Lithium Controls Central Nervous System Autoimmunity through Modulation of IFN-γ Signaling

Inhibitors of glycogen synthase kinase 3 (GSK3) are being explored as therapy for chronic inflammatory diseases. We previously demonstrated that the GSK inhibitor lithium is beneficial in experimental autoimmune encephalomyelitis (EAE), the mouse model of multiple sclerosis. In this study we report that lithium suppresses EAE induced by encephalitogenic interferon-γ (IFN-γ)-producing T helper (Th1) cells but not by interleukin (IL)-17-producing T helper (Th17) cells. The therapeutic activity of lithium required functional IFN-γ-signaling, but not the receptor for type I IFN (IFNAR). Inhibitor/s of GSK3 attenuated IFN-γ dependent activation of the transcription factor STAT1 in naïve T cells as well as in encephalitogenic T cells and Th1 cells. The inhibition of STAT1 activation was associated with reduced IFN-γ production and decreased expansion of encephalitogenic Th1 cells. Furthermore, lithium treatment induced Il27 expression within the spinal cords of mice with EAE. In contrast, such treatment of Ifngr−/− mice did not induce Il27 and was associated with lack of therapeutic response. Our study reveals a novel mechanism for the efficacy of GSK3 targeting in EAE, through the IFN-γ-STAT1 axis that is independent IFNAR-STAT1 axis. Overall our findings set the framework for the use of GSK3 inhibitors as therapeutic agents in autoimmune neuroinflammation.


Introduction
Multiple sclerosis (MS) is an autoimmune neurodegenerative disease in which both adaptive and innate immunity play a role. CD4 + T cells, believed to be early effector cells in the disease, migrate to the central nervous system (CNS), leading to demyelination, axonal loss, and neurological disability. The cells of the innate immune system are also involved both in the initiation and progression of MS by influencing the effector function of T cells [1,2]. Both Th1 and Th17 cells are involved in the pathogenesis of MS, and are the primary effector cells in experimental autoimmune encephalomyelitis (EAE), the most common animal model of MS [3][4][5][6]. These lineages have distinct effector functions and are characterized by the expression of specific transcription factors and cytokines. The differentiation of naïve CD4 + T cells to interferon-c (IFN-c)-producing T helper (Th1) cells is dependent on IFN-c and interleukin (IL)-12, activation of STAT1 and STAT4, respectively, and the transcription factor Tbet [7]. TGF-b and IL-6, and STAT3 drive IL-17-producing T helper (Th17) cell differentiation in a process that is dependent on the transcription factor ROR-ct [8,9]. Although IL-23 is not needed for differentiation, it has an essential role in pathogenicity of Th17 cells perhaps by promoting expansion and stability [10].
The IFN-c-STAT1 signaling axis has an important pleiotropic role, both pathogenic and protective, in autoimmune diseases including MS and its mouse model, EAE [11]. Both Th1 and Th17 cells are independently capable of inducing autommunity in mouse models and they not only play a role in regulating one another, but that they have a more complex, both overlapping and differential, role in tissue inflammation [4,12,13]. There is also increasing evidence of the plasticity/instability of the Th17 cell phenotype; Th17 cells may acquire Tbet expression, gaining the ability to secrete IFN-c in addition to IL-17 [14]. These dual cytokine expressing Th17 cells may ultimately lose the ability to secrete IL-17 and convert into Th1-like cells. Thus the finding that Th17 cells can turn into Th1 cells highlights the importance of controlling the effector function of Th1 cells once disease is established.
We have recently found that relapsing-remitting MS segregates into a Th1 or a Th17 disease and that each form of disease is differentially responsive to type I IFN therapy [15]. Thus the elucidation of signaling pathways regulating the production and expansion of specific Th effector cells in EAE and MS is a necessary goal to identify new specific targets for therapeutic intervention. A lot is known about the transcription factors and cytokines that are determinant for the differentiation of Th1 and Th17 effector cells, but the mechanisms regulating their production, expansion and pathogenic function in disease are still largely undefined.
GSK3 is a constitutively active serine/threonine kinase that is a critical modulator of innate and adaptive immunity through the regulation of several transcription factors important in the production of cytokines and inflammation, including NF-kB, CREB, AP-1 and STATs [16]. We have previously shown that the GSK3 inhibitor lithium is prophylactic and therapeutic in EAE [17]. Recovery from EAE in lithium treated mice was associated with reduced demyelination, reduced microglia activation, and reduced CD4 + T cell infiltration in the spinal cord. We also found that treatment of mice in vivo with the GSK3 inhibitor lithium, inhibited myelin oligodendrocyte glycoprotein peptide (MOG  )-specific T cell proliferation and significantly reduced MOG 35-55 -specific production of IFN-c, IL-6, and IL-17 from splenocytes [17]. GSK3 has been shown to facilitate IFN-c mediated activation of macrophages [18]. Furthermore inhibition of GSK3 in macrophages suppresses activation of STAT3 and STAT5, and constrains the synergistic activation by IFN-c and lipopolysaccharides (LPS) of STAT3 [19,20]. However the mechanism of the therapeutic action of lithium in neuroinflammation in vivo is still unresolved. In this study we tested the hypothesis that lithium is beneficial in EAE through GSK3 regulation of IFNc signaling. Our results show that lithium suppresses Th1 but not Th17 neuroinflammation, and through inhibition of GSK3 tunes IFN-c-STAT1 signaling for optimal therapeutic efficacy in EAE.

