Thymic regulatory T cell niche size is dictated by limiting interleukin 2 from antigen-bearing dendritic cells and feedback competition

T_reg cell development in thymic tissue slices

Previous reports have suggested that thymic T_reg cell development is limited by T_reg precursor frequency and competition for antigen, implying the existence of a limiting niche for T_reg cell development^8–11,17. To further investigate this niche, we utilized a thymic slice model in which a small number of thymocytes bearing a defined MHC class II-restricted TCR (OT-II) develop in the presence of their cognate antigen (ovalbumin). We used OT-II TCR transgenic thymocytes from a Rag2^−/− background (referred to as OT-II thymocytes) as a starting population with no detectable T_reg cells, providing a sensitive and synchronized system to track the development of new T_reg cell in response to antigen encounter within the thymic slice. We injected wild-type mice intravenously with 2 mg of soluble ovalbumin (OVA protein), an approach that was previously shown to induce antigen-specific thymic T_reg cells in vivo^18,19. One hour after injection, we sacrificed the mice, prepared thymic slices, and introduced OT-II thymocytes. Flow cytometric analysis revealed CD25 upregulation on OT-II CD4SP thymocytes after 1 day of culture, and the appearance of Foxp3⁺CD25⁺ cells after 3 days (Supplementary Fig. 1), consistent with the timing of T_reg cell development inferred from intrathymic transfer studies⁷.

T_reg cells can develop from immature T cells in the thymus, or can be derived from mature peripheral CD4⁺ T cells by TCR stimulation in the presence of the immunosuppressive cytokine transforming growth factor-β (TGF-β)²⁰. To confirm that the thymic slice system recapitulated normal thymic T_reg cell development, we compared the ability of OT-II thymocytes versus mature OT-II T cells to give rise to T_reg cells upon introduction into thymic slices (Supplementary Fig. 2a). Mature OT-II T cells showed some CD25 upregulation, but no Foxp3 induction, upon culture in thymic slices from mice injected with OVA protein. Addition of TGF-β led to a small population of Foxp3⁺ cells, however these cells expressed low amounts of the thymus-derived T_reg marker, neuropilin-1 (Supplementary Fig. 2a)^21,22. In contrast, OT-II thymocytes cultured in thymic slices gave rise to Foxp3⁺ cells without the addition of TGF-β, and expressed neuropilin-1. Neither OT-II thymocytes nor mature T cells gave rise to T_reg cells when stimulated with OVA-loaded DCs in vitro without thymic slices and without addition of TGF-β (Supplementary Fig. 2b). Moreover, depletion of mature CD4SP from OT-II thymocytes prior to addition to thymic slices delayed, but did not prevent, T_reg cell development (Supplementary Fig. 2c,d and data not shown) implying that Foxp3⁺ cells arose from an immature thymocyte population, rather than mature CD4SP thymocytes. These data indicate that the development of Foxp3⁺ cells from OT-II thymocytes on thymic slices is distinct from in vitro T_reg cell development from mature conventional CD4⁺ T cells, and suggest that the combination of the thymic environment and the developmental stage of the T_reg precursor are important for thymus-derived T_reg cell development.

To probe the role of antigen availability for T_reg cell development, we varied the antigen dose and delivery method. Injection of a high dose of soluble OVA led to OT-II T_reg cell development after 3 d on thymic slices, whereas injection of one-tenth the dose of OVA led to reduced, but detectable T_reg cell development. No CD25 or Foxp3 expression was observed when OT-II thymocytes were cultured in slices from control mice injected with PBS (Fig. 1a). As an alternative mode of antigen delivery, we utilized transgenic mice in which a membrane-associated form of ovalbumin is expressed in medullary thymic epithelial cells and the pancreas under the control of the rat insulin promoter (RIPmOVA)²³. Again, we observed a substantial population of CD25⁺Foxp3⁺ OT-II thymocytes after 3 d of culture on thymic slices from RIPmOVA transgenic mice (Fig. 1b). As a third approach to introduce antigen, we incubated bone marrow-derived DCs with ovalbumin protein and then introduced them to thymic slices containing OT-II thymocytes. Again we observed new T_reg cell development, which increased with antigen dose (Fig. 1c). In contrast to previous reports^10,11 we did not observe any correlation between the frequency of OT-II thymocytes within thymic slices and the efficiency of T_reg cell development (Supplementary Fig. 3), implying that T_reg precursor frequency is not a major limiting factor under our experimental conditions.

