CD4+CD25+FOXP3+ Treg cells require TCR engagement for suppressive function, thus ensuring that suppression occurs only in the presence of specific antigens; however, to date no studies have addressed the function of self-antigen-specific Treg in humans. These studies were designed to determine whether peripheral generation and function of islet antigen-specific adaptive Treg are defective in human subjects with type 1 diabetes (T1D). Islet antigen-specific adaptive Treg were induced in vitro by activation of CD4+FOXP3− T cells with glutamic acid decarboxylase and islet-specific glucose-6-phosphate catalytic subunit-related protein peptides in the context of T1D-associated HLA-DRβ alleles. Antigen-specific Treg were characterized using flow cytometry for FOXP3 and class II tetramer and assessed for the ability to inhibit proliferation. These adaptive Treg were then compared with influenza-specific Treg from the same study population. The function of tetramer+ cells that expressed FOXP3 was similar for both influenza and islet antigens generated from control and T1D subjects. In fact, the potency of suppression correlated with FOXP3 expression, not antigen specificity. Thus, these data suggest that development of functional adaptive Treg can occur in response to islet antigens and activation of islet-specific Treg may potentially be used as a targeted immunotherapy in T1D.
CD4+CD25+FOXP3+ T regulatory cells are vital for immune regulation. The lack of Treg results in severe autoimmunity in both mouse and man 1–3. The importance of Treg in type 1 diabetes (T1D) has been demonstrated in murine models by the acceleration of disease upon depletion of Treg as well as the cure of diabetes by transfer of Treg to animals prior to and after the onset of disease 4–7. Furthermore, adoptive transfer is more effective when the transferred Treg are specific for an islet antigen 8. In humans, defects in polyclonal Treg have been proposed as one mechanism by which individuals develop T1D and this defect appears to be in the function 9–11, as compared with the number of Treg 12. The specificity of Treg in animal models and the role of Treg in regulating disease raise the possibility that islet-specific Treg are inadequate in number or function in individuals with T1D.
Although Treg were initially believed to originate only from the thymus 13, subsequent studies in mouse and man have shown that CD4+CD25+FOXP3+ Treg can develop in the periphery under a variety of conditions and these cells are referred to as adaptive Treg 14, 15. Other types of adaptive Treg have been described including Tr1 and Th3 cells; however, these adaptive Treg are not associated with FOXP3 expression. For this study we focus on CD4+CD25+FOXP3+ adaptive Treg and use the term adaptive Treg to refer to these cells in this paper. Adoptive transfer of polyclonal adaptive Treg can prevent development of and treat autoimmunity as evidenced in several animal models 16–18. Moreover, these adoptively transferred murine adaptive Treg are long-lived and maintain FoxP3 expression, suggesting that adaptive Treg may be used as immunotherapeutic agents to treat autoimmune and inflammatory disease.
The Treg of thymic origin are termed natural Treg and in mice are known to react with self-antigens 19. Similarly, in mice induction of adaptive self-antigen-specific Treg may occur in vivo in the periphery under appropriate conditions and these antigen-specific adaptive Treg protect mice from developing autoimmunity 20, 21. In humans, induction of Treg responses has been observed in vivo where highly differentiated human CD4+CD45RO+CD25hiFOXP3+ T cells arose upon foreign antigen exposure 22. Thus far, induction of adaptive self-antigen-specific Treg has not been studied in humans. It is well established that self-reactive T cells reside in the peripheral blood of healthy individuals 23, 24, yet the cellular and molecular response to self-antigen activation in these populations is less well understood, particularly due to the possibility that this population contains low-affinity T cells 25, 26. Thus, it is important to address whether induction of a population of human self-reactive adaptive Treg occurs in humans in order to determine whether adaptive Treg may participate in the regulation of autoreactivity in vivo, whether adaptive Treg are defective in individuals with autoimmunity and whether self-antigen-specific adaptive Treg may be used as a therapeutic agent.
