Cutting Edge: Dual TCRα Expression Poses an Autoimmune Hazard by Limiting Regulatory T Cell Generation

Type 1 diabetes (T1D) is an autoimmune disease characterized by immune-mediated destruction of insulin-producing pancreatic β cells. Like all autoimmune diseases, T1D occurs due to failure of immune tolerance mechanisms to prevent or control self-reactive immune responses. In humans and NOD mice, disease is driven by intolerant insulin-specific CD4⁺ T cells (1–5). How insulin-specific CD4⁺ T cells escape tolerance is incompletely understood. We investigated dual TCR expression as a potential mechanism by which pathogenic CD4⁺ T cells escape tolerance.

Allelic exclusion ensures that most T cells only express a single recombined αβ TCR specificity. However, allelic exclusion is imperfect, and an estimated 10–30% of all T cells express two functionally recombined TCRα-chains in mice and humans (6–9). TCRβ allelic exclusion is more stringent, with 1–3% expressing two TCRβ-chains (6). Since the discovery of dual TCR T cells, there has been concern that they might promote autoimmunity (9). This typically relates to their potential to impair negative selection (8, 10–13), although dual TCR expression can also promote positive selection of otherwise unselected specificities (14–17). Many transgenic TCR mouse models support these mechanisms (10, 11, 13–17). However, most TCR transgenic models have unusually high fractions of dual TCR T cells, and earlier-than-normal expression of transgenic TCRs alters T cell development. The impact of dual TCR expression on thymic selection under normal, nonTCR transgenic conditions is unknown. In this study, we explored the contribution of dual TCR expression to spontaneous autoimmune diabetes in NOD mice with a normal polyclonal TCR repertoire, contrasting wild type (WT) NOD mice with NOD mice unable to make dual TCR T cells. To our surprise, we discovered a previously unrecognized impact of dual TCR expression on agonist selection.

TCRα knockout (KO) NOD (NOD.129P2(C)-Tcra^tm1Mjo/DoiJ) mice were from Christophe Benoist and Diane Mathis via The Jackson Laboratory (strain 004444) (18). We generated TCRβ KO NOD mice using speed congenic technology to backcross the Tcrb^tm1Mom allele from C57BL/6 TCRβ KO mice (B6.129P2-Tcrb^tm1Mom/J) onto the NOD background (IDEXX) (19, 20). The presence of known insulin-dependent diabetes loci was verified by PCR (Supplemental Fig. 1A, 1B) and the absence of dual TCR T cells was confirmed by flow cytometry (Supplemental Fig. 1C, 1D). Mice were bred and housed in specific pathogen-free colonies at the University of Minnesota. All animal use was performed per the University of Minnesota Institutional Animal Care and Use Committee–approved protocols (1206A15325 and 1503-32409A).

Female NOD mice were monitored for diabetes weekly after 10 wk of age. Mice were deemed diabetic following two consecutive blood-glucose readings >250 mg/dl (Diastix; Bayer Healthcare).

Pancreata were harvested from 10-wk old prediabetic single TCR T cell and WT NOD female mice and prepared as described (21). Insulitis was determined manually using an insulitis scoring scale: 0 = no insulitis; 1 = peri-insulitis; 2 = mild insulitis; 3 = moderate insulitis; and 4 = severe insulitis. Islet-infiltrating CD3⁺, CD4⁺_, and Foxp3⁺ T cell enumeration was determined using Imaris 8 software (Bitplane) where regulatory T (T_reg) cells were surface positive for CD4 or CD3 and nucleus positive for Foxp3.

For programmed death-1 (PD-1) blockade experiments, 10-wk old single TCR T cell and WT NOD female mice were treated i.p. with 250 μg anti-PD-1 (clone J43; Bio X Cell, West Lebanon, NH) on day 0 and with 200 μg on days 2, 4, 6, 8, and 10 (22). For transient depletion of CD25 positive cells, a single dose of 500 μg anti-CD25 (clone PC-61.5.3; Bio X Cell) was administered i.p. between 3.5 and 4 wk of age (23).

