Non-obese diabetic mice select a low-diversity repertoire of natural regulatory T cells

Abstract

Thymus-derived Foxp3⁺ natural regulatory CD4 T cells (nTregs) prevent autoimmunity through control of pathogenic, autoreactive T cells and other immune effector cells. Using T cell receptor (TCR) transgenic models, diversity within this lineage has been found to be similar to that of conventional CD4 T cells. To determine whether balanced TCR diversity may be perturbed in autoimmunity, we have analyzed receptor composition in C57BL/6 and autoimmune non-obese diabetic (NOD) mice. The natural regulatory and conventional CD4 repertoires of C57BL/6 had similar diversities. Despite the apparently normal thymic development of the NOD nTreg lineage, TCR diversity within the selected repertoire was markedly restricted. Detailed analysis of TCRα and -β chain composition is consistent with positive selection into the natural regulatory lineage being under stringent audition for interaction with MHC class II/self-peptide. The NOD MHC region, including the unique H2-A^g7 class II molecule, partly accounts for the reduction in diversity, but additional NOD genetic contribution(s) are required for complete repertoire compaction. Mechanistic links between MHC, autoimmunity, and nTreg diversity identified in this study are discussed.

Autoimmune insulin-dependent diabetes mellitus culminates in destruction of the insulin-producing β cells of the islets of Langerhans. The genetic basis of disease susceptibility is complex, with the strongest association being encoded by the MHC class II region in human and the non-obese diabetic (NOD) mouse strain. In NOD mice, T cells play an essential role in disease pathogenesis and, through the activity of regulatory T cells (Tregs), control of disease progression. Foxp3-expressing natural Tregs (nTregs), arising in the thymus, play a crucial role in suppressing the pathogenic effects of self-reactive T cells. NOD mice have been reported to be relatively deficient in nTregs (1), although more recent evaluation of thymic production, export, and peripheral levels of nTregs did not find significant differences between NOD and nonautoimmune strains (2). More generally accepted is an age-dependent waning of Treg function apparent from adoptive transfer studies and in vitro assay of suppressive activity (3, 4). Multiple mechanisms, including an age-related decline of active, membrane-bound TGF-β (5), IL-2 insufficiency (6), defective antigen presentation (7), and resistance of conventional CD4 T cells to regulation (8) may contribute to the compromised function of peripheral NOD nTreg cells.

Although thymic production of nTregs is not obviously aberrant, other aspects of T cell development and repertoire selection in NOD mice are anomalous. The TCRβ checkpoint, which drives early thymocyte proliferation and differentiation, is partially independent of β chain expression (9), and CD8⁺, CD4⁺ (DP) thymocytes have a low activation threshold (10). NOD mice exhibit partial resistance to clonal deletion of autoreactive thymocytes involving failure to induce the proapoptotic gene Bim (11). Selection of invariant natural killer T cells is inefficient, a defect attributable to retarded SLAM receptor expression (12). In fetal thymic organ culture (FTOC), NOD nTregs require higher levels of antigen for induction, but once induced, have greater capacity for expansion (13). Finally, analysis of CD4⁺ thymocytes and peripheral T cells with an A^g7/BDC-2.5 mimetope tetramer uncovered a surprisingly large population of reactive cells (14, 15).

In view of these findings, a closer examination of the NOD nTreg repertoire is warranted.

Here we show that the TCR repertoire of thymic NOD nTregs is restricted in diversity and qualitatively different from NOD conventional CD4 T cells and both regulatory and conventional C57BL/6 subsets.

Results

TCRα Chain Diversity of CD25⁻ and CD25^hi CD4 Thymocytes in NOD and C57BL/6.