To compare the efficiency of T_reg cell development across various routes of antigen delivery, we expressed the number of Foxp3⁺ OT-II T_reg cells as a ratio of the total CD4SP thymocytes in the thymic slice. The efficiency of T_reg cell development varied with antigen abundance, consistent with previous studies^9,18. However, the proportion of OT-II thymocytes that developed into T_reg cells was generally less than 20% (Fig. 1 and Supplementary Fig. 3), and never exceeded the endogenous proportion of T_reg cells within a slice (corresponding to 3–10% of total slice CD4SP thymocytes) (Fig. 1d). This upper limit on T_reg cell development was observed even under conditions in which the majority of thymocytes responded to antigen, as indicated by CD25 induction (Fig. 1c,d). Together these results suggest that there is a limited thymic niche for T_reg cell development even when antigen encounter and T_reg precursor frequency are not limiting factors.

Existing thymic T_regs cells reduce the niche size

The finding that the overall number of thymic T_reg cells seems to be limited to fewer than 10% of the total CD4SP population suggests that T_reg cells themselves may inhibit new T_reg cell development. To test whether existing T_reg cells influence future T_reg cell development, we prepared thymic slices from AND×Rag2^−/− mice for use as a thymic environment that does not contain T_reg cells²⁴. For these experiments, we introduced antigen by adding 1 μM OVA peptide to thymic slices containing OT-II thymocytes. Interestingly, we observed a consistent increase in the number of CD25⁺ Foxp3⁺ OT-II thymocytes in T_reg-deficient (AND×Rag2^−/−) compared to T_reg-sufficient (wild-type) thymic slices (Fig. 2a,b). Enhancement of new T_reg cell development in T_reg-deficient slices was also observed using a different MHC class II specific TCR (2D2 ×Rag2^−/−) as the T_reg precursor population (Supplementary Fig. 4a,b), and using a different Rag2-deficient TCR transgenic model (F5×Rag2^−/−) as a source for T_reg-deficient thymic slices (Supplementary Fig. 4c,d). Finally, the enhanced T_reg cell development in thymic slices from TCR transgenic mice could be reversed by addition of sorted thymic T_reg cells to thymic slices (Fig. 2c,d), indicating that the lack of T_reg cells, rather than other abnormal features of TCR transgenic thymic environment, was responsible for enhanced T_reg cell development. Combined these results provide evidence that existing thymic T_reg cells limit the development of new T_reg cells.

T_reg cells do not interfere with peptide recognition

T_reg cells can reduce the stimulatory capacity of antigen-presenting cells (APCs) in the periphery, suggesting that thymic T_reg cells may inhibit new T_reg cell development by interfering with the ability of thymic APCs to present agonist self-peptides²⁵. To test this idea we examined the expression of the transcription factor Nur77, which provides a quantitative readout of accumulated TCR signal strength⁸. Addition of OVA peptide to T_reg-sufficient or T_reg-deficient slices resulted in similar expression of CD25 and Nur77 induction by OT-II thymocytes at 6 h (Fig. 3a–c), suggesting that existing T_reg cells did not influence the initial TCR signal received by T_reg progenitors. As a measure of accumulated TCR signal, we also examined Nur77 expression within the OT-II T_reg cell population that developed after 3 d in the presence or absence of existing T_reg cells (Fig. 3d). Consistent with previous reports⁸, endogenous T_reg cells from the thymic slice expressed slightly more Nur77 than conventional CD4SP thymocytes (Fig. 3e), and newly developed OT-II T_reg cells expressed slightly more Nur77 than endogenous slice T_reg cells (Fig. 3e, teal and magenta lines). Although the overall proportion of OT-II T_reg cell development was enhanced in the absence of existing T_reg cells, Nur77 expression within the OT-II T_reg cell population was similar for OT-II T_reg cells that developed on wild-type or T_reg-deficient slices (Fig. 3e). Together these results provide evidence that existing T_reg cells can limit new T_reg cell development without altering TCR signaling in T_reg precursors.

