Self-antigen driven activation induces instability of regulatory T cells during an inflammatory autoimmune response

Abstract

Stable Foxp3 expression is crucial for regulatory T (Treg) cell function. We observed that antigen-driven activation and inflammation in the central nervous system (CNS) promoted Foxp3 instability selectively in the autoreactive Treg cells that expressed high Foxp3 levels before experimental autoimmune encephalitis induction. Treg cells with a demethylated Treg cell-specific demethylated region in the Foxp3 locus down-regulated Foxp3 transcription in the inflamed CNS during the induction phase of the response. Stable Foxp3 expression returned at the population level with the resolution of inflammation or was rescued by IL-2:anti-IL-2 complex treatment during the antigen priming phase. Thus, a subset of fully committed self-antigen-specific Treg cells lost Foxp3 expression during an inflammatory autoimmune response and may be involved in inadequate control of autoimmunity. These results have important implications for Treg cell therapies, and give insights into the dynamics of the Treg cell network during auto-reactive CD4⁺ T cell effector responses in vivo.

Introduction

CD4⁺CD25⁺ Foxp3⁺ regulatory T (Treg) cells are crucial for self-tolerance and for maintaining balanced immune responses. Reduced numbers or function of Treg cells have been associated with the onset of autoimmunity (Long and Buckner, 2011; Tang et al., 2008) whereas increasing the number of Treg cells has had therapeutic success in models of autoimmunity and graft versus host disease. In fact, Treg cell-based therapies are currently being tested in clinical trials (Brunstein et al., 2011 ; clinicaltrials.gov). Treg cells constitutively express the IL-2 receptor (IL-2R) and depend on IL-2 for survival and homeostasis in the periphery (Burchill et al., 2008; Cheng et al., 2011). Treg cells have attributes of activated conventional T cells with constitutively activated T cell receptor (TCR) signaling pathways and un-committed chromatin marks on many “T effector” gene loci, such as IFN-γ (Moran et al., 2011; Salomon et al., 2000; Wei et al., 2009). The Treg cell transcription factor Foxp3 activates genes required for suppressor function, and is essential for maintaining the transcriptional program of the Treg cell lineage (Gavin et al., 2007). In fact, targeted deletion of Foxp3 in Treg cells turns them into IFN-γ or IL-2 producing T effector cells (Williams and Rudensky, 2007). In this regard, Treg cells deficient for single transcription factors or key signaling proteins have identified pathways crucial for Foxp3 expression and stability (Kitoh et al., 2009; Rudra et al., 2009; Vanvalkenburgh et al., 2011; Wang et al., 2011; Yao et al., 2007; Zanin-Zhorov et al., 2010; Zheng et al., 2010). These studies support the general concept that altered signaling can lead to Treg cell instability and the possibility that, at least a subset of Treg cells may be unstable and lose Foxp3 expression once the cell lineage is fully established.

We, and others, have shown that a substantial subset of unmanipulated CD4⁺ T cells express low levels of Foxp3 in vivo, especially in lymphopenic and inflammatory settings (Zhou et al., 2009; Miyao et al., 2012). Using lineage reporter and tracer mice to characterize loss of Foxp3 expression and concomitant functional activity (Rubtsov et al., 2010; Sharma et al., 2010; Zhou et al., 2009), it was observed that Treg cell instability results in loss of the regulatory network that maintains self-tolerance. Additionally, cells that down-regulate Foxp3 expression (“exFoxp3” cells) produce proinflammatory cytokines and can act as effector cells causing tissue destruction if they are self-reactive (Zhou et al., 2009), a characteristic of thymically-derived “natural” Treg cells (Hsieh et al., 2006; Wong et al., 2007). Thus, the emergence of effector cell-like characteristics in this population could have serious repercussions for autoimmunity both in terms of loss of regulation as well as potential pathogenic activity.

More recent studies have argued that many, if not all of these exFoxp3 cells, derive from an early, “aborted” T cell differentiation process that occurs prior to full Treg cell commitment rather than from instability of bona fide nTregs (Rubtsov et al., 2010; Miyao et al., 2012). These studies were conducted under largely homeostatic conditions in the steady-state, in vitro or in the setting of acute lymphopenia, thus raising the question whether the Treg instability observed by us and others may be related to the inflammatory pathogenic setting in our studies. Indeed a number of reports have demonstrated that Treg cell reprogramming and acquisition of pathogenic potential in autoimmunity, graft versus host disease and vaccination settings (Dominguez-Villar et al., 2011; Laurence et al., 2012; McClymont et al., 2011; Sharma et al., 2010; Zhou et al., 2009), consistent with the suggestion that active immunity may have direct effects on Treg cell stability. Therefore, in this study, we set out to examine Foxp3 stability in bona fide Foxp3^hi Treg cells responding to self-antigen within a polyclonal T cell repertoire and in the context of an active CD4⁺ T cell autoimmune response. Using an experimentally-induced autoimmune encephalomyelitis (EAE) model, we observed that antigen-driven activation and inflammation promoted Foxp3 instability selectively in the autoreactive Treg cells that expressed high levels of Foxp3 before EAE induction. Transfer experiments demonstrated that bona fide Treg cells with a demethylated T regulatory cell-specific demethylated region (TSDR) in the Foxp3 locus down-regulated Foxp3 transcription during the induction phase of the response. Stimulation with cognate autoantigen induced IFN-γ production by the exFoxp3 cells in the central nervous system at the peak of the response. Stable Foxp3 expression returned with the resolution of inflammation or could be rescued by enhancing IL-2 receptor signaling with IL-2:anti-IL-2 complex treatment during the antigen priming phase. These findings suggest that a subset of antigen-specific Treg cells participating in the control of an immune response can be reprogrammed and may play a role as potentially pathogenic cells during autoimmunity.

Results

Unstable Foxp3 expression during EAE in C57BL/6 mice

Treg cells were analyzed in EAE induced in the C57BL/6 (B6) genetic background. The previously described Foxp3-lineage reporter mice (Zhou et al., 2009) were backcrossed more than 8 generations onto the B6 background. In these bacterial artificial chromosome (BAC) transgenic mice, Foxp3 promoter and regulatory elements drive Cre recombinase-green fluorescent protein (GFP) fusion protein. These mice were bred to two different independent mouse strains that express either a yellow fluorescent protein (YFP) or red fluorescent protein (RFP) transgene engineered with a stop codon flanked by lox-P sites and inserted into the Rosa26 locus. In the dual expressing (Foxp3.GFP-Cre and Rosa26.YFP or Rosa26.RFP) reporter mice, any cell expressing Foxp3 will express RFP or YFP for its lifetime, whereas GFP will be expressed only in cells that are currently expressing Foxp3. The CD4⁺ T cell compartment of 6-8 week old B6 Foxp3-Cre BAC transgenic mice crossed to Rosa26.RFP mice contains 0.5-1.5% CD4⁺ T cells that have reduced or lost Foxp3 expression (termed exFoxp3; Figure 1A) in steady state. These data were confirmed in another line of B6 mice generated with Cre recombinase expressed in the Foxp3 3’ untranslated region (UTR) (Rubtsov et al., 2008) and crossed to Rosa26.RFP mice (Supplemental Figure 1). These results demonstrated that Foxp3 down-regulation occurred within the polyclonal Treg cell population in a lymphoreplete, intact immune environment, albeit a small percentage of the cells.

