miR-181a/b-1 controls thymic selection of Treg cells and tunes their suppressive capacity

The interdependence of selective cues during development of regulatory T cells (Treg cells) in the thymus and their suppressive function remains incompletely understood. Here, we analyzed this interdependence by taking advantage of highly dynamic changes in expression of microRNA 181 family members miR-181a-1 and miR-181b-1 (miR-181a/b-1) during late T-cell development with very high levels of expression during thymocyte selection, followed by massive down-regulation in the periphery. Loss of miR-181a/b-1 resulted in inefficient de novo generation of Treg cells in the thymus but simultaneously permitted homeostatic expansion in the periphery in the absence of competition. Modulation of T-cell receptor (TCR) signal strength in vivo indicated that miR-181a/b-1 controlled Treg-cell formation via establishing adequate signaling thresholds. Unexpectedly, miR-181a/b-1–deficient Treg cells displayed elevated suppressive capacity in vivo, in line with elevated levels of cytotoxic T-lymphocyte–associated 4 (CTLA-4) protein, but not mRNA, in thymic and peripheral Treg cells. Therefore, we propose that intrathymic miR-181a/b-1 controls development of Treg cells and imposes a developmental legacy on their peripheral function.


Introduction
Regulatory T cells (Treg cells) expressing the lineage-defining transcription factor forkhead box protein P3 (Foxp3) form an integral part of the adaptive immune system and function to prevent unwanted immune responses [1,2] [3]. Accordingly, tTreg cells are generated when a TCR of a developing T cell recognizes a self-antigen with high affinity, as has been demonstrated in mouse models transgenic for both a TCR and its cognate antigen [4,5] and by analysis of polyclonal superantigen-reactive T cells [6,7]. tTreg cells can develop through two distinct precursor (prec) stages. Some Treg-cell precursors are found within a cluster of differentiation (CD) 4 single-positive (SP) Foxp3 − , glucocorticoidinduced tumor necrosis factor receptor-related protein high (GITR hi ), CD25 + population [8]. These cells are the first precursors generated in double transgenic TCR/cognate-antigen mouse models [9,10]. More recently, an additional CD4SP Foxp3 + CD25 − Treg-cell precursor has been described [11]. These cells are phenotypically less mature than tTreg cells, are generated with similar kinetics as tTreg cells upon induction of T-cell development in vivo, and efficiently become tTreg cells in vitro and in vivo [10,11]. Generation of both precursors is dependent on strong TCR signals, although on average, Foxp3 + CD25 -Treg-cell precursors have received somewhat weaker signals than their Foxp3 − GITR hi CD25 + counterparts [10]. Further differentiation into mature Foxp3 + CD25 + tTreg cells is then dependent on γc cytokines [8,[10][11][12]. The level of TCR signal strength required for tTreg cell generation in comparison to TCR signals resulting in clonal deletion have not been fully established. Data from a TCR signaling reporter as well as repertoire studies suggest that signal strength required for tTreg-cell development overlaps with that inducing clonal deletion in other autoreactive thymocytes [3,[13][14][15]. However, reduction of major histocompatibility complex (MHC) ligand levels on medullary thymic epithelial cells rescued autoreactive T cells from clonal deletion but resulted in a concomitant increase in Treg-cell development, suggesting that at least some tTreg cells are generated through weaker TCR signals than those inducing clonal deletion [16].

. Treg cells are generated during T-cell development in the thymus (thymic [t]Treg cells) as well as via peripheral induction of naive T cells (induced [i]Treg cells). Development of tTreg cells depends on strong T-cell receptor (TCR) signals
Treg cells suppress T-cell immune responses using multiple molecular mechanisms, including consumption of interleukin-2 (IL-2) and production of suppressive cytokines, as well as expression of the coinhibitory receptor cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) [17,18]. Mice specifically lacking the inhibitory receptor CTLA-4 in Treg cells succumb to fatal autoimmune disease, indicating that CTLA-4 plays a major role in suppressive function [19]. It has been proposed that CTLA-4 on Treg cell surfaces acts through capture of the costimulatory ligands CD80 and CD86 on antigen-presenting cells, thereby curtailing full activation of conventional T cells [20].
