The thymus continuously produces T lymphocytes that contribute to the maintenance of the peripheral T cell pool. Since peripheral recirculating T cells represent a very minor population among total thymocytes in normal animals, the relationship between the thymus and secondary lymphoid organs is generally considered unidirectional. Recently, several reports have described the presence of recirculating T cells in the thymus, raising issues regarding their possible function. In this article, we show that the niche for recirculating T cells in the thymus, i.e., their absolute number, is the same in lymphopenic and normal mice. Using a novel combination of TCR-transgenic mice in which the ligand necessary for positive selection of host T cells is only expressed by transferred donor T cells, we show that mature T cells recirculating back to the thymus can mediate positive selection.
For continuous T cell production, the thymus must receive multipotent progenitors derived from the bone marrow. These progenitors have recently been called thymus-settling cells (TSP)3 and enter the thymus at the corticomedullary junction (1). TSP, which initially express neither CD4 nor CD8 (double negative), migrate to the cortex and become CD4 and CD8 double positive (DP). Following TCR-positive selection, mature CD4 or CD8 single-positive (SP) cells migrate back into the medulla. On their way, they traverse the corticomedullary junction where it is thought that negative selection takes place (2). Once in the medulla, it is generally believed that all cells migrate out. However, the possibility that the medullary thymocytes constitute an equilibrium between cells produced in the cortex and recirculating peripheral T cells is rarely considered.
Previous reports had indicated that either transfer of labeled mature T cells into neonatal mice or activated T cells into adult animals resulted in some T cells entering the thymus (3, 4, 5). More recently, using new tools, such as transgenic, gene knockout and gene reporter mice, we and others have readdressed this question (6, 7, 8). These reports have shown that some SP cells in the thymus are in fact recirculating T cells that may play a role in negative selection of the TCR repertoire (9, 10). Herein, using parabiosis, T cell transfers into normal or lymphopenic hosts and a combination of TCR-transgenic mice where the ligand necessary for positive selection of host T cells is only expressed by transferred T cells, we quantify and further characterize the thymic recirculating T cell pool. In addition, we show that mature T cells recirculating back to the thymus are involved in positive selection.
Materials and Methods
All mice were on the C57BL/6 background (B6). B6 mice (Ly5.2) were obtained from Iffa-Credo. B6 Ly5.1 congenic (B6.Ly51) and B6 mice expressing GFP (B6.GFP) (11) were obtained from the Centre de Distribution Typage et Archivage animal. B6 mice deficient for the pre-Tα gene (B6 pTα−/−, referred to as B6.pTαko) were a gift from Dr. H. J. Fehling (University of Ulm, Ulm, Germany) (12). B6 mice deficient for the RAG-2 gene (B6 RAG-2−/−, referred to as B6.Ragko) were obtained from Dr. A. Rolink (University of Basel, Basel, Switzerland). OT1 TCR-transgenic mice, OVA-specific, H-2Kb-restricted, Vα2Vβ5, were provided by Drs. S. Degermann and E. Palmer; P14 TCR-transgenic mice, lymphocytic choriomeningitis virus gp33 specific, H-2Db-restricted, Vα2Vβ8, were provided by the CDTA. Hemagglutinin (HA) TCR-transgenic mice, influenza virus HA-specific, I-Ed restricted, were described before (13). All of these TCR-transgenic mice were bred with B6.Ragko or B6.Ragko.Ly51 to generate the following strains: B6 RAG-2−/− P14 TCR transgenic, referred to as B6.Ragko.P14tg, B6 RAG-2−/− OT1 TCR-transgenic, referred to as B6.Ragko.OT1tg, and B6 RAG-2−/− Ly5.1 HA TCR transgenic, referred to as B6.Ragko.Ly51.HAtg. Mice harboring the H-2bm1 haplotype in a RAG−/− background, referred to as B6.Ragko.Kbm1, have been described previously (14). To obtain B6 RAG-2−/− OT1 TCR-transgenic Kb−/− mice, referred to as B6.Ragko.OT1tg.Kbko, B6.Ragko.OT1tg were subsequently bred with Kb-deficient mice (15). All mice were bred at the animal facilities of the Commissariat á l’Energie Atomique (Grenoble, France) or at the Max Planck Institute of Immunobiology (Freiburg, Germany). Parabiosis was conducted as previously described (16). For fetal liver reconstitution, livers were dissociated by pipetting and cells were transferred into adult recipient animals that were irradiated from a 137Cs source. For intrasplenic CFSE injection, 75 μl of a 400 μM CFSE solution freshly diluted in PBS was injected into the two extremities of the spleen (150 μl total) while animals were under anesthesia with ketamine and xylazine. All animals were kept under specific pathogen-free conditions and 6- to 10-wk-old mice were used for experiments that were conducted according to institutional guidelines.