IL-2 availability influences the size of the T_reg niche

In addition to agonist self-peptide, developing T_reg cells require IL-2 to fully mature and express Foxp3^6,7. To confirm the requirement for IL-2 in the thymic slice model, we used Il2^−/− mice injected with OVA protein as a source of thymic slices for T_reg cell development. Half the number of OT-II thymocytes developed into T_reg cells after 3 d on Il2^−/− slices as compared to wild-type slices (Fig. 4a,b). This reduction in OT-II T_reg cell development is mirrored by a reduction in endogenous T_reg cells from Il2^−/− slices, and confirms the requirement for IL-2 for efficient thymic T_reg cell development and maturation^14–16. Importantly, these data indicate that any IL-2 produced by added OT-II thymocytes cannot fully compensate for the lack of IL-2 from the thymic environment, highlighting the importance of IL-2 sources within the thymic slice itself.

To determine whether IL-2 is a limiting factor for T_reg cell development, we tested the impact of IL-2 addition. We used IL-2–anti-IL-2 antibody complexes, which display an extended half-life compared to IL-2 alone^26,27. Addition of IL-2 complexes to thymic slices doubled the number of OT-II T_reg cells that developed per slice in both wild-type and Il2^−/− environments, suggesting that availability of IL-2 was a limiting factor for expansion of the T_reg niche (Fig. 4a, b). IL-2 addition also boosted the number of endogenous T_reg cells within the slice, implying that IL-2 is also limiting for T_reg precursors in the steady-state thymus (Fig. 4c). It is noteworthy that IL-2 addition did not fully restore T_reg cell numbers in Il2^−/− slices. This observation could reflect the requirement for a localized source of IL-2 for efficient T_reg cell development, or may be the result of the altered thymic environment in Il2^−/− mice. Together these data indicate that IL-2 is both necessary and limiting for thymic T_reg cell development within thymic slices.

Existing T_reg cells limit IL-2/IL-15 availability

Based on reports that peripheral T_reg cells can limit effector function via absorption of IL-2 from the environment^28,29, we considered that thymic T_reg cells may inhibit new T_reg cell development by limiting available supplies of IL-2 and IL-15, a related cytokine which also supports T_reg cell development¹⁴. To detect available cytokines in the thymus, we used the murine IL-2 (or IL-15)-dependent cell line, CTLL2³⁰. CTLL2 cells proliferated when added to wild-type thymic slices, consistent with the presence of available IL-2 and IL-15 within the tissue (Fig. 5a). Moreover, culture of CTLL2 cells on T_reg- deficient (AND×Rag2^−/−) slices lead to enhanced survival at d 2, and more robust proliferation on d 4 (Fig. 5a,b), suggesting increased abundance of available IL-2 and/or IL-15 in the T_reg-deficient, compared to the T_reg-sufficient, environment.

To further probe the role of T_reg cells in limiting IL-2 availability within the thymus, we added IL-2–anti-IL-2 complexes to thymic slices prepared from wild-type and T_reg -deficient mice that had been injected with OVA protein. Addition of IL-2–anti-IL-2 complexes to T_reg-sufficient wild-type slices increased the frequency and absolute numbers of OT-II T_reg cells 2 fold relative to that observed in T_reg-deficient slices without added IL-2 (Fig. 5c,d). Moreover addition of IL-2 complexes to T_reg-deficient slices further enhanced T_reg cell development 4 fold. These experiments demonstrate that while existing T_reg cells within the thymus limit IL-2 availability, IL-2 remains a limiting factor for new T_reg cell development even in the absence of existing T_reg cells. These data also indicate that existing T_reg cells can inhibit new T_reg cell development even in the presence of large quantities of IL-2, suggesting tight feedback control of IL-2 availability.

