Immunosuppressive drugs and cytotoxic chemotherapy agents are designed to kill or suppress autoreactive, alloaggressive, or hyperinflammatory T cells, or disseminated malignancies. However, they also cause severe immunological side effects ranging from interrupted thymopoiesis and general immunodeficiency to, paradoxically, autoimmunity. Consistent with the cross-talk between thymocytes and stromal cells, we now show that these common therapeutic agents have major effects on murine thymic epithelial cells (TEC), crucially required to rebuild immunity posttreatment. We show that the immunosuppressant cyclosporine A, which has been linked to a thymus-dependent autoimmune syndrome in some patients, causes extensive loss of autoimmune regulator (Aire+) tolerance-inducing MHC class IIhigh medullary TEC (mTEChigh). Post-cyclosporine A, Aire expression was restored within 7 days. Full recovery of the mTEChigh subset occurred within 10 days and was linked to a decrease in a relatively resistant MHC class IIlow mTEC subset (mTEClow), consistent with a previously described precursor-product relationship. Cyclophosphamide and dexamethasone caused more extensive ablation of thymocytes and stromal cells but again severely depleted tolerance-inducing mTEChigh. Together, these data show that Aire+ mTECs are highly sensitive to damage and that mTEC regeneration follows a conserved pattern regardless of the treatment regimen used.
Thymic stromal cells induce T cell differentiation and maturation through provision of growth factors, cytokines, chemokines, and through tightly regulated selective processes involving the presentation of MHC and self-peptides (1). Thymic epithelial cells (TEC)4 are the major component of thymic stroma, and phenotypically distinct TEC subsets fulfill different functions in selecting and shaping the developing T cell repertoire. Thymocytes and TECs exist in a dynamic codependence called cross-talk, whereby disruption of thymocyte development impacts on the thymic stroma and vice versa (1, 2).
The degree of affinity for self determines whether a thymocyte will mature and exit the thymus. Positive selection, mediated by cortical TECs (cTEC), ensures that mature T cells are self-MHC restricted, whereas negative selection occurs when dendritic cells or medullary thymic epithelial cells (mTEC) induce apoptosis in thymocytes with a high degree of self-avidity (1, 3, 4). Together, positive and negative selections strike a balance to create a broadly reactive TCR repertoire, with a low but not completely absent potential for self-reactivity (5). Fundamental to central (thymic) tolerance is the autoimmune regulator (Aire), a transcription factor regulating expression of many self-Ags by a defined subset of mTECs, deleting or functionally silencing thymocytes reactive to tissue-specific, sequestered, and late-onset self-Ags (6, 7).
Affinity for the lectin Ulex europaeus agglutinin-1 (UEA1) and expression of MHC class II can be used to divide TEC into four subsets (8). The mTEChigh subset expresses the highest levels of MHC class II, CD80, UEA1, and Claudin 3/4 and has the highest capacity of any TEC subset to stimulate transgenic T cells in vitro (8, 9, 10). Importantly, thymic Aire expression is exclusive to this subset (11, 12, 13).
Cyclosporine A (CsA) is a common immunosuppressant used to prevent allograft rejection. It is used primarily to prevent peripheral T cell activation, but it also causes mild thymic involution (14, 15). Paradoxically, when thymi from CsA-treated mice were transplanted to athymic nude mice, recipients developed characteristic organ-specific autoimmunity (16). Reductions in dendritic cells and FoxP3+ regulatory T cells (Treg) are likely to contribute and have been examined in depth (17, 18, 19), but symptoms are also highly reminiscent of a defect in mTEC-based tolerance (7). There has been limited investigation of the thymic stroma by histology, yielding conflicting results, ranging from a broad reduction in mTECs (20, 21) to no change at all (18). Similarly, although evidence suggests thymopoiesis is grossly interrupted after treatment with the immunoablative chemotherapy agents cyclophosphamide (22) and dexamethasone (23, 24), parallel flow cytometric studies of stromal damage and regeneration have not been thoroughly performed. Any damage to the thymic stromal microenvironment with these commonly used drugs would be expected to have a major impact on the ability of the patient to restore immune competence.