Using approaches we have established in CD8 T cells (16), the V α 9 (AV9) components of the NOD and C57BL/6 conventional (Tconv) CD4 and nTreg repertoires were analyzed. This unique AV segment is located in the 3′ region of the AV segment cluster and preferentially recombines with J segments positioned in the 5′ region of the AJ segment cluster, producing a characteristic preselection “blueprint” (16, 17). Where the stringency of α chain audition during positive selection is low, the recombination blueprint is largely preserved in the postselection repertoire. Conversely, where positive selection involves more stringent α chain audition, the recombination blueprint is lost (16). Stringent CDR3 audition can also lead to highly focused J segment usage. Tconv (CD25⁻ CD4⁺) and nTreg (CD25^hi CD4⁺) thymocytes were sorted from 5- to 6-week old female NOD and C57BL/6 mice. More than 80% of sorted CD25^hi CD4 SP thymocytes were also positive for Foxp3 by intracellular stain (not shown). AV9 rearrangements were amplified by RT-PCR, cloned, and their predicted CDR3 composition determined. Fig. 1 shows representative plots of AV9 CDR3 diversity. The thymic nTreg and Tconv repertoires of the nonautoimmune C57BL/6 strain have similar AV9 CDR3 diversity (Fig. 1 A and B). In stark contrast, the NOD strain has minimal diversity within the nTreg repertoire but has reasonably high diversity within the Tconv repertoire (Fig. 1 C and D).

Fig. 1.

Low-diversity AV9 TCR repertoire in NOD thymic nTregs. Thymic nTregs (CD25⁺CD4⁺) and conventional SP4 (CD25⁻CD4⁺) thymocytes from C57BL/6 (A and B) and NOD (C and D) mice were sorted and the frequency of individual AV9 TCR rearrangements determined. Each unique CDR3 sequence is represented by a segment proportional to its frequency. Repertoire diversity (D value) is shown for each population. Color is used for clarity only and does not represent a particular sequence. Number of sequences collected for each population is shown in Table S1. Data are derived from a single representative experiment of n = 2 (C57BL/6) and n = 3 (NOD) independent experiments.

Comparing the preselection AV9 recombination blueprint with the selected Tconv and nTreg repertoires, there is closest correspondence for C57BL/6 in which 84% and 79% of recombination events are, respectively, within the target region (Fig. 2 A and B). The CD4 Tconv repertoire of NOD (Fig. 2D) has slightly less correspondence, with 65% of events lying within the target region. By contrast, the NOD nTreg repertoire (Fig. 2C) has low correspondence, with only 36% of events using favored AJ segments. Relative diversity also correlates with the proportion of AJ segments used; both C57BL/6 repertoires use 23 of the 49 functional J segments, dropping to 18 of 49 for the conventional NOD repertoire and just 5 of 49 for the NOD nTreg repertoire. This repertoire restriction and skewing of AJ segment use were observed in 2 further experiments (Fig. S1).

Fig. 2.

J segment usage is highly restricted among AV9 NOD thymic nTregs. J segment usage within thymic nTreg (CD4⁺CD25⁺) and Tconv (CD4⁺CD25⁻) thymocytes from C57BL/6 (A and B) and NOD (C and D) mice. The percentage of the AV9 repertoire using each J segment is shown by bar height. J segments are represented on the x axis following the genomic order. Type I and type II J segments are shown as open and closed bars, respectively. Percentages of rearrangements occurring outside the recombination blueprint are indicated. Number of sequences collected for each population is shown in Table S1. Data are derived from a single representative experiment of n = 2 (C57BL/6) and n = 3 (NOD) independent experiments.

We have recently described a structural and functional classification dividing the mouse AJ segment cluster into 2 classes based on germline J segment amino acid composition and length. Type I AJ segments are longer and more flexible than type II segments; furthermore, type II AJ segments can be favored in engaging with MHC–peptide complexes (16). Type I and II J segments are represented by open and filled columns, respectively, in the J segment plots. Of the 10 repertoires analyzed, type II rearrangements dominate only in the 3 NOD nTreg examples (64%–95%). Table S1 summarizes the results of several independent experiments confirming the lack of diversity, lack of correspondence with the recombination blueprint, and preferential use of type II AJ segments by NOD AV9 nTregs.