MicroRNAs (miRNAs) play a critical role in immune homeostasis and tolerance. Global loss of miRNAs results in defective development of tTreg cells [21]. However, no individual miRNA has been demonstrated to control intrathymic generation of Treg cells. miRNA miR-181a is the most prominently expressed miRNA in double-positive (DP) thymocytes, and it has been shown in vitro that miR-181a serves as a rheostat for TCR signals in T cells and thymocytes through targeting a combination of tyrosine and dual-specificity phosphatases, including protein tyrosine phosphatase, nonreceptor type (Ptpn) 11, Ptpn22, and dual specificity phosphatase 6 (Dusp6) [22][23][24]. Deletion of miR-181a/b-1 in mice resulted in an almost complete failure in development of invariant natural killer T (iNKT) cells and Mucosal-Associated Invariant T (MAIT) cells [25][26][27] due to a defect in thymic agonist selection [26,28]. Furthermore, loss of miR-181a/b-1 caused altered selection of conventional T cells in a TCR transgenic model with a shift towards positive selection [29]. However, counterintuitively, miR-181a/b-1 −/− mice display increased resistance to experimental autoimmunity, which has not been fully explained [29].
Here, we tested the hypothesis that miR-181a/b-1 controlled intrathymic development of Treg cells. De novo production of miR-181a/b-1 −/− tTreg cells was impaired because of altered sensitivity to TCR signals during selection. Generation of Treg cells in the absence of miR-181a/b-1 resulted in elevated expression of CTLA-4, which penetrated into the periphery despite the fact that peripheral WT Treg cells express very low amounts of miR-181a. As a consequence, miR-181a/b-1 −/− Treg cells had an increased suppressive capacity.

miR-181a/b-1 controls intrathymic Treg-cell development
First, we tested the hypothesis that miR-181a/b-1 might play a role during Treg-cell development. Using a recombination activating gene 1-green fluorescent protein (Rag1 GFP ) knock-in allele to discriminate between nascent and mature thymus-resident or recirculating Treg cells, we found that frequencies and absolute numbers of de novo generated Rag1 GFP -positive Treg cells were reduced by 2-to 3-fold in miR-181a/b-1 −/− mice when compared to control (ctrl), indicating that expression of miR-181a/b-1 is required for normal Treg-cell development in the thymus (Fig 1A). Competitive bone marrow (BM) chimeras with 1:1 mixtures of donor cells from wild-type (WT) and miR-181a/b-1 −/− mice revealed a disadvantage in Treg-cell generation, but not more immature double-negative and DP as well as CD4SP populations, from thymocytes of miR-181a/b-1 −/− origin, indicating that miR-181a/b-1 controls Treg-cell formation in a cell-intrinsic manner (Fig 1B and S1A Fig). In order to test how miR-181a/b-1 influenced developmental progression towards Foxp3 + CD25 + Treg cells, we analyzed inducible Rag1 (InduRag1) miR-181a/b-1 −/− mice, in which a wave of T-cell development can be induced by transient initiation of Rag1 gene expression through a tamoxifen-inducible Cre recombinase (Cre) [30]. At day 7 after Rag1 induction, only a few CD4SP thymocytes were generated, precluding robust analysis of Treg cell development (S1B and S1C Fig). However, we noted reduced frequencies of postselection DP thymocytes as well as postselection CD4SP thymocytes at this time point in miR-181a/b-1 −/− mice when compared to ctrls (S1D Fig). These data support the notion that selection processes are altered in the absence of miR-181a/ b-1. At day 14 and day 28, frequencies of Treg cells within CD4SP thymocytes were lower in miR-181a/b-1 −/− mice when compared to ctrls (Fig 1C). Consistent with findings at steady state, these data indicate that Treg-cell formation is less efficient in the absence of miR-181a/b-1. Next, we analyzed miR-181a/b-1-dependent formation of Foxp3 − CD25 + (prec 1b) or Foxp3 + CD25 − (prec 1a) Treg-cell precursors in the InduRag1 model. At day 14 after induction, both intermediates were present at somewhat similar frequencies in miR-181a/b-1 −/− mice when compared to ctrl, whereas at day 28, we observed reduced frequencies of Foxp3 + CD25 − precursors and an accumulation of Foxp3 − CD25 + precursors, suggesting that precursor generation is not restricted by a developmental block (Fig 1C). Taken together, these data indicate that in the absence of miR-181a/b-1, Treg cells are formed with slower kinetics rather than being subject to a defined developmental block. Treg-cell development in the thymus follows a somewhat different course in the absence of a full CD4SP compartment [10]. To account for such differences, we complemented the analysis of InduRag1 mice by taking advantage of the Rag1 GFP knock-in allele described above to temporally separate Treg-cell development at steady state. Green fluorescent protein (GFP + ) cells were arbitrarily gated into 5 populations based on different GFP levels, with loss of GFP expression indicating increasing amounts of time since cessation of Rag gene expression, which occurs in DP thymocytes. Frequencies of precursor 1b were elevated in GFP hi cells from miR-181a/b-1 −/− mice when compared to ctrls. In contrast, generation of precursor 1a as well as Treg cells was delayed in miR-181a/b-1 −/− mice when compared to ctrls (Fig 1D), which is consistent with data obtained in the InduRag1 model. We conclude that loss of miR-181a/b-1 results in an overall delay of Treg-cell formation, which is likely to be initiated prior to the emergence of defined Treg-cell precursors and cannot be compensated for by increased frequencies of Foxp3 -CD25 + precursors.

