Increased number and function of FoxP3 regulatory T cells during experimental arthritis.код для вставкиСкачать
ARTHRITIS & RHEUMATISM Vol. 58, No. 12, December 2008, pp 3730–3741 DOI 10.1002/art.24048 © 2008, American College of Rheumatology Increased Number and Function of FoxP3 Regulatory T Cells During Experimental Arthritis Kristen Monte, Christina Wilson, and Fei F. Shih Objective. CD4ⴙCD25ⴙFoxP3ⴙ regulatory T (Treg) cells are critical regulators of autoimmunity. Yet the number of Treg cells is paradoxically increased in rheumatoid arthritis (RA) patients, and Treg cells show variable activity in human studies. The objective of this study was to characterize the expansion and function of Treg cells during the initiation and progression of experimental arthritis. Methods. To unequivocally identify Treg cells, we crossed FoxP3gfp mice with K/BxN mice to generate arthritic mice in which Treg cells express green fluorescence protein. We examined the expansion and function of Treg cells and effector T (Teff) cells during different stages of arthritis, using flow cytometry and cell proliferation analyses. Results. In K/BxN mice, thymic selection of KRN T cells resulted in an enrichment of forkhead box P3 (FoxP3)–positive Treg cells. Treg cell numbers increased during arthritis, with significant increases in spleens and draining lymph nodes, indicating selective tropism to sites of disease. In contrast to the in vitro unresponsiveness of Treg cells when cultured alone, substantial proportions of Treg cells proliferated in both nonarthritic and arthritic mice. However, they also underwent greater apoptosis, thereby maintaining equilibrium with Teff cells. Similarly, enhanced Treg cell–suppressive activity during arthritis was offset by greater resistance by their Teff cell counterparts and antigen-presenting cells. Conclusion. In this well-established model of RA, the interplay of Teff cells and Treg cells in K/BxN mice recapitulated many features of the human disease. We demonstrated an ordered expansion of Treg cells during arthritis and dynamic changes in Treg cell and Teff cell functions. By elucidating factors that govern Treg cell and Teff cell development in K/BxNgfp mice, we will gain insight into the pathophysiology of and develop novel therapeutics for human RA. Rheumatoid arthritis (RA) is an autoimmune disease that results in joint inflammation and destruction. Autoreactive T cells and B cells are crucial to its pathogenesis. Autoreactive T cells are predominantly deleted in the thymus, but this process is not stringent. Thus, autoreactive T cells can and do escape into the peripheral T cell pool; subsequent activation can result in autoimmune pathologic conditions. CD4⫹CD25⫹ FoxP3⫹ regulatory T (Treg) cells, which comprise 5– 10% of CD4⫹ T cells, are crucial for the maintenance of peripheral tolerance (1,2). In adult RA and juvenile idiopathic arthritis, Treg cells were found to be enriched in the synovial fluid of inflamed joints as compared with their levels in peripheral blood, suggesting active homing or expansion at sites of inflammation (3–8). These Treg cells expressed forkhead box P3 (FoxP3) and suppressed both proliferation and cytokine production by CD4⫹CD25– effector T (Teff) cells. Moreover, increased numbers of Treg cells in the synovial fluid directly correlated with limited disease (3), suggesting that Treg cells aid in disease remission. However, it is not known how Treg cells modulate immunity, because the number of Treg cells is paradoxically increased during disease, and since there is also variability in Treg cell function among different studies. Ehrenstein et al (7), for example, demonstrated that Treg cells from RA patients showed compromised function as compared with those from healthy controls, whereas other investigators (8) showed that Treg cells obtained during active RA were equally suppressive or were more suppressive than those from healthy controls. Supported by grants from the NIH (K12-HD-01487-1 and K08-AR-051980-01) and a pilot and feasibility grant from the American Cancer Society (IRG-58-010-49). Kristen Monte, BS, Christina Wilson, BS, Fei F. Shih, MD, PhD (current address: Pfizer Inc., Chesterfield, Missouri): Washington University School of Medicine, St. Louis, Missouri. Address correspondence and reprint requests to Fei F. Shih, MD, PhD, Associate Research Fellow, Indications Discovery, Pfizer Inc., 700 Chesterfield Parkway West, T1C, Chesterfield, MO 63017. E-mail: Fei.firstname.lastname@example.org. Submitted for publication November 14, 2007; accepted in revised form August 15, 2008. 3730 INCREASED FoxP3⫹ Treg CELLS DURING EXPERIMENTAL ARTHRITIS In this case, the increased suppressive function was offset by the responder Teff cells themselves being more refractory to suppression and by the presence of inflammatory cytokines (9). Because these studies examined heterogeneous patient populations, disease stages, and therapeutic regimens, these variables likely contributed to the differences between the studies. In addition, investigators have differed in the criteria they use to identify Treg cells, with some studies focusing only on CD25bright populations while others used all CD25⫹ T cells (8,10). Because expression of CD25 is also elevated in activated Teff cells, the number of which is increased during disease, varying degrees of Teff cell contamination can render interpretation difficult. It is also unclear if arthritis resulted from a primary Treg cell dysfunction or a secondary defect due to persistent inflammation. To eliminate these confounding processes, we examined Treg cell development and function in a well-characterized murine model of RA. K/BxN mice were generated by crossing KRN T cell receptor (TCR)–transgenic mice with NOD mice (11). The disease is fully penetrant in all progeny and follows a predictable course of progressive symmetric distal polyarthritis that resembles RA in humans. Arthritis results from autoreactive KRN T cells recognizing peptide 281–293 of the glycolytic enzyme glucose-6phosphate isomerase (GPI) that is bound to I-Ag7 (the NOD-specific class II major histocompatibility complex allele) (12,13). Incomplete thymic deletion allows autoreactive KRN CD4⫹ T cells to persist in the periphery and to become activated by endogenously presented GPI. KRN T cells then provide help