Mouse and human CD8⁺ CD28^low regulatory T lymphocytes differentiate in the thymus

Summary

Regulatory T (Treg) lymphocytes play a central role in the control of immune responses and so maintain immune tolerance and homeostasis. In mice, expression of the CD8 co‐receptor and low levels of the co‐stimulatory molecule CD28 characterizes a Treg cell population that exerts potent suppressive function in vitro and efficiently controls experimental immunopathology in vivo. It has remained unclear if CD8⁺ CD28^low Treg cells develop in the thymus or represent a population of chronically activated conventional T cells differentiating into Treg cells in the periphery, as suggested by their CD28^low phenotype. We demonstrate that functional CD8⁺ CD28^low Treg cells are present in the thymus and that these cells develop locally and are not recirculating from the periphery. Differentiation of CD8⁺ CD28^low Treg cells requires MHC class I expression on radioresistant but not on haematopoietic thymic stromal cells. In contrast to other Treg cells, CD8⁺ CD28^low Treg cells develop simultaneously with CD8⁺ CD28^high conventional T cells. We also identified a novel homologous naive CD8⁺ CD28^low T‐cell population with immunosuppressive properties in human blood and thymus. Combined, our data demonstrate that CD8⁺ CD28^low cells can develop in the thymus of mice and suggest that the same is true in humans.

Abbreviations

BM

bone marrow

CD8SP

CD8⁺ CD4⁻ TCR^high (CD8 single‐positive) thymocyte

Foxp3

forkhead/winged helix transcription factor

GFP

green fluorescent protein

IBD

inflammatory bowel disease

LAP

latency‐associated peptide

MFI

mean fluorescence intensity

RAG

recombinase activating gene

Tconv

conventional T lymphocyte

TCR

T‐cell receptor

TGF‐β

transforming growth factor β

Treg

regulatory T

β2m

β2‐microglobulin

Introduction

Regulatory T (Treg) lymphocytes play a major role in the modulation of innate and adaptive immune responses to self‐ and non‐self antigens. Their contribution to the maintenance of immune homeostasis is significant, ranging from prevention of autoimmune pathology to inhibition of chronic inflammation in the gut, fine regulation of immune responses to infection, and protection of the semi‐allogeneic fetus in the uterus. To date, the most extensively studied Treg cell population is characterized by expression of the co‐receptor CD4 and the forkhead/winged helix transcription factor Foxp3.1, 2 However, other regulatory subsets have recently been described within both CD4⁺ and CD8⁺ T‐cell populations. Accumulating evidence suggests that these populations play an important role in immune regulation.3, 4

In mice, reduced expression of the co‐stimulatory molecule CD28 characterizes a CD8⁺ population with potent in vitro and in vivo immunoregulatory capacity. CD8⁺ CD28^low Treg cells inhibit proliferation and interferon‐γ production by conventional T (Tconv) cells in vitro. Moreover, they prevent experimental autoimmune encephalomyelitis, myasthenia gravis and inflammatory bowel disease (IBD).5, 6, 7 Protection from IBD requires interleukin‐10 (IL‐10) production by the Treg cells and transforming growth factor‐β (TGF‐β) responsiveness of the colitogenic CD4⁺ T cells.6 Where CD8⁺ CD28^low Treg cells differentiate remains unknown. In humans, CD8⁺ Treg cells lacking expression of CD28 have been observed in the blood of patients that had accepted, but not of those that had rejected, a kidney or heart transplant,8, 9 and among tumour‐infiltrating lymphocytes as well as in the blood of cancer patients.10 However, these CD8⁺ CD28^– cells are phenotypically very distinct from mouse CD8⁺ CD28^low Treg cells, which express clearly detectable levels of CD28. A homologous CD8⁺ CD28^low population has, to our knowledge, not been described in humans. Here, we assessed the existence of human CD8⁺ CD28^low Treg cells.