Ethic Statement
All experimental animal work in this study was conducted in strict accordance with the National Institutes of Health and University of Alabama at Birmingham Institutional Animal Care and Use Committee (IACUC) guidelines. The protocol was approved by the IACUC of the University of Alabama at Birmingham (approval number 111208672). All Surgery was performed under isofluorane anesthesia, and all efforts were made to minimize suffering. Mice C57BL/6 mice were purchased from Frederick Cancer Research. B6.129S7-Ifngr1 tm/Agt /J (Ifngr1 2/2 ) mice were purchased from the Jackson Laboratory and backcrossed onto C57BL/6 background for 10-12 generations. C57BL/6 Stat1 2/2 , Ifnar1 2/2 , and IL-17F-Thy1.1 reporter mice [14] were kind gifts from R. Lorenz, J.D. Mountz, and C. Weaver, respectively (UAB). For lithium treatment, lithium was administered in pelleted food containing 0.2% lithium carbonate (Harlan-Teklad) as previously described [17]. This lithium administration is used to achieve serum levels equivalent to those attained therapeutically in human patients.

In vitro Stimulations
Macrophages were lavaged from the peritoneum on day 4 after injection with Brewer's thioglycollate. Mononuclear cells were isolated from spleen using the standard protocol of first mashing the spleen through a cell strainer, then lysing red blood cells by using ACK (Ammonium-Chloride-Potassium) Lysing Buffer, and then washing well the cells with PBS, and re-suspending them in culture medium. Mononuclear cells were stimulated with 5 U/ml IFN-c or 100 U/ml IFN-b (Biolegend) and/or 1.25 mg/ml anti-CD3 (145-2C11). Where indicated, cultures were supplemented with GSK3 inhibitors LiCl (5 mM-20 mM; Sigma) or TDZD-8 (5 mM; Calbiochem).

RNA Isolation and RT-PCR
Spinal cords were isolated from EAE mice perfused with PBS on day 20 post-immunization and snap frozen using dry ice in ethanol. RNA was extracted using Trizol reagent (Invitrogen) and cleaned up using RNeasy Mini Kit (Qiagen). cDNA synthesis was performed using SuperScript VILO cDNA synthesis kit (Invitrogen) per manufacturer's instruction. Gene expression was assayed using Taqman Gene Expression Assays (Applied Biosystems) in combination with Taqman Fast Advanced Master Mix (Applied Biosystems). Taqman assay IDs include: Hprt (Mm00446968_m1), Il10 (Mm00439614_m1), Il27 (Mm00461164_m1), Nos2 (Mm00440502_m1) and Ifnb1 (Mm00439552_s1). Expression data

Statistics
Results were analyzed by t-test or one-way ANOVA, as appropriate. For analysis of the EAE clinical scores curves and day of onset, the non-parametric Mann-Whitney test was used.