The reduction in NOD AV9 nTreg diversity was associated with selection of a highly focused AJ segment signature. In 3 independent experiments, AJ33 was the dominant AJ segment used by the NOD nTreg repertoires, and several identical AV9/AJ33 CDR3 rearrangements were present in more than 1 experiment (Fig. 2, Fig. S1, and Table S2). Pooling the 3 NOD experiments, AJ33 represented 2% and 51% of the Tconv and nTreg repertoires, respectively, a 25.5-fold skew toward selection into the nTreg lineage. Of interest, after AJ33, the next most commonly used AJ segments in the NOD AV9 nTreg repertoires, AJ27 (NYNQGKLI, experiment 1) and AJ23 (NTNTGKLT, experiment 2) are closely related to each other in primary amino acid sequence, suggesting functional similarity. AJ23, -27, and -33 favored in the NOD nTreg repertoires are all type II segments. These data show that the conventional and nTreg CD4 lineages in the NOD strain have distinct composition as well as diversity. Reduction in TCR diversity together with preferential use of a small set of type II AJ segments located outside the region of preferred recombination support the view that positive selection of NOD nTreg cells is under higher stringency than their conventional cousins. The C57BL/6 thymic nTreg repertoire does not share any of these characteristics, being similar to the corresponding conventional CD4 T cell repertoire.

To confirm and extend these findings, a second unique variable segment, AV12, which is located in the 5′ region of the AV segment cluster, was analyzed. Again, the thymic nTreg repertoire was highly restricted in the NOD strain but not C57BL/6 (Fig. 3). The preferred region of recombination for AV12 is less well defined but, in contrast with AV9, is focused on the 3′ region of the AJ cluster. Preferential use of 3′-positioned AJ segments can be clearly seen for the C57BL/6 AV12 nTreg repertoire, which uses a broad set (24 of 49) of AJ segments. Conversely, AJ segment use in NOD nTregs tended to be more central and used only 9 AJ segments (Fig. 4). As for the AV9 repertoire, AV12 type II rearrangements dominate only in the NOD Treg population (55%), in comparison with 49% in NOD Tconv and 48% and 50% in C57BL/6 Tregs and Tconv, respectively.

Fig. 3.

Low diversity and restriction of J segment use in AV12 TCR repertoire of NOD thymic nTregs. Thymic nTreg (CD4⁺CD25⁺) and Tconv (CD25⁻CD4⁺) thymocytes from C57BL/6 (A and B) and NOD (C and D) mice were sorted and the frequency of individual AV12 TCR sequences determined. The frequency of each unique CDR3 sequence correlates with segment size; color is used for clarity only and does not represent a particular sequence. Repertoire diversity (D value) is shown for each population.

Fig. 4.

Restricted J segment usage amongst AV12 NOD thymic nTregs. J segment usage within thymic nTreg (CD4⁺CD25⁺) and Tconv (CD4⁺CD25⁻) thymocytes from C57BL/6 (A and B) and NOD (C and D) mice. The percentage of the AV12 repertoire using each J segment is shown by bar height. J segments are represented on the x axis following the genomic order. Type I and type II J segments are shown as open and closed bars, respectively.

TCRβ Chain Diversity in CD25⁻ and CD25^hi CD4 Thymocytes in NOD and C57BL/6.

We next determined whether the TCRβ repertoire is also distinct in NOD nTregs. Because TCRβ diversity is magnified by incorporation of a diversity (D) segment, we focused on a specific V–J combination to limit the number of target recombination events. The TCR BV8.2–BJ2.3 combination, which we have studied extensively in C57BL/6 mice, provides a suitable level of diversity (18). Fig. 5 shows representative BV8.2–BJ2.3 diversity plots. Diversity of the conventional CD4 thymocyte subset was similar in C57BL/6 and NOD. nTreg thymocyte diversity was also similar in both strains but lower than the corresponding conventional repertoires. Table S3 summarizes data from 3 independent experiments. Although there is a trend for a greater imbalance between diversities of the conventional and nTreg TCRβ repertoires in NOD than C57BL/6, the effect is less marked than for TCRα.

Fig. 5.