to GPI-specific B cells, giving rise to arthritogenic antibodies. Treg cells are enriched in arthritic K/BxN mice (14,15), and the loss of Treg cells results in earlier and more extended disease, suggesting that although Treg cells do not prevent arthritis, they may nevertheless mitigate it (15). Similar findings have been reported for collagen-induced arthritis, in which depletion of CD25⫹ T cells exacerbated arthritis and adoptive transfer of CD4⫹CD25⫹ Treg cells or FoxP3transduced T cells ameliorated the disease (16–18). To understand how antigen-specific Treg cells develop during the course of arthritis, we crossed K/BxN mice with FoxP3gfp reporter mice to unequivocally identify Treg cells. We then analyzed Treg cell selection in the thymus and followed their expansion and function during the progression of arthritis. MATERIALS AND METHODS Mice. KRN mice have been described in detail elsewhere (11). FoxP3gfp mice were generously provided by A. 3731 Rudensky and were bred to KRN mice to generate KRNgfp mice in which FoxP3⫹ T cells expressed GFP (19). These were subsequently bred to NOD/LtJ mice (The Jackson Laboratory, Bar Harbor, ME) to generate K/BxNgfp mice. Nonarthritic controls were KRNgfp on the C57BL/6 background. Because FoxP3 is on the X chromosome, only male K/BxNgfp mice were analyzed. The K/BxNgfp mice exhibited arthritis with equivalent severity and time course as the K/BxN mice (data not shown). All mice were bred and housed under specific pathogen–free conditions in the animal facility at the Washington University Medical Center, in accordance with the standards of the Animal Studies Committee. Flow cytometry. Thymi, spleens, draining lymph nodes (LNs; popliteal, axillary, and brachial), nondraining LNs (inguinal and cervical), and mesenteric LNs were harvested from individual mice and analyzed separately. Single-cell suspensions (3 ⫻ 106) were treated with anti-CD16 (clone 2.4G2) prior to surface staining with specific antibodies according to standard protocols. The following antibodies were used: allophycocyanin (APC)–labeled anti-CD4, phycoerythrin (PE)–labeled anti-CD4 (clone GK1.5), PE-Cy7–labeled antiCD8 (clone 53-6.7), PE-Cy7–labeled anti-CD25 (clone PC61), APC-labeled antibromodeoxyuridine (anti-BrdU), APC-labeled anti-CD45RB (clone C363-16A), PE-labeled annexin V, and PE-labeled anti–Bcl-2. Antibody kits for anti-CD4, anti-CD8, anti-CD25, and anti-CD45RB were obtained from eBiosciences (San Diego, CA); all other antibodies were obtained from BD Biosciences (San Diego, CA). Streptavidin–peridinin chlorophyll A protein was obtained from BD Biosciences. TCR V␤6 was used to track the KRN-transgenic TCR. All samples were analyzed on either a FACSAria or a FACSCalibur flow cytometer (both from BD Instruments) using FlowJo software (Tree Star, Ashland, OR). Lymphocytes were gated based on forwardscatter and side-scatter characteristics, and 1–2.5 ⫻ 105 gated events were collected per sample. Lymphocyte isolation from paws. Front and rear paws were harvested from each mouse, and the skin was removed. Paws from each mouse were minced in 40 ml of RPMI 1640 with 5% calf serum, 125 l of type VIII collagenase (10,000 units/ml), and 200 l of Dispase (50 units/ml). Minced tissues were agitated for 2 hours at 37°C at a speed of 150 revolutions per minute, with vortexing every 30 minutes. Digested tissues were strained through a 70-m Nitex membrane. Lymphocytes were harvested over a Ficoll gradient and analyzed by flow cytometry. In vitro T cell cultures. CD4⫹ T cells from pooled spleens and draining LNs were enriched by positive selection using anti-CD4 microbeads (Miltenyi Biotec, Sunnyvale, CA). Purity was typically ⬎90%. Dendritic cells (DCs) and B cells were prepared from spleens as described elsewhere (20). T cells (5 ⫻ 105/well) were stimulated with 5 ⫻ 104 DCs in 1 ml of Iscove’s medium containing 10% heat-inactivated bovine growth serum (Hyclone, Logan, UT), 2 M Glutamax (GibcoBRL, Gaithersburg, MD), 2 ⫻ 10–5M ␤-mercaptoethanol, and 50 g/ml of gentamicin (referred to hereinafter as ISC-10). At various times, 1 mM BrdU was added (final concentration 10 M), and cultures were analyzed 12 hours later for BrdU incorporation and apoptosis. T cell suppression assays. CD4⫹ T cells were enriched by negative selection using anti-B220 and anti-CD8, followed by negative selection using goat anti-rat Ig–coupled paramag- 3732 MONTE ET AL Figure 1. Increased FoxP3⫹ thymocyte selection in K/BxNgfp mice. A, Thymocyte analysis by flow cytometry. Thymocytes from 4-week-old KRNgfp and K/BxNgfp mice were labeled with allophycocyanin (APC)–conjugated CD4 and phycoerythrin (PE)–conjugated CD8 and analyzed. Live cells were gated on CD4 and CD8. Values in each quadrant are the percentage of positive cells in each subset. B, Histograms showing the expression of FoxP3 (top; numbers of cells) and V␤6 (bottom; percentage of maximum) in gated CD4⫹ single-positive thymocytes, by flow cytometry. V␤6 cells were labeled with V␤6–biotin followed by streptavidin–peridinin chlorophyll A protein (PerCP). Horizontal bars show the mean percentage of FoxP3⫹ cells. C and D, Percentages (C) and numbers (D) of FoxP3⫹ thymocytes in CD4⫹ single-positive (SP) thymocytes from 8 KRNgfp mice and 10 K/BxNgfp mice. Each data point represents a single mouse; horizontal lines show the mean. P values were determined by Wilcoxon’s rank sum test. netic beads (Chemicell, Berlin, Germany). Cells were stained with PE-labeled anti-CD4 prior to sorting for CD4⫹FoxP3– and CD4⫹FoxP3⫹ T cells using a FACSAria cell sorter. Cell purity was typically ⬎98%. In some experiments, APC-labeled anti-CD45RB was used to identify naive and memory T cells. Proliferation assays were performed in triplicate with 1 ⫻ 105 CD4⫹ Teff cells/well, with varying ratios of Treg cells and 2 ⫻ 105 irradiated splenocytes and 10 M GPI peptide (Global Peptide, Fort Collins, CO) in round-bottomed plates with ISC-10. In other experiments, irradiated purified DCs (1 ⫻ 104/well) or B cells (2 ⫻ 105/well) were used as allogeneic stimulators for polyclonal CD4⫹ T cells with anti-CD3 (5 g/ml). Neutralizing anti–IL-6 (10 g/ml, clone MP5-20F3; eBiosciences) or tumor necrosis factor receptor:Fc fusion protein (etanercept; 10 g/ml) was added as indicated. Cultures were pulsed at 72 hours with 0.2 Ci of 3H-thymidine/ well and harvested 18 