FoxP3⁺ regulatory T cells increase in number and function during experimental arthritis

Abstract

Objective

CD4⁺CD25⁺FoxP3⁺ regulatory T cells (T_R) are critical regulators of autoimmunity. Yet, T_R are paradoxically increased in RA patients and show variable activity in human studies. Our objective is to characterize the expansion and function of T_R during the initiation and progression of experimental arthritis.

Methods

To unequivocally identify T_R, we crossed FoxP3^gfp mice to K/BxN to generate arthritic mice in which T_R express green fluorescence protein. We examined the expansion and function of T_R and effector T cells (T_E) during different stages of arthritis by flow cytometry and cell proliferation.

Results

In K/BxN mice, thymic selection of KRN T cells resulted in an enrichment of FoxP3⁺ T_R. T_R numbers increased during arthritis with significant increases in the spleen and draining lymph node, indicating selective tropism to sites of disease. In contrast to the in vitro unresponsiveness when cultured by themselves, substantial fractions of T_R proliferated in both non-arthritic and arthritic mice. However, they also underwent greater apoptosis thereby maintaining equilibrium with T_E. Similarly, enhanced T_R suppressive activity during arthritis was offset by greater resistance by their T_E counterparts and antigen presenting cells.

Conclusion

In this well established model of RA, the interplay of T_E and T_R in K/BxN mice recapitulated many features of human disease. We demonstrated an ordered expansion of T_R during arthritis and the dynamic changes in T_R and T_E functions. By elucidating factors that govern T_R and T_E development in K/BxN^gfp mice, we will gain insight into the pathophysiology and develop novel therapeutics for human RA.

Rheumatoid arthritis (RA) is an autoimmune disease resulting in joint inflammation and destruction. Autoreactive T and B cells are crucial for 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 periphery; subsequent activation can result in autoimmune pathology. CD4⁺CD25⁺FoxP3⁺ regulatory T cells (T_R), comprising 5–10% of CD4⁺ T cells, are crucial for the maintenance of peripheral tolerance (1, 2).

In adult RA and juvenile idiopathic arthritis (JIA), T_R were enriched in the synovial fluid of inflamed joints compared to peripheral blood, suggesting active homing or expansion at inflammatory sites (3–8). These T_R expressed FoxP3 and suppressed both proliferation and cytokine production by CD4⁺CD25⁻ effector T cells (T_E). Moreover, increased T_R in the synovial fluid directly correlated with limited disease (3), suggesting that T_R aid in disease remission. However, it is unknown how T_R modulate immunity as they are paradoxically increased during disease and there is also variability in T_R function among different studies. For example, Ehrenstein et al demonstrated that T_R from RA patients showed compromised function compared to healthy controls (7), while others showed that T_R obtained during active disease were equally or more suppressive than healthy controls (8). In this case, the increased suppressive function was offset by the responder T_E 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 differences observed between studies. In addition, investigators differed in their criteria for T_R with some studies focusing only on CD25^bright while others used all CD25⁺ T cells (8, 10). Because CD25 is also elevated in activated T_E, which are increased during disease, varying degrees of T_E contamination can render interpretation difficult. It is also unclear if arthritis resulted from a primary T_R dysfunction or a secondary defect due to persistent inflammation.

To eliminate these confounding processes, we examined T_R 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 Non Obese Diabetic (NOD) mice (11). The disease is fully penetrant in all progeny and follows a predictable course of progressive symmetrical distal polyarthritis resembling human RA. Arthritis results from autoreactive KRN T cells recognizing peptide 281–293 of the glycolytic enzyme, glucose-6-phosphate isomerase (GPI), bound to I-A^g7 (the NOD specific MHC II allele) (12, 13). Incomplete thymic deletion allows autoreactive KRN CD4⁺ T cells to persist in the periphery and become activated by endogenously presented GPI. KRN T cells then provide help to GPI-specific B cells, giving rise to arthritogenic antibodies. T_R are enriched in arthritic K/BxN mice (14, 15) and loss of T_R results in earlier and more extended disease, suggesting that although T_R do not prevent arthritis, T_R may nevertheless mitigate it (15). Similar findings are found in collagen induced arthritis in which depletion of CD25⁺ T cells exacerbated arthritis and adoptive transfer of CD4⁺CD25⁺ T_R or FoxP3 transduced T cells ameliorated disease (16–18). To understand how antigen specific T_R develop during the course of arthritis, we crossed K/BxN to FoxP3^gfp reporter mice to unequivocally identify T_R. Here, we analyzed T_R selection in the thymus and followed their expansion and function during the progression of arthritis.