Diversity of BV8.2-BJ2.3 TCR repertoire in NOD thymic nTregs. Thymic nTreg (CD4⁺CD25⁺) and Tconv (CD4⁺CD25⁻) thymocytes from C57BL/6 (A and B) and NOD (C and D) mice were sorted and the frequency of individual BV8.2-BJ2.3 TCR sequences determined. Each unique CDR3 sequence is represented by a segment proportional to its frequency; color is used for clarity only and does not represent a particular sequence. Repertoire diversity (D value) is shown for each population. Number of sequences collected for each population is shown in Table S3.

The preselection TCRβ chain repertoire emerges from positive selection with a shorter mean CDR3 length (19). For the BV8.2–BJ2.3 combination, the mean preselection CDR3 length of 10.1 aa is decreased by ≈0.5 residues in CD4 T cells (18). Fig. S2 shows CDR3 length distributions for the pooled TCRβ data. The mean CDR3 length and size distributions of the C57BL/6 conventional and nTreg repertoires are highly similar. In contrast, the NOD nTreg repertoire, in comparison with the conventional repertoire, is significantly skewed (P = 0.01) toward shorter CDR3 lengths. Thus, although not dramatically altered in diversity, NOD TCRβ CDR3 structures selected into the nTreg lineage are structurally distinct from those entering the conventional lineage.

TCRα Chain Diversity in Congenic C57BL/6^g7 CD25⁻ and CD25^hi CD4 Thymocytes.

To address the role of H2-A^g7, the candidate Idd1 gene, in limiting TCRα chain diversity, the B6^g7 strain (C57BL/6 congenic containing the NOD H2 region) was analyzed. Although some restriction in AV9 nTreg diversity relative to C57BL/6 is seen (Fig. 6), the highly stringent AV9 selection observed in the NOD strain was not reproduced (Fig. 1 and Table S1). Although the preselection AV–AJ recombination blueprint was present to similar extents in both B6^g7 repertoires, fewer J segments were used by the nTreg repertoire (8 of 49) than by the conventional repertoire (21 of 49). However, the bias to type II AJ segments and dominant use of AJ33 observed in the NOD nTreg repertoire was not seen in the B6^g7 nTreg repertoire. These data show that restriction in NOD nTreg diversity is partly but not wholly determined by MHC haplotype, but also depends on additional genetic contribution(s) from the NOD strain.

Fig. 6.

C57BL/6^g7 mice have a nTreg repertoire distinct from NOD. nTreg (CD4⁺CD25⁺) and Tconv (CD4⁺CD25⁻) thymocytes from B6^g7 were sorted (A and B) and the frequency of individual AV9 TCR sequences determined. The frequency of each unique CDR3 sequence correlates with segment size. Color is used for clarity only and does not represent a particular sequence. Corresponding J segment use is represented in the lower panels (C and D). The proportion of J segments outside the region of preferential recombination and relative use of type I (open bars) and type II (closed bars) J segments is shown.

