Metabolic control of TH17 and induced Treg cell balance by an epigenetic mechanism

Abstract

Metabolism has been shown to integrate with epigenetics and transcription to modulate cell fate and function. Beyond meeting the bioenergetic and biosynthetic demands of T-cell differentiation, whether metabolism might control T-cell fate by an epigenetic mechanism is unclear. Here, through the discovery and mechanistic characterization of a small molecule, (aminooxy)acetic acid, that reprograms the differentiation of T helper 17 (T_H17) cells towards induced regulatory T (iT_reg) cells, we show that increased transamination, mainly catalysed by GOT1, leads to increased levels of 2-hydroxyglutarate in differentiating T_H17 cells. The accumulation of 2-hydroxyglutarate resulted in hypermethylation of the Foxp3 gene locus and inhibited Foxp3 transcription, which is essential for fate determination towards T_H17 cells. Inhibition of the conversion of glutamate to α-ketoglutaric acid prevented the production of 2-hydroxyglutarate, reduced methylation of the Foxp3 gene locus, and increased Foxp3 expression. This consequently blocked the differentiation of T_H17 cells by antagonizing the function of transcription factor RORγt and promoted polarization into iT_reg cells. Selective inhibition of GOT1 with (aminooxy)acetic acid ameliorated experimental autoimmune encephalomyelitis in a therapeutic mouse model by regulating the balance between T_H17 and iT_reg cells. Targeting a glutamate-dependent metabolic pathway thus represents a new strategy for developing therapeutic agents against T_H17-mediated autoimmune diseases.

Extended Data Figure 1.

AOA reprograms T_H17 cell differentiation toward ITreg cells. a. Screening procedure showing how the screening was conducted. b. Effects of AOA on T_H17 cell differentiation. c, The effect of AOA on mTOR, AMPK, and c-Myc. Differentiating T_H17 cells (with or without 0.75 mM AOA) were collected for analyzing p-AMPK1a, mTOR downstream signaling proteins, S6K, 4E-BP1 or c-Myc. d, The effect of AOA on the proliferation of T_H17 cells. CD4 naive T cells were labeled with cell tracer violet, and then cultured under T_H17 condition or ITreg condition (the same condition as T_H17 condition except without IL-1β, IL-6 and IL-23). At the end of experiment, the cells were analyzed by intracellular staining. e) AOA promoted ITreg cell differentiation. f) AOA did not affect the survival of T_H17 cell culture. The cells were differentiated under T_H17 condition with or without AOA. At the end of differentiation, the cells were collected for analysis immediately after 7-AAD was added into cell culture.

Extended Data Figure 2.

¹⁵N-labeling analysis showed that Got1 mediates the majority of transamination and represents the main target for AOA in T_H17 cells. a) Schematic of how 15N-α-glutamine was metabolized in transamination reaction. b) The ratios of ¹⁵N-labeled amino acids to their respective intracellular amino acid pool (related to Fig. 1f). Differentiating T_H17 cells or iTreg cells (68 hours) were fed with 2 mM ¹⁵N-α-glutamine for 4 hours. The cells were collected for intracellular metabolites analysis. The ratios of 15N labeled aspartate, glycine, alanine, serine to their total respective amino acid pools were calculated. c, AOA, as a pan-tranaminase inhibitor, inhibited de novo synthesis for several amino acids (via transamination) in both cell types in addition to aspartate. The ratio of 15N-amino acid to 15N-glutamate was calculated, and this ratio was further normalized to that in T_H17 cells to reflect that AOA inhibited de novo synthesis for several amino acids. Although, AOA indeed inhibited de novo synthesis for several amino acids (via transamination) in both cell types, our rescue results clearly showed (in figure 2f and 2g) that dimethyl α-KG can largely rescued the effect of AOA on both T_H17 and iTreg cell differentiation. Thus, from a metabolic perspective, AOÀs effect can be largely attributed to its inhibitory effect on α-KG formation (carbon metabolism), rather than its inhibitory effect on amino acid synthesis (nitrogen metabolism). Therefore, we are focusing on carbon metabolism of glutamate in this study. d, Got1 is the main target for AOA in T_H17 cells. Differentiating T_H17 cells were infected with retrovirus containing shRNA against Got1 or control shRNA. The GFP+ cells were then purified on day 3, and further cultured under T_H17 condition with/or without AOA. At the end of differentiation (day 6), the cells were collected for analysis of FOXP3 and IL-17 by intracellular staining. It is clear that AOA can further inhibit T_H17 cell differentiation, however, the effect is quite subtle, which supported the conclusion that Got1 is the main target for AOA under T_H17 condition. Bar Graphs in b and c are mean±S.D. of 3 technical replicates from a representative experiment. Representative flow data are presented in d, the experiment was repeated twice.

