Genetic proof for the transient nature of the Th17 phenotype


  • Florian C. Kurschus,

    Corresponding author
    1. Institute for Molecular Medicine, University Medical Center of the Johannes Gutenberg-University Mainz, Mainz, Germany
    • Obere Zahlbacher Straße 67, 55131 Mainz, Germany Fax: +49-6131-179781
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Andrew L. Croxford,

    1. Institute for Molecular Medicine, University Medical Center of the Johannes Gutenberg-University Mainz, Mainz, Germany
    Search for more papers by this author
    • These authors contributed equally to this work.

  • André P. Heinen,

    1. Institute for Molecular Medicine, University Medical Center of the Johannes Gutenberg-University Mainz, Mainz, Germany
    Search for more papers by this author
  • Simone Wörtge,

    1. Institute for Molecular Medicine, University Medical Center of the Johannes Gutenberg-University Mainz, Mainz, Germany
    Search for more papers by this author
  • Daniele Ielo,

    1. Institute for Molecular Medicine, University Medical Center of the Johannes Gutenberg-University Mainz, Mainz, Germany
    Search for more papers by this author
  • Ari Waisman

    1. Institute for Molecular Medicine, University Medical Center of the Johannes Gutenberg-University Mainz, Mainz, Germany
    Search for more papers by this author


IL-17-producing CD4+ T cells (Th17) have been classified as a new T helper cell subset. Using an IL-17 fate mapping mouse strain, which genetically fixes the memory of IL-17 expression, we demonstrate that IL-17A/F-expressing T helper cells generated either in vitro or in vivo are not a stable T-cell subset. Upon adoptive transfer of IL-17F-reporter-positive Th17 cells to RAG-deficient or WT animals, encephalitogenic Th17 cells partially lose IL-17 expression and upregulate IFN-γ. Additionally, we show that Th1 cells can convert in vivo to IL-17A/IFN-γ-coexpressing cells in the mesenteric lymph nodes (mLN). Our data classify IL-17A and IL-17F as cytokines produced transiently in response to the local microenvironment, thus showing that IL-17 expression does not define an end-stage T helper cell subset.


Since the finding that IL-23 and not IL-12 is necessary for active induction of EAE 1, 2, the previously common dogma for the pathogenesis of the disease has changed. Th17 cells, which were soon thereafter shown to depend on IL-23 3, 4, are now regarded as major initiators of pathogenesis in a number of disease models and human conditions. Th17 achieve their pathogenic phenotype by secreting cytokines which in turn induces the surrounding tissue to secrete chemokines and other cytokines important for the immigration of potentially pathogenic leukocytes such as granulocytes and lymphocytes 5.

In a previous landmark EAE study, Th17 cells that were expanded in the presence of IL-23 were shown to be extremely efficient in inducing passive EAE 4. Low amounts of transferred cells (150 000) were able to induce EAE in SJL/J animals. This finding together with the full resistance of IL-23-deficient animals in response to active EAE induction 2 cemented the idea of Th17 cells as a major pathogenic cell population in EAE. This was further supported by the discovery that Th17 can be very efficiently generated in vitro when naïve CD4+ T cells are activated in the presence of TGF-β and IL-6 6–8 and that IL-6 is necessary for EAE induction 9–12. Furthermore, transgenic expression of TGF-β in T cells enhanced EAE severity 6. Another milestone for this hypothesis was the finding that RORγt deficiency led to a major lack of Th17 cells and to a near complete resistance against active EAE, even in the presence of extensive CNS infiltration 13. Other transfer studies in the SJL/J mouse using IL-23 expanded encephalitogenic cells found an enhanced infiltration of granulocytes concomitant with EAE development compared to transfer of IL-12 expanded T cells 5, 14, further supporting a specific role for Th17 cells in autoimmunity. Given the previous lack of suitable Th17 reporter strains, these studies relied on transfer of in vitro generated Th17 cells of a heterogenous nature, rather than a pure Th17 population.