With >1.4 million cancer diagnoses expected this year in the United States alone (25), swift restoration of thymopoiesis and tolerance induction after cytoablative or immunosuppressive therapy is of paramount clinical importance. Because TECs have a unique, critical role in restoring functional immunity and self-tolerance, regeneration of Aire and the Aire-expressing mTEChigh subset is particularly important. The kinetics of mTEC recovery and resistant cell types from which TEC subsets might develop has not, however, been effectively studied.
Using sophisticated flow cytometry, we demonstrate that thymic damage following CsA, cyclophosphamide, or dexamethasone treatment includes loss of the tolerance-inducing mTEChigh subset. TEC regeneration at the subpopulation level revealed characteristic developmental phenotypes and kinetics across several immunosuppressive regimens.
Materials and Methods
C57BL/6J (B6) mice ages 8–12 wk from Monash Animal Services were housed under specific pathogen-free conditions in accordance with institutional guidelines and with the approval of the Monash University Animal Ethics Committee.
Immunosuppression and cytoablative chemotherapy
To compare thymic recovery from different treatments, timepoints are referred to as days of recovery (RD) indicating the number of days after treatment has ceased. In all experiments, the final day of treatment was RD0. Mice were given 15 mg/kg/day CsA (Novartis) i.p. for 14 or 21 days, as indicated; or 100 mg/kg/day cyclophosphamide (BD Pharmacia) i.p. in PBS for 2 days; or a single injection, at 20 mg/kg, of dexamethasone (Lyppards). Mice were humanely killed by CO2 asphyxiation at indicated time points.
Individual thymus digestion and flow cytometry
Each thymus was individually digested in collagenase D and DNase I (Roche) as previously described (26). Supernatant fractions were pooled per thymus and filtered through a 100-μm pore size mesh. Cell counts were performed using a Z2 Coulter Counter (Beckman Coulter). Staining for plasma membrane markers was performed, using 5 × 106 cells, as previously described (26). For intracellular staining, surface-stained cells were then fixed and permeabilized using the Cytofix/Cytoperm kit (BD Biosciences) and labeled according to the manufacturer’s instructions. A FACSCalibur and CellQuest software (BD Biosciences) were used for flow cytometric analysis.
As we have previously described (26), TECs were defined as CD45−MHC class II+ and confirmed to express epithelial cell adhesion molecule. TEC subsets were defined using MHC class II and UEA1 lectin. The mTEChigh subset stains with high levels of MHC class II and UEA1 and includes a subset of cells expressing Aire (27). The mTEClow subset is defined as MHC class IIlow and UEA1low. cTEChigh and cTEClow subsets express high and low levels of MHC class II, respectively, and are UEA1 negative.
Abs and conjugates
Abs and conjugates were from BD Biosciences unless indicated. We used UEA1 lectin (Vector Laboratories), anti-Ly51 (clones 6C3 and BP-1); anti-IA/IE (M5/114.15.2); anti-CD45 (30-F11), anti-Aire (5H12-2(12)), anti-epithelial cell adhesion molecule (G8.8a, a gift from Dr. A. Farr, University of Washington, Seattle, WA), anti-Plet-1 (MTS24; in-house), anti-Ki67 (B56), anti-CD25 (PC61), rabbit anti-bovine pan-cytokeratin (DakoCytomation), anti-rat IgG2c (Southern Biotech), anti-rat IgG2a (Southern Biotech), streptavidin-PerCP Cy5.5; streptavidin-allophycocyanin.
An unpaired, two-tailed Mann-Whitney rank sum U test was used to compare nonparametric data. Statistical analysis was performed using SPSS version 15.0 software.