Extended Data Figure 3.

Metabolic profiling of T_H17 cells and iTreg cells. a) Intracellular metabolites profiling of differentiating T_H17 cells and iTreg cells performed by LC/MS. * refer to the metabolites with differential abundance between T_H17 cells and iTreg cell, which is inhibited by AOA. b) 2-HG concentration is much higher in differentiating T_H17 cells than IiTreg cells along differentiation time line. c) AOA significantly increased 2-HG, while does not affect the level of L-Lactic acid, glutathione, oxidized glutathione, L- aspartate, slightly decrease α-KG, slightly increase glutamate related to Extended Data Fig. 3a. The relative levels of 2-HG, L-lactic acid, L-glutamate, glutathione, oxidized glutathione, L-aspartate and α-KG from Extended Data Fig.3a were re-plotted in c. The data shown in b and c are mean±S.D. of three replicates from a representative experiment of three independent experiments.

Extended Data Figure 4.

Exogenously added α-KG and 2-HG rescued the effects of AOA on T_H17 and iTreg cell differentiation related to Fig. 2f and 2g. a), Cell-permeable dimethyl esters of succinate, fumarate, malate, citrate, NAC or GSH, did not rescue the inhibitory effects of AOA on T_H17 cell differentiation. Cell-permeable metabolites (0.75 mM α-KG, 0.5 mM 2-HG, 0.5 mM succinate, 50 μM fumarate, 0.5 mM malate, 0.5 mM citrate, 1 mM NAC, and 1 mM GSH) were individually added to differentiating T_H17 cells in the presence of AOA. At the end of differentiation (day 6), the cells were re-stimulated and analyzed by intracellular staining of FOXP3 and IL-17. b) Cell-permeable dimethyl esters of α-KG, 2-HG, but not succinate, fumarate, malate or citrate rescued the effects of AOA on iTreg cell differentiation. Cell permeable metabolites (0.5 mM succinate, 50 μM fumarate, 0.5 mM malate, 0.5 mM citrate) were individually added to differentiating iTreg cells in the presence of AOA. At the end of differentiation (day 5), the cells were directly analyzed for FOXP3-GFP. The representative flow data from 3 independent experiments were shown in a and b (left panel). Bar graphs in the right panel of a and b are mean±S.D. of 3 independent experiments. NS=non-significant; *P<0.05; **P<0.01; ***P<0.001 by Student`s t-test.

Extended Data Figure 5.

The effect of cell-permeable metabolites (α-KG, 2-HG, citrate, succinate, fumarate, malate) on the differentiation of T_H17 or iTreg cells. a) Indicated dimethyl metabolites were added into T_H17 culture, and the cells were analyzed on day 4 by intracellular staining of FOXP3 and IL-17. b) The indicated metabolites were added into iTreg culture, and the cells were analyzed on day 5 by intracellular staining of FOXP3 and IL-17. c, and d, 2-HG did not affect cell proliferation and survival. c, CD4 naive T cells were labeled with cell tracer violet according to manufacturès protocol. The cells were then differentiated under T_H17 condition. The cells were then stimulated and collected for intracellular staining at day 4. The example flow data was plotted in different format. d, CD4 naive T cells from double reporter mice were differentiated under T_H17 condition, and live cells were analyzed at day 4. 7-AAD was added right before analysis. The representative flow data from three independent experiments are shown in a and b (right panel). Bar graph in the right panel of a and b are mean±S.D. of three independent experiments. NS=non-significant; *P<0.05; **P<0.01; ***P<0.001 by Student`s t-test.

Extended Data Figure 6.