To date, neither the function of Treg specific to islet antigens nor the ability to generate self-antigen-specific adaptive Treg has been examined in individuals with T1D. In these studies we measured induction of adaptive antigen-specific Treg by comparing the number and function of Treg generated following stimulation with well-defined viral antigens and the representative islet autoantigens glutamic acid decarboxylase (GAD) and islet-specific glucose-6-phosphate catalytic subunit-related protein (IGRP). Antigen-specific T cells were identified by the use of HLA class II tetramers (Tmr) and regulatory cells were identified by the use of intracellular FOXP3 staining and functional assays. We establish that Treg specific for the islet antigens, IGRP and GAD, can be induced from peripheral CD4+CD25−FOXP3− T cells of healthy individuals and that these self-antigen-specific adaptive Treg function in a manner similar to those specific to the foreign antigen, influenza. When these same studies were performed with T1D subjects, we found a similar capacity to generate functional influenza and islet-specific adaptive Treg and that irrespective of the disease state, the potency of suppression correlated with FOXP3 expression.
Islet-specific CD4+CD25+FOXP3+ adaptive Treg can be generated from CD4+CD25− T cells of healthy controls
To determine whether adaptive Treg specific to self-antigens can be generated in the periphery in humans, we compared generation of foreign antigen-specific influenza reactive Treg with that of islet-specific Treg in a group of healthy subjects. We examined specificity in the context of three T1D-associated HLA class II alleles, HLA-DRβ1*0401, HLA-DRβ1*0404 and HLA-DRβ1*0301 27. A set of islet and influenza peptides was used to probe TCR specificity in the HLA class II Tmr assays based on HLA type of the subjects (Supporting Information Table 1). These peptides have been previously identified as important epitopes in healthy and T1D subjects 28–31. Low-affinity, self-antigen-specific T cells are not found at a high enough frequency to be detected directly in peripheral blood using HLA class II Tmr. Since both detection of CD4+ antigen-specific T cells by Tmr and induction of FOXP3 expression require in vitro activation, we used an in vitro induction protocol for these studies (Fig. 1A). CD4+CD25− T cells were isolated from peripheral blood and then activated with irradiated APC plus peptide in media containing human serum. Cells were kept in culture, receiving IL-2 (200 IU/mL) on day 7, but were otherwise left undisturbed. After 14 days, the CD4+ T cells were characterized for expression of FOXP3 and the ability to bind an HLA class II Tmr complexed to the peptide antigen with which these cells were activated. In these cultures, the FOXP3+ cells were present in the CD25brightCD127lo T-cell population (Fig. 1B and C). Among the Tmr+ populations both FOXP3+ and FOXP3− cells were observed (Fig. 1D). FOXP3 expression was also found among a subset of Tmr− cells. Previous studies have shown that this population of cells, Tmr−CD25+ cells, is not specific for the antigen used to stimulate the culture 32.
To confirm that Tmr+ adaptive Treg detected in our induction cultures arose from a CD25−FOXP3− population, two distinct populations of CD4+ T cells defined by CD25 and CD127 expression levels were sorted and activated with peptide antigen. These populations included CD25−CD127hi CD4+ T cells and CD25+CD127lo CD4+ T cells, which are enriched for FOXP3 expression. Following selection and activation with HA or GAD peptide, FOXP3 expression was induced in the CD25− T-cell populations regardless of the antigenic stimulation (Supporting Information Fig. 1). Consistent with the observation that FOXP3+ T cells are relatively anergic, T cells in the CD25+CD127lo population did not proliferate enough in response to peptide stimulation to detect HA or GAD-specific T cells by Tmr staining on day 14 but these cultures did contain a high frequency of FOXP3+ T cells. CD25hi T cells sorted at the end of culture of CD25−CD127hiCD4+ T cells and CD25+CD127lo CD4+ T cells were both capable of suppressing proliferation of autologous T cells (Supporting Information Fig. 1).
Persistent expression of FOXP3 is associated with regulatory function 33 while transient expression of FOXP3 is associated with activation as demonstrated by co-expression of IFN-γ 34, 35. To determine whether we were measuring persistent expression of FOXP3, we analyzed the expression of FOXP3 over time. Between days 10 and 14 of our induction culture, the frequency of FOXP3 within the Tmr+ population was constant (Fig. 1E). We also found that although a subpopulation of influenza-specific Tmr+ cells secreted IFN-γ when re-stimulated with HLA-matched Tmr or anti-CD3 antibody, the FOXP3+ cells did not secrete IFN-γ (Fig. 1F). These data suggest that FOXP3 is a marker of these adaptive Treg using this culture method.