Recently, the encephalitogenicity of Th17 cells was challenged by O'Connor et al., who showed that transferring myelin oligodendrocyte glycoprotein (MOG)-specific Th17 cells derived from a polyclonal C57BL/6 T-cell repertoire were not able to passively transfer EAE, in contrast to strong EAE induced by transfer of MOG-specific polyclonal Th1 cells 15. Also in this report, polarized TCR transgenic Th17 cells were transferred to either B10.PL or lymphopenic B10.PL animals. Under these conditions, some animals became sick, but surprisingly upon reanalysis many cells were found to express IFN-γ. In vitro polarization efficiencies often lie far under 100%, and as such it is not possible to deduce whether the observed IFN-γ-expressing cells were derived from the successfully polarized Th17 cells, which switch their expression profile, or the contaminating non-polarized cells, which were inevitably co-transferred.

In line with this, several recent publications demonstrated a surprisingly high plasticity of differentiated CD4+ T-cell subpopulations generated either in vitro or in vivo. First, a number of studies showed that Foxp3+ Treg in both mouse and human can be redirected to express IL-17 16–20. Similarly, a recent report showed that transferred natural Treg develop to follicular B-helper T cells in the Peyer's patches of T-cell-deficient hosts 21. Second, several groups demonstrated that Th17 cells generated in vitro are plastic upon exposure to Th1 cytokines and start to express IFN-γ (22–24). Finally, studies with purified in vitro generated Th17 cells transferred to NOD mice showed infiltrating cells changing their phenotype to become Th1 cells 22, 23. Very importantly, human Th17 T-cell clones were shown to be highly flexible and to co-express IFN-γ and IL-17A when stimulated in the presence of IL-12 24. Similarly a specific CD161+ subpopulation derived from human umbilical cord blood, which is prone to contain and differentiate to Th17 cells, develops strongly toward Th1 cells under the influence of IL-12 in vitro25.

Since these groups demonstrated IFN-γ production by Th17 cells following adoptive transfer, we aimed to define whether indeed trans-differentiation of IL-17 expressing cells is the cause of this finding. To address this question, we used our recently generated IL-17F fate mapping mouse line 26. When these IL-17F-Cre BAC-transgenic mice are crossed to ROSA26-EYFP reporter mice 27, IL-17F-expressing cells are irreversibly genetically tagged by Cre-mediated excision of a loxP flanked stop cassette, resulting in ubiquitous expression of EYFP in all recombined cells. We analyzed the behavior of transferred, sorted Th17 reporter cells generated either in vitro or in vivo and found that a considerable amount of these cells ceased IL-17A expression entirely, and expressed purely IFN-γ. Additionally, we found a number of previously highly pure Th1 cells co-expressing IL-17A together with IFN-γ in the mesenteric LN (mLN).


Plasticity of Th17 cells in EAE

In a first attempt to define whether in vitro generated Th17 cells maintain their cytokine phenotype upon EAE induction, we performed transfer EAE using in vitro polarized Th17 cells generated from MOG35–55-specific CD4+ cells isolated from 2D2 TCR-transgenic mice 28. After 5 days of stimulation in Th17-polarizing conditions, about 50% of cells expressed IL-17A, whereas only negligible numbers produced IFN-γ (Supporting Information Fig. S1A). We adoptively transferred 5×106 of these cells per mouse to RAG1-deficient mice (of the C57BL/6 background), resulting in severe EAE symptoms (Supporting Information Fig. S1B). In line with the findings by O'Connor et al. 15 we found during reanalysis of CNS infiltrating CD4+ cells that most of the cells lost IL-17A expression (Supporting Information Fig. S1C). A large proportion of the transferred Th17 cells expressed solely IFN-γ (11.6%). Roughly 2% of cells co-expressed both IL-17A and IFN-γ. In spleen and LN, most recovered cells were negative for IL-17A but some cells expressed IFN-γ (6 and 9% of the T cells in the spleen and the LN, respectively). Since only half of the initially transferred population was IL-17A positive (Supporting Information Fig. S1A), it was possible that IL-17-negative cells may have upregulated IFN-γ expression.

