Regulatory T cells (Tregs) use a distinct TCR repertoire and are more self-reactive compared with conventional T cells. However, the extent to which TCR affinity regulates the function of self-reactive Tregs is largely unknown. In this study, we used a two-TCR model to assess the role of TCR affinity in Treg function during autoimmunity. We observed that high- and low-affinity Tregs were recruited to the pancreas and contributed to protection from autoimmune diabetes. Interestingly, high-affinity cells preferentially upregulated the TCR-dependent Treg functional mediators IL-10, TIGIT, GITR, and CTLA4, whereas low-affinity cells displayed increased transcripts for Areg and Ebi3, suggesting distinct functional profiles. The results of this study suggest mechanistically distinct and potentially nonredundant roles for high- and low-affinity Tregs in controlling autoimmunity.
Foxp3+ regulatory T cells (Tregs) are critical for maintaining immune homeostasis and preventing the development of tissue-specific autoimmunity. Mutations in FOXP3 result in immunodysregulation polyendocrinopathy enteropathy X-linked syndrome, leading to multiorgan autoimmunity, including a high prevalence of type 1 diabetes (T1D) (1). Likewise, deletion of Foxp3+ T cells in TCR-transgenic or retrogenic (Rg) mice specific for β cell Ags leads to accelerated diabetes (2, 3). Consequently, Treg-centric immunotherapies have been vigorously pursued for prevention or treatment of T1D. However, it remains unclear whether boosting overall Treg numbers will be sufficient or whether therapeutic approaches will need to focus on a subpopulation of functional Tregs. It is largely accepted that Tregs develop in response to stronger TCR signals and are presumed to exhibit an overall higher degree of self-reactivity compared with conventional T cells (4–6). Moreover, recent work has shown that continuous TCR signaling is necessary for optimal Treg function (7, 8). Although it is tempting to assume that high-affinity T cells are generally more functional, emerging literature suggests an equal and important role for low-affinity effector T cells (Teffs) in responses against pathogens, in autoimmunity, and in tumor surveillance (9–11). However, studies addressing the role of low-affinity Tregs in immune homeostasis have not been performed; thus, it remains unclear whether TCR affinity is correlated with Treg recruitment, accumulation, and function in autoimmunity.
In our previous analysis of mice expressing eight TCRs with variable affinity for the immunodominant insulin epitope B:9-23, deletion of Tregs in mice expressing higher-affinity TCRs resulted in accelerated autoimmune diabetes, whereas the rate of disease was unaffected by Treg depletion in mice harboring lower-affinity TCRs (3). Therefore, we hypothesized that low-affinity Tregs might not be functional in autoimmune diabetes. However, because Teffs and Tregs possessed the same TCR in single-TCR Rg mice, it remains unclear whether higher-affinity Tregs were more functional or whether low-affinity Teffs were resistant to suppression. To directly compare high- and low-affinity Tregs in vivo, we used a mixed TCR Rg bone marrow (BM) chimera model. In this competitive setting, we were able to assess the relative accumulation and capacity of high- and low-affinity Tregs to control the same population of Teffs.
Materials and Methods
NOD/ShiLtJ, NOD.B6-Ptprcb (NOD.CD45.2), NOD.CB17-Prkdcscid/J (NOD.scid), NOD.129P2(C)-Tcrαtm1Mjo/DoiJ (NOD.TCRα−/−), and NOD.Cg-Foxp3sf/DoiJ (NOD.scurfy) mice were obtained directly from The Jackson Laboratory and maintained at Baylor College of Medicine. NOD.scurfy mice were crossed with NOD.scid (wild-type [wt]) mice at our facility. All mice were housed in specific pathogen–free conditions. The studies were approved by the Baylor College of Medicine Institutional Animal Care and Use Committee.
Generation of two-TCR Rg mice
Two-TCR Rg mice were generated as previously described (12). Briefly, BM was harvested from NOD.scid and NOD.scid.scurfy (scurfy) mice, transduced with retroviral TCR vectors expressing a GFP or Ametrine fluorescent reporter, and transferred into recipient NOD.TCRα−/− mice (Supplemental Fig. 1G, 1I). Mice were monitored for diabetes development or were analyzed 5–6.5 wk post-BM transfer, at which point T cell reconstitution was assessed (Supplemental Fig. 1H, 1J). For some experiments, NOD.CD45.2 BM was added at 10% of the total cell number prior to injection.
Assessment of diabetes
Diabetes incidence was monitored weekly with Diastix and confirmed with a Breeze2 glucometer (both from Bayer. Elkhart, IN). Mice were considered diabetic if their blood glucose was >400 mg/dl.