TCR signal strength determines miR-181a/b-1-dependent Treg-cell development
In order to test whether TCR signal strength differed in thymocytes from miR-181a/b-1 −/− mice, we assessed expression of Nuclear hormone receptor NUR/77 (Nur77) as a surrogate marker. Of note, miR-181a/b-1 −/− DP cells expressed lower levels of Nur77 prior to stimulation when compared to ctrls (Fig 2A). Furthermore, ex vivo stimulation of miR-181a/b-1 −/− DP cells failed to induce WT levels of Nur77, together suggesting that miR-181a/b-1 −/− DP cells received weaker TCR signals and failed to respond to TCR triggering with the same sensitivity as their WT counterparts (Fig 2A). In order to test whether TCR signaling was impaired prior to Treg-cell generation, we assessed surface expression of CD5, which correlates with TCR signal strength, on thymocyte subsets [31]. At steady state, total DP thymocytes, the majority of which have not undergone selection, displayed similar levels of surface CD5 in the presence and absence of miR-181a/b-1 ( Fig 2B). However, CD4SP thymocytes from miR-181a/b-1 −/− mice displayed lower surface levels when compared to ctrls. tTreg cells from either genotype expressed similar levels of Nur77 transcripts (Fig 2C), together indicating that miR-181-a/b-1 limits TCR signal strength prior to the emergence of Treg cells. Next, we took advantage of a system mimicking increased TCR signal strength during Treg-cell development via inducible nuclear translocation of the Nur77 family member Nr4a2 [32]. BM chimeric mice were generated to carry miR-181a/b-1-deficient or miR-181a/b-1sufficient ovalbumin-specific MHC class II-restricted alpha beta (OT-II) TCR transgenic cells expressing inducible Nr4a2. Transduction with ctrl virus resulted in generation of low frequencies of Treg cells from miR-181a/b-1-sufficient mice, and even fewer Treg cells emerged from miR-181a/b-1 −/− donor BM cells (Fig 2D). Upon activation of Nr4a2, frequencies of Treg cells generated from miR-181a +/− donors were slightly, albeit not significantly, reduced, suggesting that activation of Nr4a2 promotes a shift towards clonal deletion. In contrast, in the absence of miR-181a/b-1, Treg-cell development was rescued upon activation of Nr4a2, supporting the hypothesis that limited TCR signal strength accounts for defective Treg-cell differentiation in miR-181a/b-1 −/− mice. To corroborate these data, we analyzed chimeric mice (D) Overexpression of Nr4a2 (Nur77 family) rescues development of Treg cells deficient for miR-181a/b-1. LSK cells from OT-II × miR-181a/b-1 +/− or OT-II × miR-181a/b-1 -/mice were sorted and transduced with a retrovirus expressing a chimeric Nr4a2 molecule in which the Nr4a2 LBD is replaced by that of a mutant human estrogen receptor-α (Nr4a2-ΔLBD-ERT2) or ctrl retrovirus. Cells were injected into lethally irradiated WT recipients, and 7 weeks later, expression of Nr4a2 was induced via tamox administration for 5 consecutive days before analysis. Plots depict frequencies of Treg cells generated from precursor cells transduced with ctrl vector (upper panels) and from precursors transduced with vector carrying inducible Nr4a2 (lower panels). Graphs show summary of the results from 2 independent experiments, with n = 3 for each genotype and condition. (E) Less efficient clonal deletion and Treg-cell formation in the absence of miR-181a/b-1. OT-II × RIPmOVA chimeras were generated after lethal irradiation of RIPmOVA recipient mice (ctrl: WT/WT; transgenic: tg/WT) and injection of OT-II × miR-181a/b-1 +/− or OT-II × miR-181a/b-1 −/− BM cells. Mice were analyzed after 8 weeks. Plots depict frequencies of Vα2 + Vβ5 + OVA-specific donor cells in ctrl and tg recipients, which express OVA during negative selection. Graphs show summary of this analysis (upper panel) and absolute numbers as well as frequencies of Vα2 + Vβ5 + Foxp3 + Treg cells generated in the presence or absence of miR-181a/b-1 (lower panels). Data are representative of 2 independent experiments with n = 6-9 for each recipient/donor combination. Statistical analysis was performed using unpaired Student's t test (A, C, D). Numerical values are available in S1 Data. BM, bone marrow; CD, cluster of differentiation; ctrl, control; DP, double positive; ERT2, tamoxifen-inducible estrogen receptor 2; Foxp3, forkhead box protein P3; LBD, ligand-binding domain; LSK, lineage-negative, stem cell