hours later. Proliferation was measured as the counts per minute of 3H-thymidine incorporated. Transwell suppression assays. DCs or B cells from NOD mice or arthritic mice were activated for 1 hour with plate-bound anti-CD40 (5 g/ml) (clone 1C10; R&D Systems, Minneapolis, MN) and washed extensively before being added to the T cells. Teff cells (5 ⫻ 105), Treg cells (1.25 ⫻ 105), and irradiated splenocytes (5 ⫻ 105) were added in 0.4-m Transwell inserts or were cocultured with anti-CD40–activated DCs (5 ⫻ 104) or B cells (5 ⫻ 105) in 1 ml of ISC-10. T cells were activated with anti-CD3 (5 g/ml; clone 2C11). After 72 hours, 10 l of 1 mM BrdU was added to each well, and BrdU incorporation was assessed by flow cytometry after another 12 hours. Statistical analysis. Wilcoxon’s rank sum test and Student’s t-test were calculated using InStat 2.00 software (GraphPad Software, La Jolla, CA). RESULTS Thymic selection of FoxP3ⴙ Treg cells in K/BxN mice. We used K/BxNgfp mice to examine the development and function of FoxP3⫹ Treg cells during arthritis. In these mice, Treg cell–suppressive activity resided specifically in the FoxP3⫹ (green fluorescence protein [GFP]–positive) population, regardless of CD25 gfp INCREASED FoxP3⫹ Treg CELLS DURING EXPERIMENTAL ARTHRITIS expression (19). Therefore, we defined Treg cells as CD4⫹FoxP3⫹ T cells and Teff cells as CD4⫹FoxP3– T cells. FoxP3 expression in thymocytes was analyzed in 4-week-old KRNgfp and K/BxNgfp mice to minimize the effects of chronic inflammation on thymocyte development. In KRNgfp mice, CD4⫹ single-positive thymocytes comprised 1–3% of total thymocytes (Figure 1A). Negative selection by endogenous GPI resulted in an ⬃3fold decrease in thymic cellularity (mean ⫾ SD 21.5 ⫾ 2.37 ⫻ 107 in KRNgfp mice and 8.3 ⫾ 1.75 ⫻ 107 in K/BxNgfp mice), resulting primarily from losses in the double-positive subset (Figure 1A). Consistent with our previous observations, Treg cells were enriched 2–3-fold among CD4⫹ single-positive thymocytes in K/BxNgfp mice as compared with KRNgfp mice (Figures 1B and C) (14). To examine the effect of KRN TCR expression on FoxP3 selection, we analyzed the expression of the transgene-encoded TCR V␤6, because there is no clonotypical antibody to the KRN TCR. Consistent with the results obtained in other transgenic TCR systems, recognition of its cognate peptide appears to be a prerequisite for KRN Treg cell selection (21,22). In KRNgfp mice, which lacked GPI–I-Ag7 complexes, V␤6bright KRN T cells were seen only among the Teff cells, suggesting that Treg cells required endogenous TCR expression (resulting in lower V␤6 levels) for selection (Figure 1B). Conversely, in K/BxNgfp mice, negative selection by endogenous GPI resulted in preferential deletion of V␤6bright Teff cells. The remaining Teff cells expressed equivalently lower levels of V␤6, as in the Treg cells. Statistically higher numbers of CD4⫹ single-positive FoxP3⫹ thymocytes were found in K/BxNgfp mice (Figure 1D). The enrichment of Treg cells in K/BxNgfp mice resulted from both selective deletion of CD4⫹FoxP3– thymocytes, especially V␤6bright T cells, and induction of FoxP3⫹ thymocytes. Development of KRN Treg cells during arthritis in K/BxNgfp mice. The frequencies of Treg cells were increased similarly in the spleen and peripheral LNs. While GFP expression correlated largely with CD25 expression in KRNgfp mice, 10–15% of CD25⫹ cells were FoxP3– and were activated Teff cells (19). This proportion of CD25⫹FoxP3– Teff cells was doubled in K/BxNgfp mice, with significant overlap with FoxP3⫹ Treg cells, highlighting the power of the FoxP3-GFP reporter in this autoimmune model (Figure 2A). To show that FoxP3⫹ T cells function as Treg cells, sorted CD4⫹FoxP3⫹ T cells were tested in a suppression assay using CD4⫹FoxP3– T cells as responder Teff cells. Treg cells did not proliferate in response to NOD dendritic 3733 cells (DCs) presenting either endogenous GPI or KRN superagonist G7m (13). However, they nevertheless suppressed the proliferation of Teff cells when cocultured in a 1:1 ratio (Figure 2B). To determine Treg cell expansion during arthritis, we quantified the frequency and numbers of Treg cells in the spleen, joint draining LNs, nonjoint draining LNs, and mesenteric LNs during different phases of arthritis: nonarthritic KRNgfp control, acutely arthritic (4–6 weeks) K/BxNgfp mice, and chronically arthritic (8–10 weeks) K/BxNgfp mice. In all organs surveyed, the frequencies of Treg cells were increased in a hierarchical manner, with higher increases in the spleens and joint draining LNs relative to the nondraining LNs and mesenteric LNs in K/BxNgfp mice compared with KRNgfp mice (Figure 2C). (For spleens, P ⬍ 0.0001, nonarthritic versus acutely arthritic mice, and P ⫽ 0.0003, nonarthritic versus chronically arthritic mice, by Wilcoxon’s rank sum test. For draining LNs, P ⫽ 0.02, nonarthritic versus acutely arthritic mice, P ⫽ 0.0037, nonarthritic versus chronically arthritic mice, and P nonsignificant, acutely arthritic versus chronically arthritic mice. In both acutely arthritic and chronically arthritic mice, P ⬍ 0.005 for comparisons between each of the different organs.) There was no statistically significant difference in the distribution of Treg cells in the various organs in nonarthritic mice. When we quantified the total numbers of Treg cells in arthritic mice, Treg cell numbers were significantly increased in the spleens of acutely arthritic mice and in the spleens and draining LNs of chronically arthritic mice as compared with nonarthritic mice (P ⬍ 0.05 for spleen cells and draining LNs between nonarthritic mice and their counterparts in both acutely and chronically arthritic mice). Within the nonarthritic group, there was no statistically significant difference in the Treg cell numbers in various organs. Within the acutely arthritic group, differences between spleens and the various LNs were significant (P ⬍ 0.05). Within the chronically arthritic group, Treg cell numbers in the spleen and draining LNs were significantly different from those in nondraining LNs and mesenteric LNs. Therefore, Treg cells first expanded in the spleens and then spread to the draining LNs during disease progression (Figure 2D). Comparison of Teff cell and Treg cell numbers in the spleens and LNs showed that Treg cell expansion was met by similar