Isolation of pancreatic islets
Pancreata were digested with Collagenase IV (Worthington, Lakewood, NJ), and single islets were isolated for further analysis, as previously described (3).
Flow cytometry and Abs
Flow cytometry analyses were performed on an LSR Fortessa II (BD Biosciences), and data were analyzed with FlowJo software (TreeStar). mAbs against the following molecules were used: Foxp3 (FJK-16s), Vβ12 (MR11-1), and TIGIT (GIGD7) (all from eBioscience); CD5 (53-7.3), Ki67 (B56), and Vβ11 (RR3-15) (all from BD Biosciences); and CD3 (145-2C11), CD4 (GK1.5), CD25 (PC61), CTLA-4 (UC10-4B9), CD8 (53-6.7), GITR (YGITR 765), Vβ2 (B20.6), and IL-10 (JES5-16E3) (all from BioLegend).
Tregs were sorted from pancreatic islets and spleens of wt/wt two-TCR Rg mice based on an Ametrine or GFP TCR fluorescent reporter and CD4+CD3+GITR+CD25+ gating strategy (Supplemental Fig. 2G). Samples were sorted with an average purity of 92.6% Foxp3+ for 4-8 and 92.5% Foxp3+ for 12-4.4m1 Tregs. cDNA was synthesized using a SMARTer Ultra Low Input RNA Kit (Clontech). Library preparation was performed with an Illumina Nextera XT kit before paired-end RNA sequencing using an Illumina NextSeq 500 platform for 150 cycles (NextSeq 500 Mid Output Kit). Sequencing reads were aligned to the mm10 genome using TopHat Alignment (13, 14), and gene expression was quantified by FPKM. Cufflinks Assembly and DE were used to compute differential expression (q < 0.05) between groups, with the Benjamini–Hochberg correction for multiple testing. Heat maps and principal component analysis were generated in R (version 3.2.3) using heatmap from the gplots package (version 2.17.0) with viridis (version 0.4.0) and ggbiplot (15). The sequences presented in this article have been submitted to the National Center for Biotechnology Gene Expression Omnibus under accession number GSE106467 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE106467).
Diabetes incidence was subjected to the log-rank Mantel–Cox test. Group comparisons were performed using a two-tailed Mann–Whitney nonparametric test (Figs. 2, 3) and a Wilcoxon matched-pairs test (Fig. 4). Mean ± SEM are shown. Statistical analyses were performed using Prism (GraphPad, La Jolla, CA).
Results and Discussion
High- and low-affinity Tregs cooperate to control autoimmune diabetes
Although some studies suggest that Treg development can be supported by TCRs with a wide range of affinities for self-antigens (16), there appears to be a positive correlation between TCR affinity and Treg development (4, 6). To determine whether TCR affinity for self governs Treg function in autoimmunity, we generated two-TCR Rg mice with mixed BM from NOD.scid (wt) and scurfy mice. In this system, we transduced wt or scurfy BM with low- or high-affinity TCR and mixed the two BMs at an equal ratio prior to injection into NOD.TCRα−/− recipients (Fig. 1A). Scurfy mice carry a missense mutation in the Foxp3 gene, resulting in the complete absence of functional Tregs (17). Therefore, this mixed BM chimera allowed us to limit Treg development to the TCR that was expressed on the NOD.scid background, whereas the Teff population was derived from NOD.scid and scurfy BMs (Fig. 1B, Supplemental Fig. 1A).
We chose to study two pairs of high- and low-affinity TCRs, which were selected based on their ability to support efficient Treg development. The TCRs were stratified into “high” and “low” affinity based on their biophysical two-dimensional affinities, functional responses to wt insulin (InsB:9-23) and agonist InsB:9-23(R22E) peptides, and insulin tetramer staining (Supplemental Fig. 1B–E) (3).
We first generated two-TCR Rg mice expressing a combination of the high and the low InsB:9-23–reactive TCRs: 4-8 and 12-4.4m1 (Supplemental Fig. 1G, 1H). The two TCRs were chosen because, when expressed individually, they yield similar frequencies of islet-infiltrating Foxp3+ Tregs (12 and 9%, respectively) and have similar compositions of helioshi or thymically derived Tregs (Supplemental Fig. 1F). Importantly, the two TCRs lead to significantly different disease patterns (3). In single-TCR Rg mice, the 4-8 TCR results in spontaneous diabetes development in ∼60% of mice. In contrast, low affinity 12-4.4m1 TCR mice are free from diabetes, despite T cell infiltration of the pancreas (3). The lack of disease in 12-4.4m1 TCR mice could not be solely explained by the presence of Tregs, because Treg ablation in Foxp3DTR or in scurfy 12-4.4m1 TCR mice did not lead to diabetes development (data not shown). In contrast, Treg depletion in 4-8 TCR mice resulted in significant acceleration of diabetes (3). Therefore, we hypothesized that, unlike the high-affinity 4-8 Tregs, the 12-4.4m1 low-affinity Tregs lack sufficient levels of TCR signaling to regulate pathogenic T cells, such as 4-8 Teffs.