All MHC class II:peptide tetramers used in these experiments were generated at our institution as previously described (21). P8G and P8E tetramers were originally designed to have either glycine-glutamic acid (SHLVEALYLVCGGEG) or glutamic acid–glutamic acid (SHLVEALYLVCGEEG) at positions B:21-22 of the native InsB_9-23 sequence (SHLVEALYLVCGERG) (24).

Thymus, pancreas-draining lymph nodes (PLN), and pancreata were harvested from prediabetic 10-wk old mice. Pancreata were dissociated using the Miltenyi Biotec Tumor Dissociation Kit (product 130-096-730) and the Gentle MACs dissociator followed by debris separation with Debris Removal Solution (product 130-109-398; Miltenyi Biotec). Cell characterization was determined by surface and intracellular cell marker expression by flow cytometry. Cell viability was determined with Live/Dead-Aqua viability dye (BioLegend). Analysis was performed on a BD LSR Fortessa flow cytometer (BD Biosciences). Data were analyzed with FlowJo Software version 8 (Tree Star).

Statistical significance was determined using an unpaired Student t test or Mann–Whitney U test (pancreas flow cytometry data). Differences in diabetes incidence were determined using log-rank (Mantel-Cox) tests. Data were analyzed using Prism 5 (GraphPad Software).

We generated NOD mice with targeted deletions at TCRα, TCRβ, or both loci. We then bred female NOD mice to be hemizygous at TCRα (TCRα^+/−), TCRβ (TCRβ^+/−), or both loci (TCRα^+/− β^+/−, hereafter single TCR T cell mice). We evaluated the incidence of spontaneous diabetes in each group, using WT NOD mice as controls (Fig. 1A). Single TCR T cell and TCRα^+/− NOD mice both displayed diabetes resistance, whereas TCRβ^+/− and WT NOD mice developed diabetes similarly (Fig. 1A). These data indicate that resistance to diabetes is associated with hemizygosity at the TCRα, but not the TCRβ locus. Therefore, the effects of TCRβ hemizygosity that we have previously demonstrated, cannot be implicated (19). Two prior investigations of diabetes in TCRα^+/− NOD mice by Elliott and Altmann initially reported protection from diabetes, but their follow-up study found no difference between TCRα^+/− and WT NOD mice (25, 26). Interpretation of these studies is complicated by incomplete backcrossing and cyclophosphamide induction of disease in the first study as well as nonspecific pathogen free conditions and very low spontaneous diabetes incidence in the second; these concerns do not apply to our study (see 2Materials and Methods and Supplemental Fig. 1A). We next evaluated pancreatic islet infiltration in 10-wk old female WT and single TCR T cell NOD mice. Islet infiltration was present in single TCR T cell NOD mice, though much less severe than in WT NOD mice (Fig. 1B). Together these data demonstrate that dual TCRα expression promotes increased islet infiltration and diabetes.

Long-term tolerance to diabetes has been demonstrated to depend on the PD-1 signaling pathway (22, 27, 28). To determine whether diabetes resistance among single TCR T cell NOD mice depends on this pathway, we administered anti-PD-1 blocking Ab to 10-wk old female WT or single TCR T cell NOD mice. Blocking PD-1 induced diabetes in five of seven WT NOD mice, but only 3 of 10 single TCR T cell NOD mice (Supplemental Fig. 2A). We also observed no difference in PD-1 expression in bulk CD4⁺ or insulin-specific T cells derived from the PLN (Supplemental Fig. 2B). Further characterization of the bulk and insulin-specific PLN CD4⁺ T cells revealed an equivalent T cell Ag experience (as measured by CD44 expression) and equivalent FR4 and CD73 expression (markers of anergy) (Supplemental Fig. 2C, 2D) (29). These data demonstrate that insulin-specific CD4⁺ T cells encounter Ag and are chronically stimulated equivalently in both groups of mice. Taken together, these findings suggest that diabetes resistance in single TCR T cell NOD mice does not depend on the PD-1 pathway and is not associated with increased anergy or Ag sequestration. We therefore explored other potential mechanisms of tolerance.

