A 2020 View of Thymus Stromal Cells in T Cell Development

The thymus is a primary lymphoid organ in which bone marrow–derived lymphoid progenitor cells, with continuous support from thymic stromal cells, develop into functionally competent T cells. The developmental stages followed by the majority of thymocytes are conventionally classified by the surface expression status of CD4 and CD8, starting with double-negative (DN), immature single-positive (SP), double-positive (DP), and mature SP T cells (Fig. 1). The DN stage can be further classified, as DN1 to DN4 cells, based on the expression of CD44 and CD25, with DN1 cells expressing CD44 but not CD25, followed by appearance of CD25 during DN2 stage, loss of CD44 at DN3 stage, and finally, downregulation of CD25 in DN4 cells. The DN1 population contains a small subset (DN1a/b) that is Kit⁺ (CD117⁺), also known as early thymic progenitor cells (ETPs), and is regarded as the earliest progeny of the bone marrow–derived thymic-seeding progenitors (TSPs) (1). Throughout their development, thymocytes need to pass three major selection checkpoints: TCR β-selection at the DN3 to DN4 stages and positive and negative selection at the DP to SP stages (Fig. 1). This brief review will focus on the following: 1) the roles thymic stromal cells, which can include all non-T lineage cells, such as thymic epithelial cells (TEC), endothelial cells, mesenchymal/fibroblast cells, dendritic cells (DCs), and B cells, play during early thymocyte development; 2) transcriptional regulation of key genes involved in this process; and, 3) changes in thymic stromal cells during thymic involution and regeneration.

Thymus organogenesis in mouse embryos starts at the third pharyngeal pouch around embryonic day (E) 11, when the future thymus coexists with the future parathyroid glands in primordia. The common primordia then detach from the pharynx and migrate caudally and medially, during which thymic anlagen separate from the parathyroid anlagen between E12 and E12.5. The parathyroid rudiment then stays where they separate as the thymic anlagen continue their migration within the thorax and ends up on top of the heart (2).

Initially, the thymus, made up of two separate lobes, is solely composed of endoderm-derived TEC progenitors, which are enveloped by a neural crest cell–derived capsule (3). Homing of the first wave of TSPs to thymic rudiment occurs as early as E11.5 and is mediated primarily by the activities of two pairs of chemokine ligand/receptors, namely CCL25/CCR9 and CCL21/CCR7 (4–6). Another pair of chemokine ligand/receptor, CXCL12/CXCR4, provides partial compensation in the absence of CCL25/CCR9 or CCL21/CCR7 (6). At E11.5, CCL21 but not CCL25 is detected within the common primordia (7). A day later, Ccl21 mRNA is found primarily in the parathyroid region at remarkably high levels, whereas Ccl25 mRNA was found to be expressed exclusively within the thymic rudiment. In contrast, Cxcl12 mRNA was present in both the thymus and surrounding tissues (8). Based on their expression patterns, CCL21 appears to be the major player to recruit the first wave of TSPs to the region surrounding thymus–parathyroid primordia because of its early and higher expression levels, but it would not attract/maintain TSPs exclusively inside of the thymic anlagen. It is the role of CCL25 to attract and maintain TSPs inside of the thymic anlagen. In contrast, CXCL12 could help to recruit TSP to common primordia and further into thymic anlagen in the absence of CCL21 or CCL25, but its contribution would be relatively minor because of its lower expression levels as well as lack of a gradient between the thymic rudiment and surrounding tissues. This is consistent with the impaired thymopoiesis phenotype observed in E14.5 embryos lacking these three chemokine receptors individually or in various combinations (5, 6). Loss of CCR9/CCL25 signaling alone resulted in the greatest reduction, whereas loss of CXCR4/CXCL12 signaling alone led to the mildest reduction in the number of ETPs. Although simultaneous loss of any two signaling pathways resulted in greater reduction in ETPs than loss of corresponding individual ones, double knockout of Ccr9 and Ccr7 almost eliminated the homing of the first wave of TSPs. It should be noted that homing of first-wave TSPs gradually decreased from E12 until the second wave of TSPs homing into the vascularized thymus (9), likely as a result of the separation of thymic anlagen from parathyroid anlagen and consequently reduction in CCL21 levels surrounding thymic rudiment (2). The importance of the parathyroid anlagen in the homing of the first wave TSPs was also reflected by the reduced thymic colonization of lymphoid progenitors in mice deficient in GCM2, the transcription factor critical for parathyroid gland development (2, 5).