Differentiating T_H17 cells highly expressed IDH1 and IDH2, and shRNA against IDH1 or IDH2 effectively suppressed the expression of IDH1 and IDH2, decreased DNA methylation at FOXP3 locus, and suppressed T_H17 cell differentiation. a) Differentiating T_H17 cells or iTreg cells (day 3) were collected for mRNA expression analysis. All expression levels were normalized to β-actin, and the expression level of each enzyme was normalized to that in differentiating iTreg cells. b) Differentiated T_H17 cells and iTreg cells have similar expression of IDH3. The experiment was done exactly in a. Expression of IDH3 subunits was normalized to β-actin, and further plotted as relative level to that gene expression in iTreg cells. c. Knockdown of IDH1 and/or IDH2 efficiently suppressed the mRNA expression of IDH1 and IDH2, and Il17a, Il17f, and increases FOXP3 mRNA expression. Infected cells (GFP+ cells) containing shRNA against IDH1 or 2 were FACS sorted and re-stimulated with anti-CD3 and anti-CD28 for mRNA expression analysis. Expression was normalized to β-actin. d, knockdown of IDH1 and 2 decreased methylation level at FOXP3 locus. Differentiating T_H17 cells were infected with retrovirus containing ShGot1 or control shRNA. At the end of differentiation (day 6), the GFP+ cells were collected for DNA methylation analysis of FOXP3 promoter and its intronic CpG island by bisulfate sequencing. , methylated cytosine; , demethylated cytosine; male mice were used due to X-chromosome inactivation. e, KD of IDH1/2 suppressed T_H17 cell differentiation, which can be reversed by adding back R-2-HG. R-2-HG was added into differentiating T_H17 cells at 6 hr. The cells were infected with retrovirus containing shRNA, GFP+ cells were purified at day 3 for further culture under T_H17 condition, and R-2-HG was added into the culture till the end of experiments. Then, the cells were collected for analysis of FOXP3 and IL-17. f, KD of IDH1 or 2 has very minimal effect on HIF1α expression. Differentiating T_H17 cells were infected with retrovirus containing shRNA or shGot1. The cells were then purified at the end of experiment for analyzing IDH1, IDH2, and HIF1α expression. Mean±S.D.(n=3) of three technical replicates from a representative experiments of three experiments were shown in a, b, and c. Bar graph in e (n=3) is mean±S.D. of three independent experiments. *P<0.05; **P<0.01; ***P<0.001 by Student`s t-test.

Extended Data Figure 7.

Tet1/2 controls FOXP3 expression during T_H17 cell differentiation. a, Tet1/2 DKO promoted T_H17 cell differentiation, and largely abrogated the promoting effect of R-2-HG on T_H17 cell differentiation in WT T cells. CD4 naive T cells from Tet1/2 DKO or control mice were differentiated under T_H17 condition with or without R-2-HG. At day 4, the cells were collected for analysis of FOXP3 and IL-17 by intracellular staining. Our result clearly showed that Tet1/2 DKO accelerated T_H17 cells differentiation, consistent with previous study showing that Tet1/2 DKO regulatory T cells can be more easily and efficiently converted into T_H17 cells. b, Tet1/2 DKO partially diminished the inhibitory effect of AOA on T_H17 Cell differentiation. CD4 naive T cells from Tet1/2 DKO or control mice were differentiated under T_H17 condition with or without AOA. At the end of differentiation (day 6), the cells were collected for analysis of FOXP3 and IL-17. c) Reduced TGFβ concentration largely abolished the effect of Tet1/2 DKO on T_H17 cell differentiation, and Tet1/2 DKO even decreased IL-17 expression when TGFβ concentration is low enough, indicating Tet1/2 proteins have a dual function in the fate determination of T_H17 cell differentiation. CD4 naive T cells derived from Tet1/2 DKO mice or control mice were differentiated under T_H17 condition with varied concentration of TGFβ. The cells were collected for intracellular analysis at day 4. Previous study showed that Tet2 positively regulate T_H17 differentiation by binding to the Il17 gene locus. However, our studies showed that Tet1/2 DKO T cells enhanced T_H17 differentiation, as determined by increased IL-17 and reduced FOXP3 expression at day 4 in T_H17 culture. Considering that the Tet enzyme activity is very sensitive to various exogenous stimuli as described before. We speculate that the discrepancy between our studies and Ichiyama et al’s study could be caused by different culture condition, in which our T_H17 polarizing condition yielded high amount of FOXP3 even at later stage of T_H17 differentiation (day 4). To test this possibility, we reduced the amount of TGFβ used in our cultures, consistent with Ichiyama et al’s finding, Tet1/2 DKO caused a significant reduction of T_H17 differentiation when TGFβ concentration is low enough (Extended Data Fig 7c). A possible explanation is that high TGFβ condition induced strong and persistent activation of Smad3 and STAT5, which then recruit Tet proteins as reported to the Foxp3 gene locus and promote its expression by inducing or maintaining the demethylation status, whereas at low TGFβ condition, the recruitment of Tet enzymes to the Il17 gene locus plays a dominant role in regulating T_H17 differentiation. Our study thus identified dual-functions of Tet proteins in fate determination of T_H17 differentiation. In summary, the role of Tet1/2 during T_H17 cell differentiation is dynamic and dependent on the interplay between these targets. When FOXP3 expression is high (due to high TGFβ level or at early stage of T_H17 cell differentiation), Tet1/2 proteins suppress IL-17 expression indirectly through increased FOXP3 expression. When FOXP3 expression is low (due to low TGFβ level, therefore much less Tet1–2 will be recruited to FOXP3 locus or at the very late stage of T_H17 cell differentiation due to the selective antagonistic effect of the accumulated 2-HG on FOXP3 expression), Tet1–2 positively regulate T_H17 cell differentiation. Representative flow data from five (a), three (b) or two (c) independent experiments are shown. Bar graph in right panel of a (n=5) or b (n=3) is combination of five (a) or three (b) independent experiments and presented as mean±S.D. Mean±S.D. of 3 technical replicates (n=3) from a representative experiment of two independent experiments is shown in right panel of c. NS=non-significant, *P<0.05; **P<0.01; ***P<0.001 by Student`s t-test.