The extensively described Foxp3⁺ Treg cells have two distinct origins. Treg cells can develop in the thymus (tTreg) from haematopoietic precursors and they can also differentiate from activated Tconv cells in the periphery (pTreg) under special ‘tolerogenic’ conditions.11 The circulating pool of Treg cells is therefore composed of both tTreg and pTreg cells. Recent evidence suggests that these two populations have distinct non‐redundant functions in maintaining immune tolerance. In the IBD model, only the combination of both CD4⁺ Foxp3⁺ tTreg and pTreg cells could fully protect from the pathology, suggesting a non‐redundant division of labour between these two populations.12 These populations also acted synergistically in the prevention of the lymphoproliferative disorder and autoimmunity in Foxp3‐deficient mice.13 Recently, Zheng et al. have identified the intronic Foxp3 enhancer CNS1 as being a critical element for the differentiation of Foxp3⁺ pTreg cells.14 In CNS1‐deficient mice, reduced levels of pTreg cells led to T helper type 2 inflammation confined to the gastrointestinal tract, further underlining the distinct functions of tTreg and pTreg cells in immune homeostasis.15 Identifying the (intra‐ or extra‐thymic) origin of CD8⁺ CD28^low Treg cells is therefore an essential step towards understanding their physiological relevance. Here we investigated the potential thymic origin of CD8⁺ CD28^low Treg cells in mice and in humans.

Materials and methods

Mice

Mice were used at 6–10 weeks of age, unless indicated otherwise. NMRI and C57BL/6 (B6) mice were purchased from the Centre de Recherche et d'Elevage Janvier (Le Genest St Isle, France). B6 Thy1.1, B6 β2‐microglobulin (β2m),16 and B6 Rag‐GFP mice17 were bred in our specific pathogen‐free animal facility. All experiments involving animals were performed in compliance with the relevant laws and institutional guidelines (Regional approval # 31 09 555 45, ethical review # MP/08/14/02/12).

Human samples

Peripheral blood mononuclear cells were isolated from buffy coats obtained from the local blood transfusion service (Etablissement Français du Sang, Purpan Hospital, Toulouse). Human thymus tissue from children undergoing cardiac surgery was obtained from the Cardiology Department of the Children's Hospital, Purpan. In line with French guidelines, the French authorities authorized the study (MESR # DC‐2014‐2088) and the participants' legal representatives did not oppose the use of the thymic fragments for research purposes.

Flow cytometry analysis

The following reagents specific for mouse antigens were used: anti‐T‐cell receptor‐β (TCRβ‐FITC; anti‐TCRβ‐allophycocyanin); anti‐CD8α‐AF700; anti‐CD28‐biotin; streptavidin‐phycoerythrin (PE), all from eBiosciences (San Diego, CA); anti‐CD4‐PE‐Cy7; anti‐CD4‐Pacific Blue; anti‐CD8α‐FITC; anti‐CD25‐PE; anti‐Foxp3‐PE, all from BD‐Pharmingen (Heidelberg, Germany); anti‐Neuropilin‐PE from R&D Systems (Lille, France); anti‐Helios‐AF488 from BioLegend (San Diego, CA). The following reagents specific for human antigens were used: anti‐CD3‐PE‐Cy7; anti‐CD4‐FITC; anti‐CD8‐BV421; anti‐CD8‐AF700; anti‐CD28‐PE; anti‐CD45RA‐PE‐CF594; anti‐IL10‐PE; anti‐latency‐associated peptide (LAP)‐PE, all from BD‐Pharmingen. TO‐PRO‐3® was purchased from Life Technologies (Saint‐Aubin, France). Cells were incubated with saturating levels of antibodies. Hence labelled cells were analysed on a FACS LSRII (Becton Dickinson, Pont de Claix, France) using DIVA (Becton Dickinson) and flowjo software (Tree Star, Ashland, OR). Compensation between the distinct fluorescence channels was carefully verified using single stainings. Mature CD8⁺ T cells were electronically gated as CD4⁻ CD8⁺ TCR^high. Electronic CD28^low gates were placed to include the minimal estimate of the percentage of CD28^low cells [i.e. % of cells with CD28 level < mean fluorescence intensity (MFI) – % of cells with CD28 level > MFI].