Using this induction culture, we first compared the percentage of Tmr+ cells that co-expressed FOXP3 following in vitro activation with the influenza and islet peptides. Tmr+ cells ranged in frequency from 0.39 to 6.63 (median 1.26). When analyzing induction of antigen-specific Tmr+FOXP3+ cells, we found a wide variation in the percentage of FOXP3+ cells within the influenza and islet-specific Tmr+ populations but no difference between each antigen type studied (Fig. 2A). Variation in these cultures may be due to differences in the frequency of activated cells; however, we found no correlation between the frequency of Tmr+ cells and the percentage of FOXP3+ cells (data not shown). Alternatively, variation in these induction cultures may be due to differences between individuals. Thus, we compared the frequency of FOXP3+ cells in influenza and islet-specific T cells in individuals who reproducibly have Tmr+ T cells to both antigen types. As shown in Fig. 2B, the frequency of FOXP3+ cells within the Tmr+ populations were similar for both antigen types and, in each individual tested, variation was associated with different individuals not the antigen used for stimulation. Taken together, our data demonstrate that antigen-specific Tmr+ cells that persistently express FOXP3 can be generated from CD25−FOXP3− T cells of control subjects in response to antigenic stimulation by either foreign or self-peptides.
The potency of influenza and islet-specific CD4+CD25+FOXP3+ adaptive Treg is similar in healthy controls
We have previously shown that HLA 0401 HA TMR+CD25+ T cells isolated from induction cultures are suppressive 36. These cells require antigen-specific activation, but are capable of suppressing bystander proliferation. We utilized the same assay to address the question of whether adaptive Treg specific for self-antigen are also capable of suppressing CD4+ T-cell proliferation. Islet-specific Tmr+CD25+ T cells were isolated from induction cultures on day 14, co-incubated with autologous CD4+CD25− T cells and activated with tetanus toxoid (TT) alone or in combination with the antigen for which the CD25+Tmr+ cells were specific. These sorted GAD and IGRP Tmr+CD25+ cells were capable of suppressing CD4+CD25− T-cell proliferation in an antigen-specific manner. This is evidenced by the fact that proliferation of CD4+CD25− T cells was not suppressed when cultures containing CD4+CD25− and sorted cells were stimulated with TT alone; however, when activated with GAD or IGRP and TT, sorted Tmr+CD25+ T cells were capable of suppressing TT-specific proliferation of CD4+CD25− T cells in a bystander manner (Fig. 3A). To measure the relative potency of islet-specific Treg we utilized foreign and self-antigen-specific adaptive Treg generated from the same individual. The potency of suppression was similar for both antigen specificities over a range of CD25+Tmr+ and CD4+CD25− T-cell ratios (Fig. 3B). We used HLA 0401 GAD Tmr+CD25+ sorted T cells to determine whether islet-specific T cells require cytokine or cell contact for function. Several mechanisms of suppression have been described for FOXP3+ nTreg, adaptive Tr1 and Th3 cells, and adaptive FOXP3+ Treg in vitro including secretion of suppressive cytokines. Similar to HA-specific adaptive Treg 36, GAD-specific adaptive Treg suppress proliferation in an IL-10- and TGF-β-independent (Fig. 3C) and contact-dependent manner (Fig. 3D) when tested in vitro. Thus, influenza and islet antigen-specific Tmr+FOXP3+ adaptive Treg generated from healthy controls display similar potency.
Functions of islet-specific adaptive Treg are similar to influenza-specific adaptive Treg in T1D subjects
To address the question of whether islet-specific Treg are capable of being generated in T1D subjects, we activated CD4+CD25− T cells with islet-specific peptide as described in Fig. 1A and found that FOXP3 expression could be induced in GAD- and IGRP-specific Tmr+ populations generated from CD4+CD25− T cells of T1D subjects (Fig. 4A). This indicates that a unique defect in the induction of islet-specific FOXP3+ cells was not present in subjects with T1D.