Discussion

Recent investigations of CD4 T cell repertoires have found that TCRα diversity within the nTreg component matches or exceeds that of the conventional component (20–22). These studies have analyzed “mini” repertoires, in which β diversity is eliminated through the expression of transgenic β chains; in some cases, α diversity was also limited by the use of a transgenic AV–AJ cassette. We wished to extend these studies and determine whether changes in nTreg diversity and functional composition may be associated with predisposition to autoimmune disease. To avoid the potential for skewing of the α repertoire when coselected with a transgenic β partner (23), we analyzed mice with unmanipulated repertoires. In addition to measuring diversity, other functional parameters of the α repertoire were analyzed. We focused on 2 α variable segments (AV9 and AV12) encoded by unique genes rather than gene families spread throughout the variable segment cluster, as found for most AV genes. Examination of individual genes rather than gene families is more likely to reveal differences between repertoires for 2 reasons. First, focusing on a single gene will limit diversity of the target repertoire. The second advantage derives from the AV–AJ recombination process in DP thymocytes, by which 3′-positioned AV genes preferentially recombine with 5′-positioned J segments (17) followed by successive rearrangements joining more 5′ AV segments to more 3′ J segments. For dispersed AV gene families this allows efficient sampling of J segments located throughout the cluster, whereas V segments encoded by single genes will recombine efficiently only with a subset of J segments dictated by their position. Repertoires preferentially using limited sets of V–J combinations are more likely to reflect changes in positive selection (16). We have recently described a structural division among J segments based on length and predicted flexibility and correlated the use of inflexible, shorter J segments (type II) with peptide engagement (16). Using these parameters, we conclude that entry into the nTreg lineage in NOD mice involves highly stringent audition of AV12 and AV9 rearrangements. Positive selection of the NOD conventional CD4 repertoire is not under such stringent selection because it uses a broad range of J segments located mainly within the region of preferential recombination. A striking finding was the dominant use of AJ33 in all 3 NOD nTreg repertoires analyzed (51% of total), contrasting with its low representation in the conventional repertoires (2% of total) and providing direct evidence that entry into the 2 lineages favors different sets of TCRs. Only 6 AJ33 rearrangements were found across the 3 NOD nTreg repertoires, with 5 of 6 being 8 aa; of these, 4 were found in 2 independent experiments. Each NOD nTreg repertoire had a dominant AJ33 rearrangement, suggesting that thymocytes selected into the NOD nTreg lineage may undergo proliferation consistent with the behavior of NOD nTregs in FTOC (13). This extent of conservation and overlap, again, can only be explained by highly stringent positive selection focused on AJ33 with structurally constrained CDR3 loops. The simplest explanation for such focused TCR selection is interaction of the selected CDR3 loops with H2-A^g7 molecules complexed with a limited set of self-peptides. On the other hand, we suggest that the broad spectrum of AV9 CDR3 loops found in the NOD conventional and both C57BL/6 repertoires ensues from engagement with a correspondingly broad set of self MHC–peptide complexes.

Analysis of TCRβ repertoires focused on BV8.2–BJ2.3 rearrangements. For both strains, nTreg repertoires were less diverse than the conventional repertoires. Averaging across all experiments, the C57BL/6 and NOD nTreg TCRβ repertoires are, respectively, 1.5- and 1.9-fold less diverse than the corresponding conventional populations (Table S3). Despite the limited drop in diversity, the NOD nTreg repertoire has a different CDR3 size distribution in comparison with the Tconv repertoire, favoring shorter lengths (Fig. S2). Reduction in CDR3 length is a consequence of positive selection, reflecting productive engagement with MHC–peptide. Exaggerated CDR3 shortening suggests that positive selection of NOD nTreg cells involves distinct interactions with MHC–peptide and is consistent with the hypothesis that selection is mediated by a limited set of peptides. For AV9 and BV8.2–JB2.3, between 7% and 10% of sequences were shared between the nTreg and conventional repertoires; however, functional overlap is not inferred because partner chains are unlikely to be shared.

At least 2 parallel mechanisms could be responsible for the elevated stringency of audition into the NOD nTreg lineage. First, it may be directly related to structural aspects of the interaction of H2-A^g7–peptide complexes with the TCR, whereby few self-peptides achieve selecting affinity. Second, signal strength in the nTreg lineage may be attenuated through defective signaling, consistent with the observation that induction of NOD nTregs in FTOC requires a high level of antigen (13), again narrowing the set of peptides able to mediate positive selection. Analysis of the B6^g7 congenic, which showed some limitation of diversity but not the striking focus on AJ33 seen in NOD, confirmed the important contribution of the MHC region, likely to be due to the unique NOD MHC class II molecule H2-A^g7. Indeed, expression of a transgenic MHC class II molecule in NOD mice has previously been linked with generation of protective regulatory T cells (24). In addition, our data identify a crucial non–MHC-encoded contribution to the phenotype.

Although the antigen specificity of nTregs is poorly understood, nTreg function in anatomic locations such as the pancreas is likely to be through recognition of self. The degree to which nTreg repertoires are self-reactive is currently unclear, with reports supporting both strong bias to self-reactivity (25) and limited self-reactivity (26). An important issue is to determine whether the degree of self-reactivity within the NOD nTreg repertoire is influenced by the elevated stringency of selection.