Extended Data Figure 8

AOA and R-2-HG selectively affected DNA hydroxymethylation and methylation at FOXP3 locus, but not other important lineage specific signature gene loci examined by (h)MeDIP-seq. a, and b, Exogenous addition of dimethyl R-2-HG selectively decreases 5hmC signal and increase 5mC signal at FOXP3 locus. DNA extracted from differentiating T_H17 cells (day 4) in the absence or presence of 0.75 mM dimethyl R-2-HG were immunoprecipitated with antibodies to 5hmC (a) or 5 mC (b), followed by deep sequencing. The sequence data were analyzed as described in method, and differential peaks from hMeDIP were plotted in a. 5hmC peaks located in FOXP3 conservative region or in other conservative regions at Il4, Il5, Il13, Il10, IFNγ, Rorc, etc. were highlighted. Peak 1,2,11 were located in Gata3; peak 3 in Rorc, peak 4 in Il4, peak 5,16 in Il10, peak 6 in IFNγ, peak 7 and 10 in Il5, peak 8, and 9 in Tbx21, peak 12 in Il13, peak 13,15 in Il17a, peak 14 in Il17f. Among 330, 582 peaks detected, R-2-HG decreased 5hmC signal in 17,676 peaks, however, R-2-HG did not decrease DNA hydroxymethylation at Il4, Il5, Il10, Il13, Il17a/f, IFNγ, Rorc, Tbx21 loci. b, Further analysis of 5mC signal in these 17,676 peaks, 3,402 peaks exhibited increased 5mC signal. Apparently, R-2-HG indeed selectively decreased 5hmC signal and increased 5mC signal at FOXP3 promoter and CNS2 region, but did not affect 5hmC or 5mC signal at IFNg, Il4/5/10/13, Il17a/f, Tbx21, Rorc. Notably, exogenous dimethyl R-2-HG did not affect 5hmC or 5mC signal only at FOXP3 locus, rather it had a more broad effect at many loci. c, and d, AOA treatment selectively affect DNA hydroxymethylation and methylation at FOXP3 locus. DNA extracted from T_H17 culture in the absence or presence of 0.75 mM AOA were immunoprecipitated with antibodies to 5hmC (c) or 5mC (d), followed by deep sequencing. The sequence data were analyzed as described, and differential peaks were plotted. 5hmC peaks located in FOXP3 conservative region or in other conservative regions at Il4, Il5,Il13, Il10, IFNγ, Rorc, etc. were highlighted, the labels are the same as in a. Among 330, 582 peaks detected, AOA increased 5hmC signal in 11,896 peaks. Consistently, AOA increase hydroxymethylation at FOXP3 promoter and CNS2 region, but AOA did not affect DNA hydroxymethylation at IFNү, Il4, Il5, Il10, Il13, Rorc, Tbx21, etc. Notably, AOA treatment reduced 5hmC signal at Il17a/f loci, probably due to the antagonistic effect of FOXP3 on Rorүt to recruit Tet proteins to Il17a/f loci. d, Further analysis of 5mC signal in these 11,896 peaks, 1643 peaks exhibited increased 5mC signal. Apparently, AOA indeed decreased 5mC signal at FOXP3 promoter. Notably, the changes in 5mC signal at FOXP3 CNS2 region was unable to be detected, probably due to the fact that CNS2 region is largely methylated (around 70–80% of it is methylated in iTregs), and changes in 5mC is more subtle and difficult to be detected than 5hmC. Notably, both AOA and 2-HG affected 5hmC signal, but not 5mC signal at Gata3, indicating that AOA and 2-HG indeed have a minimal effect on DNA methylation at Gata3 locus. However, Gata3 expression is not regulated by DNA hydroxymethylation as shown in previous study, therefore 2-HG and AOA are unlikely to affect Gata3 expression. e, 17,676 peaks with increased 5hmC signal by AOA (from a, blue) is overlapped with 11,896 peaks with decreased 5hmC signal by 2-HG (from c, green). f, 3402 peaks with increased 5mC by 2-HG (from b, purple) overlapped with 1643 peaks with decreased 5mC signal by AOA (from d, red). These epigenetics analysis clearly showed that the effect of AOA and 2-HG on Foxp3 vs. other T-lineage related genes is highly selective, despite their more broad effects on genome-wide DNA methylation and hydroxymethylation. Although demethylation at FOXP3 locus promotes FOXP3 expression, expression of many other genes is not regulated by DNA demethylation, such as Rorc, Gata3. Treatment of T cells with AOA under T_H17 condition stabilizes FOXP3 expression, while not having much effect on the expression of other lineage-specific transcription factors, such as Gata3, Rorc. FOXP3 can antagonize the function of RORγt to suppress expression of T_H17 signature genes as well as recruit Dnmt1 to the gene loci of proinflammatory cytokines or signature genes to promote methylation at these loci, suppressing their expression. In addition, FOXP3 functions as both a transcriptional activator to directly activate its target genes required for ITreg cell differentiation/function, and a transcriptional repressor to directly suppress the genes associated with effector T cell function, resulting in ITreg cell fate.