antigen-1-positive, Kit-positive; MFI, mean fluorescence intensity; miR-181, microRNA-181; Nr4a1, Nuclear receptor subfamily 4 group A member 1; Nur77, Nuclear hormone receptor NUR/77 encoded by Nr4a1; OT-II, ovalbumin-specific MHC class II-restricted alpha beta TCR; OVA, chicken ovalbumin; prec, precursor; RIPmOVA, rat insulin-promoter-driven membrane-bound chicken ovalbumin; SP, single positive; tamox, Tamoxifen; TCR, T-cell receptor; tg, transgenic; Treg cell, regulatory T cell; WT, wild type. generated using OT-II TCR transgenic miR-181a/b-1-sufficient (OT-II-ctrl) or deficient (OT-II-knockout [KO]) donor cells transferred into RIPmOVA recipients. RIPmOVA mice express the cognate antigen for the OT-II TCR in the thymus, resulting in clonal deletion of OT-II TCR transgenic cells as well as generation of low numbers of Treg cells. OT-II-KO>RIPmOVA chimeras showed lower levels of clonal deletion and generated considerably lower numbers of Treg cells when compared to OT-II-ctrl>RIPmOVA chimeras (Fig 2E). OT-II-ctrl>WT as well as OT-II-KO>WT chimeras failed to generate sizeable numbers of OT-II Treg cells and showed no signs of clonal deletion of OT-II thymocytes. We conclude that impaired generation of Treg cells in miR-181a/b-1-deficient mice is due to restricted TCR signal strength during thymic selection.

Homeostatic expansion generates normal Treg-cell numbers in the periphery of miR-181a/b-1 −/− mice
To address potential consequences of impaired tTreg-cell development in the absence of miR-181a/b-1, we next determined frequencies of Treg cells in the periphery. We did not observe any differences in frequencies and absolute numbers of peripheral Treg cells in spleens from miR-181a/b-1 −/− mice compared to ctrls ( Fig 3A). Consistently, frequencies of recirculating or thymus-resident (Rag1 GFP -negative) Treg cells in the thymus were largely unaffected in the absence of miR-181a/b-1, but we observed reduced frequencies of Rag1 GFP -positive recent thymic emigrants (RTEs) among peripheral Treg cells (S2A and S2B Fig). Equilibration of Tregcell numbers in the periphery can occur through homeostatic expansion of tTreg cells or preferential peripheral induction from naive T cells. Spleens of miR-181a/b-1-sufficient and deficient mice contained comparable frequencies of RTEs (TCRβ + Rag1-GFP + ) cells, which are enriched in peripheral Treg cell precursors (S2B Fig) [33]. Furthermore, CD4 + CD25 − RTEs from miR-181a/b-1 −/− mice did not produce more iTreg cells upon transfer into lymphopenic interleukin-7 receptor α gene (Il7r) −/− hosts when compared to ctrls, suggesting that Treg-cell induction is not the primary mechanism to equilibrate peripheral Treg-cell numbers in miR-181a/b-1 −/− mice (S2C Fig). Chimeric mice generated with 1:1 mixtures of miR-181a/b-1 −/− and WT BM showed that miR-181a/b-1 −/− Treg cells had a competitive disadvantage in the periphery when WT Treg cells were present, indicating that niche availability permits homeostatic expansion of tTreg cells in miR-181a/b-1 −/− mice ( Fig 3B). This conclusion was supported by Helios staining and TCR repertoire analysis. Elevated Helios expression has been associated with Treg-cell activation and proliferation [34]. Comparison of tTreg cells from miR-181a/b-1-sufficient and deficient mice showed low and similar expression of Helios between genotypes (Fig 3C). Staining in peripheral lymphoid organs (spleen, subcutaneous lymph node (scLN), and mesenteric lymph node [mLN]) revealed elevated numbers of Helios + Treg cells in miR-181a/b-1 −/− mice, indicating that in these mice, more Treg cells are in an activated/proliferative state ( Fig 3C).
We predicted that limited de novo generation and increased peripheral expansion resulted in reduced TCR diversity in peripheral Treg cells in the absence of miR-181a/b-1. Comparison of numbers of unique CDR3 sequences as well as calculation of effective number of species as a measure for repertoire diversity showed that TCR diversity in peripheral Treg cells from miR-181a/b-1 −/− mice was significantly reduced (Fig 3D and S3A Fig). In contrast, in the thymus, Treg cells from miR-181a/b-1 −/− mice displayed comparable TCR diversity as their ctrl counterparts ( Fig 3E and S3A Fig). Thus, these data indicate that, as a consequence of less efficient generation, fewer clones egress from the thymus to be available for peripheral expansion.