Results and Discussion
Lithium Attenuates Th1 EAE and not Th17 EAE IFN-b, the major therapy for the treatment of MS, inhibits the differentiation of naïve T cells to Th17 cells [22][23][24][25]. However, we have found that it is ineffective in treating MS patients with a Th17 signature or mice with a Th17 form of EAE [15]. Thus in the present study we interrogated if lithium can equally attenuate both Th1 and Th17 disease, an important question since lithium is therapeutic in EAE [17] and inhibits in vitro differentiation of naïve T cells to Th17 cells [26]. We observed that in lithium-treated animals EAE induced by encephalitogenic Th1 cells was significantly delayed in onset (Table 1) and less severe than that in untreated mice (Fig. 1A). In contrast, Th17 EAE in lithium treated mice, although slightly delayed in onset (Table 1), had the same disease severity as in untreated mice (Fig. 1B). We confirmed this result, using highly enriched encephalitogenic Th17 from immunized IL-17F-Thy1.1 reporter mice that were sorted for Thy1.1 expression after Th17 polarization (Figs. 1C, 1D) [14]. This approach enabled us to exclude contaminating Th1 cells as well as normalize for absolute numbers of IL-17-producing cells with that of IFN-c-producing cells used for Th1 transfer. The lack of therapeutic benefit in Th17-induced EAE was surprising, considering the observation that lithium inhibits differentiation of naïve T cells to Th17 cells [26]. We suggest that this reflects a differential effect of lithium in polarized Th17 cells versus Th17 cells during differentiation and/or expansion. In fact, we did observe that lithium attenuated the expansion of Th17 cells when in vivo primed T cells were cultured with MOG 35-55 under Th17 polarizing conditions (Fig. 1E). Furthermore, lithium treatment of Th17 cells, initiated after cells were already polarized for 3 days, resulted in increased GM-CSF production and no decrease in IL-17A secretion compared to untreated controls (Fig. 1F). GM-CSF is a key pathogenic determinant of encephalitogenic Th17 cells and its elevation by lithium treatment is consistent with treatment inefficacy [27,28]. The inhibition of GSK3 by lithium may also promote pathogenicity of Th17 cells by increasing responsiveness to IL-17A and IL-17F and interrupting the GSK3-dependent feedback regulation of IL-17R signaling [29].
The attenuation of Th1-induced EAE by lithium was associated with fewer CD4 + IFN-c + cells infiltrating the spinal cord, in proportion and absolute number, compared to untreated Th1 mice (Figs. 1 G, Fig. S1). Although the number of Th1 cells was decreased, the absolute numbers of all infiltrating cells in the spinal cord or brain were slightly increased or unaltered (Fig. S1). However these differences were not statistically significant.
Lithium Attenuates IFN-c-induced STAT1-Y701 Phosphorylation in CD4 + T Cells through Inhibition of GSK3 The IFNGR-STAT1 signaling axis is essential for the differentiation and expansion of Th1 cells in a feed-forward process that involves Ifng expression [4,7,30]. We therefore investigated if GSK3 facilitated the IFN-c signaling pathway in T cells by promoting STAT1 activation. Mononuclear cells from spleens of naïve C57BL/6 mice were stimulated with IFN-c and/or anti-CD3 in the absence or presence of lithium or TDZD-8, a structurally different GSK3 inhibitor, and the extent of pSTAT1-Y701 (pSTAT1) was evaluated in different cell populations. As expected, IFN-c efficiently induced pSTAT1 in CD4 + T cells ( Fig. 2A). Anti-CD3 marginally activated STAT1, but costimulation of CD4 + T cells with anti-CD3 and IFN-c resulted in synergistic hyperphosphorylation. Although this synergism has not been previously reported, the result is inferable because T cell receptor (TCR) engagement leads to translocation of IFNGR to the TCR complex within the immunological synapse [31]. Treatment with lithium or TDZD-8, significantly attenuated pSTAT1 induced by stimulation with IFN-c in presence or absence of costimulation of the TCR (Fig. 2A). Similarly, stimulation with IFN-b also promoted STAT1 activation, and co-stimulation of CD4 + T cells with IFN-b and anti-CD3 resulted in the activation of STAT1 that was additive to that induced by IFN-b or anti-CD3 alone (Fig. 2B). Lithium also inhibited STAT1 activation induced by IFN-b with or without anti-CD3 costimulation (Fig. 2B). In agreement with a previous study [19], GSK3 inhibition has no effect on IFN-c induced activation of STAT1 in macrophages from the peritoneum or spleen (Fig. 2C). However, lithium did inhibit STAT1 activation in B-cells and CD8 + T cells (Figs. S2 A, B). As expected, the IFN-c induced pSTAT1 was absent if IFNGR signaling was genetically ablated (Ifngr1 2/2 ) (Fig. 2D). However, anti-CD3 stimulation was able to induce STAT1 activation in Ifngr1 2/2 mice (Fig, 2 D). Notably, inhibition of GSK3 had no effect on STAT1 activation induced by anti-CD3 alone ( Fig. 2A, B, and D). Overall from these data we infer that efficient activation of STAT1 in CD4 + T cells by IFN-c and IFNb requires active GSK3.