Fate mapping of sorted Th17 cells in EAE

To clarify whether Th17 cells can change their cytokine profile during the course of EAE, we made use of our IL-17F-CreEYFP (BAC-transgenic IL-17F-Cre crossed to ROSA26-EYFP) Th17 reporter mouse line, which can also serve as a fate mapping strain 26. Since Cre-mediated excision of the loxP-flanked stop cassette of the ROSA26-EYFP reporter is irreversible, cells expressing Cre (following activity of the IL-17F promoter) are EYFP+ irrespective of their subsequent cytokine expression pattern. We crossed these mice to 2D2 transgenic mice (2D2×IL-17F-CreEYFP) and generated from the latter in vitro activated MOG-specific EYFP expressing Th17 cells (Fig. 1A and Supporting Information Fig. S2). Although we found under standard Th17 differentiation conditions only 1/6 to 1/3 of the IL-17A intracellular positively stained cells to co-express the IL-17F-EYFP reporter, these cells were especially high in IL-17A expression either analyzed intracellular or by cytokine secretion assays (Supporting Information Fig. S2). We previously showed that about 95% of in vitro generated EYFP+ cells from these reporter mice express either IL-17A and/or IL-17F 26. Since the expression strength of IL-17A and IL-17F were highly correlating, EYFP+ positive cells are bona fide Th17 cells. Prior to transfer, CD4+EYFP+ cells did not express IFN-γ (Fig. 1B). We sorted EYFP+ Th17 cells (to more than 95% purity) and transferred 2×105 of these cells to RAG1−/− mice. Since these cells were too small in number to induce passive EAE, we co-transferred 1×107 2D2 Th1-polarized cells (the phenotype of which is shown in Fig. 1C). At the peak of disease (score 4 EAE), we reanalyzed the transferred cells isolated from the CNS, spleen and LN (Fig. 1D and E). Based on expression of both CD4+ and EYFP, the transferred Th17 could readily be distinguished from the transferred Th1 cells (Fig. 1D). Indeed, EYFP-expressing Th17 cells recovered from the CNS had to a large extent lost expression of IL-17A, with a sizeable proportion (17.8%) shifting to express solely IFN-γ. A minor fraction that produced both cytokines (6.4%) was also observed in the CNS (Fig. 1E). Loss of IL-17A expression was even more obvious in the cells recovered from the spleen (Fig. 1E). Interestingly, about a quarter of the cells reharvested from the LN expressed both IL-17A and IFN-γ. Compared to the propensity of Th17 cells to shift to either Th1 or Th17/Th1 cells, the transferred 2D2-Th1 cells recovered from CNS and spleen retained largely their initial Th1 orientation. Interestingly, CNS infiltrating Th1 cells kept the largest IFN-γ-positive population, probably due to the inflammatory environment or selective enrichment. Surprisingly, Th1 cells recovered from the LN (pooled peripheral LN (pLN) and mLN) showed a consistent population of IL-17A/IFN-γ double-positive cells (9.1%).

Figure 1.

Fate mapping of sorted Th17 cells in EAE. (A, B) T cells from triple transgenic animals (2D2×IL-17F-CreEYFP) were differentiated under Th17-polarizing conditions and intracellular cytokine staining (ICS) was performed. (A) CD4+ gated T cells are shown. (B) ICS for IFN-γ and IL-17A of EYFP+ CD4+ cells. (C) Th1 cells were generated from 2D2 mice in vitro and analyzed for cytokine expression by ICS. (D) EYFP+ Th17 cells from 2D2×IL-17F-CreEYFP mice were FACS sorted and 2×105 cells injected together with 10×106 of in vitro polarized 2D2 Th1 into RAG1−/− mice. EAE from transferred animals was scored daily and animals were killed at peak of disease. Transferred cells were reanalyzed by ICS gating on CD4+ and on EYFP and EYFP+ cells from CNS, spleen or LN. LN (p+m): peripheral and mesenteric LN (E) ICS for IL-17A and IFN-γ of gated EYFPCD4+ (transferred Th1 cells) and of EYFP+ CD4+ (cotransferred Th17) cells from the respective organs indicated in (D). Percentages of gated cells in the respective quadrants are indicated. Upper row with cells in green represents gated Th17 cells and lower row in red shows gated Th1 cells. A–E is representative of four similar experiments. (F) Quantitative real-time RT-PCR analysis of gene expression of the indicated genes in sorted EYFP+ cells before transfer and resorted EYFP+ cells 2 wk after transfer from mLN. Shown is the expression level relative to the house keeping gene gapdh pooled from two independent transfer and resorting experiments normalized to expression level before transfer. Data show mean+SD (n=2 per experiment).