Materials and Methods

Mice

KRN mice have been described (11). FoxP3^gfp mice were generously provided by A. Rudensky and were bred to KRN mice to generate KRN^gfp mice in which FoxP3⁺ T cells expressed GFP (19). These were subsequently bred to NOD Lt/J mice (Jackson Laboratories, Bar Harbor, ME) to generate K/BxN^gfp mice. Non-arthritic controls were KRN^gfp on the C57Bl/6 background. Because FoxP3 is on the X chromosome, only male K/BxN^gfp mice were analyzed. K/BxN^gfp exhibited arthritis with equivalent severity and time course as 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 (dLN: popliteal, axillary, and brachial), non-draining (ndLN: inguinal and cervical), and mesenteric (mLN) lymph nodes were harvested from individual mice and analyzed separately. Single cell suspensions (3×10⁶) were treated with αCD16 (2.4G2) prior to surface staining with specific antibodies according to standard protocols. The following antibodies were used: αCD4-APC, αCD4-PE, (GK1.5), αCD8-PE-Cy7 (53-6.7), αCD25-PE-Cy7 (PC61), αBrdU-APC, αCD45RB (C363-16A), Annexin-V-PE, and αBcl-2-PE kit were from BD Biosciences or eBiosciences, San Diego, CA. Streptavidin-PerCP was obtained from BD Biosciences, San Diego, CA. TCR Vβ6 was used to track the KRN transgenic TCR. All samples were analyzed on either a FACSAria or a FACScaliber flow cytometer (BD Biosciences, Mountain View, CA) with FlowJo software. Lymphocytes were gated based on forward and side scatter 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 with 5% calf serum, 125 μl Collagenase VIII (10,000U/ml) and 200 μl Dispase (50U/ml). Minced tissues were agitated for two hours at 37°C at 150 rpm with vortexing every 30 minutes. Digested tissues were strained through a 70 μM 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 dLNs were enriched by positive selection using αCD4 Miltenyi microbeads. Purity was typically >90%. Dendritic cells and B cells were prepared from spleens as described (20). T cells (5×10⁵/well) were stimulated with 5×10⁴ DC in 1 ml of Iscove’s medium containing 10% heat inactivated bovine growth serum (BGS, Hyclone, Logan UT), 2 μM Glutamax (Gibco-BRL, Gaithersburg, MD), 2×10⁻⁵ M β-mercaptoethanol, 50 μg/ml gentamicin (referred hence as ISC-10). At various times, 1 mM BrdU was added and cultures analyzed 12 hours later for BrdU incorporation and apoptosis.

T cell suppression assays

CD4⁺ T cells were enriched by negative selection using αB220 and αCD8, followed by negative selection using goat-anti-rat Ig coupled paramagnetic beads (Chemicell, Berlin, Germany). Cells were stained with αCD4-PE prior to sorting for CD4⁺FoxP3⁻ and CD4⁺FoxP3⁺ T cells using FACSAria. Cell purity was typically >98%. In some experiments, αCD45RB-APC was used to identify naïve and memory T cells. Proliferation assays were performed in triplicate with 1 ×10⁵ CD4⁺ T_E/well with varying ratios of T_R and 2×10⁵ irradiated splenocytes and 10 μM GPI peptide (Global Peptide, Fort Collins, CO) in round bottom plates with ISC-10%. For experiments in Figure 6, irradiated purified DCs (1×10⁴/well) or B cells (2×10⁵/well) were used as allogeneic stimulators for polyclonal CD4⁺ T cells with αCD3 (5 μg/ml). Neutralizing αIL-6 (10μglml, clone MP5-20F3, eBiosciences) or TNFR:Fc(10 μg/ml, etanercept, Immunex) were added as indicated. Cultures were pulsed at 72 hours with 0.2 μCi ³H-thymidine/well and harvested 18 hours later. Proliferation was measured as cpm incorporated.

Figure 6. DCs and B cells from arthritic mice reduce T_R suppression.