Isolation of T‐cell subsets

Murine splenic CD8⁺ CD28^low Treg cells were isolated as previously described.6 To isolate murine thymic CD8⁺ CD28^low and CD8⁺ CD28^high T cells, cell suspen‐sions were first depleted of CD4⁺ thymocytes through complement‐mediated lysis using an anti‐CD4 antibody (RL172.4) and rabbit complement (HD Supplies, Aylesbury, UK). The cell suspension was then labelled with anti‐FcγRII/III (2.4G2), anti‐CD4 (RL172.4), anti‐ MHC class II (M5/114), and anti‐B220 (RA3‐6B2) antibodies before negative selection with sheep anti‐rat IgG Dynabeads (Dynal biotech, Oise, France). The resulting population was labelled with anti‐CD8α‐AF700, anti‐CD4‐Pacific Blue, anti‐TCR‐allophycocyanin and anti‐CD28‐biotin followed by streptavidin‐PE, and CD8⁺ CD4⁻ TCR^high CD28^high and CD8⁺ CD4⁻ TCR^high CD28^low cells were sorted on a FACSAria II SORP (Becton Dickinson). Isolation of GFP⁺ CD8⁺ CD4⁻ TCR^high CD28^low cells from B6 Rag‐GFP mice required an additional gating on GFP⁺ cells before cell sorting. Murine responder CD4⁺ T cells used in in vitro assays were enriched from erythrocyte‐depleted splenocytes by Dynabead‐mediated depletion of FcγRIII⁺, MHC class II⁺, CD8⁺ and B220⁺ cells.

Human CD8⁺ CD28^low and CD8⁺ CD28^high T cells were isolated from peripheral blood mononuclear cells as follows. Cell suspensions were first incubated with a cocktail of mouse antibodies specific for CD4⁺ T cells, B cells, natural killer cells, monocytes, platelets, dendritic cells, granulocytes and erythrocytes (Life Technologies) and labelled cells were then eliminated using anti‐mouse IgG‐labelled Dynabeads (Dynal biotech). The resulting population was labelled with anti‐CD3‐PE‐Cy7, anti‐CD8‐BV421, anti‐CD28‐PE and anti‐CD45RA‐PE‐CF594, the CD8⁺ CD3^high CD45RA^high CD28^high and the CD8⁺ CD3^high CD45RA^high CD28^low populations were sorted using a FACSAria II SORP (Becton Dickinson). Human thymic CD8⁺ CD28^low and CD8⁺ CD28^high T cells were isolated as follows. Cell suspension was incubated with anti‐CD4 (OKT4) antibodies and labelled cells were eliminated as described above. The resulting population was labelled with anti‐CD3‐PE‐Cy7, anti‐CD4‐FITC, anti‐CD8‐AF400, anti‐CD28‐PE antibodies and CD8⁺ CD4⁻ CD3^high CD28^high and CD8⁺ CD4⁻ CD3^high CD28^low cells were sorted using a FACSAria II SORP (Becton Dickinson). A purity of > 94% was obtained routinely for both murine and human sorted T‐cell populations.

Production of immunosuppressive mediators by human T cells

Sorted cells were stimulated for 1 week with bead‐bound anti‐CD3 and anti‐CD28 antibodies (Life Technologies) in a 1 : 1 cell‐to‐bead ratio in the presence of IL‐2 (30 U/ml, Peprotech, Neuilly sur Seine, France). Cells were then stained with the indicated cell‐surface markers, anti‐LAP antibodies, and TOPRO‐3 to exclude dead cells. For intracellular detection of IL‐10, cells were re‐stimulated with PMA (50 ng/ml) and ionomycin (1 μg/ml), both from Sigma‐Aldrich (Saint‐Quentin Fallavier, France) for 5 hr at 37° and Brefeldin A (10 μg/ml, eBiosciences) was added for the last 4 hr. Cells were stained for the indicated surface markers, fixed with 2% paraformaldehyde for 30 min at 4°, permeabilized in 0·5% saponin/1% BSA in PBS for 30 min at room temperature, and then incubated for 30 min at room temperature with PE‐conjugated anti‐IL‐10 in permeabilization buffer.

Bone marrow chimeras

Bone marrow cells from femurs and tibias were prepared as previously described.18 Then, 5 × 10⁶ donor cells were injected intravenously into γ‐irradiated hosts (8·5 Gy γ; ¹³⁷Cs source, 6·3 Gy/min).

In vitro suppression assays

Suppressive activity of indicated T‐cell populations was assessed as described previously.19

In vitro thymic organ cultures

Thymic lobes were surgically removed from NMRI fetuses at gestational day 15. The lobes were placed on cell culture inserts (pore size: 0·4 μm) in six‐well tissue culture plates (Becton Dickinson) with standard RPMI complete medium supplemented with 10% FCS. At different days of culture, the thymic lobes were harvested and single cell suspensions were prepared and analysed for CD4, CD8, CD28 and TCR‐β expression by flow cytometry.

Statistical analysis

Statistical significance was determined using the Mann–Whitney and the Wilcoxon tests.