We next evaluated the function of islet-specific adaptive Treg generated from T1D subjects. We found that both GAD and IGRP Tmr+CD25+ T cells generated from T1D subjects were capable of suppressing CD4+CD25− T-cell proliferation in response to TT in a bystander manner when the autoantigen was present but not with TT stimulation alone (Fig. 4B). To evaluate whether the function of islet-specific adaptive Treg is impaired in T1D, we compared the level of suppression conferred by adaptive Treg specific for influenza and islet antigens. As shown with the healthy controls (Fig. 2B), we noted a broad range in the number of FOXP3+ cells in the CD25+Tmr+ population. On the basis of this observation and previous work in our laboratory that demonstrates a strong correlation between the percentage of FOXP3+ cells and suppression 37, we analyzed suppression with respect to the percentage of FOXP3+ cells among the sorted Tmr+CD25+ population. Regardless of antigen specificity, we found a strong linear and statistically significant (p<0.0001) relationship between the percent FOXP3+ cells and the degree of suppression. Including antigen specificity in the multivariant analysis did not significantly influence this relationship, suggesting that the function of islet-specific adaptive Treg do not differ from influenza-specific adaptive Treg generated from T1D subjects (Fig. 4C). Thus, we conclude that a specific defect in the induction or function of islet-specific adaptive Treg is not present in T1D subjects under these culture conditions.
Functions of influenza and islet antigen-specific adaptive Treg do not differ between control and T1D subjects
To determine whether self-antigen-specific Tmr+ adaptive Treg are defective in T1D subjects, we directly compared the frequency and function of islet-specific adaptive Treg generated from CD4+CD25− T cells of control and T1D subjects and found no differences (Fig. 5A and B). However, we did observe large variability in the frequency of FOXP3 in the Tmr+ population; therefore, we performed further analysis taking the frequency of FOXP3 in the sorted Tmr+ population into account. For these further analyses, we compared the potency of suppression of foreign and self-antigen-specific adaptive Treg generated from T1D and control subjects, using nested linear models. As shown in Fig. 4C, we found a strong linear and statistically significant (p<0.0001) relationship between the percent FOXP3+ cells and the percent of Tmr+CD25+ sorted cells with an associated R2 value of 0.0696. We determined that including disease status and antigen specificity did not significantly improve model fit (p=0.5882, F-test). Specifically, neither antigen specificity (Fig. 5C) nor disease state (Fig. 5D) significantly altered the model fit. These data indicate that islet-specific Treg can be induced from CD4+CD25− T cells of subjects with T1D and the function of these cells appear to be similar to that seen in healthy subjects.
Determining the role played by Treg in the development and progression of disease and the therapeutic potential of Treg are key unresolved questions in T1D. Previous studies of Treg in T1D subjects have focused on the number and function of the polyclonal CD4+CD25bright Treg found in peripheral blood 9–12. The issue of antigen specificity has not been addressed – particularly the question of whether islet-specific adaptive Treg are impaired in T1D subjects. Here, we establish that CD4+CD25+FOXP3+ Treg can be generated in response to GAD- and IGRP islet-specific self-antigens in the context of three different HLA class II molecules. The potency of islet-specific Treg is similar to that of foreign antigen-specific adaptive Treg and do not differ between controls and T1D subjects. Thus, CD4+CD25−FOXP3− T cells of individuals with T1D have the capacity to become functional islet-specific adaptive Treg. Although the in vitro studies described in this paper cannot determine whether there is an impairment of islet-specific adaptive Treg generation in vivo, the ability to induce functional GAD- and IGRP-specific Treg in vitro does raise the possibility that enhancement of this process may be feasible for therapeutic purposes.
