Thymic epithelial cells (TEC) are essential components of the thymus that guide and control the development and TCR repertoire selection of T cells. Previously, TEC have been considered as postmitotic cells that, once generated during ontogeny, were maintained in their mature state. Recently, it has become clear that TEC can be generated from common or committed medullary and cortical TEC progenitor cells in ontogeny, that stages of immature and mature TEC are phenotypically separable, and that TEC undergo a rapid turnover in a matter of a few weeks. All of these findings strongly suggest that in the adult thymus mature TEC are constantly regenerated from a pool of stem or progenitor cells, a view that renders the thymus structure potentially much more dynamic than previously thought. However, the identity of “thymus stem cells” is elusive, and developmental stages of TEC development are only beginning to be elucidated.
T-cell development within the thymus is a non-cell autonomous process, and requires constant input from the surrounding thymic microenvironment 1. Thymic epithelial cells (TEC) play key roles at multiple stages of T-cell development where they impose self-MHC restriction on the immature thymocytes via positive selection, and aid in the elimination of autoreactive thymocytes via negative selection 2. That analysis of TEC development has lagged behind our understanding of the steps regulating thymocyte differentiation is widely acknowledged, and is due in part to the paucity of suitable cellular and molecular tools to dissect TEC biology. However, in recent years, significant technical and conceptual advances have been made that shed light on pathways regulating the formation and possibly maintenance of TEC microenvironments. The aim of this review is to summarize these recent advances and to highlight key areas of future investigation (Fig. 1).
At around embryonic day 10 of gestation, the parathyroid and thymic domains of a shared anlage can be identified by expression of the transcription factors Gcm2 and Foxn1, respectively 3. The lack of compartmentalization of early embryonic thymus contrasts with the well-defined cortical and medullary areas of the adult thymus. Interestingly, the apical layer of the bilayered TEC rudiment is defined by expression of the tight junction components claudin-3 and -4, which later in thymic development are associated with medullary thymic areas 4. In addition to a lack of anatomical compartmentalization, the thymus anlage lacks expression of mature TEC markers including MHC class II 5, suggesting that cortical TEC (cTEC) and medullary TEC (mTEC) arise during thymus development from progenitors lacking these markers. Whereas morphological studies on thymus development in the nude mouse were taken as evidence for distinct ectodermal and endodermal origins of cTEC and mTEC, respectively 6, studies on birds indicated a single endodermal origin for cTEC and mTEC 7. Although this single germ layer origin has since been confirmed in the mouse 8, whether cTEC and mTEC arise from a common progenitor population, or from distinct progenitor pools of endodermal origin, was unclear.
Identification of bTEC progenitors (bTECp)
The existence of bTEC progenitors (bTECp) was first suggested by the identification of rare TEC simultaneously expressing cTEC and mTEC markers 9–11. Histological and DNA micro-satellite analysis of a series of tetraparental chimeric mice provided genetic evidence for a “late” common developmental origin of mTEC and cTEC 1, 12 and two recent studies simultaneously reported the existence of bTECp. While cell fate mapping of individual TEC from the early embryonic thymus proved the presence of a dominant population of bTECp 13, random reactivation of Foxn1 in individual TEC demonstrated the continued persistence of bTECp in the postnatal thymus 14. Earlier work had shown that the thymus medulla is composed of single cell-derived islets, providing evidence for mTEC progenitor activity during TEC ontogeny 15. Single cell-derived clusters of cTEC have also been reported 14. Collectively, the identification of bTECp, or their activity, as well as evidence for committed mTEC and cTEC progenitors can be used as a starting point for developmental stages of TEC. In the absence of Foxn1 expression, the earliest bTECp are apparently reversibly arrested in their development 14. As is the case for recent progress in the identification of other epithelial progenitors, such as those for mammary 16, prostate 17 and intestinal 18 tissues major facilitators in elucidating TEC progenitor biology are the accurate phenotypic identification of these cells, and assays to reveal their functions at the clonal level. In the E12 thymus, bTECp were purified on the basis of expression of EpCAM1 13. However, because EpCAM1 is a pan-epithelial marker, it does not separate bTECp specifically from all other TEC at this or later stages. Identifying markers that uniformly recognize TEC progenitors in the early thymus, which then become progressively restricted during thymus development is one potentially useful strategy in searching for a TEC progenitor phenotype. At E12, all TEC react with the antibody Mts24 recently shown to bind to Plet-1 19, with expression in adult thymus being limited to rare TEC in the medulla. Although initial reports suggested that Mts24 could be used to identify and most of all purify TEC progenitors, 20, 21 subsequent studies showed this not to be the case 22, 23. Thus, at present, bTECp have only been identified on the basis of their functional properties, and despite evidence for their persistence beyond the embryonic period 14, the virtual lack of appropriate markers makes it impossible to track and localize these cells in the postnatal thymus.
