The Repertoire of Newly Developing Regulatory T Cells in the Type 1 Diabetes–Prone NOD Mouse Is Very Diverse

Regulatory T lymphocytes expressing the forkhead/winged helix transcription factor Foxp3 (Treg) play a major and even vital role in the control of innate and adaptive immunity and are also involved in tissue repair (1–3). Using a large variety of suppressor-effector mechanisms, Treg inhibit immunity and thus avoid autoimmune pathology, chronic inflammation, and rejection of the semiallogeneic fetus (2). The vital character of immune regulation by Treg is best illustrated by the rapidly lethal autoimmune and inflammatory syndromes that develop in humans and mice carrying loss-of-function mutations in the locus encoding the master regulator of Treg development and function, FOXP3/Foxp3 (4,5). More subtle defects in Treg may be involved in the etiology of autoimmune pathology (6).

Treg can differentiate from T-cell precursors during T-cell development in the thymus and from mature conventional T lymphocytes during immune responses in the periphery (2). In the thymus, Treg development requires high affinity interactions of the T-cell receptor (TCR) for antigen expressed by T-cell precursors with MHC/peptide complexes expressed mainly by medullary thymic stromal cells (7). Because the latter cells present peptides derived from tissue antigens, in part through the action of the transcription factor autoimmune regulator (AIRE) (8), an autospecific Treg repertoire thus develops (9,10). Treg development in the thymus also requires cytokines, such as interleukin-2 (IL-2) and IL-15 (11), as well as signals through costimulatory receptors (12–14).

Defects in Treg development are thought to be involved in the etiology of autoimmune pathology (6). Thus, the very aggressive and rapidly lethal autoimmune syndrome developing in AIRE-deficient NOD mice appears, at least in part, due to defective selection of the Treg repertoire early in life (15). Also mice deficient in IL-2 or its receptor rapidly succumb to autoimmune pathology, but the respective roles of defective intrathymic development and peripheral survival and function of Treg remain to be assessed (11). NOD mice deficient in CD28, in which substantially fewer Treg develop in the thymus, also develop more aggressive type 1 diabetes (T1D) (16).

It has been suggested that the TCR repertoire expressed by Treg developing in the thymus of T1D-prone NOD mice is less diverse than that developing in the thymus of C57BL/6 (B6) mice, which are frequently used as a T1D-resistant reference (17–19). The results were based on analysis of CD25-expressing CD4⁺CD8⁻ thymocytes, presumably developing Treg. However, our laboratory recently demonstrated that in adult mice, actually only a fraction of thymic Treg are cells that had recently developed, the rest being cells that had recirculated from the periphery back to the thymus (20). Moreover, we recently demonstrated that the NOD thymus contains a particularly high proportion of recirculating Treg (21). Also, CD25 is not only expressed by Treg but also by immature precursors (22). To assess the potential involvement, in the etiology of T1D in the NOD mouse, of genetic defects affecting the diversity of the TCR expressed by newly developed Treg, we therefore readdressed this issue using mutant NOD and B6 mice in which newly developed Treg can be distinguished from recirculating cells and in which Treg can be isolated based on Foxp3 expression.

In Rag2-Gfp mice, the Rag2 promoter drives expression of green fluorescent protein (GFP) (23). Transcription of the GFP encoding sequence thus terminates at positive selection at the CD4/CD8 double-positive stage of T-cell development, and the accumulated GFP degrades with a half-life of 56 h (24,25). In the thymus, newly developed Treg retain sufficient green fluorescence to distinguish them from recirculating cells that had entirely lost fluorescence (20). In Foxp3-Thy1^a mice, the allelic cell-surface reporter Thy1.1 is expressed under control of the Foxp3 promoter, allowing detection of Foxp3-expressing Treg without fixation/permeabilization procedures, thus allowing for isolation of mRNA required for TCR repertoire analysis by high-throughput mRNA sequencing. Using Rag2-Gfp Foxp3-Thy1^a mice, we here firmly demonstrate that the TCR repertoire expressed by Treg newly developed in the thymus of the T1D-prone NOD mouse is as diverse as that developing in the resistant mouse strain B6.