Next, we analyzed the expression of cytokines and transcription factors by quantitative real-time RT-PCR in sorted EYFP positive cells before and after transfer and found that in accordance with the intracellular cytokine staining, tbx21 as well as ifng mRNA were highly upregulated, while the mRNA of il17a and il17f were down regulated (Fig. 1F). In contrast, we did not find a change in the expression levels of Th17-specific transcription factors rorc and irf4 (Fig. 1F). This indicates that the observed plasticity and coexpression of IL-17A and IFN-γ are based on dual expression of Th1 as well as Th17 specific transcription factors. Collectively, these data clearly illustrate that Th17 cells, once expressing IL-17A and IL-17F, are able to alter their previous cytokine expression pattern in vivo.

Th1 cells can express IL-17A

To analyze whether Th1 cells behave in a similar fashion to Th17 cells, we used a differentiation protocol in which a 2D2-Th1 population with nearly 100% IFN-γ producing cells was generated (Fig. 2A). We transferred 5×106 of these cells to RAG1−/− mice and reanalyzed their fate at the peak clinical EAE symptoms (Fig. 2B). Compared to Th17 cells, transferred 2D2-Th1 cells isolated from CNS and spleen did not shift in large numbers to express IL-17A, but either kept or lost IFN-γ expression. Surprisingly, Th1 cells recovered from the LN (pooled pLN and mLN) showed a consistent population of IL-17A/IFN-γ double-producing cells (Fig. 2C). The redifferentiation of Th1 cells in LN correlated with a rise in expression levels of IL-17A and IL-17F and a slight decrease of IFN-γ mRNA expression (Fig. 2D). In accordance with the upregulation of a Th17 phenotype, rorc expression was nearly 100-fold upregulated in Th1 cells recovered from mLN. In agreement with the relative stability of IFN-γ expression observed after intracellular staining, tbx21 remained stably expressed by Th1 cells (Fig. 2D).

Figure 2.

Th1 cells can express IL-17A. (A) MACS purified CD4+ T cells from 2D2 mice were differentiated for 5 days under Th1 inducing conditions and analyzed by ICS for IFN-γ and IL-17A expression. (B) Five million differentiated 2D2 Th1 cells were i.v. transferred to RAG1−/− mice and EAE was scored daily. The curve displays the EAE score of the mouse, of which data are shown in (C). (C) ICS of CD4+ cells from the indicated organs at day 14 of transfer EAE. A–C is representative of three similar transfer experiments. (D) Quantitative real-time RT-PCR analysis of gene expression of the indicated genes in Th1 cells before transfer and re-MACSed cells 2 wk after transfer from mLN. Pooled data from up to three independent experiments are displayed. Shown is the expression level relative to the house keeping gene gapdh normalized to expression level before transfer. Data show mean+SD (n=2 per experiment)

Plasticity of Th17 cells independent of EAE or transferred Th1 cells

Since EAE induces peripheral changes to the immune system and cellular composition, especially in the spleen and the BM, we transferred sorted, non-encephalitogenic reporter cells (IL-17F-CreEYFP) to RAG1−/− mice. Again, we found that a major part of the transferred population lost IL-17 expression and instead, upregulated the expression of IFN-γ (Fig. 3A), showing that the plasticity of the transferred Th17 population can take place independently of EAE. In this experiment, we analyzed pLN separately from mLN (Fig. 3B). Similarly to our findings with Th1 cells, we observed in the LN, most prominently in mLN, a population of cells that coexpresses IL-17A and IFN-γ (12.5 and 18%, respectively). This shows that the propensity to switch from Th17 to Th17/Th1 occurs also in a broad WT-TCR-repertoire, excluding that the observed plasticity is based on a potential bias of MOG35–55-specific CD4+ T cells to differentiate to Th1 cells 29.

Figure 3.

Th17 cells show plasticity independent of EAE or transferred Th1 cells. Sorted Th17 polarized CD4+ cells from IL-17F-CreEYFP mice were i.v. transferred into RAG1−/− animals and reanalyzed 14 days after transfer. (A) Expression of EYFP in total lymphocytes from the indicated organs. (B) CD4+ cells of gated EYFP+ cells from the organs indicated in (A) were analyzed by ICS for expression of IL-17A and IFN-γ. Data shown are representative of two experiments. pLN: peripheral LN, mLN: mesenteric LN.