Figure 1. MOG_38-49-specific Tregs down-regulate Foxp3 during EAE.

(A) Expression of GFP and RFP in lymph node and spleen CD4⁺ T cells of a 6 week old C57Bl/6 Foxp3.GFP.Cre.Rosa26.RFP mouse. Representative of 15 Foxp3.GFP.Cre.Rosa26.RFP or Foxp3.GFP.Cre.Rosa26.YFP mice. The percentage of T conventional (Tconv) (RFP⁻ Foxp3.GFP⁻), Treg (RFP⁺ Foxp3.GFP⁺) and exFoxp3 (RFP⁺ Foxp3.GFP⁻) cells are indicated. (B) High affinity MOG_38-49-specific T cells in the CD4⁺ fraction detected using I-A^b:tetramers at the indicated stages of EAE in LN (lymph node) & spleen or CNS (central nervous system: spinal cord and cerebellum). Dot-plots are gated on CD4⁺ T cells. Representative of 3 – 6 experiments. (C) CD4⁺ T cells were gated on tetramer negative (polyclonal) (top) and MOG_38-49-specific (bottom) cells and analyzed for the frequency of Tconv, Treg and exFoxp3 cells at the indicated stages of EAE in the LN and spleen or CNS. Representative of 3 – 6 experiments using Foxp3.GFP.Cre.Rosa26.RFP or Foxp3.GFP.Cre.Rosa26.YFP mice. (D, E) The proportion of Tconv, Treg and exFoxp3 cells in MOG_38-49-specific (filled square) and polyclonal (open circle) CD4⁺ populations are shown in the LN and spleen (D), and CNS (E) compartments at the indicated stages of EAE. Data points are individual Foxp3.GFP.Cre.Rosa26.RFP or Foxp3.GFP.Cre.Rosa26.YFP mice, red lines show the mean +/− SEM. The percentage of each population was not significantly different in polyclonal vs. MOG_38-49-specific cells using a paired T test. N.D.; not done. See also Figure S1.

Next, we induced EAE by immunizing B6 mice with MOG_35-55 peptide in complete Freund’s adjuvant (CFA). Lymphocytes were harvested from the draining lymph nodes (LNs) and spleen, and CNS tissues of immunized mice and examined for evidence of antigen-specific T cell expansion and differentiation using an MHC-peptide tetramer, I-A^b-MOG_38-49, which bound to MOG_35-55 peptide-specific T cells as previously described (Korn et al., 2007). Using this probe, we analyzed MOG_38-49-specific CD4⁺ T cells among the polyclonal CD4⁺ T cell population during the asymptomatic and inflammatory phases of MOG_35-55-induced EAE. Following an enrichment step, MOG_38-49-reactive cells accounted for 4% of CD4⁺ T cells in the peripheral T cell compartment after EAE induction (Figure 1B). The tetramer staining was specific as control I-A^b:HClip tetramer staining was negligible in this population (not shown). For further studies, we focused on an analysis of antigen-specific T cells within polyclonal populations.

Initial studies showed that there was virtually no detectable (0.2% of CD4⁺ T cells following an enrichment step in vitro for tetramer bound cells) I-A^b: MOG_38-49 tetramer⁺ CD4⁺ T cells in the LNs prior to immunization (data not shown). Thus, following immunization MOG_38-49-specific CD4⁺ T cells expanded in draining LNs and spleen and expressed CD44 indicative of antigen-driven activation. During the clinical phases of EAE, all CD4⁺ T cells in the CNS expressed high amounts of the activation marker CD44, and MOG_38-49-specific cells accumulated and represented >4% of CD4⁺ T cells in the CNS (without any enrichment step in vitro as required to see the cells in the draining LNs) (Figure 1B). We assayed the ‘quality’ of the self-antigen-specific Treg cells during an autoimmune response, tracing the kinetics of Treg cell during EAE and the stability of Foxp3 expression with lineage traced Treg cells. We used Foxp3.Cre.GFP × Rosa26.RFP (Figure 1C) as well as Foxp3.Cre.GFP × Rosa26.YFP and saw no differences in the percentage of Treg and exFoxp3 cells when the data was pooled from the different lineage tracer mice (Figure 1D). ExFoxp3 cells were identified as expressing low – negative amounts of GFP equivalent with RFP⁻YFP⁻ Tconv cells (Figure 1C, E). ExFoxp3 cells made up a higher proportion of MOG_38-49-specific than polyclonal cells at the preclinical stage of EAE, in the LN and spleen 8 days after immunization, and at the peak stage of the disease in the CNS (Figure 1C,D,E). During EAE resolution, the proportion of exFoxp3 cells in the MOG_38-49-specific CD4⁺ T cells was higher than polyclonal CD4⁺ T cells in the LN and spleen, but similar to polyclonal CD4⁺ T cells in the CNS (Figure 1C,D,E). The MOG_38-49-specific cells had a larger fraction of Treg than polyclonal cells at all the stages of EAE; 7 days after immunization in the LN and spleen, and in the CNS, LN and spleen during peak and resolution of EAE. The ‘enrichment’ of antigen-specific Treg cells was in agreement with a previous study that demonstrated MOG_38-49-specific Treg cell expanded from a Foxp3^hi Treg cell population that exists in C56Bl/6 mice prior to MOG immunization (Korn et al., 2007). The kinetics of the exFoxp3 cells in the MOG_38-49-specific niche mirrored Treg cells at the pre-clinical and peak stages of EAE, whereas in the CNS during EAE resolution the percentage of MOG_38-49-specific exFoxp3 cells was deflated compared with previous EAE stages when the percentage of MOG_38-49-specific Treg cells continued to rise (Figure 1D,E). The data suggests that antigen triggering during inflammation induces antigen specific Treg cells to accumulate and down-regulate Foxp3 to comprise a major population of exFoxp3 cells within the antigen specific niche.