Results

Functional CD8⁺ CD28^low Treg cells are present in the mouse thymus

We have previously observed that CD8⁺ CD28^low T cells freshly isolated from the spleen of wild‐type (wt) mice exerted suppressive activity in vitro and in vivo.6 To determine if these cells are present in the thymus of wt mice, we analysed the expression of CD28 on mature CD4⁻ CD8⁺ TCR^high (‘CD8SP’) thymocytes (Fig. 1a). If the distribution of CD28 staining of a population were Gaussian, the median and mean fluorescence intensities would be equal. However, the median fluorescence intensity of the CD28 staining on CD8SP cells was lower than its mean value (475 ± 65 versus 657 ± 74, P < 0·001, n = 8), clearly indicating the existence of CD28^low cells. Similar observations were made on CD8⁺ TCR^high splenocytes (352 ± 55 versus 486 ± 69, n = 9, P < 0·001). By contrast, the median and mean fluorescence intensities of CD28 staining on CD4⁺ TCR^high thymocytes were similar (1794 ± 308 versus 1954 ± 344, n = 8, P = 0·23), indicating a Gaussian distribution. Together, these data demonstrate the existence of CD8SP CD28^low thymocytes. The minimal estimate of the percentage of CD8SP CD28^low cells was defined as the percentage of CD8SP cells expressing CD28 at levels lower than the MFI, minus the percentage of CD8SP cells expressing CD28 at levels higher than the MFI. This strategy revealed that a substantial proportion of CD8SP cells (18·3 ± 0·4%) expressed low levels of CD28 in the thymus of wt mice, similar to what we found in the spleen (26·4 ± 1·1%).

Figure 1.

Thymic CD8SP CD28^low T cells have suppressive activity in vitro (a) Definition of mature CD8SP CD28^low cells in thymus and spleen. Left‐hand panels: CD8/CD4 flow‐cytometry profiles of electronically gated TCR ^high cells. Right‐hand panels: CD28 profiles of CD8SP T cells electronically gated as indicated in left‐hand panels. Control staining (grey line) was performed using an isotype‐matched antibody. For the phenotype‐analysis in (b), an electronic CD28^low gate (indicated) was placed to include the minimal estimate of the percentage of CD28^low cells (i.e. % of cells with CD28 level <MFI – % of cells with CD28 level >MFI). (b) Phenotype of indicated thymocyte (left) and splenocyte (right) populations, electronically gated as shown in (a). Control stainings (grey lines) were performed using isotype‐matched antibodies. Results from a typical experiment out of three performed are shown. (c) CD8SP CD28^low but not CD8SP CD28^high thymocytes inhibit proliferation of CD4⁺ effectors in vitro. Thymic CD8SP CD28^low and CD8SP CD28^high T cells isolated from wild‐type (wt) B6 mice were cultured with CFSE‐labelled responder B6 CD4⁺ T cells and antigen‐presenting cells (APC) in the presence of anti‐CD3ε antibody. Proliferation of CD4⁺ cells was assessed by FACS analysis of CFSE dilution. Results from a typical experiment out of four performed are shown.

Thymic and splenic CD8⁺ CD28^low T cells, electronically gated as described in the Materials and methods section, did not express Foxp3, CD25 and neuropilin1, characteristic markers for CD4⁺ Treg cells (Fig. 1b). Thymic CD8⁺ CD28^low T cells expressed the transcription factor Helios, another CD4⁺ Treg marker,20 at the same levels as CD8⁺ CD28^high cells, and CD8 T cells in the spleen did not express this marker (Fig. 1b). Whereas in the spleen, a small population of CD8⁺ CD122^high cells was observed, corresponding to cells with regulatory activity,21 we found no CD122^high cells among CD8SP thymocytes.

To assess if thymic CD8SP CD28^low cells are Treg cells, we next analysed the in vitro suppressive capacity of CD28^low versus CD28^high CD8SP thymocytes. FACS‐sorted thymic CD8SP CD28^high and CD8SP CD28^low T cells were cultured with CFSE‐labelled CD4⁺ responder T cells, at a 1 : 1 ratio. T‐cell stimulation was achieved using antigen‐presenting cells and anti‐CD3 antibody. Proliferation of CD4⁺ responder T cells (as assessed by CFSE dilution) was efficiently suppressed by thymic CD8SP CD28^low thymocytes but not by their CD28^high counterparts (Fig. 1c).

