No Alterations in the Frequency of FOXP3⁺ Regulatory T-Cells in Type 1 Diabetes

Peripheral blood samples were obtained from 110 individuals, including those with recent-onset (<12 weeks from diagnosis) or established type 1 diabetes, their first-degree relatives, and healthy control subjects (patient demographics outlined in Table 1). Type 1 diabetes was diagnosed after physician examination according to American Diabetes Association criteria (14). Autoantibody assays for the type 1 diabetes–associated autoantibodies anti–GAD antibody (GADA) and anti-IA2 were performed on all study participants, using commercially available immunoassay kits (Kronus, Boise, ID). Healthy control criteria included lack of an autoimmune disorder or relative with type 1 diabetes. Institutional review board–approved informed consents and assents were obtained from all study participants.

Whole blood was collected in K-EDTA S-Monovette tubes (Sarstedt, Newton, NC) and immediately subjected to cellular staining. Whole blood (100 μl) was aliquoted (per tube) along with 20 μl each appropriate test antibody or respective isotype control in four-color tests, fluorescein isothiocyanate (FITC) anti-CD3 (clone HIT3a), phycoerythrin (PE) anti-CD25 (M-A251), PE anti-CD127 (hIL-7R-M21), peridinin chlorophyll protein (PerCP) anti-CD4 (SK3), allophycocyanin (APC)-FOXP3 (clone PCH101), APC-CD45RA (HI100), APC-CD45RO (UCLH1), allophycocyanin (APC)-CD62L (DREG 56), and APC anti-HLA II DR (G46-6 [L243]). The following isotype control antibodies were used: FITC mouse IgG1 (X40), PerCP mouse IgG1 (MOPC-21), PE mouse IgG1 (MOPC-21), APC rat IgG2A (eBR2a), APC mouse IgG2b (27–35), and APC mouse IgG2a (G155-178). All antibodies for cytometric analyses were purchased from BD Biosciences (San Jose, CA), with the exception of FOXP3 (eBioscience, San Diego, CA). After surface staining for 30 min (4°C), erythrocytes were lysed and cells were fixed for 10 min at room temperature (BD FACS Lysing solution) followed by two washes with stain buffer containing 0.2% BSA. Surface-stained cells then underwent methods for intracellular FOXP3 staining using the anti-human FOXP3 staining kit according to manufacturer recommendations. Stained cells were then subjected to flow cytometric analysis using a BD FACSCalibur flow cytometer with 1.5 × 10⁵ cells acquired per test. FCS Express (version 2.200.0023; De Novo Software, Thornhill, ON, Canada) was used for analysis of cytometric data. To determine absolute cell counts, a small volume of blood was collected in potassium EDTA–containing tubes and analyzed for complete blood count (CBC) with differential when available on a Coulter Ac·T diff analyzer (Beckman Coulter, Fullerton, CA). Absolute counts were determined by multiplying the frequency of positive cells determined by cytometric analysis by the number of lymphocytes × 10³/μl as determined by CBC.

Statistical analysis used GraphPad Prizm 4.00 software (GraphPad, San Diego, CA), with Mann-Whitney and Spearman’s correlation analyses. Student’s paired t test was used when assessing the stability of cellular phenotypes. P < 0.05 was deemed significant.

To determine the frequency of Tregs, lymphocytes and CD3⁺ gated cells were analyzed for their expression of CD4, CD25, and FOXP3 (representative plots in Fig. 1A–G). The majority of FOXP3⁺ T-cells co-expressed CD4 and CD25 (Fig. 1C), supporting specificity of the staining procedure. In addition, a strong correlation was observed between the levels of surface CD25 expression and the degree of intracellular FOXP3 observed (Fig. 1G).