Treg cells down-regulate Foxp3 transcription

To gain insight into the precursor-product relationship of MOG_38-49-specific exFoxp3 cells during EAE, we analyzed Foxp3 expression in Tconv, Treg and exFoxp3 cells. Foxp3 expression is regulated at both the transcriptional and protein levels based on studies using the Foxp3.GFP-cre lineage tracer mice, which allows for three readouts of Foxp3 regulation (Bailey-Bucktrout and Bluestone, 2011). First, GFP is a direct readout of Foxp3 transcription as its expression is controlled by the Foxp3 locus promoter and enhancer activity. Second, the expression of YFP or RFP marks cells that currently or historically expressed Foxp3. Third, Foxp3 protein levels reflect translation and protein turnover at the cellular level. Together these readouts allowed us to assess Foxp3 regulation in vivo. MOG_38-49-induced Treg cells had significantly lower levels of Foxp3-driven GFP when compared to the polyclonal Treg cells based on median (not shown) or mean fluorescence intensity (Figure 2A), suggesting that the entire population of MOG_38-49-specific Treg cells down-regulated Foxp3 transcription relative to polyclonal Treg cells after antigen exposure and inflammation.

Figure 2. Reduced Foxp3 expression in antigen specific Tregs and Foxp3^low cells during EAE.

(A) Ratio of GFP expression (mean fluorescence intensity; MFI) in MOG_38-49-specific Treg versus polyclonal Treg. Mean +/− SDM of 4-6 mice per time-point. p value of unpaired T test. (B) Gating strategy for sort purification of LN/Spleen CD4⁺ T cell subsets 7 days after immunization with MOG_35-55:CFA. Representative of 3 experiments. (C) Post sort analysis of GFP and RFP expression in the sorted populations in A. (D) Post sort analysis of intracellular (i.c.) Foxp3 protein and Foxp3-GFP levels in the sorted populations in A. N.D., not done. (E) Foxp3 expression measured by quantitative PCR and normalized to GAPDH. Each symbol represents one mouse. Mean and SEM error bars, and p values are two-tailed un-paired T test.

We next analyzed the expression of Foxp3 mRNA and protein in CD4⁺ T cell populations in the LN and spleen during preclinical EAE when Foxp3-driven GFP was down-regulated in Treg cells. Foxp3 mRNA and protein was undetectable in sort purified RFP⁻ CD62L^hiCD44⁻ and CD62L^loCD44⁺ ‘naive’ and ‘effector memory’ CD4⁺ Tconv cells, respectively (Figure 2B,C,D,E). The bulk population of RFP⁺ cells was subdivided into three subsets based on GFP expression. GFP^hi Treg cells expressed the highest levels of Foxp3 mRNA and protein whereas RFP⁺ cells with lower levels of GFP expression (GFP^low cells) had significantly decreased amounts of Foxp3 mRNA in comparison to the GFP^hi Treg cells (Figure 2E). Of note, RFP⁺ GFP⁻ exFoxp3 cells contained a subpopulation of 30 – 40% cells that still expressed Foxp3 protein, although at lower levels compared to GFP^hi Treg cells (2661 MFI vs. 4534 MFI), but had significantly down-regulated Foxp3 mRNA compared with GFP^hi and GFP^lo cells (Figure 2E). The data demonstrate that Foxp3-driven GFP faithfully reflects Foxp3 transcription during EAE and further suggest a hypothesis wherein antigen-driven activation of Treg cells in an inflammatory environment results in the down-regulation of Foxp3 transcription in “bona fide” Treg cells prior to loss of longer-lived Foxp3 protein. Once the residual Foxp3 protein is lost, Treg cells become functionally unstable. Finally, the Foxp3⁺ GFP⁻ RFP⁺ cells produce similar amounts of IFNγ as the Foxp3- population (data not shown). Therefore, Foxp3 expression, per se, does not differentiate this phenotype. This result is consistent with a previous study (Ohkura N. et al., 2012).

Treg cells lose Foxp3 protein

Recent studies have suggested that Treg cells are highly stable under homeostatic conditions as Foxp3⁺ cells with high expression of IL-2Rα and demethylated marks on the CpG motifs in the Treg cell–specific demethylated region (TSDR) of the Foxp3 locus (Miyao T. et al., 2012; Ohkura N. et al., 2012). To address the hypothesis that instability is a function of exposure to an inflammatory response, we examined Foxp3 expression in antigen-specific Treg cells during the active autoimmune response in EAE. Highly purified (>96%) Treg cells from naïve mice were transferred into mice during the onset of EAE and followed for changes in Foxp3 protein. The Foxp3⁺ Treg cells expressed high amounts of CD25 and a fully demethylated TSDR (Figure 3A,B) at the time of transfer. At the peak stage of EAE (Figure 3C), we used a congenic mark to identify the transferred Treg cells in the CNS infiltrate, and analyze Foxp3 protein expression in the MOG_38-49-specific and polyclonal Treg cells (Figure 3D). The majority of the transferred polyclonal Treg cells in the CNS were stable for Foxp3 expression, as 93.5% retained Foxp3 protein (Figure 3E). In contrast, Foxp3 protein expression was significantly lower in the adoptively transferred MOG-specific Treg cells. In fact, Foxp3 protein was undetectable in 39% MOG_38-49-specific Treg cells at the peak of CNS inflammation versus <10% of the polyclonal Tregs (Figure 3D). It is important to note that even the small loss of Foxp3 expression within the polyclonal population may reflect antigen-driven effects as the I-A^b:MOG_38-49 tetramer staining does not detect all the MOG- or other neural antigen-specific Treg cells. These data formally demonstrate that a significant percentage of Treg cells, defined by high levels of Foxp3 and CD25 expression and a fully demethylated TSDR, lost Foxp3 protein following antigen exposure during an inflammatory response in vivo.

Figure 3. MOG-specific exFoxp3 cells are generated from Foxp3 high Tregs.

(A) Foxp3 and CD25 staining of sort purified CD45.1⁺ CD4⁺ CD25^hi GITR^hi Tregs. (B) CpG demethylation status of the Treg specific demethylated region in the CD45.1⁺ Tregs depicted in A. The location of each CpG in the Foxp3 locus is depicted. (C) EAE scores of mice that received 2 million Tregs depicted in A and B (circle) and un-transferred mice (square). Mean +/− SEM of 4 mice per group. The EAE score between the groups at each time-point was not significantly different in a paired T test. Difference between the groups was not significant. (D) At day 16 of EAE as depicted in C, CNS cells were isolated and Foxp3 intracellular (i.c.) protein staining shown in CD45.1⁺ CD4⁺ gated cells. Representative of 4 mice. (E) Percentage of Foxp3 loss in polyclonal (unfilled bar) and MOG_38-49-specific (filled bar) in the CD45.1⁺ transferred Tregs in the CNS at peak EAE. Individual mice and pooled data (mean +/− SEM) is shown. p value of unpaired T test. See also Figure S2.