Combined, these data demonstrate the existence of a CD8SP CD28^low population in the thymus possessing in vitro suppressive activity, suggesting that these Treg cells may develop in the thymus.

Thymic development of CD8SP CD28^low cells occurs concomitantly with that of CD8SP CD28^high thymocytes

CD8⁺ CD28^low Treg cells may differentiate from CD8⁺ CD28^high Tconv cells in e.g. peripheral lymphoid organs. If so, and if they were able to home back to the thymus, they would appear later than their CD28^high conventional counterpart in this organ of lymphopoiesis. We therefore analysed the kinetics of appearance of CD8SP CD28^low cells in the thymus during ontogeny. Very limited numbers of mature CD8SP CD28^low T cells were present in the thymus of 1‐day‐old mice and they then developed in parallel with CD8SP CD28^high T cells (Fig. 2a,b). The percentage of CD28^low cells among CD8SP thymocytes was constant (at 17·8 ± 2·1%) during the first 9 days of life. These observations indicate that CD8SP CD28^low T cells appear in the thymus in parallel with CD8SP CD28^high cells, strongly suggesting that they develop within the thymus.

Figure 2.

CD8⁺ CD28^low regulatory T (Treg) cells originate in the thymus and develop in parallel with CD8⁺ Tconv. (a,b) The proportion of CD8SP CD28^high and CD8SP CD28^low T cells among total thymocytes was analysed at indicated age of pups as indicated in the legend to Fig. 1(a). Indicated are mean values ± SD (n = 3). (c,d) Wild‐type fetal thymi (E15) were cultured in vitro for up to 2 weeks. The proportion of mature CD8SP CD28^high and CD8SP CD28^low T cells among total thymocytes was analysed at the indicated tim‐‐points. Results from a typical experiment out of three performed are shown. Note that data are represented as a percentage of total thymocytes (a, c) and as a percentage of the respective maximum values of the populations (b, d). (e) Rag‐GFP expression in electronically gated CD4⁻ CD8⁻ (‘DN’), CD4⁺ CD8⁺ (‘DP’), mature CD8 SP CD28^low and CD8 SP CD28^high thymocyte populations. Results from a typical experiment out of four performed are shown. The horizontal bar indicates the Rag‐GFP ⁺ gate used for cell‐sorting of newly developed Treg cells. (f) FACS‐sorted thymic Rag‐GFP ⁺ CD8SP CD28^low cells (gated as indicated in (e), lower panel) were cultured with CFSE‐labelled syngeneic CD4⁺ responder T cells (at indicated Treg/Tconv ratios) and antigen‐presenting cells (APC) in the presence of anti‐CD3ε antibody. Proliferation of responders was assessed by FACS analysis of CFSE dilution. A typical experiment out of three performed is shown.

We also analysed the development of CD8SP CD28^low cells in fetal thymus organ cultures in which mature T cells develop from immature precursors. CD8SP TCR^high cells started to appear after 7 days of in vitro culture of gestational day 15 fetal thymi, (Fig. 2c,d). Whereas the percentage of CD8SP cells steadily increased over the 6 days of culture (Fig. 2c), the percentage of CD28^low cells among CD8SP thymocytes was constant (at 17·7 ± 2%). CD8SP CD28^low T cells therefore differentiate simultaneously with CD8SP CD28^high cells (Fig. 2d) in the thymus.

CD8SP CD28^low T cells with in vitro suppressor activity develop in the thymus

T lymphocytes can recirculate from the periphery back to the thymus.22, 23 Hence, even if during ontogeny CD8SP CD28^low T cells develop intrathymically, in adult mice the thymic CD8SP CD28^low cells with suppressor activity may have acquired their regulatory phenotype in the periphery before migrating back to the thymus. To assess this possibility, we made use of Rag‐GFP mice. In these BAC‐transgenic mice, green fluorescent protein (GFP) is expressed under control of the Rag2 promoter.17 Expression of recombinase activating gene 2 (RAG2) in the early stages of thymocyte development is therefore concomitant with that of GFP. After thymic positive selection of T‐cell precursors, transcription downstream of the Rag2 promoter is rapidly terminated24 and the transgene‐derived GFP‐protein decays with a half‐life of approximately 56 hr.25 Hence, in the thymus, ‘freshly’ developed mature T cells express GFP while recirculating or long‐term resident T cells do not. In Rag‐GFP mice, we observed that GFP expression was bimodal in CD4 and CD8 double‐negative thymocytes and maximum in double‐positive thymocytes (Fig. 2e), which is in agreement with previously published data.25 Most of the mature CD8SP CD28^high as well as the CD8SP CD28^low T cells still expressed Rag‐GFP. Some CD8SP CD28^low thymocytes lacked Rag‐GFP and were therefore recirculating or long‐term thymus‐resident cells. The proportions of Rag‐GFP⁻ cells among the CD28^high versus CD28^low CD8SP thymocyte populations were comparable (21·7 ± 2·6% and 22·8 ± 5·1%, respectively).