Plasticity of in vivo generated Th17 cells

It was recently shown that in vivo generated Treg and Th17 cells are more stable in their phenotype than in vitro polarized cells 30, 31. We therefore aimed to analyze whether in vivo generated EYFP+ Th17 cells behave in a similar manner to in vitro generated Th17 cells. To analyze the stability of in vivo generated Th17 cells, we immunized IL-17F-CreEYFP reporter mice, sorted CD4+EYFP+ cells from draining LN and the spleen and transferred these cells to RAG1−/− mice (Fig. 4). To our surprise, these cells trans-differentiated even more than the in vitro generated Th17 cells to either express IFN-γ (about 60%) or both IL-17A and IFN-γ (up to 36% in mLN). These data show for the first time that in vivo generated Th17 cells do not represent a terminally differentiated cell population and are able to radically alter their cytokine secretion profile. To test whether Th17 cells, which differentiate under normal WT-repertoire-conditions, also change their initial cytokine bias, we induced EAE in IL-17F-CreEYFP mice and analyzed EYFP-positive cells on day eight in the draining LN, or on day 16 in the CNS of fully diseased animals (Fig. 4C). We found that whereas the early differentiated cells mostly expressed IL-17A and no IFN-γ, in the late phase in the CNS most of these cells shifted to either express IFN-γ only or IFN-γ and IL-17A. These findings strongly corroborate our previous findings using in vitro or in vivo generated and FACS-sorted Th17 cells.

Figure 4.

Plasticity of in vivo generated Th17 cells. Th17 cells (CD4+EYFP+) were sorted from spleens and draining LN from MOG35–55-CFA immunized IL-17FCreEYFP mice and i.v. transferred to RAG1−/− mice. Transferred CD4+ cells were reanalyzed 15 days post transfer by ICS. (A) Expression of EYFP in total lymphocytes from the indicated organs. (B) CD4+ cells of gated EYFP+cells from the organs indicated in (A) were analyzed by ICS for expression of IL-17A and IFN-γ. The data are representative of two similar transfer experiments. (C) Active EAE was induced directly in IL-17F-CreEYFP mice. Mice were analyzed either on day 8 (n=3) or day 16 post immunization (n=3). ICS of gated EYFP-positive CD4+CD11b cells from the draining LN (dLN) (left) on day 8 and from the CNS from day 16 with EAE score 4 (right) are shown.

Th17 cell plasticity in WT animals

To test whether plasticity of in vitro generated EYFP+ Th17 cells occurs as well in non-lymphopenic conditions, we transferred sorted in vitro differentiated Th17 cells from 2D2×IL-17F-CreEYFP mice to WT animals and reanalyzed the cells 2 wk later. Although under these conditions most transferred cells did not express IL-17 anymore, but also not IFN-γ, we could find, especially in the mLN, EYFP+ cells that expressed IFN-γ but lost IL-17A expression (Fig. 5).

Figure 5.

Th17 cell plasticity in WT animals. (A) Sorted Th17 polarized CD4+ cells from 2D2×IL-17F-CreEYFP mice were transferred into WT-C57BL/6 animals and reanalyzed 2 wk later. (B) CD4+EYFP+ cells from the organs indicated in (A) were analyzed by ICS for expression of IL-17A and IFN-γ. Data shown are representative of two similar transfer experiments.

Generation of IL-17A and IFN-γ double-positive cells in vitro

To test under which conditions T cells may either develop or shift to a double-positive IL17A/IFN-γ stage we treated naïve CD4+ cells under Th1-polarizing conditions in the presence of IL-6 for different periods with TGF-β. (Supporting Information Fig S3). We added TGF-β either from the start of culture or 18 h later. We found that TGF-β partially inhibited Th1 development depending on the time of addition and that single-positive Th17 cells as well as double-positive IFN-γ/IL17A cells were differentiating under the combined influence of IL-12, IL-6 and TGF-β. Since distinct amounts of these cytokines should be present under inflammatory conditions in the LN and/or the target organs, differentiation and plasticity of Th17 cells may well occur depending on the individual cytokine environment present at any given stage of disease.


