

Fig. S8. Stability and flexibility of GM-CSF–only T cell clones.

Fig. S9. STAT4-dependent induction of GM-CSF by IL-12.

Fig. S10. Evolutionary conserved transcription factor binding sites in the promoter region of the CSF2 gene.

Fig. S11. Antagonistic regulation of GM-CSF expression by TH17 and TH1 polarization pathways.


  1. Acknowledgments: We thank M. Levings for providing the RORC lentiviral vector, T. Kaiser and J. Kirsch for cell sorting, the physicians at the Department of Neurology at the Charité–Universitätsmedizin Berlin for acquisition of patient samples, and S. Fillatreau and C. Romagnani for scientific suggestions. Funding: This work was supported by the German Research Foundation (SFB650 to C.E.Z., A.R., and H.-D.C. and ZI 1262/2-1 to C.E.Z.), the “e:Bio—Innovationswettbewerb Systembiologie” program of the Federal Ministry of Education (M.-F.M), the ERC-2010-AdG_20100317 Grant 268987 (A.R.), the Celgene Award of the European Society for Dermatological Research (C.E.Z.), and the Wolfgang Schulze Award (C.E.Z.). Author contributions: R.N. performed and analyzed all experiments. M.-F.M., C.H., and H.-D.C. contributed to experiments and analyzed data. R.R. performed and analyzed ELISA experiments. C.H. performed and analyzed quantitative reverse transcription PCR experiments. H.R. and L.H. acquired and characterized MS patient samples and analyzed data. A.R. designed experiments, provided scientific suggestions, and helped to write the manuscript. C.E.Z. designed experiments, analyzed data, conceived and supervised the study, and wrote the manuscript. Competing interests: L.H. has received compensation from Novartis, Biogen-Idec, Merck-Serono, and Bayer for speaking and from Novartis, Sanofi, and Biogen-Idec as a member of advisory board. The other authors declare no commercial or financial conflict of interest.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article