Notably, this was also the case for key Treg-cell signature genes (Fig 4B). Comparison of transcriptomes of thymic miR-181a/b-1 −/− Treg cells with their WT counterparts also revealed no significant differences both globally as well as with regard to Treg-cell signature genes ( Fig  4C). Given that miRNAs may also act on a post-transcriptional level, we next assessed expression levels of Treg-cell signature receptors and transcription factors. Peripheral and tTreg cells from miR-181a/b-1 −/− mice showed similar expression levels of most surface receptors and transcription factors analyzed as WT ctrls (S4A, S4B and S4C Fig). Notably, we detected strongly increased levels of total CTLA-4 protein in both peripheral and tTreg cells from miR-181a/b-1 −/− mice when compared to ctrls (Fig 4D). Although the coding sequence of Ctla4 mRNA contains putative miR-181 binding sites, direct modulation of CTLA-4 expression by miR-181 could not be observed in luciferase assays (S5A and S5B Fig).
We conclude from these data that thymic generation in the absence of miR-181a/b-1 results in post-transcriptionally controlled up-regulation of CTLA-4 protein in Treg cells while other Treg-cell signature genes remain unaffected. Loss of miR-181a expression in WT peripheral Treg cells suggests that elevated expression of CTLA-4 in these cells is imprinted during development.

Elevated levels of CTLA-4 in the absence of miR-181a/b-1 are maintained via increased rates of translation
In order to understand the underlying mechanisms of how elevated levels of CTLA-4 protein are maintained in peripheral miR-181a/b-1 −/− Treg cells, we analyzed its intracellular distribution using confocal microscopy. We confirmed elevated expression of CTLA-4 protein in the absence of miR-181a/b-1 ( Fig 5A). Next, we determined intracellular localization of CTLA-4 by costaining for Ras-related in brain protein 11 (Rab11, marking recycling endosomes), lysosome-associated membrane protein 2 (LAMP2, late endosomes), early endosome antigen 1 (EEA1, early endosomes), and cis-Golgi matrix protein 130 (GM130, Golgi). We detected no differences in the extent of colocalization of CTLA-4 with early endosomes and the Golgi apparatus in Treg cells from miR-181a/b-1 −/− mice compared to ctrls ( Fig 5B). However, we noted reduced colocalization of CTLA-4 with recycling endosomes but a marked increase in colocalization with late endosomes in miR-181a/b-1 −/− Treg cells when compared to ctrls (Fig 5C).
In order to assess whether aberrant localization of CTLA-4 in miR-181a/b-1 −/− Treg cells affected protein degradation, we stimulated Treg cells in the absence or presence of the translation inhibitor cycloheximide. Inhibition of translation for 2 h reduced CTLA-4 protein to similar levels in miR-181a/b-1 −/− Treg cells and ctrls ( Fig 5D). Given the higher protein levels when translation is active, these data imply that degradation rates of CTLA-4 are higher in the absence of miR-181a/b-1, which is consistent with its preferential localization in late endosomes rather than recycling endosomes. Furthermore, these data predict that if protein degradation is intact, elevated levels of CTLA-4 protein arise as a result of increased rates of protein translation. To test this prediction, we assessed accumulation of CTLA-4 protein in Treg cells in the presence of bafilomycin, an inhibitor of lysosomal protein degradation, in vitro. Over the course of 3 hours, miR-181a/b-1 −/− Treg cells accumulated significantly more CTLA-4 protein when compared to their WT counterparts ( Fig 5E). Together, these data indicate that elevated levels of CTLA-4 in peripheral Treg cells in the absence of miR-181a/b-1 are due to increased rates of translation. Increased protein translation in the absence of altered mRNA levels may be induced by loss of miRNAs other than miR-181a/b-1. To test this possibility, we performed small RNA sequencing (small RNAseq) of miR-181a/b-1-sufficient and deficient peripheral Treg cells. Consistent with the overall small changes in the transcriptome, we identified 4 miRNAs (miR-15b, miR-150, miR-342, and lethal (let)-7g) that were moderately However, in silico analysis of Ctla4 mRNA using Targetscan7 and RNA22 provided no evidence for the existence of either canonical or noncanonical binding sites for any of these miRNAs, suggesting that elevated protein levels of CTLA-4 are not caused by reduced miRNA expression. Next, we assessed whether peripheral expansion contributed to sustained expression of CTLA-4 in peripheral Treg cells. To this end, we generated mixed BM chimeras and analyzed CTLA-4 levels on miR-181a/b-1 −/− and WT Treg cells isolated from the same mice. CTLA-4 levels were consistently higher in miR-181a/b-1-deficient Treg cells despite competition by their WT counterparts, indicating that CTLA-4 levels are regulated cell-intrinsically and do not depend on peripheral expansion (Fig 5F). Finally, we tested whether alterations in tonic TCR signaling might result in elevated expression of CTLA-4. Freshly isolated peripheral Treg cells from miR-181a/b-1 −/− and WT mice expressed comparable levels of Nr4a1 mRNA, suggesting that tonic signaling through the TCR is similar (Fig 5G). Taken together, these findings further support the hypothesis that miR-181a/b-1-dependent ctrl of CTLA-4 expression is elicited in the thymus and subsequently sustained in the periphery.