Cell preparation and flow cytometry
Single-cell suspensions were prepared from thymus, pooled inguinal, axillary, brachial, and cervical lymph nodes (LN) or spleen. For flow cytometry, cells were washed and resuspended in PBS containing 2% FCS and 0.01% sodium azide. Viable cells were counted in a hemocytometer following staining with trypan blue. The total number of T cells was calculated from the frequency determined by FACS analysis and the total number of living cells recovered per organ.
The following mAbs were used: anti-CD4 (RM4-5 or GK1.5), anti-CD5 (53-7.3), anti-CD8 (53-6.7), anti-CD8b.2 (53-5.8), anti-CD19 (1D3), anti-CD24 (human stable Ag (HSA), M1/69) anti-CD44 (IM7), anti-Ly5.1 (A20), anti-Ly5.2 (104), anti-TCRβ (H57-597) (BD Pharmingen) and anti-Vβ Abs: anti-Vβ5 (MR9-4.5), anti-Vβ8 (F23.1), the other anti-Vβ have been described previously (17). Cell surface staining was performed as previously described (16) and acquisitions were performed on FACSCalibur, FACSLSRII, or CANTO instruments (BD Biosciences). Analyses were performed with CellQuest, DIVA (BD Biosciences), or FlowJo (Tree Star) softwares. Dead cells were excluded from the analysis by a combination of light scatter and/or absence of propidium iodide (PI) staining. Cell sorting was performed either on a MoFlo (DakoCytomation) or FACSAria (BD Biosciences) and the purity of sorted cells was >98%.
GFP+CD4+ LN T cells from GFP-transgenic mice were purified by sorting and injected into B6.pTαko hosts. After 7 days, animal were perfused with PBS/4% paraformaldehyde, spleen was collected, postfixed overnight at 4°C in 4% paraformaldehyde, incubated in 30% sucrose for 24 h, and embedded in frozen medium (Cryomatrix; Therm Shandon). Immunofluorescence staining has been performed as detailed previously (18). Cryostat sections were immunolabeled with anti-CD4, anti-CD8, anti-cytokeratin (CK), and/or anti-CK18. CK5 and CK18 Abs were provided by Dr. J. Gill (University of Basel). Fluorescent images were obtained with a Leica (TCS SP2) confocal microscope. Due to perfusion with paraformaldehyde, no peripheral lymphocytes are found in vessels.
Re-entry of peripheral T cells in different thymi following parabiosis
Using parabiosis, we have recently shown that in lymphopenic B6 pTα−/− mice, referred to as B6.pTαko, a considerable proportion of SP thymocytes represent recirculating peripheral T cells (6). To investigate T cell recirculation back to the thymus in more detail, we performed parabiosis experiments combining immunocompetent mice (B6) with immunodeficient animals in which SP cells, and eventually DP cells also, are absent: 1) B6 RAG-2−/− Ly5.1 mice, referred to as B6.Ragko.Ly51, in which DP and SP cells are absent and 2) B6 RAG-2−/− Ly5.1 HA TCR transgenic, referred to as B6.Ragko.Ly51.HAtg mice, which on this MHC background (H-2b) do not positively select the HA-TCR, leading to an accumulation of DP and an absence of SP cells. Thus, we parabiosed (B6 with B6.Ragko.Ly51) and (B6 with B6.Ragko.Ly51.HAtg) and the thymi of the immunodeficient mice were analyzed after 14 days for the presence of mature T cells derived from the immunocompetent partner. Results obtained (Fig. 1) show that independently of the DP/SP composition, mature T cells from the partner mouse could be readily detected in the thymus (6). Analysis of the thymus of the B6.Ragko.Ly51 mice parabiosed with B6 mice showed that only partner-derived SP cells were present. The absence of DP cells in the thymus and of progenitor B cells in the bone marrow (data not shown) indicates that in these short-term parabiosis experiments, there was no exchange of lymphoid progenitors, including TSP.