In this manuscript, we demonstrate using a unique Th17 fate mapping approach that “Th17 cells” generated in vitro or in vivo can change their hallmark cytokine expression. Additionally, we made the surprising finding that highly pure Th1 cell populations can upregulate IL-17A, thus becoming double producing “Th1/Th17” cells. Several groups previously presented data indicating the flexibility and/or plasticity of different T helper subpopulations 16–18, 20, 22–24, 31–34 and Tc17 cells 35. These groups used either reporter mice in which the fluorescent protein was expressed under the direct control of the respective cytokine or transcription factor promoter 16, 32, 33 or cytometric cytokine secretion assays to label live cytokine producing cells 22, 31. Both methods, however, are not devoid of inherent problems. Using a direct reporter approach, cell marking is reversible and cytometric cytokine secretion assays may falsely label non-cytokine expressing cells. Alternatively, single human Th17 T-cell clones were grown and analyzed for stability of their cytokine expression under different conditions 24. Although very elegant, this system requires exposure of T cells to long-term in vitro cell culture. We complemented these recent findings using our IL-17F-CreEYFP reporter system. Since IL-17F expressing cells are irreversibly marked, one can sort live Th17 cells and follow their fate irrespective of their later cytokine expression status.

The plasticity observed using this approach may be either independent of proliferation or may occur during cell division. During the expansion phase of T helper cells, polarized cells are thought to keep their cytokine profile, which is probably maintained through epigenetic mechanisms 20, 34, 36, 37. Whether DNA methylation or histone modification patterns are altered in our system requires further clarification. Recently, genome-wide change of histone methylation patterns during in vitro trans-differentiation was demonstrated 34. Another group recently reproduced and expanded the latter finding by using in vitro generated Th17 cells trans-differentiated to Th1 by using IL-12 38. These studies showed that transcription factor genes like tbx21 or cytokine genes like ifng are especially poised for expression in Th17 cells, explaining the disposition of Th17 cells to become Th1 cells. Another potential mechanism of flexibility might be the co-expression of lineage-specific transcription factors, as was recently demonstrated for Foxp3 and RORγt in human IL-17 expressing Treg 19.

A striking but largely overlooked observation supporting plasticity in the program of T helper cells is the frequently noted IFN-γ/IL-17A double-producing T-cell populations, especially found in CNS infiltrating populations of diseased EAE animals as well as in short-term human T-cell cultures 24. Although several groups have shown in vitro that IFN-γ antagonizes Th17 differentiation 7, 8, the co-expression of IFN-γ and IL-17 is found in vivo. In line with this, we found that the combination of IL-12, IL-6 and TGF-β is able to induce Th1, Th17 and IFN-γ/IL-17A double-positive cells. One might easily envisage that these distinct cytokines are expressed under inflammatory conditions and induce the typical picture of distinct T helper effector lineages in vivo.

The data described here show that plasticity, at least on a population level, is common to Th17 and Th1 cells. Whether this plasticity occurs during natural conditions such as infections or autoimmunity needs to be defined. The data by O'Connor et al. 15 suggested that Th17-transfer EAE can only be found under circumstances where a part of the transferred population shifts toward IFN-γ-producing cells. This was not the case for Th1-transfer EAE. Our finding that in some of the highly pure transferred Th1 cell population expression of IL-17A was induced indicates that also a Th1–Th17 shift may play a role in Th1-transfer EAE. Future experiments using either IL-17A/F knockout Th1 cells or IFN-γ or T-bet knockout Th17 cells for transfer EAE should clarify the role of the cytokine shift in EAE development. In a model for airway hyperresponsiveness, another group recently showed that a shift to IFN-γ expression is necessary to induce airway hyperresponsiveness, whereas IL-17A expression was necessary for neutrophil infiltration 39. In light of the beneficial effects of IFN-γ in EAE one might speculate whether the cytokine shift to IFN-γ expression may even have a certain protective role.

Our finding that also highly pure Th1 cells are able to shift to cells that express both IFN-γ and IL-17A is new. We found these cells particularly in the mLN. Together with the finding that also Th17 cells recovered from the mLN contained a large fraction of double-expressing cells, this indicates that the gut immune system creates a specific local milieu, which favors this Th1/Th17 dichotomous response. Potential mechanisms for the bias to coexpress IL-17 might be the local presence of CD103+ and CD103 mLN DC, which may favor under certain conditions the development of Th17 cells 40, 41.