CsA causes transient loss of tolerance-inducing mTEC
CsA inhibits T cell activation by blocking calcineurin-mediated dephosphorylation of the NF of activated T cells, responsible for initiating IL-2 transcription (28). In the thymus, CsA blocks positive selection, reducing numbers of CD4+ and CD8+ single-positive (SP) and CD3highCD4+CD8+ double-positive (DP) thymocytes (14, 15). Paradoxically, in some patients and animal models, thymus-dependent autoimmunity develops after treatment, caused by autoreactive thymic emigrants released during treatment (17, 21, 29), presumably aided by a loss of peripheral Tregs (19). We therefore examined alterations to the thymic microenvironment, particularly tolerance-inducing mTEC.
As reported (14, 15), mice treated daily with CsA for 14 days (i.e., mice at RD0, just before CsA withdrawal) developed mild thymic involution, including a small reduction in DP thymocytes, but a severe loss of SP thymocytes (data not shown), with full thymocyte recovery taking 28 days post-CsA withdrawal. Consistent with the propensity to induce thymus-dependent autoimmunity (17, 21, 29), and as previously shown (19), we also found a significant reduction in thymic dendritic cells and Tregs (data not shown).
In assessing whether a direct effect of CsA on TECs was likely, we found that CD25 (IL-2 receptor α) is not expressed by CD45− cells (Fig. 1,A) and no role for calcineurin in maintaining TECs has been reported. Nevertheless, in accordance with some (20, 21) but not all (18) published histology data, we found that 2 wk of daily CsA treatment (at RD0) significantly reduced TEC, from 27.4 ± 4.9 × 104 in untreated mice to 20.3 ± 1.2 × 104 in CsA-treated mice (mean ± SD; p < 0.01). TECs were further divided into four subsets (described in Materials and Methods) based on staining with UEA1 and MHC class II (Fig. 1,B). CsA severely reduced the proportion (Fig. 1, C and D) and number (Fig. 1 E) of the mTEChigh subset after 2 wk of treatment (by RD0). These cells were replenished by RD10. mTEChigh cells were not simply down-regulating MHC class II during treatment, because the number of mTEClow cells did not increase. Nor did they down-regulate UEA1, because 96–98% of the UEA1− MHC class IIhigh cells expressed the cortical marker Ly51 (data not shown). This suggests that the mTEChigh subset was either directly damaged or lost through symbiotic attrition following the reduction in SP thymocytes.
The mTEClow subset was affected to a lesser extent than mTEChigh. After 2 wk of daily treatment (RD0), the number of mTEClow had fallen ∼25% (Fig. 1,E). This occurred during the first 7 days of treatment (data not shown), after which the number of mTEClow cells remained stable until 11 days later, between RD4 and RD10, when the number and proportion of mTEClow reduced concomitantly with mTEChigh recovery (Fig. 1, C–E). By RD10, the total number of mTEC increased, and all TEC subsets returned to normal by RD14, indicating replenishment of both the mTEChigh and mTEClow subsets (Fig. 1 E).
Unlike mTEClow, cTEChigh and cTEClow subsets increased in proportion and number during CsA treatment (before RD0). These populations returned to homeostatic numbers by RD7, before mTEChigh recovered, and did not follow the striking pattern of late numerical loss shown by mTEClow (Fig. 1 E). This late reduction in mTEClow, well after treatment withdrawal, was in accordance with their hypothesized role as progenitors of tolerance-inducing mTEChigh (11, 13, 27).
Aire expression is quickly restored
Aire was almost completely abrogated by treatment with CsA (Fig. 2,A). Given that these cells are terminally differentiated and that they pass through a number of developmental stages (11, 27), it was surprising that by RD7, the proportion of TECs expressing Aire was significantly increased compared with untreated controls, and in fact the number of Aire+ mTECs was already fully regenerated at this time point. There was a significant increase in the number of Aire+ mTECs at RD10 (Fig. 2,B). Because Aire+ mTECs develop from putative transit-amplifying (TA) Aire−mTEChigh (11, 13, 27), which in turn are purported to develop from mTEClow (13, 27), the ratio of these cell types was examined. Ordinarily, the ratio of Aire−mTEChigh to Aire+ mTEChigh is 2:1 (Fig. 2 C). The proportion of mTEChigh subsets expressing Aire normalized by RD7 but increased to a 1:1 ratio at RD10, in accordance with increased numbers of Aire+ mTEC compared with untreated controls, suggestive of either faster maturation of Aire+ mTECs, or a prolonged half-life of newly generated cells. These data show that newly generated mTEChigh cells follow a swift program of differentiation to restore Aire expression.