Our data provide the first evidence that functional islet antigen-specific CD25+FOXP3+ adaptive Treg can be generated from human peripheral CD4+CD25− T cells. Studies in mice have suggested that specificity to self-antigens is a unique characteristic of natural Treg 8. Moreover, natural Treg isolated directly from the blood are believed to be high-avidity T cells that have escaped negative selection in the thymus 38. In this paradigm, the peripheral generation of Treg would regulate immune responses to high-affinity foreign antigens, not low-affinity self-antigens. In humans, although it is well established that self-reactive T cells reside in the peripheral blood of healthy individuals 23, 24 there are distinct differences between T cells specific for self as compared with foreign antigens, particularly in TCR affinity 25, 26. Despite these differences in affinity, we demonstrate that GAD- and IGRP-specific adaptive Treg can arise from the same CD4+CD25− population as seen with adaptive Treg specific to higher affinity influenza antigens. Moreover, the level of expression of Tmr, as measured by mean fluorescence intensity, was similar in the FOXP3+ and FOXP3− populations regardless of the antigen used to stimulate the culture (data not shown). These findings are consistent with the recent observation in mice that the TCR repertoire of FOXP3+ and FOXP3− T cells is overlapping and includes T cells specific for both foreign and self-antigens 39.
We find that the expression of FOXP3 in the antigen-specific adaptive Treg population correlates with suppression. Similarly, there is a direct link between stable FOXP3 expression with suppressive function 1, 33, 40, 41. However, for peripherally generated Treg, correlation of FOXP3 expression and function remains controversial. Using non-specific activation, others have observed that induction of FOXP3 is transient and depends on the level of activation 33–35, 41. Suppressive function depends on the frequency of cells expressing FOXP3 and the assay used to measure inhibition. We find that stimulation of CD4+CD25− T cells with influenza or islet antigen induces a subpopulation of cells with stable, persistent high expression levels of FOXP3 following long-term culture (14 days) which contains functionally active Treg. The persistent FOXP3 expression and lack of correlation with IFN-γ secretion as well as the suppressive function of these cells in vitro indicate that FOXP3 is a marker for Tmr+CD25hi adaptive Treg in our culture system. In fact, here we show that the frequency of Tmr+ T cells expressing FOXP3 directly correlates with suppressive potency regardless of the specificity of the adaptive Treg.
Unlike several studies, which found the function of Treg isolated from T1D subjects to be impaired 9–11, we found that adaptive GAD- and IGRP-specific Treg generated from CD4+FOXP3− T cells of T1D subjects and controls function comparably. The in vitro assays used in this paper differ from those used in the analysis of T1D polyclonal Treg in that APC and peptide are used to stimulate CD4+CD25− T cells and Treg, not anti-CD3 antibodies. It has previously been suggested that differences in signal strength alone contribute to differences in functional studies of Treg isolated from T1D subjects 10. Antigenic stimulation likely results in a lower level of TCR stimulation than anti-CD3 antibody stimulation. Alternatively, natural Treg isolated from T1D subjects may differ from adaptive Treg in their source and maturation, resulting in functional differences 42.
The efficacy of preventing or ameliorating autoimmunity through adoptive cellular therapy using ex vivo activated and expanded Treg has been well demonstrated in mice not only with natural Treg, 7, 8, 43 but also with adaptive Treg 16–18. Several challenges remain in translating these findings into human therapies, including the frequency and availability of antigen-specific T cells. Here, we demonstrate that fully functional islet-specific Tmr+FOXP3+ cells can be induced from human peripheral blood T cells. While challenges still remain in identifying and isolating pure populations of islet-specific FOXP3+ adaptive Treg, these data provide conceptual support for the potential development of a strategy for producing clinically useful numbers of functional self-antigen-specific FOXP3+ T cells.
Materials and methods
PBMC were derived from subjects (n=27) participating in studies under the auspices of the BRI-JDRF Center for Translational Research registry. Informed consent was obtained from all subjects according to IRB approved protocols at Benaroya Research Institute and Children's Hospital and Regional Medical Center, Seattle. Control participants (n=19) were selected based on lack of personal or family history of autoimmunity or asthma. All subjects included in this study demonstrated a Tmr+ response to at least one of the peptides used in these studies (Supporting Information Table 1).