Extended Data Figure 9.

KD of Got1 using 2^nd shRNA targeting Got1 ameliorated mouse EAE disease in T_H17 polarized adoptive transfer EAE (related to Figure 4). The disease score was recorded in a and presented as mean±S.D. *P<0.05; **P<0.01; ***P<0.001 by Student`s t-test. CNS infiltrated T cells were analyzed (Gated on Thy1.1+ cells), and representative flow data were shown in b. The statistics for CNS infiltrated T cells is in c and presented as mean±S.D. *P<0.05; **P<0.01; ***P<0.001 by Student`s t-test. d, Quantification of cell number infiltrated into CNS regions. e, Diagram shows our working mechanistic model. The cell number in Fig 4d–g was retrieved from the FACS data. In this experiment, we did not count the number of individual populations, but we did notice that the total cells infiltrating into the CNS was reduced by shGot1 KD, based on our flow cytometry data that recovery of shGot1 group cells were below 50,000 cells, while every control mice recovered more than 50,000 cells. Thus, for mice receiving T cells infected with shRNA control virus, cell number was calculated from FACS-acquired 50,000 total live cells (this group of mice have much higher number of total infiltrated cells, and we stopped acquiring after the live cells reached 50,000 threshold during FACS running). Mice receiving T cells infected with Got1 shRNA virus have much less cells infiltrated into CNS, and harvested cells from CNS did not reach 50,000 threshold when run out all the samples during FACS running, therefore, the cell numbers represent all the cells infiltrated into CNS region. Despite not very accurate, shGot1 decreased total absolute number of cells infiltrated into CNS. For other two EAE experiments, we counted cell numbers right after we isolated cells from CNS.

Figure 1.