mTEChigh cells are actively cycling during thymic regeneration
Given the intense interest in mechanisms of mTEC generation and regeneration, likely mechanisms behind the swift increase in mTEChigh were addressed. Ki67 expression studies determined that the remaining mTEChigh entered a state of relative quiescence by RD0 but entered cell cycle at normal proportion within 4 days after treatment ceased (Fig. 3). After RD4, and in untreated mice, the highest proportion of cells positive for Ki67 was mTEChigh (Fig. 3,B). Increased numbers of cTEChigh were entering cell cycle at RD0 (Fig. 3,A), consistent with their early increase in number (Fig. 1,E), but the proportion of cTEChigh expressing Ki67 was no different from that of untreated mice (Fig. 3 B), suggesting that the cells were accumulating, not actively undergoing expansion.
Persistent loss of mTEClow coincides with increase in mTEChigh
A large proportion of mTEChigh fit the profile of TA cells, supported by a body of experimental evidence (11, 13, 27). By definition, TA cells develop from a progenitor cell, before proliferating to amplify cell numbers, and subsequently maturating into a terminally differentiated cell type (30, 31). Evidence in other systems, using adult or embryonic TECs placed into a permissive embryonic environment (11, 13, 27), suggests that mTEClow subsets represent a quiescent progenitor of the purported TA mTEChigh subset. However, direct differentiation studies following adult TECs transplanted into adult thymi are not yet available.
The high level of mTEChigh cell cycling seen soon after treatment withdrawal does not explain the observed numerical loss of mTEClow cells as mTEChigh numbers recovered (Fig. 1 C). To assess the fate of mTEClow lost during mTEChigh recovery, we tracked the loss and recovery kinetics of mTEClow and mTEChigh cells with a different treatment time course.
The experimental protocol was varied to address two hypotheses (depicted schematically in Fig. 4 A). Hypothesis i states that the late reduction in mTEClow is intrinsically linked to the recovery of mTEChigh. Here, regardless of the length of treatment, the number of mTEClow should reduce during the recovery phase, as mTEChigh are replenished. Because mTEChigh should recover at a similar time point after treatment withdrawal, regardless of the regimen, it would be expected that mTEClow should always further diminish posttreatment withdrawal. Hypothesis ii states that the late reduction in mTEClow occurs independent to the recovery of mTEChigh, most likely due to damage to an unknown mTEClow-specific precursor cell, resulting in attrition as mTEClow cells reach the end of their lifespan. According to hypothesis ii, the reduction in mTEClow cells should always occur at the same time point after treatment begins.
To test these alternatives, we extended the original treatment protocol, this time treating mice with CsA every day for 21 days rather than 14 (Fig. 4,B). Our data supported hypothesis i. The number of mTEClow always further reduced between RD4–10 as mTEChigh recovered, regardless of whether CsA had previously been administered for 3 wk (Fig. 4,B), 2 wk (Fig. 1 C), or 1 wk (data not shown). The loss of mTEClow from RD4 to RD10 did not occur at a particular time after treatment was started; rather, it occurred at a defined time point after treatment was withdrawn.
It is possible that an as yet unknown progenitor of both mTEClow and mTEChigh skews toward replenishing mTEChigh in damage situations, resulting in a coincident loss of mTEClow from attrition. This would result in an experimental profile that would also fit hypothesis i. We believe that it is more likely, as suggested by other experimental systems (11, 13, 27), that residual mTEClow up-regulate MHC class II and become mTEChigh cells as a developmental progression, followed by extensive proliferation and then Aire expression.