B6 mice expressing the Foxp3-Thy1^a mutation and a Rag2-Gfp transgene were previously described (26). The two mutations were introduced into the NOD genetic background by speed-backcrossing (27). After all Idd-loci were of the NOD allotype (6 generations), the mice were crossed to NOD mice for another 10 generations. Wild-type (WT) NOD.CD45.2 and B6.CD45.1 congenic mice were bred in our colony.

Thymocytes from female 8-week-old WT NOD and B6 mice were incubated with anti-FcRIII/FcRII antibody (2.4G2) and then labeled with EF450-labeled anti-CD8β (H35-17.2), phycoerythrin (PE)-labeled anti-CCR7 (4B12; both from eBioscience), allophycocyanin-Cy7–labeled anti-CD4 (GK1.5), PE-Cy7–labeled CD25 (PC61; both from BD Biosciences), and BV605-labeled anti-CD73 (TY/11.8, Biolegend). The labeled cells were stained with a TCR-Vβ screening FITC-labeled antibody panel (BD Biosciences). Cells were then stained with EF660-labeled anti-Foxp3 (FJK-16s; eBioscience) according to the manufacturer’s instructions. The labeled cells were analyzed using a Fortessa flow cytometer (BD Biosciences), and data were analyzed using FlowJo software (Tree Star).

Pooled thymocytes from 8-week-old female Foxp3-Thy1^a Rag2-Gfp NOD and B6 mice (n = 8-10) were complement depleted of CD8⁺ cells using an anti-CD8α hybridoma (31M) supernatant. Remaining cells were stained with PE-labeled anti-CD4 (GK1.5), EF450-labeled anti-CD8β (H35-17.2; both from eBioscience), and allophycocyanin-labeled anti-Thy1.1 (OX7; BD Biosciences). CD4⁺CD8⁻Thy1.1⁺GFP⁺ Treg (n = 10⁵) were sorted using the FACSAria II cell sorter (BD Biosciences). TCR sequencing (TCRseq) was performed as previously described, with minor modifications (28,29). mRNA was extracted by NucleoSpin RNA XS (Macherey-Nagel) according to the manufacturer’s instructions and was quality controlled (RNA integrity number >8) using the Agilent 2100 BioAnalyzer.

cDNA was prepared as follows: mRNA was reverse transcribed using primers aligning to the sequences encoding the constant regions of TCRα (CTCAGCGTCATGAGCAGGTTAAAT; CAGGAGGATTCGGAGTCCCATAA; TTTTACAACATTCTCCAAGA; TTCTGAATCACCTTTAATGA; ATGAGATAATTTCTACACCT; TTTGGCTTGAAGAAGGAGCG; TTCAAAGCTTTTCTCAGTCA; and TGGTCTCTTTGAAGATATCT) and TCRβ (GGTAGCCTTTTGTTTGTTTG; CCCCTGGCCAAGCACACGAG; TGCCATTCACCCACCAGCTC; GCTATAATTGCTCTCCTTGT; TTGCGAGGATTGTGCCAGAA; CTTGTCCTCCTCTGAAAGCC; and GCCTCTGCACTGATGTTCTG). Then, 5' adapters containing unique molecular identifiers (UMI) were added using a template-switch reaction (TACACGACGCTCTTCCGATCUNNNNUNNNNUNNNNUCTTrGrGrGrGrG).