We previously reported the frequency of CD4⁺ T-cells expressing low to intermediate levels of CD25 (CD4⁺CD25^LOW; defined by CD25 staining above isotype control and <100 fluorescence intensity units) increases with age, and the frequency of this cell type is independent of type 1 diabetes (10). One question that was difficult to address previously was whether increases in CD25⁺ T-cells were accompanied by increasing frequencies of FOXP3⁺ T-cells. Using the previous commonly used markers of CD3, CD4, and CD25, and now including FOXP3 as a more specific indicator of Tregs (Fig. 2), we have now identified a significant correlation (using all study subjects, n = 110) between the frequency of CD4⁺CD25⁺FOXP3⁻ T-cells (Fig. 2A, top left quadrant) and age (r = 0.69, P < 0.0001). This population corresponds to the CD4⁺CD25^LOW T-cell population previously noted to increase with subject age (Fig. 2C). In contrast, the frequency of CD4⁺CD25⁺FOXP3⁺ T-cells (Fig. 2A, top right quadrant) appears independent of subject age (Fig. 2F; r = 0.12, P = 0.2). As previously observed with the frequency of CD4⁺CD25⁺ T-cells, the influence of age was independent of disease state (data not shown).

To establish the stability of these cellular phenotypes and determine the reproducibility of our testing procedures, we performed a longitudinal analysis on a subset of individuals participating in this study. No significant variations were observed in a period up to 8 months, because the frequency of CD25⁺ and FOXP3⁺ cells (alone or in combination) remained relatively stable (Fig. 3, A–C, Student’s paired t test, NS). These findings, taken together with the relative stability of CD4⁺CD25⁺FOXP3⁻ T-cells over time (Fig. 3B; mean coefficient of variation ± SD 16.6 ± 7.9%), suggest that long-term alterations in T-cell phenotype rather than a brief increase in acutely activated cells underlie the observed age-associated increases in the frequency of CD4⁺CD25⁺ cells and CD4⁺CD25⁺FOXP3⁻ T-cells.

Because lymphopenia has been reported in association with type 1 diabetes of NOD mice (15), we assessed both the frequency and absolute number of FOXP3⁺ T-cells in our study populations. The frequency of Tregs defined by the frequency of total FOXP3⁺ T-cells, CD25⁺FOXP3⁺ T-cells, or the absolute number of FOXP3⁺ T-cells did not significantly differ as a function of type 1 diabetes state or disease risk (Fig. 4A–C). Significant differences were, however, observed in the frequency of CD25⁺FOXP3⁻ T-cells for first-degree relatives (Fig. 4D; 20.5 ± 8.4%) compared with control subjects, new-onset type 1 diabetes, and established type 1 diabetic subjects (16.5 ± 6.3%, P = 0.02; 13.02 ± 4.2%, P = 0.002; and 15.7 ± 5.6%, P = 0.04, respectively). No other groups were significantly different in CD25⁺FOXP3⁻ T-cell frequency (all NS). This facet was most likely related to the influence of subject age on CD25⁺FOXP3⁻ T-cell frequency (Fig. 2E), as opposed to disease state. This patient population contained a large number of parents of affected offspring (see Table 1) and thus contained a larger proportion of older subjects than the other three subject groups. Likewise, neither the duration of type 1 diabetes nor the autoantibody titer associated with the frequency of CD25⁺FOXP3⁺ T-cells (NS). For reasons of limited subject availability, we elected to study a random sampling of first-degree relatives of type 1 diabetic patients independent of their autoantibody status in order to identify potential genetic associations between Treg frequency and disease risk. Although none were identified, future efforts could benefit from analysis of individuals at high (i.e., autoantibody-positive) versus low (i.e., autoantibody-negative) risk for the disease.

In an effort to understand what factors influence the development of CD25⁺FOXP3⁻ T-cells, we sought to monitor additional markers associated with T-cell developmental state and antigen exposure and lymphocyte homing capacity. Therefore, markers of naive (CD45RA) and memory (CD45RO) T-cells and the homing marker CD62L were analyzed in association with CD4 and CD25 expression levels (Fig. 5). Analysis of these markers indicated that CD4⁺CD25⁺-gated T-cells (Fig. 5A, top right quadrant) predominantly express the memory T-cell marker CD45RO (Fig. 5B; 82.3 ± 9.8% for CD45RO for the control group vs. 40.4 ± 14.8% for CD45RA). Conversely, CD4⁺CD25⁻-gated T-cells (Fig. 5A, bottom right quadrant) predominantly express the naive T-cell marker CD45RA (Fig. 5C; 76.3 ± 11.0% for CD45RA vs. 41.3 ± 17.6% for CD45RO). Although some significant differences were observed for the subject groups in CD45RA and CD45RO expression levels (Fig. 5B and C), once again, these results most likely represent differences in cohort age as opposed to disease-associated differences.