We demonstrated that Treg cells enriched for self-reactivity are biased for loss of Foxp3 expression during an autoimmune response, but the exFoxp3 cells that accumulate during EAE (Figure 1) could also have developed in a subset of antigen-specific cells with de novo Foxp3 expression due to antigen priming. It’s been suggested in epigenetic tracing studies that exFoxp3 cells arise from a Foxp3⁻ T cell that transiently expresses Foxp3 (Miyao T. et al., 2012). In vitro-induced ‘adaptive’ CD4⁺ Treg (aTreg) cells that express Foxp3 during antigen recognition in extrathymic compartments have been described as being unstable for Foxp3 expression and suppressor activity (Chen et al., 2011; Josefowicz et al., 2012; Yadav, M., et al. 2012). Therefore, we addressed the possibility that exFoxp3 cells generated during EAE arose from a T conventional population expressing Foxp3 for a transient but sufficient period of time to express cre-recombinase thus RFP. Congenically-marked CD4⁺ GFP⁻RFP⁻ cells were transferred into mice with active EAE and followed over time to determine whether CD4⁺ GFP⁻RFP⁺ could develop within the antigen-specific Tconv population. In four experiments, none of the MOG_38-49⁺ tetramer cells expressed RFP (Supplemental Figure 2). Thus, only the transfer of GFP⁺RFP⁺ Tregs into the inflamed EAE setting lead to loss of GFP (i.e. GFP⁻ RFP⁺) while similar transfer of GFP⁻RFP⁻ Tconv cells does not result in a ‘transient’ GFP⁻RFP⁺ Tconv population among the MOG antigen-specific T cells. In contrast, MOG_38-49⁺ exFoxp3 cells in the inflamed CNS at the peak of EAE arise from Treg cells. Thus, although the exFoxp3 cells can be comprised of both a population of uncommitted and previously committed cells, in the context of an autoimmune inflammatory response, such as EAE, the antigen-specific Foxp3^lo/− population is derived overwhelmingly from previously committed, bona fide, Treg cells. These results are consistent with MOG_38-49-specific Foxp3⁺ cells expanding from an antigen-specific Treg population that existed prior to autoimmunity, and with the MOG_38-49-specific exFoxp3 cells being derived from an established Treg population and not a transient Foxp3-expressing cell.

MOG_38-49-specific and polyclonal exFoxp3 cells in the CNS are differentially demethylated at the Treg specific demethylated region

Next, we examined the methylation status of CpG motifs in the TSDR of the Foxp3 locus (Huehn et al., 2009). Previous studies have shown that Treg cells exhibit demethylated CpG sites in the TSDR, whereas the CpG sites in this intron are fully methylated in the overwhelming majority of Tconv cells and aTreg cells with unstable or transient Foxp3 expression (Floess et al., 2007; Haribhai et al., 2011; Ohkura et al., 2012).

Using an assay developed for low cell numbers (Supplemental Figure 3), we determined the TSDR methylation status in polyclonal and MOG_38-49 specific Tconv, Treg and exFoxp3 cells in the CNS at peak EAE. As expected both polyclonal and MOG_38-49-specific RFP⁻ Foxp3.GFP⁻ Tconv cells were fully methylated at the TSDR, and both polyclonal and MOG_38-49-specific RFP⁺ Foxp3.GFP^high Treg cells were fully demethylated at the TSDR (Figure 4). The MOG_38-49-specific RFP⁺ Foxp3.GFP⁻ exFoxp3 cells in the CNS at peak EAE were more demethylated at the TSDR than the polyclonal exFoxp3 cells at the same location (Figure 4). In addition, the TSDR was predominantly demethylated in MOG_38-49-specific exFoxp3 cells from the CNS in two out of three mice. Thus, these data support the hypothesis that Treg cells with a fully demethylated TSDR can lose Foxp3 expression and become exFoxp3 cells.

Figure 4. Distinct T regulatory cell-specific demethylated region (TSDR) CpG demethylation marks in auto-reactive and polyclonal exFoxp3 cells in the CNS during EAE.

Percent demethylated TSDR CpG motifs in sort purified MOG_38-49-specific (filled bar) and polyclonal (open bar) CD4⁺ RFP⁻ Tconv, CD4⁺ RFP⁺ Foxp3.GFP⁺ Treg and CD4⁺ RFP⁺ Foxp3.GFP⁻ exFoxp3 cells infiltrating the CNS at the peak stage EAE. The results of 3 mice analyzed are shown. See also Figure S3.

Autoreactive exFoxp3 cells produce potentially pathogenic cytokines

T effector cells in the CNS of mice with EAE produce cytokines including IFNγ and IL-17A (Ivanov et al., 2006) that induce a cascade of inflammatory cell recruitment, glial cell activation and death of myelin-producing oligodendrocytes and their precursors. To address whether at least some of the cytokine producing, and thus potentially pathogenic T cells might be derived from the MOG-specific exFoxp3 population, we examined the production of IFN-γ by Tconv and lineage tracer⁺ cells isolated from the CNS following MOG_35-55 antigen recall in vitro. During peak EAE, 3.4% of the exFoxp3 cells produced IFN-γ without antigen restimulation in vitro. This percentage increased to 9.6% of the exFoxp3 cells in response to MOG_35-55 stimulation, a percentage only slightly lower than the 11.6% of Tconv cells that produced IFNγ in the same conditions (Figure 5A,B). IFNγ production was never detected in antigen-stimulated Foxp3⁺ Treg cells during EAE. Interestingly, IL-17A was undetectable in exFoxp3 T cells in this assay, and was detected only after PMA + ionomycin stimulation (not shown). Therefore, Foxp3 loss bestows MOG-specific Treg cells with IFNγ production comparable to pathogenic T effector cells similar to that observed in the diabetes setting (Zhou et al., 2009). These results further argue against the hypothesis that the exFoxp3 cells are derived from a transient Foxp3-expressing population as others have shown that these cells are marked by co-expression of Foxp3 and RORγt, the lineage-specific transcription factor for the IL-17-producing Th17 subset (Zhou et al., 2008).

Figure 5. CNS exFoxp3 cells produce IFNγ similar to Tconv cells.