To assess if de novo developing CD8SP CD28^low thymocytes were Treg cells, we sorted Rag‐GFP⁺ CD8SP CD28^low T cells and analysed their suppressive activity in an in vitro assay. These cells efficiently suppressed proliferation of CD4⁺ responder T cells in a dose‐dependent manner (Fig. 2f). These data firmly demonstrate that CD8SP CD28^low T cells with regulatory function can develop in the thymus of mice.

Development of CD8SP CD28^low Treg cells requires MHC class I expression on radioresistant stromal cells

Thymic differentiation of most T‐cell populations requires MHC expression on radioresistant epithelial cells (reviewed in ref. 26). However, a small population of natural killer T cells with strong immunomodulatory function develops upon interaction with non‐classical MHC class I molecules expressed by radiosensitive immature thymocytes (i.e. developing T cells, reviewed in ref. 27) and CD4⁺ Foxp3⁺ Treg cells can develop upon interaction with radiosensitive thymic dendritic cells.28, 29 We therefore assessed if MHC class I expression by radioresistant and/or haematopoietic cells is sufficient for development of CD8SP CD28^low Treg cells. Absence of these cells in β2m‐deficient (β2m) mice established, as expected, that they required MHC class I for their development (data not shown). We then generated bone‐marrow chimeras in which MHC class I expression was confined to radioresistant and/or radiosensitive cells. Lethally irradiated wt or β2m hosts were reconstituted with bone marrow from wt or β2m donors. Eight weeks after reconstitution, we analysed the presence, proportion and suppressive function of the CD8SP CD28^low Treg cell population in the thymi of these chimeras (Fig. 3). We did not detect mature CD8SP CD28^low T cells in the thymus of β2m hosts reconstituted with wt bone marrow (‘wt→β2m’ chimeras, Fig. 3a). On the other hand, we found similar proportions of thymic CD8SP CD28^low T cells in β2m→wt and wt→wt chimeras (Fig. 3a,b). Moreover, CD8SP CD28^low T cells isolated from the thymus of β2m→wt chimeras exhibited a dose‐dependent in vitro suppressor function comparable to that of Treg cells from wt→wt controls (Fig. 3c). Similar results were obtained for splenic CD8⁺ CD28^low Treg cells from these chimeras (data not shown). Altogether, our data indicate that development of CD8SP CD28^low Treg cells requires MHC class I expression by radioresistant thymic cells. MHC class I expression by haematopoietic elements is not required and does not increase the proportion of these cells.

Figure 3.

MHC class I expression on radioresistant but not radiosensitive thymic stromal cells is required for the development of functional CD8SP CD28^low regulatory T (Treg) cells. (a) Lethally irradiated hosts were reconstituted with T‐cell‐depleted donor bone marrow (‘donor → host’). Eight weeks after reconstitution, proportions of mature CD8SP CD28^low T cells were calculated by flow cytometry as indicated in the legend to Fig. 1(a). Flow cytometry profiles (CD8/CD4 on electronically gated TCR ^high thymocytes, CD28 on electronically gated TCR ^high CD8SP thymocytes) are representative of five independent experiments. Numbers indicate percentages within indicated electronic gates. (b) Percentage of CD28^low cells among CD8SP thymocytes in wild‐type (wt) → wt and β2 microglobulin (β2m) → wt chimeras, calculated as described in the legend to Fig. 1(a). Mean values ± SD (n = 9, five independent experiments) are indicated. (c) Thymic CD8SP CD28^low T cells electronically sorted from wt → wt and β2m → wt chimeras were cultured with CFSE‐labelled syngeneic CD4⁺ responder T cells (at the indicated regulatory T cells (Treg) to conventional T cells (Tconv) ratios) and antigen‐presenting cells (APC) in the presence of anti‐CD3ε antibody. Proliferation of responders was assessed by FACS analysis of CFSE dilution. Results from a typical experiment out of five performed are shown.