Regulatory T (T_reg) cells are essential mediators of peripheral immune tolerance and protection from T1D in humans and mice (30–35). We asked whether T_reg cells are responsible for diabetes protection in single TCR T cell mice using early ablation of CD25-expressing cells (32). Anti-CD25 treatment induced diabetes in all the WT NOD mice by 20 wk of age. Importantly, five of seven single TCR T cell NOD mice also developed diabetes by 20 wk of age; in sharp contrast to untreated (1/10) and anti-PD-1 treated (3/10) single TCR T cell NOD mice (Fig. 1A, 1C, Supplemental Fig. 2A). These findings demonstrate 1) that single TCR T cell NOD mice can develop diabetes; and 2) that T_reg cells help to maintain tolerance. We then focused on how dual TCR expression affects Ag-specific T_reg cell populations.

We characterized T cell lineage commitment in the PLNs of 10-wk old single TCR T cell and WT NOD mice by flow cytometry. Both groups had similar numbers of bulk, insulin-specific, and foreign-specific (HEL_11-25) CD4⁺ T cells, indicating resistance to diabetes is not a result of a smaller insulin-specific T cell population (Fig. 2A). In contrast, the insulin-specific T_reg population was 3-fold larger in single TCR T cell mice relative to WT NOD mice, whereas bulk and HEL_11-25-specific T_reg cell numbers were similar (Fig. 2B). When considered together these data reveal a 3-fold increased insulin-specific T_reg to conventional CD4⁺ T cell ratio. Diabetes is largely driven by insulin-specific conventional T (T_conv) cell responses (1–5), and many studies have shown that expansion of the insulin-specific T_reg population can prevent diabetes (36–38). These studies indicate that the balance between T_reg:T_conv insulin-specific cells is crucial to the disease. Altogether, these data favor insulin-specific T_reg cells as a primary mediator of diabetes resistance in single TCR T cell NOD mice.

We then enumerated T_reg cells in the pancreas by immunohistochemistry. Single TCR T cell NOD mice had significantly fewer CD3⁺ T cells present in or around islets and a significantly higher T_reg:CD3⁺ T cell ratio per islet relative to WT NOD mice (Fig. 3A, AB). We further verified the presence of T_reg cells in and around islets via immunostaining of tissue sections with CD4 and FOXP3 Abs (Fig. 3C). Collectively, the data presented in Figs. 2 and 3 support a model in which skewing of insulin-specific T cells to the T_reg lineage promotes diabetes resistance in single TCR T cell NOD mice.

We postulated that the T_reg lineage skewing in the single TCR T cell mice originates in the thymus. We therefore contrasted thymocyte development stages of the single TCR T cell, TCRα^+/−, and WT NOD mice. Inclusion of TCRα^+/− NOD mice in these experiments allowed us to separate the effects on T cell development resulting from hemizygosity at TCRβ locus, and therefore not involved in diabetes pathogenesis, from those of hemizygosity at the TCRα locus. We found that single TCR T cell and TCRα^+/− NOD mice had a decreased proportion of single positive (SP) thymocytes relative to WT NOD mice (Fig. 4A). We separated the SP population into CD4⁺ (nonT_reg), CD8⁺, and T_reg cells and found that both single TCR T cell and TCRα^+/− NOD mice had lower frequencies of CD4⁺ and CD8⁺ cells than WT NOD mice, but an equivalent proportion of T_reg cells. This resulted in a higher T_reg:SP ratio, similar to the pancreas and insulin-specific T cell population in the PLN, suggesting that this ratio is established early in the thymus (Fig. 4B, 4C). We scrutinized positive selection using TCRβ and CD5 expression as surrogate markers for TCR signaling and progression through positive selection (where TCRβ^lo CD5^lo -preselection, TCRβ^lo CD5^int-initiating positive selection, TCRβ^int CD5^hi-undergoing positive selection, and TCRβ^hi CD5^hi-postselection) (Fig. 4D) (39). We found that single TCR T cell and TCRα^+/− double positive (DP) thymocytes were more likely to be classified as preselection and less likely to initiate and undergo positive selection relative to WT DP thymocytes, indicating less efficient positive selection (Fig. 4D). Similarly, the proportion of postselection DP thymocytes was lower in TCRα^+/− NOD mice, and all but one single TCR T cell NOD mouse trended lower than WT (Fig. 4D). The reduction in total SP thymocytes could be attributed to decreased efficiency of positive selection during the DP stage due to an increased failure to recombine a functional TCR, consistent with prior work from our group and others (14, 19). However, because T_reg committed thymocytes are derived from the CD4 SP thymocyte subset, this population would be similarly affected. We suggest that this effect in mice lacking dual TCRα expression is counteracted by the increased efficiency of agonist selection into the T_reg lineage among medullary thymocytes.