(A) CNS cells were isolated at the peak of EAE and incubated for 16 hours with MOG_35-55, and Tconv, Treg and exFoxp3 cell subsets were identified based on staining for CD4, YFP/GFP and Foxp3. IFNγ (tinted histogram) and isotype control (dashed line) staining in the live CD4⁺ population is shown. Representative of 3 experiments. (B) Percentage of CNS CD4⁺ T cells expressing IFNγ as in (A) with (filled bar) or without (open bar) MOG_35-55 stimulation in vitro. Mean +/− SDM of 3 experiments with 2-3 mice pooled in each.

exFoxp3 cells are pathogenic causing EAE

In EAE, pathogenic T cells induce an inflammatory cell infiltrate in the spinal cord and cerebellum resulting in tail and hind limb dysfunction (Bailey et al., 2006). MOG_35-55-reactive CD4⁺ Tconv and exFoxp3 cells produced equivalent levels of IFNγ in the CNS at peak (Figure 5), suggesting they may be pathogenic. The implication of Treg instability leading to pathogenicity in autoimmunity is of major significance, because Treg cells are enriched for self-reactivity (Hsieh et al., 2006; Wong et al., 2007) and can express homing receptors allowing preferential migration to inflamed tissues Dominguez-Villar, M., 2011; Josefowicz, S.Z., 2012). To test the pathogenicity of exFoxp3 cells in EAE, isolated exFoxp3 cells were compared with Tconv and Treg cells to determine their ability to induce EAE following adoptive cell transfer. Congenically marked CD4⁺ RFP⁻ Foxp3.GFP⁻ Tconv, CD4⁺ RFP⁺ Foxp3.GFP⁺ Treg, and CD4⁺ RFP⁺ Foxp3.GFP⁻ exFoxp3 cells were purified from the LN and spleen of MOG_35-55:CFA immunized mice by fluorescent-activated cell sorting and expanded in vitro using MOG_35-55 to enrich antigen-reactive cells, then with anti-CD3 plus anti-CD28 polyclonal stimuli to generate enough cells for the experiment. Expansion was performed using IL-2 for Tconv and Treg cells, and IL-2 and IL-7 for exFoxp3 cells. IL-7 was needed to expand exFoxp3 cells ex vivo, likely because they express low levels of CD25 (IL-2Rα) and express CD127 (IL-7Rα) at higher levels (Zhou et. al. 2009). No Th1 or Th17 cell-driving cytokines were added, that might influence the effector function of the T cells. The individual cell populations remained stable for Foxp3.GFP expression during the expansion, and were resorted to purify from feeder cells during the antigen-driven expansion (Supplemental Figure 4). Individual cell populations were transferred into lymphodeficient recipients which were immunized with MOG_35-55:CFA to induce EAE. Immunization was necessary to induce EAE because the cells were expanded in ‘neutral’ conditions, and no population induced EAE without immunization. As expected, Treg cells did not induce EAE, whereas MOG-specific exFoxp3 and Tconv cells induced EAE with similar incidence and severity (Figure 6A,B). The onset of disease was slightly delayed in recipients of exFoxp3 cells compared to Tconv in the experiment depicted but the difference was not significant when the experiments were pooled (not shown). Infiltrates were isolated from the spinal cord and cerebellum (CNS) of recipients at the peak of EAE (Figure 6A), and the abundance and composition of the inflammatory cells compared. The transferred exFoxp3 cells had a stable RFP⁺ GFP⁻ phenotype after the induction of EAE and the majority of Tconv cells remained RFP⁻, although 4.7% transiently expressed Foxp3 (Supplemental Figure 4). This transient Foxp3 expression by a minor population of Tconv cells in a lymphodeficient environment is in agreement with previous studies (Miyao et al., 2012). The number of macrophages (CD45^hi, CD11b⁺ CD11c⁻), myeloid DC (mDC; CD45^hi, CD11b⁺ CD11c⁺), plasmacytoid DC (pDC; CD45^hi, CD11b⁻, CD11c⁺, CD45R⁺), microglia (CD45^med, CD11b⁺) and T cells in the CNS after EAE induction with Tconv or exFoxp3 was similar (Figure 6C), demonstrating that Tconv and exFoxp3 cells orchestrate similar cellular immune responses during autoimmunity. The majority of Tconv and exFoxp3 cells in the inflamed CNS produced IFNγ (Figure 6D), which could have initiated the inflammatory cascade that caused EAE in the recipients. These results demonstrate that exFoxp3 cells can function as pathogenic effector cells producing IFNγ in the CNS, and MOG-specific exFoxp3 cells induce EAE similar to Tconv cells.

Figure 6. ExFoxp3 cells cause EAE.

(A) The proportion of mice that developed EAE after transfer of CD4⁺ RFP⁻ GFP⁻ Tconv, RFP⁺ GFP⁺ Treg and RFP⁺ GFP⁻ exFoxp3 cells. 3 mice per group with 2 experiments pooled. Mean +/− SEM. (B) EAE clinical score of mice that developed EAE as described in (A) in 1 of 2 experiments. (C) The number of the indicated inflammatory cell populations in the spinal cords of mice shown in (B), representative of 2 experiments. (D) Staining and percentage of IFNγ in CNS T cells from (B). Also see Figure S4.

IL-2R signaling over-rides Foxp3 instability

IL-2 is critical for expression of Foxp3 in Tregs and the homeostasis of Tregs in the periphery. It has been previously shown that targeting IL-2 to Treg cells using low concentrations of IL-2:anti-IL-2 complexes prevents or reverses autoimmune diabetes, and delays onset and severity of EAE by causing Treg cell expansion (Tang et al., 2008; Webster et al., 2009). We hypothesized that supplementing antigen-experienced Treg cells with exogenous IL-2 would stabilize Foxp3 expression in MOG-specific Treg cells, preventing the de novo generation of exFoxp3 cells. Treatment with IL-2:anti-IL-2 complexes after immunization induced a unimodal increase in phosphorylated STAT5 levels in Treg cells, but did not affect phosphorylated STAT5 levels in Tconv or the majority of exFoxp3 cells (Supplemental Figure 5). IL-2:anti-IL-2 complex treatment resulted in a significant reduction in the proportion and number of MOG_38-49-specific Treg cells that lost Foxp3 expression and became exFoxp3 cells, but it did not affect the abundance of polyclonal exFoxp3 cells (Figure 7 B,C,D) which is consistent with the absence of phosphorylated STAT5 in these cells in response to treatment (Supplemental Figure 5). There was a trend towards increased numbers of MOG_38-49-specific and polyclonal Treg cells, as expected (Tang et al., 2008; Webster et al., 2009), however, it was striking that the only significant difference was stabilized Foxp3 expression in MOG_38-49-specific Treg cells after treatment with IL-2:anti-IL-2 complexes (Figure 7). We conclude that Treg cell expansion per-se does not result in Foxp3 instability, because the increase in number of MOG_38-49-specific Treg cells did not result in increased numbers of exFoxp3 cells but rather a sharp reduction in the frequency of these cells. We also conclude that Foxp3 instability in Treg cells responding to in vivo antigen is IL-2 sensitive. In a previous study, Webster et al. (2009) treated EAE prophylactically with IL-2 complexes to protect from disease, the conclusion was the IL-2-expanaded polyclonal Tregs suppressed the development of the auto-reactive effector T cell response during priming for EAE. The goal of our experiments was to target Treg instability, which, based on the enrichment of MOG_38-49-reactive exFoxp3 cells compared with polyclonal cells, occurred within the initial 7 days following EAE induction (Figure 1C,D). Tregs were targeted with IL-2:anti-IL-2 complexes in the 3 days following EAE induction, the MOG_38-49-reactive Treg cells retained high level Foxp3 expression and the mice were protected from disease (Figure 7G), suggesting that the stability of antigen-specific Tregs may be important for the course and control of autoimmunity.