expansion by Teff cells, such that the ratios of Teff cells to Treg cells were maintained during arthritis (Figure 2E). We also examined Teff cells and Treg cells in the paws. In nonarthritic mice, ⬍1% were CD4⫹ T cells. Of these, 5.9 ⫾ 3.2% (mean ⫾ SD) were Treg cells 3734 MONTE ET AL Figure 2. Site-specific expansion of FoxP3⫹ Treg cells during arthritis. A, Splenocyte analysis by flow cytometry. Splenocytes from 4-week-old KRNgfp and K/BxNgfp mice were labeled with allophycocyanin (APC)–conjugated CD4 and phycoerythrin (PE)–conjugated CD25 and analyzed. Shown are percentages in live cells gated on CD4 and CD25 (top) and expression of CD25 and FoxP3–green fluorescence protein (GFP) in gated CD4⫹ splenocytes (bottom). Values in each quadrant are the percentage of positive cells in each subset. B, T cell suppression assay. Treg cells (TR; CD4⫹FoxP3⫹) and Teff cells (TE; CD4⫹FoxP3–) were sorted and stimulated with irradiated NOD splenocytes, with or without 1 M G7m peptide, either separately or in coculture (1:1 ratio). C, Percentages of CD4⫹FoxP3⫹ Treg cells in spleens and lymph nodes (LNs; draining [dLNs], nondraining [ndLNs], and mesenteric [mLNs]) from 8 nonarthritic, 10 acutely arthritic, and 6 chronically arthritic mice. D, Total numbers of CD4⫹FoxP3⫹ Treg cells in spleens and LNs from 8 nonarthritic, 10 acutely arthritic, and 6 chronically arthritic mice. E, Total numbers of CD4⫹FoxP3– Teff cells in spleen and LNs from 8 nonarthritic, 10 acutely arthritic, and 6 chronically arthritic mice. Values in B–E are the mean ⫾ SD. F, Analysis of lymphocytes from inflamed paws. Lymphocytes were isolated from inflamed paws and analyzed for PE-labeled CD4 and GFP-labeled FoxP3 by flow cytometry. Live lymphocytes were identified by forward-scatter and side-scatter characteristics, as well as propidium iodide exclusion. Values in each quadrant are the percentage of positive cells in each subset; numbers in parentheses are the percentage of CD4⫹ T cells. Results are representative of 5 nonarthritic, 6 acutely arthritic, and 6 chronically arthritic mice. Med ⫽ medium. INCREASED FoxP3⫹ Treg CELLS DURING EXPERIMENTAL ARTHRITIS 3735 Figure 3. Higher rates of Treg cell proliferation than Teff cell proliferation in vivo. A, Fraction of bromodeoxyuridine (BrdU)–positive Treg cells (TR) and Teff cells (TE) in spleens, draining lymph nodes (dLNs), and nondraining LNs (ndLNs) from KRNgfp mice and acutely arthritic K/BxNgfp mice. Values are the mean and SD of 4 mice per group. Results are representative of 4 separate experiments. B, Plot of the numbers of BrdU⫹ Treg cells versus Teff cells in draining LNs from 20 arthritic mice. The bisecting diagonal line indicates 1:1 correspondence. With a few exceptions, more Treg cells than Teff cells incorporated BrdU. C, Bcl-2 expression in splenic Treg cells and Teff cells from arthritic mice. Cells were gated on FoxP3–green fluorescent protein (GFP) and CD4–allophycocyanin (right) and analyzed for intracellular Bcl-2 expression (left). Values in each quadrant are the percentage of each subset in splenocytes. Dark line shows phycoerythrin (PE)–labeled Bcl-2; shaded histogram shows the isotype control. Horizontal bars show the percentage of Bcl-2 cells. D, Frequency of Bcl-2⫹ cells and mean fluorescence intensity (MFI) of Bcl-2 expression in Teff cells and Treg cells from draining LNs obtained from 4 arthritic mice. Horizontal bars show the mean. Results are representative of 2 independent experiments. P values were determined by Fisher’s t-test. E, Percentage of annexin V⫹ (Ann V⫹) Teff cells and Treg cells in draining LNs from 5 arthritic mice. Horizontal bars show the mean. Results are representative of 2 independent experiments. P value was determined by Fisher’s t-test. (Figure 2F). During acute arthritis, CD4⫹ T cells comprised 2.1 ⫾ 0.7%, with 58.5 ⫾ 10.9% being Treg cells. As the inflammation waned in the paws, Treg cells became less frequent, comprising 39.9 ⫾ 10.1% of CD4⫹ T cells. Our data were consistent with findings in human RA showing an accumulation of Treg cells during active inflammation. Greater in vivo turnover of Treg cells than Teff cells in K/BxNgfp mice. Because increased numbers of Treg cells in the spleen and LNs during arthritis can arise from increased proliferation or migration, we directly examined Teff cell and Treg cell proliferation by in vivo BrdU incorporation. Nonarthritic KRNgfp mice and acutely arthritic K/BxNgfp mice were injected intra- 3736 peritoneally with 1 mg of BrdU 12 hours prior to killing. The proportions of BrdU⫹ V␤6⫹CD4⫹FoxP3– (KRN Teff cells) and V␤6⫹CD4⫹FoxP3⫹ (KRN Treg cells) were quantified. In nonarthritic KRNgfp mice, higher proportions of Treg cells (⬃10%) than Teff cells (1–3%) incorporated BrdU (Figure 3A). This tonic proliferation by Treg cells suggests active surveillance at steady-state, even in nonlymphopenic mice. In arthritic K/BxNgfp mice, recognition of GPI increased the proportion of cycling Teff cells and Treg cells by 3-fold, resulting in 5–10% of Teff cells and 30–35% of Treg cells incorporating BrdU. Consistently higher proportions of cycling Treg cells were observed in the LNs (Figure 3A) and accounted for the greater increase in T cell numbers seen in the LNs as the disease progressed from an acute to a chronic phase (Figure 2D). To directly compare the numbers of cycling Teff cells and Treg cells, we plotted the numbers of BrdU⫹ Teff cells versus the numbers of BrdU⫹ Treg cells in draining LNs from arthritic mice (Figure 3B). With few exceptions, more Treg cells than Teff cells incorporated BrdU. In some mice, the number of BrdU⫹ Treg cells was 3-fold higher than the number of BrdU⫹ Teff cells. We did not, however, find any correlation between our gross measurements of paw inflammation with the numbers of BrdU⫹ Treg cells or Teff cells. Similar results were seen in the spleens and other LNs (data not shown). Given the higher rate of Treg cell proliferation, the proportions of Treg cells should increase over time. However, this was not the case (Figure 2C). To reconcile this discrepancy, we hypothesized that the increased turnover is balanced by increased death. In support of this, we found that Treg cells from arthritic mice expressed significantly less antiapoptotic factor Bcl-2 (Figures 3C and D) and higher levels of apoptosis (annexin V⫹ cells) than did Teff