In the 4-8 and 12-4.4m1 NOD.scid mixed BM chimeras, in which both TCRs gave rise to Tregs, mice were partially protected from diabetes, with only 40% developing disease by 20 wk post-BM transfer (Fig. 1C, black line). However, when both TCRs were expressed on the scurfy background, we observed an accelerated disease course, with 100% penetrance (Fig. 1C, green line). Surprisingly, the absence of low- or high-affinity Tregs resulted in similar acceleration of disease, with ∼70% of mice developing diabetes (Fig. 1C, blue and red lines), indicating that both Treg populations were critical for protection and acted in a cooperative manner.
To confirm our observations with a different set of InsB:9-23–specific TCRs, we used a combination of the high-affinity TCR 1-10 and low-affinity 8-1.1, both of which displayed similar frequencies of Tregs when expressed individually (9 and 8%, respectively), and both were pathogenic (Supplemental Fig. 1I, 1J) (3). As with the first pair of TCRs, we observed similar levels of helios expression, suggesting equal distribution of thymically and peripherally derived cells within the two Treg populations (Supplemental Fig. 1F). Importantly, as observed with the first set of TCRs, disease onset was accelerated when both or either of the two Treg populations were absent (Fig. 1D). Taken together, these data indicate that high- and low-affinity Tregs can contribute to regulation of autoimmunity.
Treg frequencies are regulated by TCR-intrinsic mechanisms
Because elimination of either Treg population accelerated diabetes development comparable to the scurfy/scurfy group with no functional Tregs, we considered the possibility that the absence of one Treg population might have a negative effect on the survival and homeostasis of the remaining Tregs in the increasingly inflammatory environment. Therefore, we assessed the frequencies of Foxp3+ Tregs in the spleens and pancreatic islets of 4-8/12-4.4m1 two-TCR chimeras. Indeed, we observed an overall decrease in Treg frequencies in wt/scurfy chimeras; however, in general, the decrease in Treg frequencies and numbers did not exceed 50% loss (Fig. 2A, Supplemental Fig. 2A). Intra-TCR analysis confirmed that the absence of either Treg population had minimal effect on the frequencies or numbers of the other population (Fig. 2B, Supplemental Fig. 2B). This suggests that Treg frequencies are likely determined by the strength of TCR signaling during thymic development and tonic TCR signaling in the periphery, and are not affected by the size or composition of the Treg compartment. Overall, these data indicate that a net Treg/Teff ratio, rather than Treg TCR affinity, is more important for controlling the pathogenic Teff population in autoimmunity.
Tregs have been known to take on specialized functional characteristics appropriate for suppression of specific Th subsets (18, 19). We considered that low- and high-affinity Tregs might be specialized for controlling high- or low-affinity effector populations. Therefore, we asked whether deletion of high- or low-affinity Tregs resulted in preferential expansion of 4-8 or 12-4.4m1 effectors. Contrary to our expectations, we observed a relative increase in low-affinity 12-4.4m1 Teffs when either Treg population was removed, indicating perhaps that low-affinity Teffs were generally more susceptible to regulation by either Treg population (Fig. 2C).
Low-affinity Tregs are competitive in a polyclonal environment
Although high- and low-affinity Tregs infiltrated the pancreas, high-affinity TCRs supported an overall greater frequency of Tregs in the periphery (Fig. 3A). Moreover, when we analyzed the relative contribution of high- and low-affinity Tregs to the whole Treg population by first gating on Foxp3+ T cells and then separating high- and low-affinity cells based on Vβ expression, the relative proportion of high-affinity 4-8 Tregs was increased at the site of inflammation (Fig. 3B). To examine whether the increase in high-affinity Tregs is reflected in their increased proliferation, we compared the expression of Ki67, a marker of cell cycle, between the two populations. Only the high-affinity Tregs exhibited signs of activation and proliferation in the draining pancreatic lymph nodes (PLNs), the site of initial Ag exposure, based on the increase in Ki67+ cells (Fig. 3C). However, once in the islets, both Treg populations reached similar high levels of proliferation. This observation suggested that competition for Ag between the two Treg populations was limited to the site of initial Ag exposure: draining PLNs.