How might reduced complexity of the nTreg repertoire influence in vivo regulatory activity in the context of tissue-specific autoimmunity? In comparison with C57BL/6, reliance on the recognition of a smaller set of self-peptides for the activation of NOD nTregs may lead to “holes” in the repertoire and compromise protection from autoimmunity if expression of the relevant activating MHC–peptide complex(es) is limiting. In addition to differential availability of the nTreg activating peptides, the presented peptide repertoire is likely to be modified by the milieu of self-peptides competing for access to MHC class II. These parameters will be different for each tissue, suggesting a novel mechanism that may contribute to tissue-specific failure of nTreg-mediated regulation. In the context of diabetes, the increased burden of apoptotic material for processing during β cell remodeling may be especially relevant (27). Furthermore, relative insensitivity to antigen during selection may result in poor activation and regulatory function in the periphery.

Given the low complexity of the NOD thymic nTreg repertoire, it will be of interest to determine whether its composition is shaped further in the periphery, leading to an outgrowth of initially rare specificities that may become prone to clonal exhaustion. A recent report did not find any indication that induced Tregs can arise from islet-specific conventional CD4 T cells in NOD mice, underlining the paramount importance of the nTreg population (22).

In summary, we find that in the autoimmune NOD mouse, despite the quantitatively normal appearance of the nTreg population, there is an imbalance between diversity of the thymic nTreg repertoire and that of the conventional CD4 repertoire. We propose that this contributes to the development of autoimmunity and suggests a novel mechanism contributing to the strong association between MHC and disease susceptibility. It will be of interest to determine whether other murine, MHC-linked, autoimmune models show similar repertoire skewing, and to extend this study to human repertoires. Finally, also highlighted in this study is the role of non-MHC factors in shaping the regulatory repertoire. Identifying these factors might be a key step in the understanding of disease etiology.

Materials and Methods

Mice.

NOD and C57BL/6 mice were housed and bred under specific pathogen-free conditions in Imperial College animal facilities. C57BL/6^g7 (B6^g7) mice were housed and bred under specific pathogen-free conditions at the University of Bristol, United Kingdom. Mice were analyzed at age 5 to 6 weeks. Procedures were approved by local ethical review and performed under Home Office authority.

Cell Sorting.

Thymocytes from 2 females aged 5 to 6 weeks were pooled and sorted into CD4⁺CD8⁻CD25⁻ or CD4⁺CD8⁻CD25^high. Approximately 100,000–150,000 cells were collected for each population. Anti-CD4, anti-CD8, and anti-CD25 antibodies were purchased from BD PharMingen. Sorting was performed on a FACSAria or a FACSDiva (BD Biosciences).

RT-PCR and Sequence Analysis.

RNA was extracted using TRIzol (Invitrogen). cDNA was synthesized with SuperScript II RNase H⁻ Reverse Transcriptase (Invitrogen) and random hexamers (Amersham Biosciences). For TCRα analysis, cDNA was amplified by PCR using either the forward AV9 or AV12 primer and reverse ACα primer. For TCRβ analysis, cDNA was amplified using the forward BV8.2F and reverse BJ2.3R primers. PCR products were cloned into the pCR2.1 vector using the TOPO TA Cloning Kit (Invitrogen). PCR was then performed on individual colonies using the same primers and sequenced using the internal α constant region ACseq or variable β region BV8.2S primers. Nomenclature for TCRAJ is according to the ImMunoGeneTics Internet-based resource (http://imgt.cines.fr). To quantify diversity (D), the Shannon entropies (H) were calculated resulting in values ranging from 0 to 1, expressed in terms of percentage (28, 29). TCRβ nTreg and conventional CD4 thymocyte CDR3 length distributions were statistically compared using the stratified Wilcoxon-Mann-Whitney test. Primer sequences (5′–3′) were as follows: AV9: ACACCGTTGTTAAAGGCACC; AV12: CTCTGTTTATCTCTGCTGACC; ACα: CTTTCAGCAGGAGGATTCG; ACseq: CATAGCTTTCATGTCCAGC; BV8.2F: GGTGACATTGAGCTGTAAT; BJ2.3R: AGTCAGTCTGGTTCCTGAG-3′; BV8.2S: TTCATATGGTGCTGGCAGC.