Dendritic cells (A) and B cells (B) were purified from NOD and K/BxN arthritic mice using Miltenyi microbeads. T_E and T_R were sorted from Foxp3^gfp mice. T_E (1×10⁵/well) were cultured with either 1×10⁴ DCs (A) or 2×10⁵ B cells (B) and 5 μg/ml αCD3. T_R were added to T_E cultures at the indicated ratios. Proliferation was quantitated at 72 hours by ³H-thymidine incorporation. Proliferation by T_E to APCs were 6874±752 cpm (NOD DC), 17455±788 cpm (K/BxN DC), 4059±298 cpm (NOD B cells), and 8094±1442 cpm(K/BxN B cells). Results represent mean of triplicate wells ± SD. Data is representative of 2 experiments. C) T_E and T_R were cultured at a 4:1 ratio and stimulated by αCD3 and irradiated splenocytes. DCs and B cells were purified from NOD or arthritic mice and activated by platebound αCD40 for 1 hour and washed. Activated APCs were added to the T cell cultures either directly or separated by a transwell membrane. In some T cell cultures, conditioned supernatant from αCD40 activated DC and B cell 48 hour cultures were added to a 1:1 vol/vol. BrdU incorporation was assayed at 72 hours. Results represent mean of duplicate wells ± SD. D) T_E (1×10⁵/well) ± T_R (5×10⁴/well) were stimulated by αCD3 and DC (1×10⁴/well) from either NOD or K/BxN arthritic mice in the presence of neutralizing αIL-6 antibodies (10 μg/ml) or TNFR:Fc (10μg/ml). Proliferation was quantitated at 72 hours by ³H-thymidine incorporation. Data represent mean of triplicate wells ± SD. Data is representative of 2 experiments.

Transwell suppression assays

DCs or B cells from NOD or arthritic mice were activated with plated bound αCD40 (5μg/ml, 1C10, R&D) for 1 hour and washed extensively before adding to T cells. T_E (5×10⁵), T_R (1.25×10⁵) and irradiated splenocytes (5×10⁵) were added in 0.4 μm transwell inserts or co-cultured with αCD40 activated DCs (5×10⁴) or B cells (5×10⁵) in 1 ml ISC-10. T cells were activated by αCD3 (5 μg/ml, 2C11). After 72 hours, 1 mM BrdU was added to each well and BrdU incorporation was assessed by flow cytometry after 12 hours.

Statistical analysis

Wilcoxon Rank Sum test and Student T test were calculated using InStat 2.00.

Results

Thymic Selection of FoxP3⁺ T_R in K/BxN^gfp mice

We have used K/BxN^gfp mice to examine the development and function of FoxP3⁺ T_R during arthritis. In these mice, T_R suppressive activity resided specifically in the FoxP3⁺ (GFP⁺) population, regardless of CD25 expression (19). Therefore we defined T_R as CD4⁺FoxP3⁺ T cells and T_E as CD4⁺FoxP3⁻ T cells.

FoxP3 expression in thymocytes was analyzed in 4 week old KRN^gfp and K/BxN^gfp mice to minimize effects of chronic inflammation upon thymocyte development. In KRN^gfp mice, CD4⁺ SP thymocytes comprised 1–3% of total thymocytes (Figure 1A). Negative selection by endogenous GPI resulted in a 3 fold decrease in thymic cellularity (21.5±2.37 ×10⁷ in KRN^gfp and 8.3±1.75×10⁷ in K/BxN^gfp mice), resulting primarily from losses in the DP subset (Figure 1A). In agreement with our prior observation, T_R were enriched 2–3 fold among CD4 SP thymocytes in K/BxN^gfp mice compared to KRN^gfp mice (Figure 1B,C) (14). To examine the effect of KRN TCR expression on FoxP3 selection, we analyzed the expression of the transgene encoded TCR Vβ6 because there is no clonotypic antibody to the KRN TCR. In agreement with results from other transgenic TCR systems, recognition of its cognate peptide appears to be a pre-requisition for KRN T_R selection (21, 22). In KRN^gfp mice, which lacked GPI-I-A^g7 complexes, Vβ6^bright KRN T cells were only seen in T_E, suggesting that T_R required endogenous TCR expression (resulting in lower Vβ6 levels) for selection (Figure 1B). Conversely, in K/BxN^gfp mice, negative selection by endogenous GPI resulted in preferential deletion of Vβ6^bright T_E. The remaining T_E expressed equivalently lower levels of Vβ6 as T_R. Statistically higher numbers of CD4⁺ SP FoxP3⁺ thymocytes were found in K/BxN^gfp mice (Figure 1D). The enrichment of T_R in K/BxN^gfp mice resulted from both selective deletion of FoxP3⁻ CD4 thymocytes, especially Vβ6^bright T cells, and induction of FoxP3⁺ thymocytes.

Figure 1. Increased FoxP3⁺ thymocytes selection in K/BxN^gfp mice.