Treg cells from miR-181a/b-1 −/− mice display increased suppressive capacity
In order to test the functional consequences of elevated levels of CTLA-4 in miR-181a/b-1 −/− Treg cells, we assessed their suppressive capacity. First, loss of miR-181a/b-1 resulted in reduced levels of tumor necrosis factor (TNF)α, IL-4, and IL-2 by conventional CD4 + T cells (Fig 6A). In contrast, expression of IL-17 and IL-10 remained unaffected. Whereas levels of TNFα were also reduced in miR-181a/b-1 −/− Treg cells, we did not observe additional significant miR-181a/b-1-dependent alterations in cytokine production by Treg cells or CD8 + T cells (Fig 6A and S6A Fig). Such alterations in cytokine profiles might be due to cell-intrinsic effects or due to modulation of Treg-cell suppressive capacity. Therefore, we next assessed in  co-transferred with miR-181a/b-1 −/− Treg cells when compared to those co-transferred with miR-181-a/b-1-sufficient Treg cells (Fig 6B). This reduction in frequency was reflected by lower absolute numbers of Tconv cells recovered in the presence of miR-181a/b-1 −/− Treg cells, whereas absolute numbers of Treg cells recovered were similar in both conditions ( Fig  6C). Together, these data indicate that in the absence of miR-181a/b-1, Treg cells have a stronger capacity to suppress lymphopenia-driven expansion of Tconv cells in vivo. Of note, we did not observe significant alterations in suppressive capacity of miR-181a/b-1 −/− Treg cells in vitro (S6B Fig). Taken together, our data indicate that miR-181a/b-1 controls intrathymic Treg cell development in a TCR-dependent manner. Impaired Treg-cell development in the absence of miR-181a/b-1 is associated with post-transcriptional up-regulation of CTLA-4, which penetrates into the periphery and results in increased suppressive capacity. Low levels of miR-181a in peripheral WT Treg cells suggest that the effects of loss of miR-181a/b-1 are imprinted during intrathymic development.

Discussion
Here, we demonstrated that intrathymic generation of Treg cells depends on miR-181a/b-1 via establishing signaling thresholds to adequately respond to strong TCR signals. In the absence of miR-181a/b-1, de novo generation of Treg cells was attenuated and resulted in Treg cells expressing elevated levels of CTLA-4. Homeostatic expansion resulted in a completely filled peripheral Treg-cell compartment while CTLA-4 levels remained elevated via a post-transcriptional mechanism, resulting in Treg cells with increased suppressive capacity.
Treg cells develop from CD4SP thymocytes through two possible intermediates, Foxp3 − CD25 + and Foxp3 + CD25 − [8,11]. It has been suggested that generation of these precursors occurs through a TCR-dependent step, whereas further maturation into mature Foxp3 + CD25 + Treg cells is dependent on the cytokines IL-2 and IL-15 [8,10,11]. Analysis of InduRag1 mice as well as a Rag1 GFP virtual timer indicated that miR-181a/b-1 predominantly affects formation of Foxp3 + CD25 − precursors, whereas Foxp3 − CD25 + are more frequent in miR-181a/b-1-deficient mice. Nevertheless, these precursors cannot compensate for the partial loss of Foxp3 + CD25 − precursors, suggesting that the major route of Treg-cell development is through the latter. Indeed, it has been shown that in WT mice, only approximately 20% of CD4 + Foxp3 − CD25 + cells ultimately give rise to Treg cells [8,10]. Treg-cell development via Foxp3 − CD25 + intermediates predominantly occurs in double transgenic mouse lines expressing a transgenic TCR plus its cognate antigen [9]. Furthermore, these cells express higher levels of a Nur77 GFP reporter than either Foxp3 + CD25 − precursors or mature Treg cells [10]. Thus, it has been suggested that Foxp3 − CD25 + intermediates arise at the extreme end of the TCR affinity spectrum and might increase in frequency by an influx of cells otherwise targeted for clonal deletion. Accordingly, reduction in MHCII levels on thymic epithelial cells in a monoclonal system diverted thymocytes from clonal deletion into the Treg cell lineage [16]. TCR repertoire analyses and autoreactivity suggest that TCR signal strength required for tTreg-cell generation overlaps both with that of positively selected thymocytes as well as that of cells normally undergoing clonal deletion [3,13,14]. Rescue experiments performed in this study agree with both non-mutually exclusive models. Fewer donor-derived miR-181a/b-1 −/− OT-II Treg cells developed in RIPmOVA antigen