Lithium Inhibits Th1 Expansion by Tuning IFN-c-induced STAT1-Y701 Phosphorylation
From the above data we predicted that lithium might modulate encephalitogenic Th1 cells by altering activation of STAT1. In order to test this hypothesis, cells from spleens and dLNs from MOG 35-55 -immunized mice were stimulated with IFN-c in the presence or absence of TCR engagement (anti-CD3), or with anti-CD3 alone. Both IFN-c and anti-CD3, each significantly induced pSTAT1 in encephalitogenic T cells, but the co-stimulation with anti-CD3 and IFN-c resulted in synergistic hyperphosphorylation (Fig. 3A) as observed in naïve cells ( Fig. 2A). Lithium attenuated IFN-c-induced pSTAT1 in the presence or absence of costimula- tion with anti-CD3 (Figs. 3 A). We next tested the activation of STAT1 in in vivo primed T cells restimulated with antigen (MOG 35-55 ) (Figs. 3 B, S3 C). The levels of pSTAT1 in unstimulated cells from MOG 33-55 -immunized mice were very low, and did not differ from unstimulated cells from naïve mice (Figs. S3 A, B). Restimulation with MOG 35-55 for 24 h induced STAT1 activation (Fig. S3 C). Lithium treatment during the 24 h stimulation lowered p-STAT1 levels with respect to untreated control, but the decrease was not statistically significant (Fig. 3B).
Remarkably, acute treatment with LiCl (1 h) was sufficient to significantly decrease pSTAT1 (Fig. 3B). This rapid down modulation of pSTAT1 suggests that the target of GSK3 might be a phosphatase, such as SHP2 [32]. Exogenous IFN-c was unable to further enhance pSTAT1 indicating that STAT1 was maximally activated by endogenous IFN-c produced during restimulation (Fig. 3B). Notably lithium also dramatically inhibited the TCR-activation dependent production of IFN-c (Fig. 3C). We then evaluated if lithium-dependent inhibition of pSTAT1 would affect the generation/expansion of Th1 cells by MOG  restimulation under Th1 polarizing conditions (Figs. 3 D, S3 D). We found that the pSTAT1 was much higher in Th1 polarized T cells compared to unstimulated naïve CD4 + T cells and that it was not increased further by acute stimulation with IFN-c (Fig. S3 D). Both continuous and acute lithium treatment significantly inhibited STAT1 activation in Th1 cells (Fig. 3D). Importantly, lithium inhibited the generation of IFN-c + Th1 cells by more than 50% (Fig. 3E). Acute treatment with 10 mM of lithium at the end of Th1 polarization also significantly inhibited the production of IFN-c (Fig. 3F). Treatment with 5 mM of lithium, a dose suboptimal for acute exposure, did not have any significant effect  on IFN-c production. Overall, these results demonstrate that lithium suppresses Th1 expansion by modulating IFN-c-STAT1 activation pathway.