A. Thymocytes from 4 week old KRN^gfp and K/BxN^gfp mice were labeled with CD4-APC and CD8-PE and analyzed by flow cytometry. Live cells were gated by forward and side scatter and analyzed by CD4 and CD8. Numbers indicated the % in each subset. B. Histograms of FoxP3 (top) and Vβ6 (bottom) expression in gated CD4⁺ SP thymocytes. C and D. The percentage (C) and numbers (D) of FoxP3⁺ thymocytes seen in CD4⁺ SP thymocytes in KRN^gfp mice (circles, n=8) and K/BxN^gfp mice (squares, n=10). Each dot represented an individual mouse and lines indicated means. P values were calculated by Wilcoxon Rank Sum test.

Development of KRN T_R during arthritis in K/BxN^gfp mice

The frequencies of T_R were similarly increased in the spleen and peripheral lymph nodes (LN). While GFP expression correlated largely with CD25 expression in KRN^gfp mice, 10–15% of CD25⁺ were Foxp3⁻ and were activated T_E (19). This fraction of CD25⁺FoxP3⁻ T_E was doubled in K/BxN^gfp mice with significant overlap with FoxP3⁺T_R, highlighting the power of FoxP3-GFP reporter in this autoimmune model (Figure 2A). To show that FoxP3⁺ T cells function as T_R, sorted CD4⁺FoxP3⁺ T cells were tested in a suppression assay using CD4⁺FoxP3⁻ T cells as responder T_E. T_R did not proliferate to NOD dendritic cells (DCs) presenting either endogenous GPI or KRN superagonist G7m (13). However, they, nevertheless, suppressed the proliferation of T_E when co-cultured in a 1:1 ratio (Figure 2B).

Figure 2. Site specific expansion of FoxP3⁺T_R during arthritis.

A. Splenocytes from 4 week old KRN^gfp and K/BxN^gfp mice were labeled with CD4-APC and CD25-PE and analyzed by flow cytometry. Live cells were analyzed by CD4 and CD25 (top panels). Lower panels showed the CD25 and FoxP3 expression in gated CD4⁺ splenocytes. B. T_R (CD4⁺FoxP3⁺) and T_E (CD4⁺FoxP3⁻) were sorted and stimulated with irradiated NOD ± 1 μM G7m peptide in separately or in a co-culture at a 1:1 ratio. Results are mean ±SD. C. Percentages of FoxP3⁺ CD4⁺ T_R in spleens and lymph nodes from non-arthritic (n=8), acutely arthritic (n=10), and chronically arthritic mice (n=6). Results are presented as means ± SD. P values were calculated based on the Wilcoxon Rank sum. For spleen cells, p<.0001 between non-arthritic and acute mice, and p=.0003 between non-arthritic and chronic mice, and NS between acute and chronic mice. For dLN, p=.02 between non-arthritic and acute mice, and p=.0037 between non-arthritic and chronic, and NS between acute and chronic mice. In both acute and chronic mice, P<.005 for comparisons between each of the different organs. There was no statistically significant difference in distribution of T_R in the various organs in non-arthritic mice. D. Total numbers of FoxP3⁺ CD4⁺ T_R cells in spleens and lymph nodes. Results are presented as means ± SD. P <0.05 between non-arthritic spleen and their counterparts in both acute and chronically arthritic mice, and between non-arthritic dLN and chronically arthritic dLN. Within the non-arthritic group, there was no statistically significant difference in the T_R numbers in various organs. Within the acutely arthritic mice group, p<0.05 for differences between spleen and dLN, ndLn, and mLN. Within the chronically arthritic group, T_R numbers in the spleen and dLN were significantly different from ndLN and mLN. E. Total numbers of CD4⁺FoxP3⁻ T_E in spleen and lymph nodes. Results are presented as means ± SD. F. Lymphocytes were isolated from inflamed paws and analyzed by flow cytometry for CD4-PE and FoxP3-gfp. Live lymphocytes were identified by forward and side scatter, as well as PI exclusion. Numbers indicated % of cells within each quadrant and the number in parenthesis indicated % of CD4⁺ T cells. Data are representative of 5 non-arthritic, 6 acutely arthritic, and 6 chronically arthritic mice.