Antibodies and reagents
Antibodies used included FITC CD25, PE CD25 and CD127, PerCP CD4, and APC CD4 and CD25 (BD Pharmingen), AlexaFlour® 647 FOXP3 (clone 259D) and matching isotype control (BioLegend), purified anti-CD3 (UCHT1) and anti-CD28 (CD28.2)(BD Bioscience).
HLA-DRβ1*0401 Tmr loaded with influenza HA antigen (306–318), islet-specific GAD (555–567I), or glucose-6-phosphate catalytic subunit-related protein – IGRP (247–259) peptides, HLA-DRβ1*0404 Tmr loaded with influenza matrix protein (60–73) or GAD (555–567) peptides, and HLA-DRβ1*0301 Tmr loaded with influenza non-structural protein (32–45) or IGRP (226–238) peptides were produced as described previously 24, 32.
Cell preparation and phenotyping
Fresh PBMC were prepared by centrifugation over Ficoll-Hyplaque gradients. CD4+ T cells were purified with a CD4+ no-touch T-cell isolation kit (Miltenyi) followed by negative selection with Miltenyi CD25 microbeads. APC were obtained from the positive fraction of the CD4+ no-touch selection. FOXP3 expression in CD4+CD25− cells was 0.1–1.2%. For some experiments, enrichment was performed by sorting based on CD25 and CD127 expression.
CD4+CD25− T cells were activated with peptide in the presence of irradiated (5000 rad) APC at a 1:2 ratio with 6×106 total cells/well in a 24-well plate. IL-2 (Chiron) (200 IU/mL) was added at day 7. Cells were cultured for 14 days and stained for expression of Tmr as described previously 32. CD25 and FOXP3 staining was performed according to the manufacturer's instructions (BioLegend). When assaying co-expression of FOXP3 and IFN-γ secretion, cells were activated overnight with matching Tmr or irrelevant Tmr as a control. For non-specific activation cells were stimulated for 6 h with 5 μg/mL of αCD3 plus irradiated APC in the presence of 1 nM Brefeldin A and stained using the BioLegend FOXP3 staining protocol. FOXP3 isotype in conjunction with CD25 and CD127 expression on activated T cells were used as FOXP3 staining controls. Tmr staining controls utilized Tmr of the matched HLA class II type bound to an irrelevant peptide. Positive Tmr staining was determined to be responses at least 4-fold greater than control Tmr stains. An example of these staining controls is shown in Supporting Information Fig. 2.
Sorted Tmr+CD25hi cells (2.5×104), thawed CD4+CD25− cells (2.5×104) isolated from autologous PBMC, or both were incubated with irradiated APC (2.5×104), TT and 5 μg/mL of peptide antigen in a 96-well round-bottom plate as described previously 36. An aliquot of 1 μCi [3H]thymidine was added during the final 16 h of a 6–7 day assay and proliferation was measured by a scintillation counter. Cytokine blocking experiments were performed by adding 10 μg/mL of anti IL-10 or TGF-β to cultures in a 96-well plate. For transwell experiments, cells were cultured with or without a 4 μm transwell separating CD4+CD25+ (2.5×105 cells/well) cells from CD4+CD25− (2.5×105/well) in a 48-well plate. All culture conditions were performed in triplicate.
For analysis of experiments comparing a single variable, statistical significance was analyzed using an independent students t-test, unless otherwise noted, requiring a p value of <0.05 for the data to be significantly different. For experiments involving multiple variables comparing percentage FOXP3 of the sorted Tmr+ population and percentage suppression, regression analysis was performed and statistical significance between experimental groups was determined by measuring the improvement in goodness of fit, as defined by the residual sum of squares across nested linear models using F-tests.
The authors thank K. Arugamanathan for cell sorting. The authors acknowledge the investigators and staff of the BRI-JDRF Center for Translational Research at Children's Hospital and Regional Medical Center, Seattle and the BRI Diabetes Clinical Research Program for subject recruitment as well as the BRI Translational Research Clinical Core for sample processing and handling. This work was supported by grants from the NIH (DK07245), JDRF (The Center for Translational Research at BRI), NCRR (M01-RR-00037) and NCRR (1UL1RR025014).
Conflict of interest: The authors declare no financial or commercial conflict of interest.