Libraries were prepared as follows: The first PCR reaction was performed with 5' primer TACACGACGCTCTTCCGATC and 3' primers AAGTCGGTGAACAGGCAGAG for Tcra and TGATGGCTCAAACAAGGAGACC for Tcrb (2 min at 95°C, 10 cycles of 20 s at 95°C, 15 s at 59°C, and 45 s at 70°C, and final incubation for 3.5 min at 70°C). The amplicons were then purified using Agencourt AMPure XP beads. The second seminested PCR was performed with 5' primer TACACGACGCTCTTCCGATC and 3' primers AGCAGGTTCTGGGTTCTGGA for Tcra and GGGTGGAGTCACATTTCTCAGAT for Tcrb (2 min at 95°C, 20 cycles of 20 s at 95°C, 15 s at 59°C, and 45 s at 70°C, and final incubation of 3.5 min at 70°C). The third PCR was performed using 5' primer AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATC and 3' primers CAAGCAGAAGACGGCATACGAGATXXXXXXXXGTGACTGGAGTTCAGACGTGTGCTCTTCCGATCTAGCAGGTTCTGGGTTCTGGA for Tcra and CAAGCAGAAGACGGCATACGAGATXXXXXXXXGTGACTGGAGTTCAGACGTGTGCTC TTCCGATCTGGGTGGAGTCACATTTCTCAGAT for Tcrb (in which XXXXXXXX is a sequencing index) (2 min at 95°C; one cycle of 20 s at 95°C, 15 s at 59°C, and 45 s at 70°C; five cycles of 20 s at 95°C, 15 s at 75°C, and 45 s at 70°C; and a final incubation of 3.5 min at 70°C). For the fourth PCR, the 5' primer AATGATACGGCGACCACCGA and 3' primer CAAGCAGAAGACGGCATACGA were used (2 min at 95°C; five cycles of 20 s at 95°C, 15 s at 60°C, and 45 s at 70°C; and a final incubation of 3.5 min at 70°C). The quality of each library was verified using the Agilent 2100 BioAnalyzer (mean peak size 640 base pairs). The samples were indexed and sequenced with 300 base pairs paired-end on a MiSeq sequencer (Illumina). Thus, three Tcra and three Tcrb libraries per mouse strain were generated.

The reads were preprocessed with the toolkit pRESTO (30) as follows: FilterSeq was used to select reads with a quality >20. The sequences corresponding to the TCRα constant region (AGCAGGTTCTGGGTTCTGGA) or TCRβ constant region (GGGTGGAGTCACATTTCTCAGAT) and indicating location of the UMI (CTTGGGGG) were searched for using MaskPrimers and PairSeq algorithms and indexed to the head of the paired reads. BuildConsensus was used to construct consensus-sequences of the reads with the same UMI. Next, the forward and reverse reads were aligned to assemble the Tcra and the Tcrb sequences (AssemblePairs), and the UMI groups containing at least two reads were selected. The sequenced fragments from each selected UMI were aligned to the Tcra or Tcrb genomic region using the toolkit MiXCR (31) with the tools “align” and “assemble.” The aligned fragments were exported as “clonotype-tables” using the tool exportClones.

VDJtools (32) and customized R scripts were used to process these clonotype tables, and graphs were generated. The V-J use plots were generated using the PlotFancyVJUsage command. The CDR3 lengths were calculated based on the information in the clonotype table. The numbers of “P” and “N” nucleotides were extracted from the clonotype tables using the command exportClones VGeneWithP. Rarefaction plots and diversity statistics were calculated and graphed using the R package iNEXTm. We used three different diversity measures: Chao1 estimate, Shannon diversity, and Simpson diversity. All three provide an estimate of the number of clonotypes in the repertoire but differ in the extent to which they discount rare clonotypes. The more rare and therefore difficult to detect clonotypes are discounted, the more precisely the diversity measure can be estimated. By jointly considering these three diversity measures, we cover the full diversity information contained in the sample data (33). All custom scripts used are available at https://github.com/arielgalindoalbarran/NOD_TCR_repertoire.

TCRseq data generated during the current study are available in the Gene Expression Omnibus database (https://www.ncbi.nlm.nih.gov/gds) under accession number GSE159001. The other data sets are available from the corresponding author upon reasonable request. The mutant mice used in the current study are available from the corresponding author upon reasonable request.

We and others previously showed that the pool of Treg in the thymus is constituted of newly developed and of fully mature cells. The latter Treg are mostly cells that had recirculated from the periphery back to the thymus. We therefore analyzed NOD mice carrying a Rag2-Gfp transgene and the Foxp3-Thy1^a mutation in which we distinguished newly developed GFP⁺ from recirculating GFP⁻, surface Thy1.1⁺ Treg (20). Using flow cytometry, we found that 32 ± 11% of thymic CD4⁺CD8⁻Foxp3⁺ Treg were GFP⁺ in 8-week-old NOD females (Fig. 1A). This extends our previous finding that in the thymus of adult B6 mice, only a minority of Treg are newly developed cells (20,21). In this study we focused our attention on these newly developed Treg.