To determine T_R expansion during arthritis, we quantified the frequency and numbers of T_R in the spleen, joint draining LN (dLN), non-joint draining LN (ndLN), and mesenteric LN (mLN) during different phases of arthritis: non arthritic KRN^gfp control, acutely arthritic (4–6 weeks) and chronically arthritic (8–10 weeks) K/BxN^gfp mice. In all organs surveyed, the frequencies of T_R were increased in a hierarchical manner with higher increases in the spleens and joint dLN relative to ndLN and mLN in K/BxN^gfp mice compared to KRN^gfp mice (Figure 2C). When we quantified the total numbers of T_R in arthritic mice, T_R numbers were significantly increased in the spleen of acutely arthritic mice and in the spleen and dLN of chronically arthritic mice compared to non-arthritic mice. Therefore, T_R first expanded in the spleens and spread to the dLN during disease progression (Figure 2D). Comparison of T_E and T_R numbers in the spleen and LN showed that T_R expansion was met by similar expansion by T_E: such that the ratios of T_E:T_R were maintained during arthritis (Figure 2E).

We also examined T_E and T_R in the paws. In non-arthritic mice, <1% were CD4⁺ T cells. Of which, 5.9±3.2% were T_R (Figure 2F). During acute arthritis, CD4⁺ T cells comprised 2.1±0.7% with 58.5±10.9% T_R. As the inflammation waned in the paws, T_R became less frequent comprising 39.9±10.1% of CD4⁺T cells. Our data resonated with findings in human RA showing accumulation of T_R during active inflammation.

T_R experienced greater turnover in vivo than T_E in K/BxN^gfp mice

Because increased numbers of T_R in the spleen and LN during arthritis can arise from increased proliferation or migration, we directly examined T_E and T_R proliferation by in vivo BrdU incorporation. Non arthritic KRN^gfp and acutely arthritic K/BxN^gfp mice were injected ip with 1 mg of BrdU 12 hours prior to sacrifice. The fractions of BrdU⁺ Vβ6⁺CD4⁺FoxP3⁻ (KRN T_E) and Vβ6⁺CD4⁺FoxP3⁺ (KRN T_R) were quantified. In non-arthritic KRN^gfp mice, higher fractions of T_R (~10%) incorporated BrdU compared to T_E (1–3%) (Figure 3A). This tonic proliferation by T_R suggests active surveillance at steady state even in non-lymphopenic mice.

Figure 3. T_R proliferate at higher rates than T_E in vivo.

A. Fraction of BrdU⁺ T_R and T_E in spleens, dLN, and ndLN in KRN^gfp and acutely arthritic K/BxN^gfp mice. Bars represent mean from 4 mice and error bars show SD. Results are representative of 4 separate experiments. B. The numbers of BrdU⁺ T_R in dLN were plotted against the numbers of BrdU⁺ T_E in each mouse. Each dot represents an individual arthritic mouse. The bisecting diagonal line indicates 1:1 correspondence. N=20. C. Bcl2 expression is increased in T_E. Splenic CD4⁺ T_R and T_E from arthritic mice were gated and analyzed for intracellular Bcl-2 expression. Black line=Bcl-2 PE and filled gray= isotype control. Numbers in the gate represent % of cells. D. The frequency of Bcl-2+ and MFI of Bcl-2 expression of T_E (filled circles) and T_R (filled diamonds) in dLN from 4 arthritic mice are depicted. Bars represent mean and p values are calculated by Fisher t-tests. Experiment is representative of two independent experiments. E. The percentage of Annexin V+ T_E (filled circles) and T_R (filled diamonds) in dLN from 5 arthritic mice are represented with the bar denoting mean. P values are calculated by Fisher t-tests. Experiment is representative of 2 independent experiments.

In arthritic K/BxN^gfp mice, recognition of GPI increased the fraction of cycling T_E and T_R by three fold resulting in 5–10% of T_E and 30–35% T_R incorporating BrdU. Consistently higher fractions of cycling T_R were observed in the LN (Figure 3A) and accounted for the greater increase in T cell numbers seen in the LN as the disease progressed from acute to chronic arthritis (Figure 2 D).

To directly compare the numbers of cycling T_E vs T_R, we plotted BrdU⁺T_E vs BrdU⁺T_R in dLNs from arthritic mice (Figure 3B). Similar results were seen in the spleens and other LN (data not shown). With few exceptions, more T_R incorporated BrdU than T_E. In some mice, BrdU⁺T_R were 3 fold higher than BrdU⁺T_E. We, however, did not find any correlation between our gross measurements of paw inflammation with numbers of BrdU⁺ T_R or T_E.