Recirculation in non parabiosed mice
In the above parabiosis experiments, we could show that mature T cells recirculate back to the thymus. To quantify the number of mature T cells capable of entering the thymus of lymphopenic and normal mice, an alternative experimental approach was undertaken. Genotypically distinct (GFP+) mature T cells were adoptively transferred into either B6.pTαko or B6 host mice and their thymi were analyzed after 14 days. In lymphopenic B6.pTαko hosts, after injection of 25 × 106 GFP+ lymph node cells, the level of T cell chimerism in the spleen was 47 ± 5% (proportion of GFP+ cells among TCRβ+ cells). The proportion of GFP+TCRβ+ cells in the thymus was 1.9% (range, 0.5–6.7, n = 5). Assuming that injected GFP+ and host-derived (GFP−) mature T cells recirculate equally well to the thymus, we calculated that the thymus of B6.pTαko mice contains ∼105 recirculating, peripheral T cells (Fig. 2,A). As expected, in nonlymphopenic B6 mice, the degree of T cell chimerism in the spleen was only 1.4 ± 0.5% and the proportion of GFP+TCRβ+ cells in the thymus was 0.00097 ± 0.00049% (i.e., frequency of 0.97 ± 0.49 × 10−5). Using these values, we calculated that the total number of recirculating T cells in the thymus of normal B6 mice was approximately the same as in lymphopenic B6.pTαko mice, namely, 105 cells (Fig. 2 A). That this number represents an increased proportion in B6.pTαko mice reflects the reduced thymus cellularity of these mice. Further phenotypic analysis indicated that more CD4+ than CD8+ T cells recirculated to the thymus.
To complement the above results, an independent experimental approach was devised to increase the proportion of peripheral cells that would be labeled. To this end, CFSE was directly injected into the spleen of mice and, 7 days later, the proportion of CFSE+ cells in lymphoid organs was analyzed (Fig. 2 B). Assuming that mature T cells recirculate independently of CFSE labeling, we estimated that the total number of mature recirculating T cells in the thymus of B6 mice was 8.6 ± 2.0 × 104. Similar experiments were conducted with B6.pTαko mice and the calculated number of recirculating cells was similar. Importantly, these results indicate that recirculation of T cells back into the thymus is not a unique property of lymphopenic mice and, irrespective of the thymus size or genotype, the thymus can accommodate ∼105 recirculating peripheral T cells (19).
Two groups using RAG-2 promoter-driven GFP mice (RAG2p-GFP) have also recently shown that mature phenotype thymocytes contain a population of recirculating cells (7, 8). In RAG2p-GFP mice, thymocytes are initially generated as GFP+ cells, which upon migration to the periphery rapidly lose GFP expression; thymus recirculating cells are identified being GFP−. Based on two different methods, namely, GFP+ cell transfer and intrasplenic CFSE injection, we found that in wild-type (WT) animals on average ∼1% of thymic SP cells are recirculating lymphocytes, whereas Hale et al. (7) suggest that it is ∼5% in 5- to 9-wk-old mice. However, our approaches, in which peripheral cells are positively stained, might slightly underestimate the phenomenon if a proportion of peripheral T cells in the thymus have a slow turnover. In contrast, use of RAG2p-GFP mice might lead to an overestimation due to premature loss of GFP expression in newly generated SP thymocytes before emigration from the thymus and nondiscrimination of αβ and γδ T cells within the SP populations.
In certain mice displaying a partial block in T cell production and concomitant peripheral lymphopenia, the absolute number of SP thymocytes is relatively high compared with the number of DP cells; i.e., the ratio SP:DP is higher than in WT controls. This is the case for mice deficient for genes involved in T cell development (12, 20), in mice in which genes encoding molecules expressed by epithelial cell supporting T cell maturation have been deleted (21), as well as for mouse models of T cell lymphopenia (6). Interestingly, the SP:DP ratio is also elevated in cervical thymi (22). Instead of being caused by an increased rate of survival, positive selection, or cell division of SP cells, in these situations a considerable fraction of SP cells may represent recirculating mature T cells, as in B6.pTαko mice.