Plet-1+ TEC are proportionally resistant to CsA treatment, but Plet-1+ mTEClow are lost during thymic regeneration
The Ab MTS24 recognizes the Plet-1 Ag (32), expressed in the emerging E10.5 thymic anlage (33, 34, 35) as well as a progenitor population of follicular keratinocytes (36). In the embryo, Plet-1 expression does not exclusively identify TEC progenitors, but does preferentially encompass the high-efficiency progenitors (34, 35, 37). We therefore examined MTS24 staining in these adult thymus models as a means of potentially enriching for progenitor TEC populations.
In the postnatal thymus, as previously reported, a large subset of TEC express Plet-1, including both MHC class IIhigh and MHC class IIlow cells (35) and Fig. 5,A). When assessed more specifically on gated TEC subsets, it was found that the majority of MTS24+ TECs were mTEClow (Fig. 5 A).
At RD0, after 2 wk of CsA treatment, TECs expressing Plet-1 were found to be proportionally and numerically more resistant than Plet1− cells (Fig. 5,B). Furthermore, cTEChigh and cTEClow Plet-1+ cells also increased in both proportion and number following CsA treatment (Fig. 5 C), showing expression of this Ag by more cells in response to thymic damage, over and above the selective sparing of Plet-1+ cells. Plet-1 expression returned to normal in all subsets by RD10. The loss of Plet-1+ mTEClow cells at RD7 suggests that these cells either suddenly become susceptible to cell death 7 days after treatment was withdrawn (despite being more resistant than other cells throughout 2 wk of treatment), or that progenitor cells residing within the mTEClow subset lose expression of Plet-1 upon differentiation to mTEChigh.
In further characterizing mTEClow cells in untreated mice, we noted a distinct subpopulation coexpressing Ly51, a marker commonly used to identify cortex (Fig. 5,D, histogram). It has recently been shown that a minor proportion of Aire+ cells (exclusively UEA1+) also express Ly51 (12); we therefore examined Ly51 expression across TEC subsets. In untreated mice, analysis of UEA1 and Ly51 expression indicates 3 cell populations (Fig. 5 D, large dot plot). As expected, UEA1+Ly51− cells (population i) were primarily mTEChigh and mTEClow, where UEA1−Ly51+ cells (population iii) fell within the cTEChigh and cTEClow gates. However, cells coexpressing both markers (UEA1low, Ly51+; population ii) were primarily MHC class IIlow, falling within the mTEClow gate (see small dot plots and bar graph, each gated from large dot plot). Any distinct function or further phenotype of this new TEC subset remains to be determined, but certainly Ly51 and UEA1 are not mutually exclusive, and the mTEClow subset is more heterogeneous than previously reported.
Conserved regeneration patterns across more severe models of thymic damage
We studied other models of thymic damage to see whether Aire+ mTECs were similarly sensitive, and also to determine whether mTEChigh regeneration under more severe ablative conditions followed the same pattern, proportionally increasing at the expense of mTEClow. The clinically relevant cytotoxic agent cyclophosphamide induced severe thymic involution (Fig. 6,A), as previously reported (22). Similarly, the proportion (Fig. 6, B and C) of all TEC subsets was profoundly altered following treatment as epithelial involution occurred, causing a sustained reduction in TEC number across all subsets (Fig. 6,D). Total TEC fell to a low of 7.2 ± 0.4 × 104 compared with 21.4 ± 6.8 × 104 in untreated controls (mean ± SD, p < 0.01). As with CsA treatment, cyclophosphamide caused the greatest numerical reduction within mTEChigh subset (Fig. 6,D). This reduction again resulted in a profound loss of Aire-expressing cells (Fig. 6,E). Although thymic cellularity was restored to near untreated levels by 10 days posttreatment (Fig. 6,A), the proportion (Fig. 6, B and C) and number (Fig. 6 D) of TEC subsets took between 14 and 28 days to return to normal.