Figure 7. IL-2 stabilizes Foxp3 during EAE.

(A) Foxp3.GFP levels in polyclonal and MOG_38-49-specific CD4⁺ RFP⁺ T cells from LN and spleen 6 days after EAE induction and treatment with IL-2:anti-IL-2 complex or saline control on day 1, 3 and 5. (B-F) CD4⁺ MOG_38-49–reactive and polyclonal RFP⁺ GFP⁻ exFoxp3 and RFP⁺ GFP⁺ Treg cells in LN and spleen from mice receiving saline or IL-2: αIL-2 complex treatment as in (A). Each point represents one sample, lines represent mean +/− SEM of the pooled data. P values are of two-tailed T tests. (G) EAE course with IL-2:anti-IL-2 complex treatment day 1, 3 and 5. Triangle’s were control (saline) treated animals, and squares IL-2 treated animals. Mean +/− SEM of 5 animals. Representative of two experiments. Also see Figure S5.

Discussion

Using lineage reporter and “fate-mapping” mice to characterize Foxp3 expression and down-regulation within the same cell overtime, we previously showed that a subset of cells that expressed Foxp3 could become Foxp3^lo (exFoxp3 cells) and Previous studies by our group (Zhou et al., 2009) have led to the hypothesis that a subset of Tregs could demonstrate highly reduced levels of Foxp3 and become functionally unstable in the context of autoimmunity and inflammation.. However, this interpretation has been challenged by other studies, which concluded that “terminally differentiated” Treg cells are stable and exFoxp3 cells would instead derive from loosely committed Tregs that have not fully acquired many features of bona fide Treg cells such as high amounts of Foxp3 and CD25 and demethylation at the TSDR (Rubtsov et al., 2010; Miyao et al., 2012). Here, we address these apparently contradictory results in a model of autoimmunity by demonstrating that loss of Foxp3 and the resulting instability do occur in bona fide Treg cells but predominantly in a subset of self-antigen-specific Treg cells during responses initiated by immunization with this self-antigen. There is indeed evidence for a percentage of exFoxp3⁺ cells derived from cells transiently expressing Foxp3 but these cells are not derived from the autoantigen-specific T cell population induced during EAE.

Rather, the Treg cells followed in this autoimmune setting expressed high amounts of CD25 and had a demethylated TSDR, markers that identify Treg cells that in the steady state maintained high levels of Foxp3 expression under lymphoreplete conditions in other studies (Miyao et al., 2012; Rubtsov et al., 2010). However, these studies were largely performed in steady-state and a homeostatic setting where the immune system had not been perturbed by autoimmunity. Importantly, these studies did not focus on selective, antigen-specific Treg cells expanded in response to an inflammatory autoimmune reaction. We addressed whether committed Treg cells lose Foxp3 during a mouse model of autoimmunity, EAE, by interrogating antigen-specific Treg cells within the polyclonal repertoire. Using lineage tracing and cell transfers, we found that although the majority of polyclonal and antigen-specific Treg cells stably expressed Foxp3 during the autoimmune response in this mouse model of autoimmunity, a substantial fraction of antigen-specific Treg cells with “signature” features (Foxp3^hi, CD25^hi, demethylated TSDR) down-regulated Foxp3 transcription, lost Foxp3 protein and acquired characteristics of effector T cells such as IFNγ production and pathogenic potential in vivo. It should be highlighted that the exFoxp3 cells were only apparent in adoptive transfer experiments where bona fide CD4⁺GFP⁺RFP⁺ Treg cells were transferred into mice with ongoing EAE. In some animals up to a third of the MOG tet⁺ antigen-specific Tregs lost Foxp3 expression. In contrast, transfer of highly purified CD4⁺ GFP⁻RFP⁻ Tconv cells did not result in any antigen-specific exTregs suggesting that this population did not derive from a differentiating transient Foxp3⁺ subset. These findings link Foxp3 instability in an antigen-specific subset of Treg cells to pathogenic responses during autoimmunity. In this regard, circulating Tregs that have altered characteristics including reduced suppressive ability, production of IFNγ, and other features of Th1 cells are enriched in type 1 diabetes and multiple sclerosis in humans (Dominguez-Villar et al., 2011; McClymont et al., 2011). Interestingly, stable Foxp3 expression resumed during the remission phase of disease or after treatment with IL-2:anti IL-2 complexes that also ameliorate disease suggesting potential therapeutic opportunities in these areas.