Focusing on Ag-specific cells, we found that insulin-specific CD4⁺ thymocytes accounted for a similar proportion of the repertoire in the three groups of mice (Fig. 4E). We measured the mean fluorescence intensity (MFI) of CD5 (a reliable surrogate for TCR signal strength) (40), on CD4⁺ SP, T_reg, and insulin-specific thymocytes from all groups. Insulin-specific thymocytes, much like T_reg thymocytes, expressed higher levels of CD5 than CD4⁺ SP thymocytes, indicating that they have received strong TCR signals during thymic selection (Fig. 4F). Thymocytes that receive strong TCR signals are either deleted by apoptosis or undergo agonist selection to become T_reg cells, invariant NKT cells, or intraepithelial lymphocytes. We therefore sought clues to determine whether these CD4 SP thymocytes were undergoing apoptosis or agonist selection. Caspase 3 is a member of the cysteine-aspartic acid family of proteases that is activated during apoptosis and therefore serves as a marker for apoptotic cell death. Importantly, agonist selection of strongly self-reactive thymocytes occurs in the medulla. As such, we measured activated caspase 3 in signaled (bulk CD4⁺ CD5⁺ TCRβ⁺) medullary (CCR7⁺) thymocytes by flow cytometry. We observed a lower proportion of caspase 3 activation in single TCR T cell and TCRα^+/− NOD mice relative to WT NOD mice in the medullary fraction (Fig. 4G). The higher T_reg:SP ratio present in^- NOD mice hemizygous for TCRα and our observations that CD5 expression is high among the insulin-specific thymocytes, yet apoptosis is decreased among the signaled CD4 SP medullary thymocytes, support the hypothesis that the increase in insulin-specific T_reg cell commitment arises due to enhanced agonist selection into the T_reg lineage, and that this effect, rather than enhanced negative selection, underlies their diabetes resistance.

Our findings suggest that a subset of insulin-reactive CD4⁺ T cells express TCRs that, when expressed alone, favor T_reg cell commitment. The notion that certain TCRs preferentially direct T_reg cell commitment has previously been described, including TCRs with high affinity for InsB_9-23 (41–43). Increasing evidence suggests that thymically derived T_reg cells are predominantly selected by Aire-expressing medullary thymic epithelial cells (41, 42). InsB is known to be both regulated by Aire and expressed by medullary thymic epithelial cells in the thymus (44). We hypothesize that when T_reg-biasing, insulin-specific TCRs are expressed in the context of a second TCR, the overall TCR avidity is reduced, favoring commitment to the T_conv lineage over the T_reg lineage. This would appear to be at odds with the findings of Tuovinen et al. (45), which suggest the peripheral T_reg population is enriched for dual TCR T cells in humans. However, their observations are based on one dual Vα population (Vα2⁺ Vα12⁺) and it is not known if they are thymically derived or peripherally induced T_reg cells. If other thymocyte populations reactive against Aire-regulated self-antigens behave similarly, one would expect a similar skewing toward a higher T_reg:CD4⁺ cell ratio among these populations. Future studies will need to be performed to investigate how dual TCR expression and thymic Ag expression patterns combine to influence T cell lineage commitment.

Altogether, our findings thus reveal a novel pathway of dual TCR T cell–mediated induction of autoimmune disease resulting from reduced agonist selection into the T_reg cell lineage, altering the composition of the self-reactive T cell population to a less tolerogenic, more autoimmune-prone phenotype. Furthermore, our findings support the idea that interventions designed to improve self-antigen–specific T_reg cell commitment and maintenance are a logical approach to prevent or treat T1D and other autoimmune diseases.

We thank the University of Minnesota’s flow cytometry core and Jason Mitchell for assistance with microscopy.

The authors have no financial conflicts of interest.