Because we observed unequal Treg expansion in the draining lymph nodes, potentially driven by local competition for Ag, we needed to determine whether there was a similar competition within the single-islet microenvironment that was obscured by pooling the islets for analysis. We considered the possibility that low-affinity Tregs preferentially accumulated and expanded in the islets with lower numbers of high-affinity cells, an environment with reduced competition for Ag and IL-2. To this end, we analyzed single pancreatic islets and observed the consistent presence of both Treg populations within the same microenvironment (Supplemental Fig. 2C). Of 27 islets analyzed from seven mice, both Treg populations were detected in 24 islets (88.9%). The observed coexistence of high- and low-affinity Tregs in individual islet microenvironments suggested that low-affinity Tregs are competitive at the site of inflammation. Next, we considered the possibility that, in a limited two-TCR system, low-affinity Tregs had an artificial advantage and would be less competitive in a polyclonal environment. Alternatively, a diverse repertoire could result in reduced competition for Ag, thus expanding the insulin-reactive Treg-developmental niche, potentially favoring low-affinity T cells (5, 20). Therefore, we cotransferred two-TCR Rg BM with congenically marked NOD.CD45.2 polyclonal BM cells (Supplemental Fig. 2D). By 5 wk posttransfer, the subpopulation of insulin-specific Rg T cells accumulated at the site of Ag in the draining lymph nodes and pancreatic islets (3.8% in spleen versus 9.2% in PLNs and 10.4% in the islets) (Supplemental Fig. 2D). The expansion of the antigenic niche resulted in a preferential increase in low-affinity Tregs. Compared with the lymphopenic environment, the presence of polyclonal T cells resulted in a 1.59-fold increase in low-affinity Tregs in PLNs (from 3.7 to 5.9%) and a 1.74-fold increase in pancreatic islets (from 2.6 to 4.6%) (Fig. 3A, Supplemental Fig. 2E), although the relative proportion of high-affinity 4-8 and low-affinity 12-4.4m1 Tregs was largely unchanged by the expansion of the antigenic niche (Supplemental Fig. 2F). These data suggest that, in a polyclonal setting, Tregs with low TCR affinity can successfully compete for the developmental niche and accumulate in the inflammatory tissue.
High- and low-affinity insulin-specific Tregs are poised to use distinct suppressive functions
To elucidate the suppressive mechanisms used by high- and low-affinity Tregs in the tissue site, we assessed the transcriptional landscape of 4-8 (high-affinity TCR) and 12-4.4m1 (low-affinity TCR) Tregs isolated from spleens and pancreatic islets. Principal component analysis showed tight clustering of the two Treg populations in the spleens, whereas the islet-infiltrating Treg populations were more variable in their genetic profile and significantly distinct from the spleen (Fig. 4A). Further analysis revealed similar expression of Treg functional genes, including Foxp3, Tnfrsf18 (GITR), Ctla4, and Tgfb1 (Fig. 4B) (21). Interestingly, several genes associated with Treg suppressive function had distinct expression in either high- or low-affinity Tregs. High-affinity Tregs preferentially expressed Il10, Gzmb, Lag3, and Tigit, all previously described to be important for Treg suppression of Th1 responses, including autoimmune diabetes (22–25). In contrast, low-affinity Tregs exhibited significantly higher levels of Areg and Ebi3 (a subunit of heterodimeric IL-35) transcripts, which are known to be important for tissue repair, Treg survival, and suppression of autoimmune responses (26–29). To confirm the RNA sequencing results, we assessed protein expression of GITR, CTLA-4, TIGIT, and IL-10. Interestingly, the slight difference observed in Gitr and Ctla4 transcript expression was significantly enhanced at the protein level (Fig. 4C, 4D, Supplemental Fig. 2H). Although both Treg populations upregulated GITR and CTLA-4 upon entry into the pancreas, high-affinity Tregs displayed enhanced expression of these functional markers. Consistent with the transcriptional analysis, expression of TIGIT and IL-10 was significantly higher in 4-8 high-affinity Tregs (Fig. 4E–G, Supplemental Fig. 2H). Although there was some variability in the frequency of IL-10–expressing Tregs, high-affinity Tregs generally expressed greater levels of IL-10 based on mean fluorescence intensity (Fig. 4F, 4G). Taken together, these data suggest that neither Treg population alone is sufficient to control autoimmune diabetes, and high- and low-affinity Tregs have the potential to use distinct nonredundant suppressive mechanisms for combined effective control of tissue-specific autoimmune responses.