A common caution when interpreting studies involving patients with type 1 diabetes is whether age influences the factor(s) under assessment. To determine the influence of age on the frequency of CD45RA⁺, CD45RO⁺, and CD62L⁺ T-cells, we correlated the frequency of these markers, determined by flow cytometry at the time of testing versus subject age. These analyses indicated long-term alterations in T-cell immune phenotype associated with subject age (Fig. 6; n = 110, all subjects analyzed). Specifically, whether assessed on CD4⁺CD25⁺-gated T-cells (Fig. 5A, top right quadrant gate) or CD4⁺CD25⁻ T-cells (Fig. 5A, bottom right quadrant gate), the general trend indicates a reduction in CD45RA expression with age (Fig. 6A and B; r = −0.48, P < 0.0001, and r = −0.54, P < 0.0001, respectively). Assessment of the memory T-cell marker CD45RO exhibited the inverse relationship, with the frequency of CD45RO⁺ T-cells increasing with subject age on both CD4⁺CD25⁺ and CD4⁺CD25⁻ T-cells (Fig. 6C and D; r = 0.68, P < 0.0001, and r = 0.66, P < 0.0001, respectively). In terms of the lymphocyte homing marker CD62L, which has been reported to influence Treg tracking and function (16), a negative correlation was again observed with subject age (Fig. 6E and F; r = −0.24, P = 0.01, and r = −0.51, P < 0.0001 for CD4⁺CD25⁺ and CD4⁺CD25⁻ T-cells, respectively).

To lend support to our abilities to provide staining procedures identifying Tregs, and to verify recent efforts defining markers of Tregs, the cellular distribution of CD127 (the α-chain of the IL-7 receptor) was analyzed along with the markers CD25, FOXP3, and the major histocompatibility complex (MHC) class II molecule DR on CD3⁺ lymphocytes (Fig. 7D–F). This analysis was consistent with the findings of Liu et al. (17) and Seddiki et al. (18), in that FOXP3⁺ cells were CD127⁻ to CD127^LOW, whereas CD25⁺ cells were less restricted in their expression of this marker, with CD127 expressed across a continuum of levels. On the other hand, MHC class II DR expression appears across the continuum of cells expressing CD127, indicating that class II DR expression is not unique to Tregs.

The robust interest in Tregs over the past decade likely results from the prevailing paradigm that autoimmunity results from an eventual imbalance in CD4⁺ T-cells, favoring pathogenic effector T-cell phenotypes over a regulated response (19). To these studies, we add our findings that there does not appear to be a simple deficiency in the peripheral blood frequency or absolute number of FOXP3⁺ T-cells in patients with type 1 diabetes versus healthy control subjects. Inclusion of the Treg lineage marker FOXP3 did, however, identify important age-related influences over CD4⁺ T-cells, which likely confounded previous studies of Tregs that relied primarily on the markers CD4 and CD25. As part of these analyses, our studies identify several key phenotypic properties of CD4⁺ T-cells influenced by subject age. Specifically, we note that CD4⁺CD25⁺FOXP3⁻ but not CD4⁺CD25⁺FOXP3⁺ T-cells appear to increase with age. Furthermore, this increase in CD4⁺CD25⁺FOXP3⁻ T-cells is accompanied by a concomitant increase in CD45RO⁺ memory T-cells and a reduction in CD45RA⁺ and CD62L⁺ T-cells. We speculate that this represents a combination of reduced thymic output with age as well as an accumulation of antigen exposures over time. In line with these findings, a recent report tracking T-cell populations in vivo indicated relatively stable frequencies of CD4⁺CD25^HI cells despite their rapid turnover rate (20). From this observation, Vukmanovic-Stejic et al. (20) suggest that Treg populations are continuously replenished by pre-existing CD4⁺CD25^HI cells themselves or replenished by conversion of CD4⁺CD25⁻ cells into the CD4⁺CD25^HI T-cell pool.