Identification of CD8⁺ CD28^low Treg cells in human blood and thymus

We next searched for CD8⁺ CD28^low T cells in the blood of healthy humans. Analysis of peripheral blood mononuclear cells showed a heterogeneous pattern: most individuals had a discrete population of CD28^neg cells (9 out of 11 samples analysed) whereas few lacked this population. In contrast to CD28^pos cells, which were mostly CD45RA^high, most CD28^neg cells were CD45RA^low and were therefore activated/memory cells (Fig. 4a). When we focused on naive CD8⁺ CD45RA^high CD28^pos T cells, we found that the median fluorescence intensity of CD28 staining was significantly lower than its mean value (P < 0·01, n = 9). The minimal estimate of the percentage of CD8⁺ CD28^low cells was defined as the percentage of CD8⁺ CD28^pos cells expressing CD28 at levels lower than the MFI, minus the percentage of CD8⁺ CD28^pos cells expressing CD28 at levels higher than the MFI (Fig. 4b). This strategy revealed a rather similar proportion of CD28^low cells among naive CD8⁺ T lymphocytes in the several samples analysed (10·1 ± 3·2%). We next analysed if CD8⁺ CD28^low peripheral T cells produce the immunosuppressive mediators TGF‐β and IL‐10, which are known to play a non‐redundant role in the in vivo function of these cells in experimental mouse models. After in vitro activation, a substantial fraction of CD28^low (but less CD28^high) cells expressed cell‐surface TGF‐β (as measured by surface expression of LAP) or intracellular IL‐10 (Fig. 4c, Table 1). These data show that the naive CD8⁺ CD28^low lymphocyte population in human blood is enriched in cells producing immunosuppressive mediators.

Figure 4.

Identification of human CD8⁺ CD28^low Treg in blood and thymus. (a) A CD28 profile (left panel) of a peripheral blood mononuclear cell (PBMC) sample in which a discrete CD28^neg population was observed. CD45RA expression on CD28^neg (open curve) and CD28^pos (black curve) CD4⁻ CD8⁺ TCR ^high PBMC (right panel, gated as in left panel). (b) Definition of mature CD28^low cells from PBMC and thymus. Top panels: CD8/CD4 profiles of PBMC and thymocytes, electronically gated as indicated. Bottom panels: CD28 profiles of CD8 T cells electronically gated as shown in top panels. The percentage of CD28^low cells was determined as described in the legend to Fig. 1(a). For cell‐sorting, an electronic CD28^low gate was placed that included the calculated percentage (as indicated). Cells expressing higher levels of CD28 were sorted as CD28^high (as indicated). (c) CD28^low and CD28^high CD8⁺ TCR ^high CD45RA ^high PBMC were activated in vitro and then stained for surface expression of latency‐associated peptide (LAP; top panels) or intracellular expression of interleukin‐10 (IL‐10; bottom panels). (d) The percentage of CD28^low cells among CD8SP thymocytes was analysed in samples from patients of indicated age. (e) Thymic CD8SP CD28^low and CD8SP CD28^high cells were activated in vitro and then stained for surface expression of LAP. In (c) and (e), results from a typical experiment out of three performed are shown (see also Table 1). Numbers within cytometry plots indicate % of cells within indicated electronic gates.

Table 1.

Proportion of cytokine‐producing cells among human CD8⁺ T‐cell subsets

	LAP+ cells (%) among		IL‐10+ cells (%) among
	CD8+ CD28high	CD8+ CD28low	CD8+ CD28high	CD8+ CD28low
Peripheral blood mononuclear cells
donor 1	8·6	23·0	4·3	9·7
donor 2	5·4	10·2	4·0	8·5
donor 3	20·5	31·4	0·3	4·1
Thymus
sample 1	11·7	42·0	n.d.	n.d.
sample 2	3·0	7·0	n.d.	n.d.
sample 3	7·4	16·3	n.d.	n.d.

n.d., not done.