Phenotype, diversity, and localization of recirculating T cells
To determine whether T cells recirculating to the thymus represented a particular subpopulation of peripheral T cells, we used the CFSE labeling approach described above to determine the proportion of naive vs memory phenotype T cells among recirculating cells. As shown in Fig. 3 A, the CD44 profile of splenic and thymic CFSE+CD4+ cells differed with 29% splenic and 52% thymic CD4+ cells being CD44high. In a series of five such experiments, the proportion of CD44highCD4+ splenic cells was 24 ± 4% while in the thymus it was 50 ± 2%. Thus, there was an enrichment for CD44high cells among T cells recirculating to the thymus.
To determine whether there was a bias in TCRVβ usage by recirculating cells, LN cells from B6.Ly5 1 mice were transferred into B6.pTαko hosts. After 11 days, cells from spleen, LN, and thymus were stained with anti-Ly5 1 and a panel of Vβ-specific mAbs. These experiments were performed using B6.pTαko hosts because the proportion of donor-derived cells in the thymus required for a complete Vβ repertoire analysis was insufficient using B6 mice. Results obtained from a pool of animals (Fig. 3 B) showed that, first, the Vβ repertoire of recirculating T cells in the thymus was diverse and, second, that there was no bias in the representation of different Vβ specificities. This result was confirmed by parabiosis experiments between B6 RAG-2−/− P14 TCR-transgenic mice (Vβ8+), referred to as B6.Ragko.P14tg, with B6 RAG-2−/− OT1 TCR-transgenic mice (Vβ5+), referred to as B6.Ragko.OT1tg: T cells from the partner mouse were clearly present in each contralateral thymus 7 days after surgery (data not shown).
Function of recirculating T cells
To localize the recirculating T cells in the thymus, LN T cells from GFP-transgenic mice were purified by cell sorting and injected into B6.pTαko hosts. After 7 days, thymic sections were prepared and immunolabeled for cytokeratin-5 and cytokeratin-18 to stain medullary epithelial cells and cortical epithelial cells, respectively. Three-color confocal microscopy (Fig. 3 C) showed that donor GFP+ cells were found predominantly in the medulla. However, GFP+ cells could also be found throughout the cortex. Because positive selection of DP thymocytes takes place in the cortex, we designed experiments to determine whether recirculating mature T cells could support positive selection of DP cells, a function commonly attributed to cortical epithelial cells only.
In line with this, our analysis of positive selection of OT1 TCR-expressing cells in recipient mice lacking the restricting class I MHC molecule H-2Kb had led to the surprising finding that the presence of H-2Kb on cells of the hematopoietic lineage was already sufficient for positive selection. Thus, when B6.Ragko.OT1tg fetal liver cells were used to reconstitute irradiated mice of H-2bm1 haplotype (B6.Ragko.Kbm1) or mice deficient for Kb (B6.Ragko.Kbko) (14, 15), the recipient thymi contained a reduced, but significant, compartment of mature SP thymocytes (CD4−CD8+Vα2+HSA−), in comparison to thymi of control mice that express H-2Kb also on thymic epithelial cells (Fig. 4,A). By analogy to published data, the hematopoietic cell type supporting positive selection in this situation may include thymocytes (23, 24) and/or recirculating T cells. Indeed, this type of positive selection on bone marrow-derived cells is efficient in trans, as shown by reconstituting irradiated B6.Ragko.Kbko mice with a mixture of B6.Ragko and B6.Ragko.OT1tg.Kbko fetal liver cells (Fig. 4,B). In contrast, when these mice were reconstituted with a mixture of B6.Ragko.Kbko and B6.Ragko.OT1tg.Kbko fetal liver cells, there was no evidence of positive selection, as expected (Fig. 4 B).