cells ex vivo (Figure 3E). Different kinetics of proliferation and death of FoxP3ⴙ Treg cells and FoxP3– Teff cells. To test our hypothesis that Treg cells proliferated and died at a faster rate than Teff cells, we directly examined the kinetics of Treg cell and Teff cell proliferation and attrition in vitro. KRNgfp CD4⫹ T cells were purified from pooled spleens and LNs and were cultured with NOD DCs. This approach allowed us to specifically examine intrinsic differences between Treg cells and Teff cells separately from the inflammatory milieu in arthritic mice. We used naive T cells to synchronize the exposure to antigenic stimuli. At various times, BrdU was added to the cultures, and BrdU incorporation was analyzed 12 hours later. MONTE ET AL Figure 4. Kinetics of Treg cell and Teff cell proliferation and cell death upon antigen stimulation. A, Bromodeoxyuridine (BrdU) incorporation by Treg cells (TR) and Teff cells (TE) analyzed at the indicated times. Purified CD4⫹ T cells from KRNgfp mice were cultured with NOD dendritic cells (DCs). BrdU (1 mM) was added at various times, and incorporation was analyzed by flow cytometry 12 hours later. Values are the mean ⫾ SD of duplicate wells. Results are representative of 3 experiments. B and C, Percentages of dead and apoptotic Treg cells and Teff cells in arthritic and nonarthritic mice. Purified CD4⫹ T cells from KRNgfp and arthritic K/BxNgfp mice were cultured with NOD DCs, and dead and apoptotic CD4⫹ T cells were analyzed with 7-aminoactinomycin D (7-AAD) and annexin V, respectively, at the indicated times. Values are the mean of duplicate wells (SDs were ⬍15%). The experiment was performed 3 times, and the results were similar. In contrast to the in vitro anergy noted when Treg cells were cultured alone, Treg cells showed significant proliferation when cultured at a physiologic ratio with Teff cells and stimulated with physiologic amounts of GPI. By day 1, 20% of FoxP3⫹ Treg cells had entered cell cycle and incorporated BrdU, as compared with ⬍2% of FoxP3– Teff cells. BrdU incorporation by Treg cells peaked on day 2 and declined over days 3 and 4 (Figure 4A). In contrast, Teff cells did not proliferate to INCREASED FoxP3⫹ Treg CELLS DURING EXPERIMENTAL ARTHRITIS any appreciable degree in the first 24 hours of culture but proliferated vigorously in the subsequent days. Our data showed that Teff cells and Treg cells proliferated with distinct kinetics and magnitude. Treg cells proliferated earlier, but the response was short lived. Teff cell proliferation, though delayed, was more sustained. In a parallel experiment, cell death and apoptosis of KRNgfp T cells were analyzed using 7-aminoactinomycin D and annexin V to identify dead cells and apoptotic cells, respectively. As shown in Figure 4B, the proportions of dead and apoptotic Treg cells were roughly twice the proportions of dead and apoptotic Teff cells observed during this time course. Interestingly, the time course of Treg cell apoptosis correlated with Treg cell proliferation: both peaked on day 2 and declined on days 3 and 4. This finding contrasted with the kinetics of Teff cell proliferation and apoptosis and provided support that Bcl-2 expression in Teff cells conferred protection from apoptosis. To determine if prior activation conferred protection from apoptosis, Teff cells and Treg cells from arthritic K/BxNgfp mice were also analyzed (Figure 4C). Naive and activated Teff cells displayed similar rates of cell death. Compared with naive Treg cells, Treg cells from K/BxNgfp mice displayed less apoptosis and death. However, relative to their Teff cell counterparts, both naive and activated Treg cells still showed higher rates of apoptosis. Together, the data confirmed our hypothesis that increased Treg cell proliferation was offset by increased Treg cell apoptosis, resulting in Teff cell/Treg cell homeostasis. Increasingly suppressive Treg cells during arthritis. To interrogate Treg cell function during arthritis, we sorted Teff cells and Treg cells from acutely and chronically arthritic mice. In addition, to determine if there were site-specific effects, Teff cells and Treg cells from spleens and draining LNs were analyzed separately. We first compared the proliferation of Teff cells from the spleens and LNs during acute and chronic disease. LN Teff cells proliferated better than splenic Teff cells in response to GPI, but there was no statistically significant difference in the proliferation of Teff cells between acute and chronic disease (Figure 5A). We next examined the ability of Treg cells to suppress Teff cell proliferation. Because Teff cells from chronically arthritic mice would be targets for Treg cell immunotherapy, we used LN Teff cells from chronically arthritic mice as responders. As shown in Figure 5B, Treg cells obtained from different sites and during different disease states suppressed Teff cell proliferation with differing efficacy. While all Treg cells were effective at a 1:1 ratio, splenic Treg cells from acutely arthritic 3737 Figure 5. Dynamic changes in Treg cell and Teff cell activity during arthritis. A, Teff cells (CD4⫹FoxP3–) were sorted from spleens (Spl) or pooled lymph nodes (LNs) obtained from acutely or chronically arthritic mice. Teff cells (1 ⫻ 105/well) were stimulated with irradiated NOD dendritic cells (DCs), and cell proliferation was assayed by 3 H-thymidine incorporation during the last 16 hours of a 72-hour culture. Values are the mean ⫾ SD of triplicate wells. B, Treg cells (TR) from different sites and different disease states were analyzed for their suppressive activity against Teff cells (TE). Values are the mean ⫾ SD of triplicate wells. C and D, CD4⫹FoxP3–CD45RBlow (memory) and CD4⫹FoxP3–CD45RBhigh (naive) Teff cells from arthritic mice were cultured at varying ratios of Treg cells with NOD DCs. Values are the mean ⫾ SD of either the cpm of triplicate wells (C) or normalized as a fraction of the maximum cpm of Teff cells alone (D). P ⬍ 0.02 for CD4⫹FoxP3–CD45RBhigh versus CD4⫹FoxP3– CD45RBlow cells at each of the Teff cell–to–Treg cell ratios in C, by Fisher’s t-test. NS ⫽ not significant. Results are representative of 3 similar experiments. mice were the least suppressive. Conversely, LN Treg cells from chronically arthritic mice were the most effective. Thus, Treg cell activity increased during