Even in the inflamed setting, a large number of MOG_38-49-specific Treg cells retained high levels of Foxp3 protein. This likely reflects positive feedback of Foxp3 on its own expression (Zheng et al., 2010), and we hypothesize that once a low threshold of Foxp3 transcription is reached, the Treg transcriptional landscape can be lost and characteristics of exFoxp3 cells become apparent, for example loss of Foxp3 and acquisition of IFNγ expression (Zheng et al., 2007). Induced Tregs or loosely committed Foxp3⁺ cells are prone to Foxp3 instability and demonstrate a failure to imprint stable epigenetic marks at the Foxp3 locus (Haribhai et al., 2011; Josefowicz et al., 2012, Ohkura, N., et al., 2012). We thus investigated whether the Treg cells that down-regulate Foxp3 transcription during EAE may be pre-disposed because of a methylated TSDR. In two out of three mice, Foxp3 down-regulation occurred predominantly in antigen-specific Treg cells with a demethylated TSDR suggesting that lack of epigenetic imprinting in this region of the Foxp3 locus is not a leading cause of loss of Foxp3 expression in Treg cells during EAE. In the third mouse, the antigen-specific exFoxp3 cells had a methylated TSDR suggesting that the exFoxp3 cells derived from loosely committed Treg perhaps as a consequence of the local inflammatory environment, as previously reported (Miyao et al., 2012). Thus, exFoxp3 cells derive from bona fide Tregs as well as loosely committed cells in individual mice, which could reflect differences in the local milieu that influence lineage differentiation and stability at the population level. The TSDR data further suggest that the majority of polyclonal exFoxp3 cells arise from transient or unstable Foxp3 expression (Miyao et al., 2010). The factors inducing Foxp3 instability in a subset of antigen-specific Treg cells remain unclear. Low CD25 expression and a loss of STAT5-driven Foxp3 expression have been implicated in Foxp3 instability during homeostasis (Miyao et al., 2012), but this does not appear to be the case in the generation of MOG-specific exFoxp3 cells during EAE. Treg cells responded to the IL-2 administered with unimodal phosphorylation of STAT5, suggesting that Treg cells are able to respond to IL-2 during EAE. Rescue of Foxp3 expression with IL-2/anti-IL-2 complexes suggests a deficiency in IL-2 available to Tregs in vivo. Treg cells compete with T effector cells for local IL-2 during an immune response (O'Gorman et al., 2009), which is required for efficient Treg cell homeostasis (Barron et al., 2010; Zorn et al., 2006). The development of a strong effector cell response that consumes IL-2 may tip the balance of stability in Treg cells responding to antigen in a microenvironment with T effector cells. Consistent with this hypothesis, other studies have shown that Treg instability occurs during strong Th1-polarized responses (Oldenhove et al., 2009; Takahashi et al., 2011), with IL-2 deficiency being implicated in one study (Oldenhove et al., 2009). IL-2:anti-IL-2 complex stabilization of Foxp3 in Treg cells correlated with reduced CNS inflammatory disease, further supporting that Treg cell instability is detrimental for tissue specific tolerance. In this regard, autoimmune diabetes has also been associated with a local deficiency in Treg-mediated immunoregulation and low Foxp3 expression in islet-infiltrating Tregs (Tang et al., 2008; Zhou et al., 2009). Moreover, IL-2 therapy restored Foxp3 expression in pancreatic Tregs and could prevent or cure new-onset diabetes in NOD mice (Grinberg-Bleyer et al., 2010; Tang et al., 2008), suggesting that local deficits in IL-2 signaling may be a general mechanism leading to down-regulation of Foxp3 and Treg instability in autoimmune diseases.

Our data suggest that Foxp3 instability in Treg cells is induced and pronounced following antigen triggering and during inflammatory responses. Treg cell stability mirrors the resolution of autoimmune inflammatory disease. It is tempting to speculate that Treg cell stability drives the regulation of autoimmune responses. Recent studies of antigen-specific Treg cells has demonstrated a rapid expansion of Treg cells with a second wave of antigen availability (Rosenblum et al., 2011, Rowe, J. H., et al., 2012). In line with a ‘priming and memory’ model for Treg cells, we have shown Treg cell instability occurs during the acute phases of the inflammation, which may remove inefficient, unstable Treg cells from the pool. The stable Treg cells that remain may be better equipped to control further flares of inflammation, indeed Treg stability correlated with EAE resolution. We determined that the frequency of MOG-specific exFoxp3 cells declined during peak and the resolution phases, and increased in the peripheral lymph nodes during the same time-frame. It is possible that the unstable Treg cells emigrate from the CNS during the decline of inflammation and take up residence in the secondary lymph node organs, as do effector cells during relapsing remitting EAE, although they have no further pathogenic consequence to the progression of the disease (Vanderlugt et al., 2000). Thus, instability at the individual cell level in a subset of Tregs could be followed by improved stability and regulation at the population level in certain circumstances.

Treg-derived MOG-specific exFoxp3 cells presented an effector cell phenotype with IFNγ production and pathogenic potential after re-stimulation with their cognate antigen in vitro or in vivo. Although exFoxp3 cells are 10% of the frequency of Tconv cells that produce IFNγ in the CNS during EAE, the local production of IFNγ can provide amplifying feedback on local conventional Th1 cells (Takahashi et al., 2011). Indeed, exFoxp3 cells producing IFNγ cause EAE with similar severity of Tconv cells. Teleologically, it is possible that Foxp3 instability during normal immune responses may aid the generation of an anti-pathogen response, and an increase in protective cytokine production early in the response could provide a boost to protective immunity. However, as highlighted in this study, loss of Foxp3 correlates with, and may contribute to the development of autoimmunity. Treg cell therapies are being tested in clinical trials in type 1 diabetes and GVHD, raising concerns that Treg instability may lead to unwarranted effects in patients and indicating that additional studies will be needed to determine the factors leading to instability. Moreover, Treg cell therapy may require the cover of a stabilizing factor such as IL-2 that has been used to directly modulate the frequency and proliferation of Treg cells in patients (Long et al., 2012; Koreth et al., 2011; Zorn et al., 2006).

Experimental Procedures

Mice

Foxp3.GFP-Cre.ROSA26-YFP reporter mice have been described (Zhou et al., 2009), and were backcrossed greater than 8 generations onto the B6 background. To generate new lineage reporter mice that had less spectral overlap than GFP vs YFP, allowing improved purification by flow cytometric sorting, B6 Foxp3.GFP-Cre mice (Zhou et al., 2009) and Foxp3.YFP.Cre mice (Rubtsov et al., 2008) were crossed with B6 ROSA26-RFP reporter mice (Luche et al., 2007). B6 TCRα were obtained from Jackson Laboratories (Bar Harbor, ME). All mice were housed and bred under specific pathogen-free conditions at the University of California, San Francisco Animal Barrier Facility. The Institutional Animal Care and Use Committee of the University of California, San Francisco, approved all animal experiments.

Antibodies

Labeled antibodies specific for CD4 (RM4-5), CD8 (Ly-2), CD25 (PC61), CD44 (pgp-1), FoxP3 (FJK-16s), IL-17A (eBio17B7), IFNγ (XMG1.2), phosphorylated STAT5 (pY694), and specific isotype controls were purchased from BD PharMingen (San Jose, CA) or eBioscience (San Diego, CA). Intracellular GFP/YFP was stained with anti-GFP (rabbit polyclonal eBioscience catalogue 14-6774-81) and Fab’2 anti-rabbit (goat polyclonal eBioscience catalogue 11-4839-81).

Flow cytometry

Stained single cell suspensions were analyzed with a LSRII flow cytometer running FACSDiva (BD Biosciences, San Jose, CA), and FSC 2.0 files analyzed and presented with FLOWJO Software (Treestar Inc. Ashland, OR). T cells were sorted using a MoFlo cytometer high speed cell sorter (DakoCytomation, Glostrup, Denmark) or FACSAria (BD Biosciences, San Jose, CA).

Phosphorylated STAT5 Flow

Spleen cells were harvested directly into fixative for ex vivo analysis of STAT5 phosphorylation and staining was performed as described in O'Gorman et al. (2009).