In our transfer experiments, the driving force of trans-differentiation in the lymphopenic environment might be homeostatic proliferation of the transferred cells. Evidence against that is a recent report demonstrating that shifting of Th17 cells to IFN-γ expression was independent of IL-7 blockage 33, which largely inhibited proliferation of the injected cells. Whether, and which, other factors present in the lymphocyte-deficient lymphoid compartments trigger the reprogramming of Th17 cell populations needs to be determined. In transfers to RAG1−/−, and more strikingly in transfer experiments using WT mice, we found a strong downregulation of cytokine expression of the donor cells. In immunodeficient animals, this loss of cytokine expression especially in the spleen is not explained by high levels of the homeostatic cytokine IL-7, since this was shown to rather specifically enhance cytokine expressing cell populations 42. Transfer of Th17 cells to WT mice showed some cells changing their cytokine expression to express IFN-γ. The stronger loss of cytokine expression in WT mice may at least in part be due to the presence of Treg in WT mice, which are lacking in the transfer experiments to RAG1-deficient animals. The difference of cytokine expression in CNS, LN and spleen may be explained by a previously recognized sequential homing of transferred myelin specific cells and their differential expression of activation markers 43. In addition, the transfer of cytokine expressing cells in the absence of Treg in RAG1-KO mice might induce subclinical autoimmunity also in the case of non-encephalitogenic T-cell transfers, similar as in T-cell-mediated colitis experiments. This inflammatory milieu might be needed to maintain cytokine expression and might also contribute to the shift from Th17 to Th1.

The very initial description of Th1 and Th2 cells by Mosmann et al. 44 was based on repetitive stimulations of in vivo primed T-cell lines, which were further cloned by limiting dilution. These T-cell clones were stable in their cytokine secretion pattern for 18 months. We either stimulated Th17 cells once for 5 days or twice for a total of 9 days but we did not find differences in their plasticity. Also others who repetitively stimulated Th17 cells over several wk were able to trans-differentiate Th17 cells to Th1 cells in vitro32. In vivo, such a repetitive stimulation might only take place in the case of chronic infections or chronic autoimmunity. In a normal immune response, stability is maintained by memory T cells. Recently, memory CD4+ T cells were described to reside as Ly6C+ cells in the BM 45. When we analyzed BM-memory CD4+ T cells, we found practically no IL-17A expressing Ly6C+ helper T cells, whereas IFN-γ was expressed by a low but reproducible number of this memory population (data not shown). Additionally, it was extremely difficult to detect EYFP positive cells in the BM several months after immunizations. This indicates that the IL-17 response is transient and is quickly lost, most likely due to its highly dangerous nature. This finding is in line with a recent report by Pepper et al. who showed that Listeria monocytogenes-specific Th17 cells are short lived in comparison to long-lived Th1 cells 46. Earlier and more recent findings that human Th17 clones express in part also IFN-γ, or also shift to become Th1 cells, further substantiate our findings of the transient nature of the IL-17 response by T helper cells 24, 47.

During recent years, many reports claimed the necessity of Th1 and Th17 cells for autoimmunity, using transfer models of in vitro generated T-cell populations. In light of the findings presented here, showing that transferred cells populations change their cytokine profile in vivo, these claims should be taken with some caution. We and others further demonstrated that several of the major cytokine players expressed by Th17 cells, such as IL-17A and IL-17F 48, IL-22 49 and IL-21 50, are not essential for EAE induction. Together this hints to a role of IL-23 independent from Th17 cell differentiation 51.

It is evident that formally sought “terminally-differentiated” cell types can keep a certain “stemness” or pluripotency. Recently, fibroblasts were demonstrated to dedifferentiate under appropriate manipulations 52 and to regain induced pluripotent stem cell potency (iPS cells). The expression of only four transcription factors was sufficient to induce this cell fate change. We propose that flexibility in differentiation and trans-differentiation of distinct T helper lineages is necessary to cope with the multiple and differential demands the immune system encounters during its combat against a multitude of infectious agents 53.

Materials and methods

Mice and animal experiments

Generation of IL-17F-CreEYFP mice is described 26. ROSA26-EYFP mice were previously published 27. 2D2 mice have been described 28. All strains used were backcrossed to the C57BL/6 background. All animal experiments performed were in accordance with our license of the government agency for animal welfare of Rheinand-Pfalz (Mainz, Germany).

All animal procedures used were in accordance with guidelines of the committee on animals of the Max Planck Institute of Neurobiology and with the license of the Regierung von Oberbayern (Munich, Germany).