arthritis and at sites of active inflammation. To determine if the Teff cell activation state influences its susceptibility to suppression, we sorted FoxP3–CD45RB high (naive) Teff cells, FoxP3– CD45RBlow (memory) Teff cells, and FoxP3⫹ Treg cells from pooled splenic and draining LN T cells obtained from arthritic mice. Naive CD45RBhigh Teff cells were 3738 MONTE ET AL Figure 6. Reduced Treg cell suppression in dendritic cells (DCs) and B cells from arthritic mice. A and B, Dendritic cells and B cells were purified from NOD and K/BxN arthritic mice using microbeads. Teff cells (TE) and Treg cells (TR) from FoxP3gfp mice were sorted. Teff cells (1 ⫻ 105/well) were cultured with either 1 ⫻ 104 DCs (A) or 2 ⫻ 105 B cells (B) and 5 g/ml of anti-CD3. Treg cells were added to Teff cell cultures at the indicated ratios. Proliferation was measured at 72 hours by 3H-thymidine incorporation. Mean ⫾ SD proliferation of Teff cells in response to antigenpresenting cells was 6,874 ⫾ 752 cpm in NOD DCs, 17,455 ⫾ 788 cpm in K/BxN DCs, 4,059 ⫾ 298 cpm in NOD B cells, and 8,094 ⫾ 1,442 cpm in K/BxN B cells. Values are the mean ⫾ SD of triplicate wells. Results are representative of 2 experiments. C, Teff cells and Treg cells were cultured at a 4:1 ratio and stimulated with anti-CD3 and irradiated splenocytes. DCs and B cells from NOD or arthritic mice were purified and activated by plate-bound anti-CD40 for 1 hour and then washed. Activated antigen-presenting cells were added to the T cell cultures either directly (coCX) or separated by a Transwell membrane (TW). In some T cell cultures, conditioned supernatants from 48-hour cultures of anti-CD40–activated DCs and B cells were added (1:1 volume/ volume). Bromodeoxyuridine (BrdU) incorporation was assayed at 72 hours. Values are the mean ⫾ SD of duplicate wells. D, Teff cells (1 ⫻ 105/well), with or without Treg cells (5 ⫻ 104/well), were stimulated with anti-CD3 and DCs (1 ⫻ 104/well) from either NOD or K/BxN arthritic mice in the presence of neutralizing anti–interleukin-6 (␣IL-6) antibodies (10 g/ml) or tumor necrosis factor receptor:Fc fusion protein (TNFR-Fc) (10 g/ml). Proliferation was measured at 72 hours by 3Hthymidine incorporation. Values are the mean ⫾ SD of triplicate wells. Results are representative of 2 experiments. less proliferative and more easily suppressed than experienced CD45RBlow Teff cells (Figure 5C). To determine if this was due to differences in the magnitude of the T cell response, we normalized the response against proliferation in the absence of Treg cells. At Teff cell–to–Treg cell ratios of 2:1 and 4:1, Treg cell suppres- INCREASED FoxP3⫹ Treg CELLS DURING EXPERIMENTAL ARTHRITIS sion was 2-fold less effective with CD45RBlow Teff cells (Figure 5D). Thus, during arthritis, both Teff cells and Treg cells enhanced their response to GPI, such that increased Treg cell suppression was met by increased Teff cell resistance, resulting in an immunologic détente. Contribution of inflammatory antigenpresenting cells to resistance to Treg cell suppression. Proinflammatory cytokines produced by monocytes and DCs can abrogate Treg cell function by increasing Treg cell apoptosis or by increasing Teff cell resistance (9,23,24). In K/BxN mice, DCs and B cells comprised 2 major antigen-presenting cells for the activation of KRN T cells (20). DCs were the most efficacious at stimulating KRN T cells, whereas B cells were the most abundant. We therefore compared the efficacy of Treg cell suppression in the presence of DCs and B cells from chronically arthritic K/BxN mice or control NOD mice. We have previously reported that antigen-presenting cells from arthritic K/BxN mice showed greater presentation of GPI peptide compared with nonarthritic antigen-presenting cells (20). To eliminate the role of the GPI dose, we used polyclonal T cells from FoxP3gfp mice on the H-2k background in which the antigenpresenting cells would provide a strong allostimulation in addition to anti-CD3. As expected, DCs from arthritic mice attenuated the suppressive effect of Treg cells (Figure 6A). With NOD antigen-presenting cells, Treg cells suppressed Teff cell proliferation by 50% at a Teff cell–to–Treg cell ratio of 8:1. Four-fold higher numbers of Treg cells were necessary in the presence of arthritic DCs. Arthritic B cells similarly inhibited the suppressive effect of Treg cells (Figure 6B). To determine the mechanism by which DCs and B cells inhibit Treg cell suppression, we added antiCD40–activated DCs and B cells to the Teff cell/Treg cell suppression assays, either in the same wells or separated by Transwell membranes. Proliferation of Teff cells was assayed by flow cytometry and BrdU incorporation. As shown in Figure 6C, anti-CD40–activated DCs and B cells abrogated Treg cell suppression only when they were cocultured with the Teff cells and Treg cells. Restoration of Teff cell proliferation was absent when DCs and B cells were separated by Transwell membranes or when supernatants from DC and B cell cultures were added. The finding that inflammatory DCs can reverse Treg cell suppression has been previously described (9,24). However, those reports attributed the abrogation of Treg cell activity to soluble factors such as tumor necrosis factor ␣ (TNF␣) and interleukin-6 (IL-6). Here, we found that neutralization of IL-6 or TNF␣ had no 3739 effect on the suppressive activity of Treg cells (Figure 6D). Similar results were seen when we used B cells as antigen-presenting cells (data not shown). Together, our data demonstrated that both arthritic DCs and B cells decrease Treg cell activity by a cell–cell contact–dependent or close-contact–dependent mechanism. The finding that arthritic B cells can also inhibit Treg cell suppression provides a potential mechanism for the efficacy of B cell depletion in controlling RA. DISCUSSION There is considerable interest in harnessing CD4⫹CD25⫹ Treg cells to control autoimmune diseases (25,26). In RA patients, the number of CD4⫹CD25⫹ Treg cells is increased in inflamed joints, and these cells show variable activity against their Teff cell counterparts. Results have differed among studies partly due to differences in the criteria used for Treg cell identification, as well as the disease states and patient populations studied. It is also not clear what factors control their tropism, function, growth, and