Given the higher rate of T_R proliferation, the fractions of T_R 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, we found that T_R from arthritic mice expressed significantly less anti-apoptotic factor Bcl-2 (Figure 3C,D) and higher levels of apoptosis (Annexin V⁺) than T_E ex vivo (Figure 3E).

FoxP3⁺T_R proliferate and die with different kinetics than FoxP3⁻ T_E

To test our hypothesis that T_R proliferated and died at a faster rate than T_E, we directly examined the kinetics of T_R and T_E proliferation and attrition in vitro. KRN^gfp CD4⁺ T cells were purified from pooled spleens and LN and cultured with NOD DCs. This approach allowed us to specifically examine intrinsic differences between T_R and T_E separate from the inflammatory milieu in arthritic mice. We used naïve 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. Contrary to the in vitro anergy noted when T_R were cultured alone, T_R showed significant proliferation when cultured in physiologic ratio with T_E and stimulated with physiologic amounts of GPI. By day 1, 20% of FoxP3⁺ T_R have entered cell cycle and incorporated BrdU compared to <2% of FoxP3⁻ T_E. BrdU incorporation by T_R peaked at day 2 and declined over days 3 and 4 (Figure 4A). In contrast, T_E did not proliferate to any appreciable degree in the first 24 hours of culture but proliferated vigorously in the subsequent days. Our data showed that T_E and T_R proliferated with distinct kinetics and magnitude. T_R proliferated earlier but the response was short lived. T_E proliferation, though delayed, was more sustained.

Figure 4. Kinetics of T_R and T_E proliferation and cell death upon antigen stimulation.

A. Purified CD4⁺ T cells from KRN^gfp mice were cultured with NOD DCs and 1 mM BrdU was added at varying times. BrdU incorporation was analyzed 12 hours later by flow cytometry. Time indicates time of cell analysis. Results are presented as means of duplicate wells ± SD. Experiment is representative of 3 experiments. B. Purified CD4⁺ T cells from KRN^gfp and arthritic K/BxN^gfp mice were cultured with NOD DCs. At various times, apoptotic and dead CD4⁺ T cells were analyzed by 7-AAD and Annexin V respectively. Results are mean of duplicate wells. SD were <15%. Experiment has been performed 3 times with similar results.

In a parallel experiment, cell death and apoptosis of KRN^gfp T cells were analyzed using 7-AAD and Annexin V to identify dead and apoptotic cells respectively. As shown in Figure 4B, the fractions of dead and apoptotic T_R were roughly twice that of T_E cells observed during this time course. Interestingly, the time course of T_R apoptosis correlated with T_R proliferation: both peaked at day 2 and declined at days 3 and 4. This finding contrasted with the kinetics of T_E proliferation and apoptosis, and provided support that Bcl2 expression in T_E conferred protection from apoptosis.

To determine if prior activation conferred protection from apoptosis, T_E and T_R from arthritic K/BxN^gfp mice were also analyzed (Figure 4C). Naive and activated T_E displayed similar rates of cell death. In comparison to naïve T_R, T_R from K/BxN^gfp mice displayed less apoptosis and death. However, relative to their T_E counterparts, both naïve and activated T_R still undergo higher apoptotic rates. Together the data confirmed our hypothesis that increased T_R proliferation was offset by increased T_R apoptosis, resulting in T_E:T_R homeostasis.

T_R become increasingly suppressive during arthritis

To interrogate T_R function during arthritis, we sorted T_E and T_R from acutely and chronically arthritic mice. Additionally, to determine if there were site specific effects, T_E and T_R from spleen and dLNs were analyzed separately. We first compared proliferation of T_E from the spleen and LN during acute and chronic disease. LN T_E proliferated better to GPI compared to splenic T_E but there was no statistically significant difference in the proliferation of T_E between acute and chronic disease (Figure 5A).

Figure 5. Increased T_R activity during arthritis is countered by increased T_E resistance to suppression.

A. T_E (CD4 ⁺FoxP3⁻) were sorted from spleens or pooled LN from acutely or chronically arthritic mice. T_E (1×10⁵/well) were stimulated with irradiated NOD DCs and cell proliferation was assayed by ³H-thymidine incorporation during the last 16 hour in a 72 hour culture. Data denotes mean of triplicates wells ± SD. B. T_R from different sites and disease state varied in their suppressive activity against T_E. Data denotes mean of triplicates wells ± SD. C and D. CD4⁺FoxP3⁻CD45RB^lo(memory) and CD4⁺FoxP3⁻CD45RB^hi (naïve) T_E from arthritic mice were cultured with varying ratios of T_R with NOD DCs. Data represents mean cpm of triplicate wells ± SD (C) or normalized as a fraction of max cpm of T_E alone (D). P values were <.02 for CD4⁺FoxP3⁻CD45RB^hi vs CD4⁺FoxP3⁻CD45RB^lo at each of the T_E:T_R ratios in (C) and other P values for D are shown. N.S. not significant. Data is representative of three similar experiments.