APCs provide a local source of IL-2 to T_reg precursors

We next investigated the cellular source of IL-2 for thymic T_reg cell development. While T cells are typically considered the major producers of IL-2, our observation that wild-type OT-II thymocytes could not rescue T_reg cell development in Il2^−/− thymic slices suggested that thymocytes might not provide an important source of IL-2 for T_reg cell development. Previous reports have suggested that DCs are capable of expressing low amounts of IL-2 in certain settings^12,13,31. To investigate whether antigen-bearing DCs might provide a local source of IL-2, we compared T_reg cell development promoted by wild-type versus Il2^−/− antigen-loaded DCs. Dendritic cells derived in vitro from the bone marrow of young, asymptomatic Il2^−/− mice expressed similar cell surface markers compared to wild-type bone marrow derived DCs (data not shown). Moreover, when Il2^−/− DCs were incubated with OVA protein and introduced into thymic slices, they led to comparable upregulation of TCR activation markers on OT-II thymocytes compared to wild-type DCs (Supplementary Fig. 5a–d), indicating that they have normal capacity to process and present antigen. However, OT-II T_reg cell development was reduced by nearly half using OVA-loaded Il2^−/− DCs as compared to OVA-loaded wild-type DCs (Fig. 6 a–c). As expected, there was no change in overall numbers of endogenous slice T_reg cells, regardless of whether added DCs were IL-2 deficient or sufficient, indicating that addition of wild-type DCs did not globally increase IL-2 abundance in the environment (Fig. 6d). Addition of IL-2-sufficient OVA-loaded DCs to IL-2-deficient thymic slices increased T_reg cell development over that observed when both DC and slices were deficient for IL-2 (Supplementary Fig. 6). Combined, these results indicate that antigen-bearing DCs can provide a relevant local source of IL-2 to promote T_reg cell development.

Our data, together with evidence that thymic DCs are important as APCs for T_reg cell induction^18,19,32,33, suggest that thymic DCs may express IL-2. Although we were unable to detect IL-2 protein in the thymus (data not shown), we were able to measure Il2 mRNA from thymus using a sensitive qRT-PCR assay (Fig. 6e). We found that purified thymic DCs expressed approximately 5 fold more Il2 mRNA compared to total, dissociated cells from the thymus, whereas the adherent thymic stromal cell compartment is depleted for Il2 mRNA compared to total thymus (Fig. 6e left hand plot). While thymic DCs express more IL-2 compared to thymocytes as a whole, it is noteworthy that depletion of CD11c⁺ cells from dissociated thymus samples did not significantly reduce the overall amount of Il2 mRNA in the samples. These data indicate that DCs are a substantial but not the sole, source of IL-2 within the thymus. Interestingly, while Il2 mRNA was also detectable over background in bone marrow derived DC, the level of expression was ~250 fold lower than the levels found in thymic DCs (Fig. 6e, right hand plot). This indicates that very low levels of IL-2 can have a potent impact on T_reg cell development when expressed by the same cell that also displays the antigenic peptide. Altogether our data support a model in which limited APC-derived IL-2, together with competition by existing T_reg cells, define the size of the thymic niche for new T_reg cell development (Supplemental Fig. 6c).