Dexamethasone is a potent, clinically relevant glucocorticoid used to suppress immune responses. It induced a severe degree of thymic involution comparable with cyclophosphamide (Fig. 7,A), and again extensively depleted total TECs (5.8 ± 2.0 × 104, compared with 28.9 ± 4.6 × 104 in untreated mice; mean ± SD, p < 0.01). All TEC subsets were ablated, to varying degrees (Fig. 7, B–D), but again, the mTEChigh subset proved most sensitive to treatment. Although cyclophosphamide, dexamethasone, and CsA each affected mTEChigh to a comparable extent, these cells took longer to recover after cyclophosphamide and dexamethasone treatment (compare Figs. 6,D and 7,D with Fig. 1 E).
During recovery from both cyclophosphamide and dexamethasone, the proportion of mTEClow again steadily reduced in proportion concomitant with mTEChigh recovery (Figs. 6,C and 7 C), as seen during mTEChigh regeneration post-CsA. This pattern of mTEC regeneration held true regardless of the method used to achieve mTEC ablation or the severity of damage sustained.
Numerically, a different pattern of recovery was noted following severe ablation, where mTEClow and mTEChigh both increased in number between RD7 and RD10, with mTEChigh expanding proportionally faster, again befitting a TA population (Figs. 6,D and 7 D). The same proportional skewing of mTEC subsets, with the same kinetics of recovery, were noted regardless of the sex of the mice, suggesting similar regenerative patterns across both sexes. Aire was also severely ablated post-dexamethasone (S. Sakkal, manuscript in preparation). Unlike CsA, a significantly increased rebound in numbers and proportion of Aire+ TEC beyond untreated levels was not observed in these models where damage to all TEC subsets occurred.
The thymus undergoes acute, naturally reversible involution during pregnancy (38) or stress (39), demonstrating an ability to regenerate. Similarly, thymic damage induced by cytoablative chemotherapy or radiotherapy causes acute involution followed by eventual regeneration (22, 40, 41), at least in young animals. TEC regeneration is essential for the restoration of an immunocompetent, self-tolerant T cell pool in such conditions. This is particularly pertinent for CsA, which is commonly used to manage allograft rejection, yet can cause a T cell-mediated graft-vs-host-like autoimmune disease (17, 20, 29, 42) or organ-specific autoimmunity (16) to develop. Loss of peripheral suppressor cells is contributive (20, 21, 29, 43), but our results also clearly demonstrate a significant loss of Aire-expressing mTEChigh. The mTEChigh subset is also exquisitely sensitive to at least two cytoreductive therapies: cyclophosphamide and dexamethasone.
The well-recognized cross-talk symbiosis between thymocytes and stromal cells makes it difficult to ascertain whether CsA directly affects mTEC number. Because TECs do not express IL-2Rα, it is more likely that mTEChigh are lost secondary to an effect on SP T cells, which is supported by robust evidence that SP T cells maintain the mTEC maturation program through provision of ligands for CD40 and TNF-related activation induced cytokine (44, 45, 46), but we cannot exclude a contributive role for calcineurin in mTEC maintenance.
mTEChigh and Aire+ mTEC recovered fully before SP thymocytes reached steady-state numbers, in principle providing appropriate tolerance-inducing mechanisms. However, SP T cells were never completely depleted. Although evidence clearly states that the pathogenic T cells arise from poorly censored, autoreactive recent thymic emigrants emerging during treatment (47), further work should aim to definitively show that loss of Aire (in addition to established effects on Tregs and dendritic cells; Refs. 17 and 20) contributes to their escape. Certainly, the reported pattern of organ-specific autoimmunity (16) in CsA-treated mice shows some hallmarks of Aire deficiency (gastritis, oophoritis, pancreatic infiltration; Ref. 7).