To assess the diversity of TCR expressed by newly developed Treg, we first analyzed their TRBV (i.e., TCR-Vβ) repertoire by flow cytometry. The commercially available anti–TCR-Vβ antibody panels are FITC labeled, which is incompatible with the GFP-based identification of newly developed Treg in Rag2-Gfp transgenic mice. We therefore identified newly developed nuclear Foxp3⁺ Treg in WT mice using their CCR7^highCD73^−/low phenotype that we recently described (21) (Supplementary Fig. 1). Using antibodies to 15 distinct TRBV, we found that the TRBV repertoires expressed by newly developed NOD and B6 Treg were distinct (Supplementary Fig. 2A), but similarly diverse, even if the very small difference in the diversities (3.5%) was statistically significant (Fig. 1B).

To study the TRAV repertoire expressed by newly developed Treg, we sorted CD4⁺CD8⁻Thy1.1⁺GFP⁺ thymocytes from Foxp3-Thy1^a Rag2-Gfp NOD and B6 mice, isolated mRNA, and then performed high-throughput TCRseq of Tcra mRNA. We used unique UMIs to identify sequences derived from individual mRNAs. Whereas we observed substantial differences between the TRAV-repertoires of NOD and B6 Treg, they were similarly diverse (Fig. 2A and Supplementary Table 1). Using TCRseq, we also confirmed the similar diversities of the TRBV repertoires expressed by newly developed NOD and B6 Treg (Fig. 2B and Supplementary Table 2). The diversities (i.e., D-values, Shannon diversity) of the TRBV repertoires as determined by TCRseq were higher than those determined by flow cytometry analyses (compared with Fig. 2B vs. Fig. 1B), because whereas existing antibodies do not cover the entire range of TRBV segments, TCRseq detects all of them. Comparison of the TRBV repertoires as determined by flow cytometry versus TCRseq revealed, as expected, substantial although imperfect similarity (Supplementary Fig. 3). Differences are probably because of the two techniques analyze individual cells and mRNAs, respectively, and each TRBV has its own promoter region, potentially leading to distinct levels of mRNAs in individual cells.

The V-J combination used in a TCR is a critical parameter in its specificity. We therefore assessed the distinct V-J combinations used in TCR expressed by newly developed Treg in NOD and B6 mice. We found some substantial differences in V-J associations used in TCRα and TCRβ expressed by newly developed NOD versus B6 Treg (Supplementary Tables 3 and 4), but the diversities of these associations appeared very large and comparable (Fig. 2C and D). Rearrangements of the Tcra locus continue until positive selection occurs, thus with time, using more upstream V and more downstream J segments (24). To assess the “stringency” of positive selection, we therefore assessed the position-weighted use of V and J segments and found no statistically significant differences between the NOD and B6 Tcra sequences (Fig. 2C and Supplementary Fig. 4).

During the course of the distinct thymic selection events, the CDR3s of TCRα and TCRβ chains gradually shorten (34–37). To assess whether newly developed Treg have undergone more or less stringent selection In NOD versus B6 mice, we therefore next used the TCRseq data to assess the CDR3 lengths of the TCRα and TCRβ chains. The average CDR3 lengths of the TCRα and TCRβ chains were identical between NOD and B6 mice (Fig. 2E and F, left panels). Also the CDR3 length distributions were practically identical for Treg newly developed in NOD and B6 mice, even if we observed some very small but statistically significant differences (Fig. 2E and F, right panels).

Non–germline-encoded CDR3 amino acids appear to reduce the affinity of the TCR’s interactions with MHC molecules (38). These amino acids are mostly encoded by N and P nucleotides. To assess this parameter, we extracted from the TCRseq data the average numbers and distributions of N and of P nucleotides in the CDR3 regions of the TCRα and TCRβ chains expressed by newly developed Treg. We found that they were practically identical in NOD and B6 mice, even if we observed some very small but statistically significant differences (Fig. 2G–J).