Studies performed in polyclonal and single-TCR systems expressing Treg-derived TCRs revealed that Tregs preferentially express TCRs with higher affinity for self-antigens, and unperturbed TCR signaling is critical for optimal Treg function (30–32). The intensity of TCR signaling during Treg activation dictates the expression levels of several key genes involved in Treg homeostasis and function, including CD25 and CTLA4 (7, 30). Thus, it has been widely accepted that the functional potential of Tregs is directly correlated with their affinity for self. However, there is little direct evidence to show whether TCR affinity has a direct effect on Treg function in autoimmunity. Importantly, we found that the simultaneous presence of high- and low-affinity Tregs was necessary to delay the onset of diabetes (Fig. 1C, 1D). This unexpected observation suggests that, although tissue specificity and affinity are necessary for optimal Treg infiltration (Fig. 3B) (33), TCR affinity for tissue Ag might be important in activating distinct regulatory programs. Upon entry into the site of autoimmune inflammation, both populations increased the expression of Treg functional mediators, suggesting at least a partial, if not equal, contribution of low-affinity Tregs to the regulation of autoimmunity.
Interestingly, in the recipients of haplodeficient BM, disease kinetics and incidence were similar to scurfy BM recipient mice, which were completely devoid of functional Tregs. These results are in contrast to previous studies of polyclonal haplodeficient or insufficient systems in which the remaining Tregs are able to expand and compensate for the deficiency (34). Treg frequencies are regulated by TCR-intrinsic factors and the availability of IL-2 (8, 20, 35). Although in a polyclonal system partial deletion of Tregs presumably relieves IL-2 sources that drive Treg expansion to fill the niche, in our two-TCR system Treg frequencies do not change within each TCR population in response to reductions in the overall Treg compartment (Fig. 2B). Therefore, within the context of fixed Ag availability, intra-TCR Treg frequencies seem to be limited by TCR-intrinsic parameters and remain stable.
Recently, we began to appreciate the heterogeneity of the Foxp3+ Treg population, which often mirrors Teffs in their ability to use a wide range of tissue-specific and context-dependent responses (36, 37). Heterogeneity of Teff responses is primarily regulated at the level of TCR signaling, which dictates the level of activation, as well as instructs the type and relative proportion of helper lineage development (38, 39). It is likely that the range of Treg phenotype and function is similarly dependent on the level of TCR activation. Moreover, regulatory mechanisms used by Tregs are differentially dependent on TCR signaling, and some of these are induced by inflammatory cytokines rather than TCR activation. Perhaps not surprisingly, one of these genes, Areg or amphiregulin (26), was preferentially upregulated in low-affinity insulin-reactive Tregs (Fig. 4B), suggesting that, in the absence of strong TCR signaling, Tregs are more likely to use non-TCR–dependent suppressive functions. In contrast, high-affinity Tregs preferentially upregulated TCR-dependent regulatory molecules, including CTLA-4, TIGIT, and IL-10 (Fig. 4D–G) (7, 30). Given the dramatic disease acceleration in mice devoid of the high- or low-affinity Tregs, it is tempting to postulate that the two Treg populations use distinct regulatory mechanisms, and both are necessary for regulation of autoimmunity (37). An alternative explanation is that the regulation of autoimmunity is highly dependent on the Teff/Treg ratio; once that ratio is compromised, the regulation fails completely. Collectively, our data suggest that functional Tregs span a range of TCR affinities, and high- and low-affinity populations cooperatively prevent autoimmune pathology. These results might have important implications for the development of Treg-based approaches for the monitoring and treatment of autoimmune diseases.
We thank the National Institutes of Health Tetramer Core Facility for providing peptide and MHC monomers, the Baylor College of Medicine Cytometry and Cell Sorting Core for assistance with cell sorting, and Richard (Aaron) Cox for helpful discussions.
This work was supported by National Institutes of Health Grants R01 AI125301-01A1 and P30 DK079638-05, American Diabetes Association Grant 7-14-JF-07, JDRF Grant 1-FAC-2014-243-A-N, National Institutes of Health Immunology Scientist Training Grant T32 AI053831 (to M.L.S.), and The Robert and Janice McNair Foundation. I.S. was the recipient of an American Association of Immunologists Career Development Fellowship.
The sequences presented in this article have been submitted to the National Center for Biotechnology Information Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE106467) under accession number GSE106467.
The online version of this article contains supplemental material.
Abbreviations used in this article:
pancreatic lymph node
type 1 diabetes
effector T cell
regulatory T cell
The authors have no financial conflicts of interest.