TEC regeneration took longer in cyclophosphamide- and dexamethasone-treated mice compared with CsA, reflecting a more severe ablation of all thymocyte and TEC subsets. Unlike recovery from CsA, all TEC subsets began to numerically recover simultaneously, reflecting an expedited program of emergency expansion and differentiation, in parallel with thymocyte subsets, to restore the thymic microenvironment. The increase in cTEC during CsA treatment was striking, and the possibility that these cells (or a subset) contribute to mTEC regeneration cannot be ruled out. However, arrested DP thymocyte emigration from cortex to medulla during treatment may be contributive, and cTEC returned to normal number before either the increase in mTEChigh or decrease in mTEClow. Recent work defined cTEC differentiation in embryonic thymi by up-regulation of surface markers (48) analogous to the maturation program for mTEC proposed here and by others (11, 13, 27), but there is as yet no experimental evidence to suggest that cTEC routinely generate mTEC in adult mice.
Recent data show Aire expression as a postmitotic, end-stage differentiation step (27). Combined with our present results showing the loss of mTEChigh and Aire following cytotoxic chemotherapy and immunosuppression, the fact that Aire recovered relatively quickly raises important questions regarding the origin of mTEC, in particular the differentiation kinetics and stimuli for relevant progenitor cells. It has been suggested, using data generated in vitro, that the minimum amount of time taken for an mTEClow cell to differentiate and up-regulate Aire is 3 days (27). In these in vivo studies, the process appears to take at least 7 days, probably reflecting the additional time required for differentiation and trafficking of mature SP thymocytes and/or lymphoid tissue inducer cells (11, 27, 44, 45, 46). However, there was a statistically significant rebound in Aire+ mTEC differentiation, such that more cells expressed Aire at day 10 posttreatment than at day 7 or in untreated mice, pointing to an early, prioritized restoration compared with other mTEC subsets. The significantly increased proportion of Aire+ cells at day 10, even though Aire+ mTEC were fully recovered at day 7, may suggest an early commitment to Aire expression, or increased half-life of newly generated cells. The return to normal number and ratio in just 4 further days suggests that these cells were not replaced after cell death to restore their homeostatic balance.
Unlike Gray et al. (27), we did not find a proportion of Aire+ cells expressing low levels of MHC class II, suggesting the possibility of artifacts during TEC development in in vitro culture systems. However, our results do support the overall findings of this and similar studies (11, 27), which showed that mTEClow, when sorted and either injected into fetal thymic lobes, or reaggregated with embryonic TEC, could differentiate into Aire+ mTEChigh.
As the mTEChigh subset regenerated, there was consistent proportional loss of mTEClow regardless of the treatment used or degree of thymic damage sustained. Following CsA, the total number of mTEC remained relatively constant through RD0, 4, and 7, during which the number of mTEClow cells reduced in direct proportion with the increase in mTEChigh. We then tested two hypotheses pertaining to the fate of the mTEClow. Hypothesis i suggested that the loss of mTEClow and recovery of mTEChigh were intrinsically linked, suggesting either a precursor-product relationship, or a common progenitor of both cell types which skewed its potential post-CsA to replenish mTEChigh, causing mTEClow attrition. Hypothesis ii stated that the reduction in mTEClow and recovery of mTEChigh were not linked, suggesting probable attrition at the end of their natural lifespan due to nonreplacement during the treatment period. The data were clearly consistent with Hypothesis i. The reduction did not occur because of mTEClow attrition following a defined mTEClow lifespan, because the reduction in mTEClow always occurred 4–10 days after treatment ceased, even if the treatment period was extended.
Importantly, up-regulation of MHC class II was accompanied by full acquisition of Aire expression. If the mTEClow population merely up-regulated MHC class II until mTEChigh cells developed from a separate lineage, Aire expression would lag until just before mTEClow recovery. If, however, mTEClow was a relatively quiescent population that sits static or turns over slowly until required to repopulate the actively cycling mTEChigh subset, the Aire ratio should recover early and quickly, as observed.