transgenic mice compared to WT OT-II Treg cells. Concomitantly, clonal deletion of miR-181a/b-1 −/− OT-II cells was also impaired in RIPmOVA mice, test. Numerical values are available in S1 Data. CD, cluster of differentiation; IFN-γ, interferon gamma; IL, interleukin; miR-181, microRNA-181; PMA, phorbol 12-myristate 13-acetate; pos., positive; Tconv, conventional T; TCR, T-cell receptor; TNF, tumor necrosis factor; Treg cell, regulatory T cell. suggesting that in this particular paired TCR-antigen model, TCR signal strength is reduced through loss of miR-181a/b-1 to limit both Treg-cell formation and clonal deletion. Conversely, induced expression of Nr4a2 promoted Treg-cell production in the absence of miR-181a/b-1 but resulted in somewhat limited production of Treg cells in the presence of miR-181a/b-1, suggesting that in the latter case, clonal deletion might be favored over Treg-cell development.
Intrathymic development of Treg cells depends on CD28-mediated costimulatory signals [36]. Thus, it might be possible that elevated expression of CTLA-4 by Treg cells in the thymus contributes to impaired development. Costimulation via CD28 is required for efficient generation of Foxp3 − CD25 + Treg-cell precursors but less so during later Treg-cell development, suggesting that CD28 signaling protects thymocytes from clonal deletion [36][37][38][39]. Loss of CD28 signaling does not result in export of autoreactive cells into the periphery, indicating that it does not simply act as an amplifier of TCR signal strength [39]. Consistently, loss of CD28-mediated costimulation and loss of miR-181a/b-1 generate phenotypically distinct developmental defects, also supporting the notion that elevated levels of CTLA-4 in miR-181a/b-1 −/− Treg cells are a consequence rather than cause of inefficient generation of Treg cells in these mice.
Consequences of altered TCR signal strength in the thymus have been previously analyzed in mice carrying hypomorphic mutations of key signal mediators, such as zeta-chain-associated kinase of 70 kD (Zap-70), or reduced numbers of immunoreceptor tyrosine-based activation motifs (ITAMs) within CD3z molecules [40][41][42]. Collectively, these studies showed that alterations in TCR signal affected positive and negative selection as well as Treg-cell formation, albeit in a manner that is not easily predictable. Thus, these data indicate that the quantitative relationship between proximal TCR signaling and effcient thymic selection needs to be tightly balanced. Furthermore, mutations characterized in these studies equally affect T-cell activation and tonic signaling in the periphery, precluding analysis of developmental consequences of altered TCR signaling exclusively occurring in the thymus.
In contrast, expression levels of miR-181a/b-1 in peripheral Treg cells are very low and should therefore allow WT-like levels of tonic TCR signaling. Notably, peripheral TCR signaling controls Treg-cell homeostasis and helps to maintain functional Treg cells [43,44]. For instance, in the absence of peripheral TCR expression, levels of CTLA-4 protein are reduced, and suppressive capacity is compromised [44].
Our data suggest that alterations in thymic selection caused by the absence of miR-181a/b-1 have long-term impact and are translated to increased suppressive activity of peripheral Treg cells. We therefore propose that the developmental legacy of TCR signal strength during agonist selection determines Treg-cell function in the periphery. Thus, altered TCR thresholds during selection might affect a Treg cell's responsiveness to tonic signaling. Similar observations have previously been reported for both CD4 and CD8 Tconv cells [45,46]. The avidity of positively selecting self-peptides and thus strength of the TCR signal during selection determines the outcome of a T-cell immune response even of T cells recognizing the same foreign antigen with an identical avidity [45]. In contrast to Treg cells, differential reactivity to selfpeptide by CD8 Tconv cells was accompanied by clear differences in gene expression profiles [46]. Although in Tconv cells, the capacity for tonic signaling in the periphery contributes to distinct responsiveness to pathogens, thymically predetermined levels of the feedback regulator of TCR signaling CD5 are likely to help control tonic signals [45]. Thus, the quality of protective T-cell responses as well as Treg-cell mediated suppression appear to be preset during thymic selection.