We next assessed the total diversity of the TCRα and TCRβ clonotypes (as characterized by V segment use, CDR3 sequence, and J segment use) expressed by newly developed Treg in NOD and B6 thymi. Rarefaction plots showed similar diversities of TCRα and TCRβ clonotypes expressed by NOD and B6 Treg (Fig. 3A and Supplementary Fig. 5A). Chao1, Shannon, and Simpson indexes confirmed the similar diversities of TCRα and TCRβ clonotypes in NOD and B6 samples (Fig. 3B and Supplementary Fig. 5B).

Previous reports indicated that compared with thymic Treg in B6 mice, those in NOD mice expressed a limited diversity of TCRα chains containing TRAV12 and TRAV9 (TRAV2 and TRAV17, respectively, according to the nomenclature of the international ImMunoGeneTics information system, www.imgt.org) (17,18) We therefore also separately analyzed the diversity of TCRα clonotypes using these TRAV. As shown in Fig. 3C and D and Supplementary Fig. 5C, we observed similar diversities in NOD and B6 samples.

These data indicate that the diversities of TCRs expressed by Treg newly developed in the NOD and B6 thymi are similar. This conclusion disagrees, in part, with earlier reports (17,18). One of the several potentially underlying reasons is that, in contrast to the earlier studies, we studied newly developed cells. A difference in the diversities of the very numerous recirculating cells in the pools of Treg in NOD versus B6 thymi (i.e., GFP⁻ cells in Fig. 1A) might provide a partial explanation. To assess this possibility, we performed TCRseq analyses on recirculating CD4⁺CD8⁻Thy1.1⁺GFP⁻ thymocytes from Foxp3-Thy1^a Rag2-Gfp NOD and B6 mice. Recirculating NOD versus B6 Treg expressed similarly (though not identically) diverse TRAV and TRBV repertoires (Supplementary Fig. 6A and B) and V-J associations (Supplementary Fig. 6C and D), and they had similar distributions and average values for CDR3 lengths and numbers of N and P nucleotides (Supplementary Fig. 6E–J). They also expressed similarly diverse TCRα and TCRβ repertoires (Supplementary Fig. 7A and B). Finally, the TRAV2 and TRAV17 repertoires expressed by recirculating Treg were also similarly diverse in NOD versus B6 thymi (Supplementary Fig. 7C and D).

The data we present here indicate that the CDR3 characteristics and the diversities of the TCRα and TCRβ repertoires expressed by Treg newly developed in the thymus of the T1D-prone NOD mouse are very similar to those found in the autoimmune disease-resistant mouse strain B6. This conclusion is principally based on our deep TCRseq analysis of Tcra- and Tcrb-derived mRNAs that allowed for an unbiased analysis of thousands of cells.

Importantly, TCRseq analysis requires numbers of cells that are very difficult to obtain from single mice. We therefore pooled 8–10 mice per sample analyzed. Whereas pooling several mice would have reinforced the statistical significance of differences found, it eliminates potential interindividual differences that may be biologically significant. It would therefore be of substantial interest to measure the interindividual variation of the diversities and the private and public TCR repertoires expressed by newly developed cells. To address such questions, the TCRseq procedure will need to be further substantially refined.

Despite the generally high levels of similarity, we observed some small differences between the TCR repertoires expressed by Treg newly developed in the thymi of NOD versus B6 mice. Thus, whereas they were similarly diverse, the TRBV and TRAV usages were distinct. This result was most probably due to the different alleles of MHC class II molecules expressed by the two mouse strains, I-A^g7 for NOD and I-A^b for B6 and, for TRBV, to the distinct MMTV-encoded superantigens encoded in their genomes (39). We also observed some minor but statistically significant differences in CDR3β length distribution between TCRs expressed by newly developing Treg in NOD versus B6 mice. Together with some minor differences in the numbers of N and P nucleotides, these observations may again be due to the distinct I-A alleles expressed in NOD versus B6 mice and/or to potentially slightly lower stringencies of the positive and negative selection processes in NOD compared with B6 mice (34–37). Analyses involving MHC-congenic mice will be required to shed light on these different possibilities.