Despite our finding that the frequency of Tregs, defined by expression of the markers CD25 and FOXP3, does not appear to be deficient in type 1 diabetes, several caveats exist before we can definitely state that no differences in Tregs are associated with type 1 diabetes. First, these studies were conducted in peripheral blood and may not reflect the T-cell repertoire present at the site of inflammation (i.e., the pancreas and pancreatic draining lymph node). Furthermore, although the markers of CD25 and FOXP3 represent appropriate markers identifying a lineage of Tregs, additional phenotypic properties responsible for the functional capacity of Tregs may be deficient in the case of type 1 diabetes.

It has been reported that aging is associated with a progressive decline in immune function and speculated that increasing CD4⁺CD25⁺ T-cells may contribute to these defects (21). In terms of alterations in the functional capacity of Tregs with age, Tsaknaridis et al. (22) report declining Treg function with age in humans. However, a report by Sakaguchi and colleagues (23) reported equivalent Treg function in aged mice but noted a relative hyporesponsiveness of CD4⁺CD25⁻ T-cells. These events may not be mutually exclusive. One can envision a situation in which defects in Treg function may also be linked to improper activation of CD4⁺CD25⁻ effector T-cells (Teffs) (24). Teffs represent the primary source of IL-2, and Tregs require IL-2 signaling for proper development and function (25). In the case of type 1 diabetes, we and others have previously reported a hyporesponsiveness of T-cells in NOD mice and type 1 diabetes, characterized by reduced T-cell proliferation and production of IL-2 (10,26–28). Genetic association studies have identified type 1 diabetes susceptibility loci related to IL-2 in NOD mice and in the gene encoding the IL-2Rα (CD25) in humans (29–31). Furthermore, downstream elements associated with the IL-2 signaling pathway, including that of STAT5, also appear to be defective in type 1 diabetes (32–35). Clearly, more research in this area is required.

In line with the notion of a need for additional lineage-specific identifiers, two additional markers, CD127 and HLA class II DR, have recently been reported to identify distinct functional populations of Tregs. The surface marker CD127 (IL-7 receptor α-chain) appears to be expressed at the highest levels on effector T-cells, whereas functional Tregs are CD127^LOW or negative (17,18). HLA II-DR has also been reported to be preferentially expressed on functional Tregs with the greatest suppressive capacity in vitro (36). Our own phenotypic analysis of these markers does in fact identify a population of CD127^HICD25^LOW T-cells that might influence suppression when selecting solely based on CD25 expression (Fig. 7E, bottom right quadrant) and should be subject to further investigation. In addition, the influence of a CD25^LOW population would be more apparent in people of older age (Fig. 2E). These recent advances in the capacity to segregate functional Tregs from Teffs should help to address the current limitations on isolation and analysis of functional human Tregs in studies of type 1 diabetes.

Although the exact mechanisms of suppression used by Tregs remains subject to debate, it appears clear that Tregs require IL-2 signaling and FOXP3 expression for normal development and function (37–39). Previous studies of Tregs in type 1 diabetes (8–10) lack agreement with respect to the question of whether functional differences exist between type 1 diabetic patients and healthy individuals in the capacity of their CD4⁺CD25⁺ T-cells to suppress proliferation of autologous Teff cell proliferation after polyclonal stimulation. One key aspect in need of further investigation is to correlate the degree of suppression with the percentage of FOXP3⁺ cells isolated for use in suppression assays. In a partial yet indirect step toward this question, we have observed that the degree of FOXP3⁺ T-cells detected in peripheral blood does not dictate the degree of in vitro suppression (data not shown). Like efforts should also be addressed by testing for the degree of suppression in association with the aforementioned makers of HLA-DR and CD127.

In summary, these studies provide substantive evidence that altered Treg frequencies are not specifically associated with type 1 diabetes. This finding should not be viewed as eliminating the potential mechanistic impact these cells might have on the pathogenesis of type 1 diabetes. Instead, we believe future efforts should be directed toward the identification of factors capable of influencing the functional activity of Tregs and Teffs in a setting of type 1 diabetes (e.g., CTLA-4 or TGF-β), to elucidate how these cells act at the site of pancreatic inflammation, and to determine how treatments aimed at augmenting the frequency and/or functionality of Tregs (40) modulate the pathogenesis of type 1 diabetes.

These studies were supported, in part, by grants from the National Institutes of Health (AI42288 and AI39250) and by the Keene and Sebastian Family Professorships.

Special thanks are extended to all study subjects who generously participated in this study along with the diabetes research program staff at the University of Florida.