Mice

All mice were bred and maintained under pathogen-free conditions within American Association of Laboratory Animal Care approved facilities at the University of California, Berkeley. The internal review board and the Animal Use and Care Committee at UC Berkeley approved all procedures. C57BL/6, B6.SJL-Ptprca Pepcb/BoyJ (CD45.1), 2D2 TCR transgenic, AND TCR transgenic, B6.129P2-Il2tm1Hor/J(IL2−/−), C57BL/6-Tg(Ins2-TFRC/OVA)296Wehi/WehiJ (RIPmOVA), and Foxp3-GFP mice were purchased from Jackson. 2D2×Rag2^−/− mice were generated by crossing 2D2 mice with Rag2^−/− mice purchased from Taconic. OT-II×Rag2^−/− TCR transgenic mice were purchased from Taconic. F5×Rag2^−/− mice have been described previously⁴⁵. Endotoxin-free soluble ovalbumin was purchased from Invivogen, and injections were performed via i.v. injection, followed by sacrifice 1 h later.

Thymocytes, T_reg cell isolation and CTLL2 culture

Thymocytes were isolated from TCR transgenic (OT-II×Rag2^−/−, 2D2×Rag2^−/−) thymi via mechanical disruption. Cells were filtered through nylon mesh, and resuspended in DMEM complete media for overlay onto thymic slices. Thymocytes were distinguished from slice by use of congenic markers for CD45.1 and CD45.2, used in conjunction with staining for the specific TCRα chain corresponding to the transgenic mice. Thymic T_reg cells were isolated from Foxp3-GFP mice. Thymic cell suspensions were depleted of CD8SP and DP thymocytes using anti-CD8 microbeads (Miltenyi), then the remaining CD4 single positive thymocytes were sorted by flow cytometry based on GFP expression to obtain >95% pure thymic T_reg cells. Sorted thymic T_reg cells were resuspended in DMEM complete media for use with thymic slices. As a bioassay for available IL-2 and IL-15, we used a transformed IL-2/IL-15-dependent cell line derived from C57BL/6 mice called CTLL2. CTLL2 cells (a gift from N. Shastri lab, UC Berkeley, verified mycoplasma free prior to use) were cultured in RPMI complete media containing 10% FBS along with 50 U/ml IL-2 until being washed and overlaid on thymic slices. 1 × 10⁶ CTLL2 cells were added per slice, and allowed to migrate into slices for 4 h prior to washing.

Thymic slices

Thymic slice preparation has been described previously^46,47. Briefly, whole thymic lobes were isolated from mice, embedded in 4% low melt agarose and cut into 400 μm sections using a vibratome. Slices were cultured in 0.4 μm tissue culture inserts set in DMEM complete media. 1 × 10⁶ thymocytes resuspended in 10 μl of DMEM were overlayed on the slice and allowed to migrate into thymic slices for 2–4 h, followed by gentle pipetting to remove thymocytes that did not migrate into the slice. Under these conditions, 3000–10,000 thymocyte enter the slice, and migrate to their normal locations following chemokine gradients and other guidance cues within the tissue slice, as we and other have previously reported^48,49. Media was changed daily during the 3-day culture period. For sorted T_reg cell experiments, OT-II thymocytes were first overlaid for 4 h, followed by washing and addition of 1 × 10⁵ Foxp3-GFP T_reg cells for 4 h. Thymic slices derived from each thymus were randomized by order of slicing after the slicing process, prior to use for experiments. To minimize the impact of variation in the size of thymic slice and the proportion of cortex and medulla, we used adjacent thymic slices for experimental comparisons. In some cases, we normalized T_reg cell numbers to the number of total CD4SP (including slice-resident CD4SP), allowing us to pool data from multiple experiments into a single graph. For mature OT-II T cell experiments, T cells were isolated from the spleens of OT-II×Rag2^−/− mice and 1 × 10⁶ cells were overlaid onto thymic slices. In some cases, 5 ng/ml of TGF-β (Peprotech) was added along with the mature T cells to the slice.

Bone marrow derived dendritic cell cultures

Bone marrow was flushed from femurs and tibias of mice, followed by RBC lysis using ACK buffer. Cells were resuspended in 10 ng/ml GM-CSF and IL-4 (Peprotech), plated at 0.5 × 10⁶ /ml and cultured for 7 d. A half media change was performed on d2, and full media change with GM-CSF + IL-4 was performed on d5. Semi-adherent cells were isolated on d7, washed, and loaded with soluble OVA protein (1 mg/ml, Invivogen) for 30 min at 37 °C. Cells were washed 3x with HBSS, and overlaid onto thymic slices 4 h after addition of OT-II thymocytes. Cells were allowed to migrate into thymic slices for at least 4h prior to washing off excess. Bone marrow from Il2^−/− mice was harvested when mice were < 4 weeks of age, before the onset of pathology.