death during disease. In the present study, we used K/BxNgfp mice to model the development and function of arthritogenic Teff cells and Treg cells in a well-characterized murine model of RA. The K/BxN mouse model has many advantages with regard to RA in humans. The progression of arthritis has been extensively characterized and follows a predictable course: K/BxN mice develop overt joint inflammation at 4–5 weeks, which peaks at 7–8 weeks, and is followed by a chronic phase in which joint destruction and deformity predominate (27). In addition, the arthritis shows regional involvement, with the distal joints being more affected than the proximal and axial joints. This feature allowed us to determine if there is regional expansion of Treg cells at sites of disease. Moreover, by examining GPI reactivity, we focused specifically on the arthritogenic KRN Teff cells and Treg cells in this system. Last, we used FoxP3gfp to unequivocally identify Treg cells. These experiments are the first to demonstrate the expansion and function of antigen-specific Treg cells during the initiation and progression of spontaneous arthritis. In these experiments, we confirmed our previous findings that incomplete deletion of KRN thymocytes by endogenous GPI resulted in an increased frequency of GPI-specific Treg cells in K/BxNgfp mice through selective deletion of V␤6bright FoxP3– Teff cells and induction of FoxP3⫹ Treg cells in the thymus (14). In contrast to other models of autoimmune disease in 3740 which Treg cell deficits resulted in pathologic changes, K/BxN mice displayed enhanced populations of GPIspecific Treg cells (14,15,28–30). These Treg cells proliferated quite well, both in vitro and in vivo, in response to endogenously presented GPI, they were functional, and they suppressed Teff cell proliferation in vitro. Treg cell numbers expanded in parallel with Teff cell numbers in the spleens and LNs during arthritis. In K/BxN mice with a ubiquitous self antigen, both Teff cell and Treg cell responses were initiated in the spleen, with subsequent extension to the LNs and with a preference for the draining LNs, indicating selective tropism to inflamed joints from the increased inflammation and/or GPI presentation during disease. However, this enhanced activity was offset by an increased resistance of GPIspecific Teff cells and antigen-presenting cells to suppression. A recent study described similar dynamic changes in the spatiotemporal expansion of Teff cells and Treg cells in the draining LNs during the induction and progression of adjuvant-induced arthritis that presumably reflected changes in the concentration of the inciting antigen (31). Those findings complement the findings of our studies, with the exception that their waxing/ waning Treg cell course likely reflected the monophasic course in adjuvant-induced arthritis compared with the progressive expansion of Teff cells and Treg cells in the K/BxN model. Despite the enrichment in Treg cell activity, K/BxN mice nevertheless developed arthritis. Our in vitro studies demonstrate effective Teff cell suppression at ratios that approximate the physiologic ratio found in vivo. Why, then, are Treg cells ineffective in vivo? There are several possible explanations. First, Treg cells and Teff cells are differentially regulated during ontogeny. CD4⫹FoxP3– single-positive thymocytes reached mature levels during the first week of life. In contrast, CD4⫹FoxP3⫹ single-positive thymocyte development was substantially delayed and did not achieve mature levels until 21 days of age (32). The delayed appearance of Treg cells in K/BxN mice would result in unopposed Teff cell reactivity during the initiation of arthritis. Second, the precursor frequency of GPI-specific T cells is supraphysiologic in K/BxN mice. The self antigen is also ubiquitously expressed and can be presented by all antigen-presenting cells, including B cells. Under such conditions, autoreactive B cells can be driven into memory B cells despite abundant Treg cells (33). Thus, for a humorally driven arthritis, Treg cells may be less effective than for a T cell–driven disease such as colitis. We propose that the developmental regulation and numeric superiority of KRN Teff cells overcome Treg cell sup- MONTE ET AL pression at the initiation of disease. Treg cells might simply arrive too late and in too few numbers. During the ensuing arthritis, Treg cells are highly activated, with ⬎30% of Treg cells proliferating in a 12-hour period. However, this vigorous Treg cell activity is attenuated by 3 factors: Treg cells undergo shorter proliferative bursts and higher apoptosis rates than Teff cells upon antigen stimulation, Teff cells become more refractory to Treg cell suppression as they differentiate into memory T cells, and antigen-presenting cells from arthritic mice abrogate Treg cell suppression. This inhibition of Treg cell suppression required close contact, but it is unknown whether this is due to increased resistance of antigen-presenting cells to Treg cells or to enhanced antigen presentation to Teff cells. It is also unclear whether these features are due to changes in the costimulatory molecules or to secretion of paracrine cytokines in activated antigen-presenting cells from arthritic mice. Studies to dissect these mechanisms are ongoing. Our findings parallel observations in human RA and provide an explanation for the paradoxical occurrence of disease in the presence of abundant Treg cells. By elucidating the factors and mechanisms that augment Treg cell activity, it may be possible to achieve control of disease. K/BxNgfp mice thereby provide a valuable model in which to examine Teff cell/Treg cell homeostasis during disease and therapy. ACKNOWLEDGMENTS We are grateful to Drs. D. Mathis and C. Benoist and to INSERM/IGBMC for the KRN mice. We thank Dr. A. Rudensky for the FoxP3gfp mice. We are grateful to Drs. P. Tarr, C. Pham, and J. P. Atkinson for critical review of the manuscript. AUTHOR CONTRIBUTIONS Dr. Shih had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study design. Shih. Acquisition of data. Monte, Wilson. Analysis and interpretation of data. Shih. Manuscript preparation. Shih. Statistical analysis. Shih. REFERENCES ⫹ 1. Shevach EM. CD4 CD25⫹ suppressor T cells: more questions than answers. Nat Rev Immunol 2002;2:389–400. 