How TCR signal strength during thymic agonist selection confers long-term changes in CTLA-4 protein expression remains unclear. Our study supports a model in which expression of CTLA-4 is cell-intrinsically sustained in the periphery through a post-transcriptional mechanism controlling its translation rate. Translational control can be exerted at multiple different levels, including changes in mRNA composition through alternative splicing and miR-NAs as well as RNA binding proteins. Given the lack of umambiguous evidence for one of these mechanisms being predominant, our data suggest that multiple factors may act in concert to control CTLA-4 protein. Tight control of CTLA-4 expression is likely to be paramount for tuning suppressive capacity of Treg cells.
Our study establishes miR-181a/b-1 as a central regulator of agonist-selected αβT cells. Earlier studies showed that miR-181a/b-1 is critical for the development of innate-like T cells expressing semi-invariant TCRs, such as iNKT cells and MAIT cells (but not their γδTCRexpressing counterparts) [25,26,47]. Here, we demonstrated that the role of miR-181a/b-1 can be extended to highly diverse polyclonal T-cell populations. This finding was not anticipated because it might be expected that a shift in integrated signal strength might be compensated for by a shift in the repertoire. Such compensation might partly explain why the effect of miR-181a/b-1 deletion on Treg cells is somewhat milder when compared to other agonist selected T cells. Finally, based on the dramatic down-regulation of miR-181a after the DP stage, our study implies that the lineage fate decision to become a Treg cell manifests itself early during selection.

In vivo induction of Treg-cell development
Treg-cell development was induced in 6-to 8-week-old InduRag1 × miR-181a/b-1 −/− and InduRag1 × miR-181a/b-1 +/− mice by oral administration of tamoxifen (Ratiopharm, Ulm, Germany) in corn oil (Sigma Aldrich, St. Louis, MO, USA), 0.6 mg/400 μl/mouse every 4 days. To ensure proper tamoxifen solubility, in the first step, it was dissolved in 100% ethanol prewarmed to 55˚C and later in prewarmed corn oil. Before each oral administration, the mixture was always prewarmed to 37˚C.

Overexpression of Nr4a2 in BM chimeras
Six weeks after injection, mice were orally administered 0.6 mg/400 μl of tamoxifen (Ratiopharm) dissolved in corn oil (Sigma Aldrich) per recipient mouse for 5 consecutive days. Tamoxifen preparation was performed as for induction of Treg-cell development in InduRag1 mice. Twelve hours after the final administration, mice were analyzed.

RIPmOVA × OT-II BM chimeras
BM cells were isolated from the tibia and femur of age-matched (7-10 weeks) OT-II × miR-181a/b-1 −/− and OT-II × miR-181a/b-1 +/− mice. Red blood cells were lysed, and cells were counted and injected (5 × 10 6 /recipient) into the lateral tail vein of lethally (2 × 4.5 Gy) irradiated RIPmOVA (CD45.1/2 + ) recipient mice. Mice were provided with antibiotic-containing water and were housed in sterile microisolator cages. Analysis of BM chimeras was performed 8-12 weeks after transplantation. For intracellular stainings, an intracellular staining buffer set and a Foxp3/ transcription factor staining buffer set (both eBioscience) were used according to the manufacturer's protocol. RNA flow cytometry was performed using the PrimeFlow system (Thermo Fisher Scientific). Cells were stained with Nr4a1-AF647. Samples were acquired on LSRII (BD Biosciences) cytometers and sorted on a FACS Aria II (BD Biosciences). Data were analyzed with FlowJo software, v.9.4.9 (Tree Star, Ashland, OR, USA). For analysis, dead cells and debris were excluded by gating of forward and side scatter. Sorted cells were of 98% or higher purity, as determined by reanalysis.

Microarrays
RNA isolation, cDNA preparation, and DNA microarray analysis of gene expression were performed at the Microarray Genechip Facility of the University of Tübingen (MFT Services). Fluorescent images of hybridized microarrays (MOE-430 version 2.0; Affymetrix, Santa Clara, CA, USA) were obtained using an Affymetrix Genechip Scanner. Microarray data were analyzed using BioConductor Suite 2.1 software. All samples were repeated two times with individually sorted cells and averaged.

Small RNAseq
Treg cells (1 × 10 5 ) sorted from 3 pooled WT and miR-181a/b-1 −/− thymi were stored in RNAprotect Cell Reagent (Qiagen). Small RNAseq was performed by Admera Health (South Plainfield, NJ, USA) using the SMARTer smRNA-Seq Kit (Takara, Kusatsu, Japan). Adapters were trimmed with Flexbar 3.4, and rRNA was removed using Bowtie 2. The remaining reads were aligned using STAR aligner and counted using HTSeq. Differential expression analysis was performed in R using the DESeq2 package. Three biological replicates per genotype were analyzed.