AOA reprograms T_H17 cell differentiation toward iTreg cells by inhibiting Got1. a) AOA reprograms T_H17 cell differentiation toward FOXP3+ iTreg T cells. b) mRNA expression in cells from a. c) AOA promotes iTreg cell induction under iTreg condition. d) FOXP3 mRNA expression from cells in c. e) Got1/2 is the major transaminase catalyzing glutamate flux into α-KG. Mean±S.D. of three replicates from a representative experiment was shown. f) Got1 is highly up-regulated under T_H17 condition. g) Knockdown of Got1 reduced T_H17 cell differentiation and reciprocally increased iTreg cell differentiation. h) Statistics of cell population from f. i) the effect of knockdown of Got1 on gene expression. Representative flow data from five (a) or three (c, and g) independent experiments were shown. Data are presented as mean±S.D. of five (a, right panel), or three (right panel of c, and h) independent experiments. Mean±S.D. of 3 replicates from a representative experiment of three independent experiments was shown in b, d and i. NS=non- significant; *P<0.05; **P<0.01; ***P<0.001 by Student`s t-test.

Figure 2.

2-HG derived from glutamine/glutamate is highly elevated under TH17 condition, and facilitates TH17 cell differentiation. a) The relative abundance of significantly changing metabolites from Extended Data Figure 3a was normalized to that in iTreg cells. b) Schematic of labeling patterns for U-13C glutamine feed for TCA cycle intermediates. c) U-13C-Glutamine labeling shows that carbon of glutamine enters into TCA cycle and is responsible for 2-HG synthesis. d) Abundance of 2-HG+5 and α-KG+5 (normalized to intracellular U-13C-glutamine). e) Exogenously added α-KG and 2-HG rescued the effects of AOA on TH17 cell differentiation. f) Cell-permeable dimethyl esters of α-KG, 2-HG rescued the effects of AOA on iTreg cell differentiation. Mean±S.D. of three replicates from a representative data of three (a) or two independent experiments (c and d) are presented in a, c, and d. Representative flow data from 3 independent experiments were shown in e and f. Mean±S.D.(n=3) of three independent experiments is presented in right panel of e and f. *P<0.05; **P<0.01; ***P<0.001 by Student`s t-test.

Figure 3.

2-HG promotes T_H17 cell differentiation by promoting methylation of FOXP3 locus. a, Dimethyl R-2-HG promoted T_H17 cell differentiation. b, mRNA expression from cells in a. c, Dimethyl 2-HG inhibited iTreg cell differentiation. d, FOXP3 mRNA from cells in c. e, Knockdown of IDH1/2 decreased 2-HG production in T_H17 culture. f, Knockdown of IDH1/2 suppressed T_H17 cell differentiation and reciprocally promoted iTreg cell differentiation. g, Exogenous dimethyl R-2-HG promotes methylation of FOXP3 locus during T_H17 or iTreg cell differentiation. h, AOA promoted hypomethylation of FOXP3 locus. In g and h, , methylated cytosine; , demethylated cytosine. i and j) The effect of 2-HG or AOA on hydroxymethylation/ methylation at FOXP3 locus examined by (h)MeDIP-seq. Peaks of 5hmC (blue) or 5mC peaks (red) at FOXP3 locus were shown. In g-j, Male mice were used due to the X-chromosome inactivation. Flow data from a representative experiment were shown in a, c, and f. The experiments were repeated at least three times. Bar graph in right panel of a, c, f are mean±S.D. of three independent experiments. Bar graph in e (n=12) is combination of four independent experiments, and presented as mean±S.D.. Mean±S.D. of 3 replicates from a representative experiment of three independent experiments is presented in b and d. NS=non-significant; *P<0.05; **P<0.01; ***P<0.001 by Student`s t-test.

Figure 4.

AOA produced significant recovery from EAE diseases mainly through targeting Got1. a-c, AOA ameliorated mouse EAE diseases, and disease scores were recorded in a. b, the representative flow data of T cells (gated on Thy1.1+) infiltrated into CNS. c, statistics of each population from b, n=11 (control), n=10 (AOA). The results in a-c are combinations of two experiments. d-g, T_H17 polarized adoptive transfer EAE showed that knockdown of Got1 ameliorated mice EAE diseases. The disease score was recorded in d, and disease incidence is in e. CNS infiltrated T cells were analyzed (Gated on Thy1.1+ cells), and representative flow data were shown in f. The statistics for CNS infiltrated T cells is in g. d-g was from one representative experiment of two experiments. The data in a, c, g were presented as mean±S.D. The disease score in d is represented as mean±s.e.m. NS=non-significant; *P<0.05; **P<0.01; ***P<0.001 by Student`s t-test.