We next examined the ability of T_R to suppress T_E proliferation. Because T_E from chronically arthritic mice would be targets for T_R immunotherapy, we used LN T_E from chronic mice as responders. As shown in Figure 5B, T_R from different sites and disease states suppressed T_E proliferation with differing efficacy. While all T_R were effective at a 1:1 ratio, splenic T_R from acutely arthritic mice were the least suppressive. Conversely, LN T_R from chronically arthritic mice were the most effective. Thus, T_R activity increased during arthritis and at sites of active inflammation.

To determine if the T_E activation state influences its susceptibility to suppression, we sorted FoxP3⁻CD45RB^hi (naïve) T_E, FoxP3⁻CD45RB^lo (memory) T_E, and FoxP3⁺ T_R from pooled splenic and dLN T cells from arthritic mice. Naïve CD45RB^hi T_E cells were less proliferative and more easily suppressed than experienced CD45RB^lo T_E (Figure 5C). To determine if this was due to differences in the magnitude in T cell response, we normalized the response to the proliferation in the absence of T_R. At T_E:T_R ratios of 2:1 and 4:1, T_R suppression was two fold less effective with CD45RB^lo T_E (Figure 5D). Thus, during arthritis, both T_E and T_R enhanced their response to GPI. Such that increased T_R suppression is met by increased T_E resistance, resulting in an immunologic détente.

Inflammatory APCs contribute to Resistance to T_R Suppression

Pro-inflammatory cytokines produced by monocytes and DC can abrogate T_R function by increasing T_R apoptosis or by increasing T_E resistance (9, 23, 24). In K/BxN mice, DCs and B cells comprised two major APC for the activation of KRN T cells (20). DCs were the most efficacious at stimulating KRN T cells while B cells were the most abundant. We therefore compared the efficacy of T_R suppression in the presence of DC and B cells from chronically arthritic K/BxN mice or control NOD mice. We have previously shown that APCs from arthritic K/BxN mice showed greater presentation of GPI peptide compared to non-arthritic APCs (20). To eliminate the role of GPI dose, we used polyclonal T cells from FoxP3^gfp mice on the H-2^k background in which the APCs would provide a strong allostimulation in addition to αCD3. As expected, DCs from arthritic mice attenuated the suppressive effect of T_R (Figure 6A). With NOD APCs, T_R suppressed T_E proliferation by 50% at a T_E:T_R 8:1. Four fold higher numbers of T_R were necessary in the presence of arthritic DCs. Arthritic B cells similarly inhibited the suppressive effect of T_R (Figure 6B).

To determine the mechanism by which DC and B cell inhibit T_R suppression, we added αCD40-activated DC and B cells to T_E:T_R suppression assays either in the same wells or separated by transwell membranes. Proliferation of T_E was assayed by BrdU incorporation by flow cytometry. As shown in Figure 6C, αCD40-activated DCs and B cells abrogated T_R suppression only when they are co-cultured with the T_E and T_R. Restoration of T_E proliferation was absent when DCs and B cells were separated by transwell membranes or when cultured supernatants from DC and B cell cultures were added. The finding that inflammatory DCs can reverse T_R suppression has been described (9, 24). However, these reports attributed abrogation of T_R activity to soluble factors such as TNFα and IL-6. Here, we found that neutralization IL-6 or TNFα had no effect on T_R suppressive activity (Figure 6D). Similar results were seen with B cells as APCs (data not shown). Together, our data demonstrated that both arthritic DCs and B cells decrease T_R activity in a cell-cell contact or close contact dependent mechanism. The finding that arthritic B cells can also inhibit T_R suppression provides a potential mechanism for the efficacy of B cell depletion in controlling RA.