Induction of EAE

B6 mice were immunized subcutaneously with 100 μl of emulsified complete Freund’s adjuvant (CFA) (BD Difco^TM, San Jose, CA) supplemented with 4 mg/ml Mycobaterium tuberculosis H37Ra (BD Difco^TM, San Jose, CA) and 200 μg of MOG_35-55 peptide (MEVGWYRSPFSRVVHLYRNGK. Genemed Synthesis, San Antonio, TX), and received intraperitoneal injections of 200 ng pertussis toxin from Bordetella pertussis (Sigma Aldrich, St. Louis, MO) at the time of immunization and 48hr later. Clinical disease was assessed by scoring ascending hind limb paralysis as follows: no signs - score 0, flaccid tail - score 1, hind limb weakness - score 2, partial hind limb paralysis - score 3, complete hind limb paralysis - score 4, moribund mouse - score 5.

Transfer of CD4⁺ T regulatory and T conventional cells into EAE to track Foxp3 stability

Donor mice congenic for CD45.1 were immunized with MOG_35-55 peptide emulsified in CFA similar to EAE induction, no pertussis toxin was administered. 5 days after immunization, regulatory T cells were sort purified from axillary, inguinal and brachial LNs and spleen, and were gated as CD45.1⁺, CD4⁺, CD25^high, GITR^high. For Foxp3 lineage tracing T conventional cells, 5 days after immunization of Foxp3.GFP.RFP reporter mice CD45.1⁺ CD4⁺ GFP⁻ RFP⁻ T cells were sort purified from axillary, inguinal and brachial LNs and spleen. 2×10⁶ CD45.1⁺ Tregs or CD4⁺ T conventional cells were transferred intravenously into CD45.2 mice at the onset of EAE. At the peak of EAE, the CD45.1⁺ transferred cells were analyzed for MOG-tetramer staining in the CNS as described below.

CD4⁺ T conventional, T regulatory and exFoxp3 cell expansion and adoptive transfer for EAE

8 days after EAE induction in Foxp3-GFP.RFP reporter mice, CD4⁺ T cell populations were sort purified from axillary, inguinal and brachial LNs and spleen. Viable T cells were gated as CD4⁺, Tregs were gated as GFP⁺ RFP⁺, exFoxp3 cells gated as GFP⁻ RFP⁺ and Tconv gated as GFP⁻ RFP⁻. T cells were cultured @ 2×10⁵/ml + 5X the number of lethally irradiated congenically marked splenocytes as feeder cells, 20μg/ml MOG_35-55 and growth factor cytokines: Tconv, 100U/ml IL-2; Treg, 2,000U/ml IL-2; exFoxp3, 2,000U/ml IL-2 and 10ng/ml IL-7. Growth factor cytokines were refreshed every 2 days for 11 days. Cells were re-plated at 2×10⁵/ml and polyclonally expanded for 7 days with 1:1 bead:cell T-activator CD3/CD28 beads (Dynabeads^®, Life Technologies, Invitrogen^TM, Carlsbad, CA) and growth factor cytokines as above. After 7 day, viable CD4⁺ T cells were re-sorted for transfer as described above and shown in Supplemental Figure 7. Sub-lethally irradiated (350 rads) B6 TCRα^−/− mice received 350,000 – 500, 000 T cells intravenously. Recipients were immunized with MOG_35-55/CFA on the day of cell transfer and 7 days later. Mice were scored for EAE in a blinded fashion every 2 days.

IL-2-anti-IL-2 complex treatment

Mice immunized for EAE received an intraperitoneal injection of 1.5 μg recombinant mouse IL-2 (eBioscience, San Diego, CA) complexed with 5 μg anti-IL-2 (clone JES6-1A12, eBioscience, San Diego, CA) for 15mins at 37°C, on days 1, 3 and 5.

MHC II Tetramer staining

I-A^b:MOG_38-49 and I-A^b:HClip tetramers were provided by the NIH tetramer core facility. Central nervous system samples were stained with tetramer and analyzed directly, whereas tetramer labeled lymph node and spleen samples were enriched for analysis as published (Moon et al., 2007).

TSDR methylation analysis

Methylation at the TSDR was evaluated by sequencing as described in Zhou et al., (2009) and by qPCR as described in Yadav et al. (2012).

Foxp3 quantitative PCR (qPCR)

To assay endogenous Foxp3 transcription, we used a qPCR probe that recognizes a sequence in the 10^th exon of the Foxp3 gene, which is not expressed by the BAC-Foxp3.GFP-Cre construct as it expresses a stop-codon in exon 1 (Zhou et al., 2008). RNA was isolated by trizol (Life Technologies, Invitrogen^TM, Carlsbad, CA) and cDNA obtained with first-strand cDNA synthesis kit (GE Healthcare Biosciences, Pittsburg, PA). qPCR was performed with Taqman^® gene expression assays (Mm 00475165-m1, Life Technologies, Applied Biosystems, Grand Island, NY) and analyzed with a 7500 Fast real-time PCR system (Foster City, CA).

Isolation of Central Nervous System Leukocytes

Mice were sacrificed with CO₂ and immediately perfused through the left ventricle with PBS until the effluent ran clear. Spinal cords were extruded by flushing the vertebral canal with PBS and cerebellum removed. Spinal cord and cerebellum were diced and placed in HBSS containing 25 mM Hepes, 300 Wunsch Units/ml type-D clostridial collagenase (Life Technologies, Invitrogen^TM, Carlsbad, CA) and 50 μg/ml DNase I (Roche, San Francisco, CA) and incubated for 30 min (37°C). The homogenate was re-suspended in 30% isotonic Percoll (Pharmacia, Piscataway, NJ), under-laid with 70% Percoll and centrifuged at 1,160 g at room temperature for 30 min. Mononuclear cells were collected from the Percoll interphase and washed 2X in HBSS 2% FCS.

Cytokine production

CNS mononuclear cells were incubated at 5×10⁶/ml with 1 μg/ml anti-CD28 (PV-1) with or without 10 μg/ml MOG_35-55 for 16 hours at 37°C. Cells were activated for 2.5 hours with 50 ng/ml phorbol 12-myristate 13-acetate and 2 μM monensin before labeling with LIVE/DEAD fixable dead stain (Life Technologies, Invitrogen^TM, Carlsbad, CA), and staining for CD11b (dump), CD4, , Foxp3, IFNγ and GFP/YFP and performing flow cytometric analysis.

Highlights.

• In EAE, antigen-specific Tregs down-regulate Foxp3 more than polyclonal Treg cells.

• Foxp3 loss occurs in ‘bona fide’ Treg cells during inflammatory EAE response.

• A significant fraction of MOG-specific exFoxp3 cells produce IFNγ and transfer EAE.

• MOG-specific exFoxp3 cells decrease with EAE resolution and IL-2 treatment.