To directly investigate whether recirculating peripheral T cells expressing H-2Kb class I molecules could mediate positive selection of OT1 thymocytes, we transferred peripheral P14 TCR-transgenic H-2Kb-expressing cells into B6 RAG-2−/− OT1 TCR-transgenic Kb−/− mice (referred to as B6.Ragko.OT1tg.Kbko). Since P14 T cells are restricted to H-2Db, a MHC class I molecule expressed normally in H-2Kb-deficient recipient mice, their survival would be ensured. Three weeks after injection of P14 cells, lymphoid organs of recipient mice were collected, stained for CD4, CD8, HSA, PI, and Vβ (Vβ5 or Vβ8) and the number of Vβ5+ OT1 (host) and Vβ8+ P14 (donor) CD8+ cells was determined. As expected, Vβ8+ P14 T cells were detected in the spleen (8.0 ± 0.4 × 105), pooled LN (1.6 ± 0.3 × 106), and thymus (8.2 ± 0.3 × 104) of recipient mice (Fig. 5,A). In the thymus, we also detected 3.1 ± 0.9 × 104 CD4−CD8+HSA−Vβ5+ cells suggesting that P14 T cells that recirculate into the thymus provided the H-2Kb-restricting element necessary for the positive selection of OT1 T cells (Fig. 5, B and C). In the peripheral lymphoid organs of B6.Ragko.OT1tg.Kbko mice, we observed few CD4−CD8+HSA−Vβ5+ cells (1.5 ± 0.6 × 106) and their number was not elevated after transfer of P14 LN cells (1.0 ± 0.5 × 106). This suggested that OT1 SP cells arising in the thymus of B6.Ragko.OT1tg.Kbko mice after engraftment with P14 LN cells are newly generated SP thymocytes and not rare preexisting cells that underwent homeostatic expansion. In the above experiment, the vast majority of the injected cells were CD8+ P14tg T cells and positive selection of OT1 cells is likely to be mediated by recirculating peripheral T lymphocytes. However, to exclude that positive selection was mediated by other cells types, a similar experiment was conducted with purified T cells. Results obtained showed that purified P14.Kb+ T cells mediate the positive selection of OT1 thymocytes; as expected, expression of Kb on the transferred P14 T cells is essential (Fig. 6, A and B). No positive selection was observed when peripheral mature Kb-expressing non-B and non-T cells (from B6.Ragko donors) were transferred into B6.RAGko.OT1tg.Kbko recipients, suggesting that under these experimental conditions, T cells are those cells that migrate back to the thymus most efficiently to mediate positive selection. Positive selection of OT1 cells was detected in B6.Ragko.OT1tg.Kbko mice injected with 2 × 107 CD3+TCRβ+ purified T cells from B6.Ly51 donors, showing that polyclonal T cells recirculating back into the thymus can also mediate thymocyte-positive selection (data not shown).
Whether mature, peripheral T cells can recirculate back to the thymus has been the subject of debate (8). Early observations indicated that this phenomenon might be restricted to the neonatal period, a time when the periphery is lymphopenic, or to the recirculation of some activated T cells back to the adult thymus (5). Clearly, during an ongoing antiviral immune response, some degree of priming is seen among mature thymocytes (25, 26). Using either parabiosis (6) or transfer of CFSE-labeled cells (27), we have previously shown that in a situation of peripheral lymphopenia, the contribution of recirculating peripheral T cells to the pool of mature “thymocytes” was significant. Using complementary approaches, including adoptive transfer of genetically marked T cells and the intrasplenic injection of CFSE, in this study we quantify this recirculation in detail and analyze the phenotype and TCR Vβ repertoire of recirculating cells. As expected, T cells recirculating to the thymus had a diverse TCR Vβ repertoire and were enriched in CD44high, so-called “memory-like” cells (28). Because it had already been shown that recirculating T cells can mediate negative selection of developing thymocytes (9, 10), their localization to the medulla might have been expected (3, 4). In this study, we show that medullary recirculating cells were clearly detectable outside of blood vessels and intermingled with host thymocytes. Moreover, we also observed recirculating T cells in the cortex, the site where positive selection is presumed to occur (1, 2). This observation led us to investigate whether recirculating T cells might be involved in positive selection by using novel TCR-transgenic strains. In B6.RAGko.OT1tg.Kbko mice, because of the absence of expression by thymic epithelial cells of the H-2Kb molecule necessary for the positive selection of OT1 TCR, thymocyte development is arrested at the DP stage (14). When these mice are injected i.v. with B6.Ragko.P14tg CD8+ T cells, some recirculate to and enter the recipient thymus, thereby providing H-2Kb molecules necessary for the positive selection of host OT1Tg thymocytes.