Discussion

There is considerable interest in harnessing CD4⁺CD25⁺ T_R to control autoimmune diseases (25, 26). In RA patients, CD4⁺CD25⁺ T_R are increased in inflamed joints and showed variable activity against their T_E counterparts. Results differed among studies in part due to differences in the criteria for T_R, disease states, and patient population. It is also not clear what factors control their tropism, function, growth, and death during disease. We have used K/BxN^gfp mice to model the development and function of arthritogenic T_E and T_R in a well characterized murine model of RA. K/BxN mice displayed many advantages relative to human RA patients. The progression of arthritis has been extensively characterized and follows a predictable course: K/BxN mice develop overt joint inflammation at 4–5 weeks, peaking at 7–8 weeks, followed by a chronic phase in which joint destruction and deformity predominates (27). Additionally, the arthritis shows regional involvement with the distal joints more affected than proximal and axial. This feature allowed us to determine if there is regional expansion of T_R at sites of disease. Moreover, by examining GPI-reactivity, we focus specifically on the arthritogenic KRN T_E and T_R in this system. Lastly, we have used FoxP3^gfp to unequivocally identify T_R.

These experiments describe for the first time the expansion and function of antigen specific T_R during the initiation and progression of spontaneous arthritis. Here, we confirmed our prior findings that incomplete deletion of KRN thymocytes by endogenous GPI resulted in increased frequency of GPI-specific T_R in K/BxN^gfp mice through selective deletion of Vβ6^bright FoxP3⁻ T_E and induction of FoxP3⁺T_R in the thymus (14). In contrast to other models of autoimmune disease in which T_R deficits resulted in pathology, K/BxN mice displayed enhanced populations of GPI-specific T_R (14, 15, 28–30). These T_R proliferated quite well both in vitro and in vivo in response to endogenously presented GPI, were functional, and suppressed T_E proliferation in vitro. T_R numbers expanded in parallel with T_E during arthritis in the spleens and lymph nodes. In K/BxN mice with an ubiquitous self-antigen, both T_E and T_R responses initiated in the spleen with subsequent extension to the lymph nodes with preference for the dLN, indicating selective tropism to inflamed joints from increased inflammation and/or GPI presentation during disease. However, this enhanced activity was offset by increased resistance by GPI-specific T_E and APCs to suppression. A recent study described similar dynamic changes in the spatio-temporal expansion of T_E and T_R in the draining LNs during the induction and progression of adjuvant arthritis that presumably reflected changes in the concentration of the inciting antigen (31). Their findings complemented our studies here with the exception that their waxing/waning T_R course likely reflected the monophasic course in adjuvant arthritis compared to the progressive expansion of T_E and T_R in the K/BxN model.

Despite the enrichment in T_R activity, K/BxN mice nevertheless developed arthritis. Our in vitro studies demonstrate effective T_E suppression at ratios that approximate the physiologic ratio found in vivo. Why then are T_R ineffective in vivo? There are several possible explanations. First, T_R and T_E are differentially regulated during ontogeny. FoxP3⁻ CD4 SP thymocytes reached mature levels during the first week of life. In contrast, FoxP3⁺ CD4⁺ SP thymocytes development was substantially delayed and did not achieve mature levels until 21 days of age (32). The delayed appearance of T_R in K/BxN mice would result in unopposed T_E reactivity at the initiation of arthritis. Second, the precursor frequency of GPI specific T cells is supra-physiologic in K/BxN mice. The self antigen is also ubiquitously expressed and can be presented by all APCs, including B cells. Under such conditions, autoreactive B cells can be driven into memory B cells despite abundant T_R (33). Thus for a humorally driven arthritis, T_R may be less effective than for a T cell driven disease like colitis. We propose that the developmental regulation and numerical superiority of KRN T_E overcome T_R suppression at the initiation of disease. T_R might simply arrive too late and in too few numbers.

During the ensuing arthritis, T_R are highly activated with >30% of T_R proliferating in a 12 hour period. However, this vigorous T_R activity is attenuated by three factors: T_R undergo shorter proliferative bursts and higher apoptosis rates than T_E upon antigen stimulation, T_E become more refractory to T_R suppression as they differentiate into memory T cells, and APCs from arthritic mice abrogate T_R suppression. This inhibition of T_R suppression required close contact but it is unknown whether this is due to increased resistance of APCs to T_R or enhanced antigen presentation to T_E. It is also unclear whether these are due to changes in the co-stimulatory molecules or secretion of paracrine cytokines in activated APCs from arthritic mice. Current studies are on-going to dissect these mechanisms. Our findings parallel observations in human RA and provide an explanation for the paradoxical occurrence of disease in the presence of abundant T_R. By elucidating factors and mechanisms that augment T_R activity, it may be possible to achieve disease control. K/BxN^gfp mice thereby provide a valuable model to examine T_E:T_R homeostasis during disease and therapy.