2. Sakaguchi S. Naturally arising Foxp3-expressing CD25⫹CD4⫹ regulatory T cells in immunological tolerance to self and non-self. Nat Immunol 2005;6:345–52. 3. De Kleer IM, Wedderburn LR, Taams LS, Patel A, Varsani H, Klein M, et al. CD4⫹CD25bright regulatory T cells actively regulate INCREASED FoxP3⫹ Treg CELLS DURING EXPERIMENTAL ARTHRITIS 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. inflammation in the joints of patients with the remitting form of juvenile idiopathic arthritis. J Immunol 2004;172:6435–43. Ruprecht CR, Gattorno M, Ferlito F, Gregorio A, Martini A, Lanzavecchia A, et al. Coexpression of CD25 and CD27 identifies FoxP3⫹ regulatory T cells in inflamed synovia. J Exp Med 2005;201:1793–803. Cao D, Malmstrom V, Baecher-Allan C, Hafler D, Klareskog L, Trollmo C. Isolation and functional characterization of regulatory CD25brightCD4⫹ T cells from the target organ of patients with rheumatoid arthritis. Eur J Immunol 2003;33:215–23. Cao D, van Vollenhoven R, Klareskog L, Trollmo C, Malmstrom V. CD25brightCD4⫹ regulatory T cells are enriched in inflamed joints of patients with chronic rheumatic disease. Arthritis Res Ther 2004;6:R335–46. Ehrenstein MR, Evans JG, Singh A, Moore S, Warnes G, Isenberg DA, et al. Compromised function of regulatory T cells in rheumatoid arthritis and reversal by anti-TNF␣ therapy. J Exp Med 2004;200:277–85. Van Amelsfort JM, Jacobs KM, Bijlsma JW, Lafeber FP, Taams LS. CD4⫹CD25⫹ regulatory T cells in rheumatoid arthritis: differences in the presence, phenotype, and function between peripheral blood and synovial fluid. Arthritis Rheum 2004;50: 2775–85. Van Amelsfort JM, van Roon JA, Noordegraaf M, Jacobs KM, Bijlsma JW, Lafeber FP, et al. Proinflammatory mediator–induced reversal of CD4⫹,CD25⫹ regulatory T cell–mediated suppression in rheumatoid arthritis. Arthritis Rheum 2007;56:732–42. Cao D, Borjesson O, Larsson P, Rudin A, Gunnarsson I, Klareskog L, et al. FOXP3 identifies regulatory CD25brightCD4⫹ T cells in rheumatic joints. Scand J Immunol 2006;63:444–52. Korganow AS, Ji H, Mangialaio S, Duchatelle V, Pelanda R, Martin T, et al. From systemic T cell self-reactivity to organspecific autoimmune disease via immunoglobulins. Immunity 1999;10:451–61. Matsumoto I, Staub A, Benoist C, Mathis D. Arthritis provoked by linked T and B cell recognition of a glycolytic enzyme. Science 1999;286:1732–5. Basu D, Horvath S, Matsumoto I, Fremont DH, Allen PM. Molecular basis for recognition of an arthritic peptide and a foreign epitope on distinct MHC molecules by a single TCR. J Immunol 2000;164:5788–96. Shih FF, Mandik-Nayak L, Wipke BT, Allen PM. Massive thymic deletion results in systemic autoimmunity through elimination of CD4⫹ CD25⫹ T regulatory cells. J Exp Med 2004;199:323–35. Nguyen LT, Jacobs J, Mathis D, Benoist C. Where FoxP3dependent regulatory T cells impinge on the development of inflammatory arthritis. Arthritis Rheum 2007;56:509–20. Morgan ME, Sutmuller RP, Witteveen HJ, van Duivenvoorde LM, Zanelli E, Melief CJ, et al. CD25⫹ cell depletion hastens the onset of severe disease in collagen-induced arthritis. Arthritis Rheum 2003;48:1452–60. Morgan ME, Flierman R, van Duivenvoorde LM, Witteveen HJ, van Ewijk W, van Laar JM, et al. Effective treatment of collageninduced arthritis by adoptive transfer of CD25⫹ regulatory T cells. Arthritis Rheum 2005;52:2212–21. Ohata J, Miura T, Johnson TA, Hori S, Ziegler SF, Kohsaka H. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 3741 Enhanced efficacy of regulatory T cell transfer against increasing resistance, by elevated Foxp3 expression induced in arthritic murine hosts. Arthritis Rheum 2007;56:2947–56. Fontenot JD, Rasmussen JP, Williams LM, Dooley JL, Farr AG, Rudensky AY. Regulatory T cell lineage specification by the forkhead transcription factor foxp3. Immunity 2005;22:329–41. Shih FF, Racz J, Allen PM. Differential MHC class II presentation of a pathogenic autoantigen during health and disease. J Immunol 2006;176:3438–48. Jordan MS, Boesteanu A, Reed AJ, Petrone AL, Holenbeck AE, Lerman MA, et al. Thymic selection of CD4⫹CD25⫹ regulatory T cells induced by an agonist self-peptide. Nat Immunol 2001;2: 301–6. Van Santen HM, Benoist C, Mathis D. Number of T reg cells that differentiate does not increase upon encounter of agonist ligand on thymic epithelial cells. J Exp Med 2004;200:1221–30. Valencia X, Stephens G, Goldbach-Mansky R, Wilson M, Shevach EM, Lipsky PE. TNF downmodulates the function of human CD4⫹CD25hi T-regulatory cells. Blood 2006;108:253–61. Pasare C, Medzhitov R. Toll pathway-dependent blockade of CD4⫹CD25⫹ T cell-mediated suppression by dendritic cells. Science 2003;299:1033–6. Tang Q, Henriksen KJ, Bi M, Finger EB, Szot G, Ye J, et al. In vitro-expanded antigen-specific regulatory T cells suppress autoimmune diabetes. J Exp Med 2004;199:1455–65. Kretschmer K, Apostolou I, Jaeckel E, Khazaie K, von Boehmer H. Making regulatory T cells with defined antigen specificity: role in autoimmunity and cancer. Immunol Rev 2006;212:163–9. Mandik-Nayak L, Wipke BT, Shih FF, Unanue ER, Allen PM. Despite ubiquitous autoantigen expression, arthritogenic autoantibody response initiates in the local lymph node. Proc Natl Acad Sci U S A 2002;99:14368–73. Sakaguchi S, Sakaguchi N, Asano M, Itoh M, Toda M. Immunologic self-tolerance maintained by activated T cells expressing IL-2 receptor ␣-chains (CD25): breakdown of a single mechanism of self-tolerance causes various autoimmune diseases. J Immunol 1995;155:1151–64. Wu AJ, Hua H, Munson SH, McDevitt HO. Tumor necrosis factor-␣ regulation of CD4⫹CD25⫹ T cell levels in NOD mice. Proc Natl Acad Sci U S A 2002;99:12287–92. Powrie F, Read S, Mottet C, Uhlig H, Maloy K. Control of immune pathology by regulatory T cells. Novartis Found Symp 2003;252:92–8. Nolte-’t Hoen EN, Boot EP, Wagenaar-Hilbers JP, van Bilsen JH, Arkesteijn GJ, Storm G, et al. Identification and monitoring of effector and regulatory T cells during experimental arthritis based on differential expression of CD25 and CD134. J Leukoc Biol 2008;83:112–21. Fontenot JD, Dooley JL, Farr AG, Rudensky AY. Developmental regulation of Foxp3 expression during ontogeny. J Exp Med 2005;202:901–6. Guay HM, Larkin J III, Picca CC, Panarey L, Caton AJ. Spontaneous autoreactive memory B cell formation driven by a high frequency of autoreactive CD4⫹ T cells. J Immunol 2007;178: 4793–802.