The expression of Plet-1 (identified with the MTS24 Ab) by the CsA-resistant population is interesting given that it identifies a population containing high-efficiency TEC progenitor cells in the embryo (34, 35, 37). However, although transplantation of as few as 2500 MTS24+ embryonic TECs generates a complete thymic microenvironment (35), a recent study showed that both MTS24+ and MTS24− TEC could generate thymic grafts if much higher cell numbers (100,000 cells) were transplanted together with E12 thymic mesenchyme. These studies show that although Plet-1 is not restricted to embryonic progenitors, MTS24 can be used to enrich for these cells. After E18, MTS24+ TECs no longer reaggregate for transplant (37), but Plet-1 is still expressed by a minority of TEC throughout life (35). It has been suggested that Plet-1 expression is progressively restricted to mature, differentiated cells (37).
Conversely, we report that the mTEClow subset, which is suggested in this study and others (11, 13, 27) to contain a subset of quiescent mTEC progenitor cells, expressed the highest levels of Plet-1. A similar finding can be inferred from an earlier study (35), which showed that most MTS24 staining in 4-wk-old (CBA × C57BL/6)F1 mice was found on MHC class IIlow cells, although the enriching stromal cell isolation protocol used in that study, itself biased toward mTECs (26), results in slightly different flow cytometry profiles. We further found that MTS24+ cells were proportionally more resistant to cytoreductive therapy, as would befit quiescent cells, and that MTS24+ mTEClow numbers were significantly reduced during mTEChigh regeneration. A lower proportion of the mTEChigh subset was found to express Plet-1 compared with other TEC. This observation is in keeping with the hypothesis that Plet-1 is expressed by less mature epithelium, because the mTEChigh subset has been shown to encompass a large terminally differentiated Aire+ population (27).
In summary, although requisite lineage tracking studies following adult TEC subsets injected into adult thymi are yet to be performed, our data support the theory (11, 13, 27) that mTEClow generate mTEChigh. Under this model, these data suggest that a quiescent subset of UEA1lowMHC class IIlowCD80− cells are capable of swift differentiation into the proliferating mTEChigh subset. A proportion of these cells also express cortical marker Ly51 forming a triple-low population (MHC class IIlowUEA1lowLy51+) and showing further heterogeneity within TEC subsets. These cells now form a basis for transcriptome analysis and more detailed functional studies.
These studies document, for the first time, the regenerative kinetics of Aire+ mTEC and other TEC subsets in vivo following immunosuppression and cytotoxic chemotherapy using common clinically relevant agents. Although we show evidence herein to suggest the existence of a developmental hierarchy among mTEC subsets during thymic regeneration, a true TEC stem cell in adult mice, capable of differentiating into both cortical and medullary TECs and their subsets is yet to be unequivocally defined.
We thank Jade Homann and Jade Barbuto for expert animal care; Daniel Layton, Tomoo Ueno, Adele Barnard, Melanie Hince, and Adam Uldrich for helpful discussions; and Geza Paulovic, Andrew Fryga, and Darren Ellemor for cell sorting.
R.L.B. is chief scientific officer of and A.P.C. is a consultant to Norwood Immunology, Ltd.
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.
This work was supported by the Australian Stem Cell Centre, Norwood Immunology, and the National Health and Medical Research Council of Australia. H.S.S. was supported by National Health and Medical Research Council Fellowships 171601 and 461204, National Health and Medical Research Council Program Grants 257501 and 264573, and Eurothymaide, the 6th Framework Programme of the European Union.
Abbreviations used in this paper: TEC, thymic epithelial cell; Aire, autoimmune regulator; CsA, cyclosporine A; RD, recovery day; cTEC, cortical TEC; mTEC, medullary TEC; mTEChigh, medullary TEC expressing high levels of MHC class II; mTEClow, mTEC expressing low levels of MHC class II; TA, transit amplifying; TRA, tissue-restricted Ag; UEA1, Ulex europeaus agglutinin 1; Treg, regulatory T cell; SP, single positive; DP, double positive.