Our results are in part different from previously published data (17,18). In the latter studies, thymic Treg were defined as CD4⁺CD8⁻CD25⁺ cells. This population is heterogeneous and contains, besides newly developed Treg, also immature Treg precursors and recirculating Treg (20,40). The recirculating Treg population in the thymus grows steadily with age and is particularly prominent in NOD mice (21). In 5- to 6-week-old NOD and B6 mice, analyzed in the previous reports, this population represents 47 ± 10% and 37 ± 9% of thymic Treg, respectively (21). However, our analysis of the TCR repertoires expressed by recirculating Treg indicated that they were similarly diverse in NOD versus B6 mice, thus excluding this potentially confounding factor. Interestingly, we found that the diversity of the TCR expressed by recirculating Treg in the thymus is very close to that of newly developed cells. Because mainly activated Treg migrate back to the thymus (21), this observation suggests a massive activation of autospecific Treg in the periphery, consistent with previously published data (41). Moreover, approximately half of the previously analyzed CD4⁺CD8⁻CD25⁺ cells are CD25⁺Foxp3⁻ Treg precursors (22). Thus, the newly developed Treg population actually represents a small minority of the cells previously analyzed. Using tools that have more recently become available, we were able to sort a pure population of newly developed Foxp3⁺ Treg. Moreover, the protocol used previously to determine the TCR-Vα encoding sequences was very different from ours. Whereas in the former case TRAV2- and TRAV17-targeted and non–UMI-based amplification of TRAV-cDNAs, followed by cloning and sequencing was used, we used a nonbiased and UMI-based protocol and high-throughput sequencing. The combined differences between the approaches taken previously and in the current study somehow explain the discrepancy of the results. Whatever the precise reason(s), our data indicate that the diversities of the TCRα and TCRβ repertoires expressed by newly developed Treg are very similar in NOD and B6 mice. Also when focusing our analysis on the two TRAVs analyzed previously, we found similar diversities between the repertoires expressed by newly developed NOD versus B6 Treg.

We therefore conclude that the repertoire of TCRs expressed by newly developed Treg in the NOD mouse is of similar diversity as that expressed in the control B6 strain. Importantly, this does not exclude the possibility that essential antigenic specificities may be lacking in the NOD Treg repertoire. Thus, the presentation of tissue-restricted antigens by medullary thymic stromal cells is known to be involved in shaping the TCR repertoire expressed by Treg (15). Genetic differences in expression of tissue-restricted antigens, as previously reported, may, for example, be involved in the susceptibility of the NOD mouse to T1D (42).

Acknowledgments. The authors are very grateful to the following individuals for excellent technical assistance: Fatima L’Faqihi, Valérie Duplan-Eche, Anne-Laure Iscache, Lidia De la Fuente, and Paul Menu of the Center for Pathophysiology of Toulouse Purpan Cytometry Platform, Adrien Castinel and Gwenolah Annonay of the GeT-PlaGE Genotoul, and the personnel of the INSERM US006 Anexplo/Centre Régional d'Exploration Fonctionnelle et Ressources Expérimentales animal facility. The authors are grateful to the members of the “T-cell mediated immune Tolerance team” and to Sylvie Guerder (Center for Pathophysiology of Toulouse Purpan) for discussion and input in the project. J.P.M.v.M. is grateful to the staff of the Biochemistry Institute of the Lausanne University, Epalinges, Switzerland, for its hospitality.

Funding. This work was financially supported by the Fondation pour la Recherche Médicale (FDT201904008280 to J.C.S. and DEQ20160334920 to J.P.M.v.M.), the IdEx Toulouse (to P.R.), the Conseil Régional Midi Pyrénées (15/06/12.05 to J.P.M.v.M.), and the Agence Nationale de la Recherche (ANR-16-CE15-0015-01 to P.R.). A.G.-A., S.C., J.C.S., O.P.J., and J.P.M.v.M. dedicate this work to the memory of their late, much respected, and regretted colleague P.R., who is deceased.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. A.G.-A. designed and performed bioinformatic analysis and helped to write the manuscript. S.C. designed and performed experiments and helped to write the manuscript. J.C.S. designed and performed experiments, performed analysis, and helped to write the manuscript. O.P.J. helped in designing the study and analyzing the data and reviewed the manuscript. B.H. analyzed data and helped to write the manuscript. P.R. helped in designing the study and analyzing the data. J.P.M.v.M. designed and supervised the study and wrote the manuscript. J.P.M.v.M. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.