Despite being able to readily detect CD8+HSA−Vβ5+ (OT1) cells in the thymus, we did not detect increased number of these cells in the periphery, presumably because the H-2Kb ligand required for the survival of OT1 T cells was not expressed at sufficient frequency or not on the relevant cells to allow for their survival. The failure of donor T cells to survive in the absence of their selecting ligand in the periphery may also explain why the transfer of B6.Ly51 T cells into β2-microglobulin-deficient mice did not result in an efficient positive selection of host CD8+ thymocytes: in the absence of peripheral MHC class I molecules, injected cells will not survive and hence not migrate in sufficient numbers to the thymus to support positive selection (data not shown). The above results indicate that class I molecules and, by implication, other self -Ags expressed by recirculating T cells may be involved in positive selection.
Indeed, there has been an increasing realization that thymocyte-positive selection can be mediated by cells other than thymic epithelial cells. Thus, and as recently reviewed by Berg (29), such “unconventional” positive selection may apply to CD4+ NKT cells (30), FoxP3+ regulatory T cells, and so-called “innate” CD8+ T cells (31). Because in our case, the number of positively selected cells was quite small, this precluded detailed analysis of their innate vs conventional phenotype. The relative “efficiency” of epithelial vs T cell positive selection is difficult to directly estimate. However, by intrathymic injection of 107 MHC class I-bearing fibroblasts into β2-microglobulin-deficient mice, Pawlowski et al. (32) demonstrated a complete restoration of CD8+ SP-positive selection with normal numbers, TCR repertoire, and effector function of the selected cells.
More recently, two groups (23, 24), developed mouse models in which thymocytes were the only MHC class II-expressing cells. In these mice, CD4+ SP T cells were selected as efficiently as in control mice, although the functionality of these cells was later questioned (33). In a normal mouse adult thymus, ∼50,000 thymic epithelial cells are present and presumably able to mediate most thymocyte-positive selection by contacting several cells simultaneously and because of their high MHC expression density. Numerically, this leads to the selection of ∼107 CD4+ and 2 × 106 CD8+ mature SP T cells per WT thymus (and ∼5–7 × 106 Vβ5+ CD8+ mature SP T cells per B6.Ragko.OT1tg thymus). In our experimental system, the presence of ∼105 transferred T cells intrathymically generated only 2 × 104–2 × 105 mature Vβ5+CD8+ SP T cells. Thus, the efficiency of positive selection by recirculating T cells is relatively low in our experimental system and might be very modest under physiological conditions, but this does not preclude functional consequences for this process.
In order for central tolerance to be maximally efficient, the thymus must ensure that developing cells are exposed to the maximum repertoire of self-Ags. This is probably ensured, to a large extent, by AIRE-mediated regulation of peripheral tissue-specific self-Ag expression by epithelial cells (34). However, we would propose that this may be complemented by the direct import of self-Ags expressed by recirculating lymphocytes including B cells (3, 9), T cells (6, 9) or, as recently shown, dendritic cells (35). In practical terms, recirculation of lymphocytes to the thymus could be envisioned as a group of Trojan horses in that they could be used to introduce Ags necessary for both positive and negative selection of the T cell repertoire.
J. Kirberg thanks Yvonne Joos for expert technical assistance; F. Agenès and R. Ceredig thank Institut National de la Santé et de la Recherche Médicale for continual support. We express our gratitude to Pr. A. Rolink (University of Basel) for indispensable help with sorting and scientific input. F. Agenès is greatly indebted to Dr. R. P. Sekaly for hosting him at the INSERM U743 (Montreal, Canada). We thank Drs. C. Perreault (Institute for Research in Immunology and Cancer, Montreal, Canada) and C. Viret (Institut National de la Santé et de la Recherche Médicale Unité 563, Toulouse, France) for critical reading of this manuscript.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Abbreviations used in this paper: TSP, thymus-settling cell; DP, double positive; LN, lymph node; SP, single positive; tg, transgenic; HA, hemagglutinin; PI, propidium iodide; CK, cytokeratin; WT, wild type; HSA, human stable Ag.