We then assessed if CD8⁺ CD28^low Treg cells develop in the human thymus. Flow cytometry analysis revealed that the median fluorescence intensity of CD28 staining on CD8SP thymocytes is significantly lower than its mean value (P < 0·01, n = 8). The proportion of CD28^low cells among thymic CD8SP cells was 26·3 ± 1·4% (Fig. 4b) and was remarkably constant in children of different ages (Fig. 4d). A substantial fraction of CD28^low CD8SP thymocytes, and much less CD28^high cells, produced the critical immunosuppressive mediator TGF‐β (Fig. 4e, Table 1). Collectively, these data therefore strongly suggest that CD8⁺ CD28^low cells with immunosuppressive potential develop in the human thymus.

Discussion

In this report we address the thymic versus peripheral origin of mouse CD8⁺ CD28^low Treg cells and their potential existence in humans. We show that functional Treg cells are found in the murine thymus and that they develop locally and are not recirculating from the periphery. We also identify in human blood and thymus a homologous population of CD8⁺ CD28^low cells with a naive phenotype and producing the critical immunosuppressive mediators TGF‐β and IL‐10. Our results therefore demonstrate that in mice CD8⁺ CD28^low Treg cells develop intrathymically and strongly suggest that the same is true in humans.

The origin of CD8⁺ CD28^low Treg in mice had until now remained elusive. Immunization with a myasthenogenic peptide induces a myasthenia gravis‐like disease, which can be prevented by injection of a ‘dual altered peptide’ ligand. An in vivo expanded CD8⁺ CD28^low population was involved in this phenomenon,5 suggesting that these cells were peripherally induced Treg cells. On the other hand, we previously showed that peripheral CD8⁺ CD28^low T cells isolated from unmanipulated wt mice had a naive phenotype and efficiently inhibited proliferation of and interferon‐γ production by CD4⁺ responder T cells in mixed lymphocyte cultures. CD8⁺ CD28^low (but not CD8⁺ CD28^high) cells also effectively prevented IBD in an experimental mouse model.6 Similar results were obtained in the multiple sclerosis model experimental autoimmune encephalomyelitis.7 These data showed that CD8⁺ CD28^low Treg cells can differentiate without intentional manipulation of the immune system and suggest that they are a naturally occurring population.

We here report that CD8⁺ CD28^low T cells develop in organ cultures of mouse fetal thymi and that CD8⁺ CD28^low and CD8⁺ CD28^high T cells simultaneously develop during ontogeny. Moreover, newly developing CD8SP CD28^low cells isolated from adult thymi had in vitro immunosuppressive activity. These data demonstrate a thymic origin of CD8⁺ CD28^low Treg cells. By contrast, we did not observe any CD8⁺ CD122^high Treg cells in the thymus, strongly suggesting that this Treg cell population differentiated in the periphery. We recently reported that CD8⁺ CD28^low Treg cells from Auto Immune REgulator‐deficient mice failed to prevent experimental IBD in the mouse.19 The data reported here suggest that this transcription factor (which is expressed by stromal cells in the thymus and in lymph‐nodes) plays a role in the thymic selection of the TCR‐repertoire of CD8⁺ CD28^low Treg cells, even if further selection in lymph nodes may also be involved. Our data do not exclude the possibility that some CD8⁺ CD28^low Treg cells can also differentiate in the periphery from CD8⁺ CD28^high Tconv cells, similar to the activation‐ and cytokine‐dependent peripheral differentiation of CD4⁺ Foxp3⁺ Treg cells.30

In humans, Treg cells of CD8⁺ CD28^neg phenotype have previously been described.9 This population lacks expression of CD28, has an activated CD45RO⁺ and HLA class II⁺ phenotype and produces neither IL‐10 nor TGF‐β.8, 9 It is therefore distinct from the human CD8⁺ CD28^low Treg cells that we describe here because the latter cells express clearly detectable levels of CD28 and have a naive CD45RA^high phenotype. Identified in the blood, we also detected substantial proportions of these cells in the human thymus. Whereas no specific markers exist to distinguish between newly developing and recirculating CD8 thymocytes, the proportion of CD28^low cells among CD8SP thymocytes was constant during early life. The latter observation, combined with their naive phenotype, provides supplementary support for our conclusion that CD8⁺ CD28^low Treg cells probably develop within the human thymus.

In conclusion, the data presented here firmly demonstrate a thymic origin for CD8⁺ CD28^low Treg cells. The identification of a homologous population in humans opens the way to studying the potential role of these cells in immunopathology. Further work will need to identify more distinctive markers, assess the selection‐criteria of these developing Treg cells, their antigen‐specificity, and their physiological role in the maintenance of immune homeostasis.

Disclosure

The authors declare having no competing interests.