To induce Th17 cells in IL-17F-CreEYFP reporter mice, mice were immunized s.c. with 100 μL CFA, containing 1.1 mg of heat killed Mycobacterium tuberculosis and 50 μg of MOG35–55 peptide. CD4+ cells were recovered from draining LN and spleen and CD4+ cells were enriched by MACS beads (Miltenyi Biotech, Bergisch Gladbach, Germany) and thereafter sorted for EYFP expression.

In vitro differentiation

T cells were differentiated to either Th1 cells or Th17 cells in RPMI medium containing 10% FCS, 2 mM L-glutamine, 100 units/mL penicillin, 100 μg/mL streptomycin, 1 mM sodium pyruvate, 50 μM 2-mercaptoethanol, 10 mM HEPES and 1% non-essential amino acids (MEM). 2D2 cells were stimulated during differentiation either using MOG35–55 peptide (20 μg/mL) for 9 days with two stimulations (d0 and d5) or with anti-CD3 (1 μg/mL)/CD28 (6 ng/mL) for 5 days. Polarization for Th1 cells was performed using IL-12 (20 ng/mL) and IL-18 (20 ng/mL) and IL-2 (10 ng/mL). Th17 cells were differentiated using rh-TGFβ1 (2 ng/mL) IL-6 (20 ng/mL), IL-23 (20 ng/mL) and anti-IFN-γ (10 μg/mL). For sorting of Th17 cells, cells were stained and thereafter sorted for CD4+ and EYFP expression. Naïve CD4+ T cells were purified by MACS-sorting using the naïve CD4+ T-cell purification kit from Miltenyi Biotech.

Transfer EAE

Transfer EAE was induced by i.v. transfer of the indicated number of cells and i.p. injection of 200 ng of pertussis toxin (Sigma-Aldrich) at days 0 and day 2. Mice were scored daily as described 54 and organs were analyzed for transferred cells at peak EAE (days 13–15).

Flow cytometry

Antibodies were from BD-Biosciences or eBioscience. Infiltrating CNS cells were purified similarly as described 55. For intracellular cytokine staining cells were activated for 4 h in PMA (50 ng/mL) and Ionomycin (750 ng/mL) in the presence of Brefeldin A (1 μg/mL). Thereafter, cells were surface stained for CD4+ (CD4+-PerCP), washed and fixed in 3% PFA in PBS for 10 min on ice. Cells were then permeabilized using a saponin buffer (SB): 0.1 % saponin, 1% BSA and 0.02 % NaN3. To block unspecific binding sites, Rat IgG was added to the permeabilization step. Thereafter, cells were stained for IL-17A (APC) and IFN-γ (PE) in SB for 30 min's on ice and then washed with SB buffer. Alternatively, Th17 cells were analyzed by cytokine secretion assay according to the manufacturers' instructions (Miltenyi Biotech). Cells were analyzed using a Calibur Flow cytometer or a Canto II flow cytometer (BD-Bioscience; FZI, Mainz, Germany).

Real-time RT-PCR

RNA of sorted or MACSed cells was prepared by using QIAshredder Mini spin columns and by using the RNeasy Mini or the RNA-Micro kit from Qiagen with a DNA digestion step included. cDNA was prepared using the first strand synthesis kit from Invitrogen supplemented with 4 U/μL of RNAsin. One microliter of cDNA was used for a quantitative real-time reaction using the QuantiTect SYBR Green reaction mixture from Qiagen on white 96-well plates from Roche. Primer mixes were from Qiagen or in the case of rorc synthesized by Metabion (Martinsried, Germany) according to published sequences 56. Real-time PCR was performed on a Roche Lightcycler 480 II. Shown are relative expression levels of the respective samples to GAPDH calculated by the delta-delta Ct method of the Roche software. The data shown were further normalized to expression levels before cell transfer.


The authors thank Julia Altmaier, Sebastian Attig and Magdalena Brkic for cell sorting. This work was supported by the DFG grants SFB490 and SFB/TR 52 to A. W., who is supported by funds from the Böhringer Ingelheim Stiftung and by the German Ministry for Education and Research (BMBF, Consortium UNDERSTANDMS, as part of the “German Competence Network of MS”).

Conflict of interest: The authors declare no financial or commercial conflict of interest.