Intact IFN-c Signaling is Required for Lithium's Therapeutic Effectiveness in EAE
The IFN-c-STAT1 signaling axis has an important pleiotropic role in autoimmune diseases including MS and its mouse model, EAE [11]. IFN-c is both pathogenic and protective in autoimmune diseases [4,11]. These studies suggest that too much or too little IFN-c signaling can exacerbate autoimmune diseases such as MS; therefore, we posited that lithium balances IFN-c signals and suppresses EAE by attenuating STAT1 activation. To test this model, we compared the ability of lithium to suppress EAE in wild-type (WT) mice to that in Stat1 2/2 mice (Figs. 4 A, 4 B). In agreement with our previous study [17], lithium treatment in WT mice beginning day 10 after immunization significantly suppressed clinical disease (Fig. 4A). Lithium treatment caused a more moderate, although significant, attenuation of EAE in Stat1 2/2 mice (Fig. 4B), and did delay the onset of disease (Table 2). This suggests the existence of a minor STAT1 independent activity of lithium, a possibility since IFNcR can signal through other STATs [11]. STAT1 is activated by signaling through multiple cytokine receptors, including the IFNGR and the IFNAR [33]. We therefore interrogated if intact IFNGR-STAT1 signaling was necessary for lithium to attenuate EAE. We found that lithium did not significantly suppress EAE in Ifngr1 2/2 mice (Fig. 4C). This result indicates that intact IFN-c signaling is necessary for therapeutic effectiveness of lithium. Unlike in Ifngr1 2/2 mice, lithium delayed the onset ( Table 2) and effectively attenuated EAE in Ifnar1 2/2 mice (Fig. 4D). This result demonstrates that although lithium inhibits STAT1 activation initiated by IFNAR engagement (Fig. 2B), the type I IFN receptor is dispensable for lithium's therapeutic activity in EAE (Fig. 4D).

GSK3 Regulates IFN-c Signaling Dependent Expression of Il27 in EAE
IL-27 is an IFN-c response cytokine that has an essential role in limiting EAE by a mechanism that includes, but not limited to the induction of IL-10 in CD4 + T cells [34][35][36]. We recently reported that IFN-b promotes the expression of IL-27 in an IFN-c dependent manner and consequently induces the expression of IL-10 in Th1 cells to limit Th1 EAE [15]. We therefore investigated if lithium treatment induces Il27 expression in the CNS of mice with EAE. We found that Il27 expression was upregulated in the spinal cord of WT mice with EAE at the peak of disease and lithium treatment significantly elevated it by approximately 2 fold with respect to untreated mice (Fig. 4E). The induction of Il27 was much diminished in Ifngr1 2/2 mice and was not altered by lithium treatment. IFN-b, a cytokine expressed in the CNS during EAE, can also induce Il27 expression [37]. We found upregulation of Ifnb at peak of disease, but this was not further increased by lithium (Fig. 4E). Therefore our data indicate that GSK3 regulates IFN-c induced, but not IFN-b-induced, IL-27 expression. Il10 expression in CNS of WT mice correlated with Il27 expression (Fig. 4E). Although Il10 expression in Ifngr1 2/2 mice was elevated, its expression levels did not increase with lithium treatment. We also observed that lithium treatment enhanced Nos2 (inducible nitric oxide synthase) in WT but not in Ifngr1 2/2 mice. However the increases in Il10 and Nos2 were not statistically significant (Fig. 4E). Overall our data support the tenet that lithium attenuates EAE by an IFN-c-dependent expression of IL-27.
In conclusion, we discovered a novel mechanism for the therapeutic activity of lithium in EAE that involves the tuning of STAT1 in an IFN-c-signaling dependent manner. Our study provides new insights in limiting CNS autoimmune disease by regulation of IFN-c signaling, that is independent of signaling through the receptor for IFN-a/b. These findings will contribute to development of new treatment strategies for neuroinflammatory diseases, as well as offer the opportunity to use GSK3 inhibitors as combined therapy with IFN-b for the treatment of MS to enhance therapeutic effectiveness.  with anti-CD3 and anti-CD28 (1 mg/ml each) for 8 h in the absence or presence of LiCl was assessed by ELISA. Representative results shown (n = 2). *p,0.05, as determined by t-test or one-way ANOVA, as appropriate. doi:10.1371/journal.pone.0052658.g003 Table 2. Analysis of disease parameters for active EAE induced in untreated and lithium-treated WT, Stat1 2/2 , Ifngr1 2/2 and Ifnar1 2/2 mice.