Regulation of bTECp development
While thymocyte-TEC crosstalk influences some aspects of TEC development, the initial specification of bTECp into the cTEC and mTEC lineages can occur independently of thymocytes 24, 25. Whether persisting bTECp in the postnatal thymus can also differentiate in a thymocyte-independent manner is not clear. This initial lack of requirement for thymocyte-derived signals for TEC differentiation in the embryonic thymus suggests that TEC differentiation is either cell-autonomous, or involves input from additional non-haemopoietic and/or non-T-lineage haemopoietic cells. In this regard, removal of perithymic mesenchyme from early embryonic thymus does not prevent emergence of cTEC and mTEC, indicating that any requirement for mesenchyme in TEC differentiation does not require prolonged contact 26. In contrast, TEC progenitor expansion requires sustained interactions with mesenchyme, and their production of FGF 26. Much of the mesenchymal contribution to the embryonic thymus is neural crest-derived, which persists into the adult thymus 27, 28. Whether these cells influence TEC proliferation at this stage is not clear, although it is interesting to note that adult thymic mesenchyme continues to express FGF7 and FGF10, factors which influence survival of adult TEC 29. Finally, it is important to note that, in addition to the suspected de novo generation of cTEC and mTEC from bTECp, the studies of Gray et al. 30 also demonstrate the capacity for cTEC and mTEC generation via expansion of pre-exisiting subsets, suggesting that the generation and turnover of TEC compartments is likely to be a result of bTECp activity together with subsequent expansion of cTEC and mTEC restricted progeny.
It has become increasingly clear that mTEC, notably the subset expressing the Aire gene, play a key role in T-cell tolerance 31. Although anti-Aire antibodies 32 and subsequently Aire-GFP reporter mice 33, 34 have been useful in identifying Aire+ mTEC, until recently the mechanisms regulating their development was unclear. Relevant to this, we and others 35–37 showed that Aire+ mTEC derive from a Aire−CD80− progenitor population, findings in-line with a Terminal Differentiation Model38 of mTEC development, which has recently been suggested to culminate in the appearance of Involucrin+ mTEC 34. Identifying precursor–product relationships of mTEC is important in elucidating the stages and checkpoints that regulate formation of thymic microenvironments, and provides an opportunity to identify the molecular mediators of these stages. With this is mind, recent studies have shown the importance of TNF receptors in mTEC development. Indeed, whereas lymphotoxin (LT) was initially thought to induce development of Aire+ mTEC 39, we and others showed this was not the case 35, 40, 41 although LTβR signaling clearly influences some aspects of mTEC development 42, 43. Instead, we showed receptor activator of NF-κB (RANK) expression by mTEC, which is directly responsible for Aire+ mTEC differentiation 35. In addition, lymphoid tissue inducer cells were identified as a key population of RANK ligand (RANKL)-expressing cells in fetal thymus. Subsequent studies have shown a role for RANK and CD40 in development of adult mTEC 44–46, which is consistent with the absence of Aire+ mTEC in TNF receptor associated factor 6-deficient mice 47. Interestingly, in adult thymus RANKL can also be provided by CD4+CD8− thymocytes 44, which also express CD40L 48. However, even in adult thymus, mature thymocytes are still not essential for Aire+ mTEC generation, as these cells are present in adult MHC class I/class II double-deficient mice 48 and TCR-α-deficient mice 44 where thymic selection is absent. Thus, it may be the case that in the adult, several RANKL+ cell types influence Aire+ mTEC development.
There are few markers that enable isolation and study of cTEC, which are typically identified by an EpCAM1+Ly51+ phenotype. Interestingly advances have been made in understanding the specialization of cTEC for positive selection (reviewed in 49). A novel proteasome subunit, β5t, was reported 50 which is expressed specifically within cTEC, and is linked to their ability to mediate positive selection of CD8+ T cells. No doubt, generation and use of anti-β5t antibodies and β5t-reporter mice will provide useful tools to study cTEC heterogeneity and development. Most recently, Gommeaux et al. 51 showed that deletion of Prss16, the gene encoding thymus-specific serine protease that is specifically expressed by cTEC, is important in MHC class II restricted positive selection of CD4+ T cells. Interestingly, Boehm shows that unlike β5t, the emergence of Prss16 in vertebrate evolution predates adaptive immunity, suggesting that this pre-existing gene was adopted by cTEC 49.
We recently set out to define stages in cTEC development, and proposed the existence of a proliferating EpCAM1+CD205+CD40−MHC class II−/low cTEC progenitor population 52. Development of this cTEC subset can occur independently of thymocyte-derived signals, whereas further progression is blocked in the absence of thymocyte development, suggesting a stage-specific requirement for thymocyte crosstalk in development of the cTEC lineage. Collectively, the identification of β5t, together with the use of markers such as CD205, and a recently described antibody that detects the cTEC-expressed Notch ligand Delta-like 4, that is essential for T-cell development 53–56, should shed light on the stages of cTEC development and in understanding events underlying their regulation.
Bipotent TEC progenitors provide a starting point to dissect the mechanisms regulating lineage choice within the TEC compartment, and this represents a clear goal for future studies. Precursor–product relationships have been identified within the mTEC lineage, and RANK and CD40 have been shown to be key regulators in the generation of Aire+ mTEC. The identification of CD205+β5t+CD40− cTEC progenitors should also help provide insight into regulation of cTEC development. New experimental models enabling genetic manipulation and gene reporter analysis in TEC 14, 33, 34, 7–61 should prove to be useful tools in further studies aimed at a better understanding of the formation of mature and functional cTEC and mTEC microenvironments.
G.A. and E.J.J. are supported by an MRC Programme Grant, H.R.R. is supported by grants from the Deutsche Forschungsgemeinschaft (SFB 497-B5 and KFO 142-P8).
Conflict of interest: The authors declare no financial or commercial conflict of interest.