Quantitative PCR

Thymi from wild-type mice were dissociated using collagenase (Sigma-Aldrich.) digestion for 1 h at 37°C. Some sample was reserved for RNA isolation (“whole thymocyte” samples), and the remainder was used for DC enrichment using MACS CD11c microbeads according to manufacturer’s instructions (Miltenyi), followed by flow cytometry to reach a final purity >85% CD11c⁺ F4/80⁻ population (“thymic DC” samples). The thymus sample after removal of CD11c⁺ cells was also used for RNA isolation (“CD11c depleted” samples). Cell samples were lysed in Trizol (Life Technologies), and total RNA prepared using the RNAeasy Kit (Qiagen) as per manufacturer’s instructions. Reverse transcription was performed using the Quatitect RT Kit (Qiagen) and cDNA was utilized to run qPCR reactions using Taqman probes along with TaqMan Real-Time PCR master mix (Life Technologies). PrimeTime probes, (Mm.PT.58.11478202 (IL-2) and Mm.PT.39a.1 (GAPDH)), were synthesized by IDT. All reactions were run on an Applied Biosystems 7300 RT PCR machine. IL-2 expression data were normalized to GAPDH and expressed as fold increase over the background values from Il2^−/− bone marrow-derived DCs using ΔΔC_t values. (For Il2^−/− DCs, no IL2 signal was observed after 40 cycles of PCR, therefore values reported are upper estimates (indicated by grey shading), and were used to determine the lower limit of detection of the assay.

Flow cytometry

The following antibodies were used for cell surface staining: anti-mouse CD4-Alexa700 (clone GK1.5), CD8-APC-eFluor780 (clone 53–6.7), CD25-eFluor450 or PerCPCy5.5 (clone PC61.5), Vα2TCR-PerCP eFluor710 or APC (clone B20.1), Vα3.2TCR-FITC (clone RR3-16), CD45.1-PE-Cy7 (clone A20), CD45.2-FITC (clone 104) (eBioscience), or CD4-redfluor710, CD8-APC-Cy7 (Tonbo Bioscience), and Neuropilin-1-PE (clone 3E12) (Biolegend). Cell proliferation dye eFluor450 staining was performed at 2 μM for 10 min at 37 °C (eBioscience). For intracellular staining: Nur77-PE (clone 12.14), and Foxp3-PE (clone FJK-16s) or -APC were used (eBioscience). Cells were first surface stained for 15 min on ice, followed by fixation and permeabilization according to manufacturer’s instructions using Foxp3 Transcription Factor staining buffer set (eBioscience). Flow cytometry was performed using a BD Fortessa or Fortessa X20 analyzer (BD Bioscience), and data were analyzed using FlowJo software (Treestar).

Antibody complexes and peptides

IL-2:anti-IL-2 mAb complexes were described previously^26,27. Briefly, 1.5 μg IL-2 recombinant protein (eBioscience) was incubated with 7.5 μg of functional grade anti-IL-2 mAb (clone JES6-1, eBioscience) for 30 min at 37 °C. Complexes were added directly to thymic slices (10 μl/slice) daily over the course of the experiment. Mouse MOG 35–55 and chicken OVA 323–339 peptides were purchased from Anaspec. Peptide was overlaid on thymic slices for 10–30 min after thymocyte migration into the slice.

Statistics

Unpaired Student’s t-test and one-way analysis of variance (ANOVA) test with Tukey’s post-test were calculated using Prism software (Graphpad). A P value <0.05 was considered statistically significant. No data points were excluded and the number of mice used was consistent with previous experiments using similar experimental designs.