After the thymus becomes vascularized around E15.5 (5), chemokines produced by thymic endothelial cells appear to also play an important albeit less essential role in TSP homing. The expression profiles of chemokines in thymic endothelial cells have not been systematically investigated, but mRNA encoding Ccl25, Cxcl12, and Ccl19, another CCR7 ligand, have all been detected in thymic endothelial cells in neonatal or adult mice (10–12). Among them, CCL25/CCR9 signaling plays a dominant role in TSP homing, as its deficiency alone but not that of CCR7 and/or CXCR4 led to a detectable reduction in TSP homing, although additional disruption of CCR7 and/or CXCR4 resulted in further ETP reduction in E17.5 or adult thymus (6, 13–15). It should be pointed out that chemokines produced by TEC regulate intrathymic trafficking of developing thymocytes (16). In particular, it has recently been shown that CCL21 protein encoded by the Ccl21a gene in medullary TEC (mTEC) is required for recruiting positively selected thymocytes to the medulla for effective negative selection (17).

In addition to chemokines, adhesion molecules expressed on thymic endothelial cells play a critical role in TSP homing. For example, P-selectin on thymic endothelial cells and its ligand PSGL-1, expressed on bone marrow–derived lymphoid progenitor cells, have been demonstrated to be critical for TSP homing (13, 18). In addition, ICAM-1 and VCAM-1, expressed on thymic endothelial cells, were also shown to be required for efficient adhesion and extravasation of TSPs when assayed using short-term homing assays (13).

Thymic endothelial cells likely also produce molecules other than those discussed above to facilitate TSP homing to the thymus. A recent study found that more than half of P-selectin⁺ thymic endothelial cells did not express Ly-6C at the cell surface, a phenotype similar to high endothelial venules found in lymph nodes (11). These Ly-6C⁻P-selectin⁺ endothelial cells were preferentially associated with the perivascular space at the corticomedullary junction (CMJ) region (11), where TSP were reported to enter the thymus (19, 20). Disruption of LTβR signaling, which was initiated by engagement of lymphotoxin/LIGHT produced by positively selected SP thymocytes to LTβR expressed on endothelial cells, resulted in a decrease in Ly-6C⁻ and an accompanying increase in Ly-6C⁺ population among P-selectin⁺ endothelial cells without significantly affecting the total number of P-selectin⁺ endothelial cells (11). This led to a reduction in the ETP population. RNA-sequencing analysis revealed, among the molecules discussed above, only an ∼2-fold increase of Vcam1 mRNA and ∼5-fold less of Cxcl12 mRNA in Ly-6C⁻P-selectin⁺ endothelial cells compared with Ly-6C⁺P-selectin⁺ endothelial cells, suggesting molecules other than those discussed above contribute to the activity of Ly-6C⁻P-selectin⁺ endothelial cells to facilitate TSP homing (11). Significant differences in the expression levels of many genes involved in cell adhesion, extracellular matrix organization, and blood vessel morphogenesis were observed (11), although which of these are additionally involved in TSP homing remains to be identified and elucidated.

In the postnatal thymus, the extremely low frequency (<0.001%) of newly arrived TSPs, now called ETPs, highlights the challenges of thymic entry by these rare cells. Nonetheless, ETPs must survive and proliferate to generate a significant pool of cells that will be induced to fully commit to the T cell lineage and eventually give rise to thymocytes bearing a diverse TCR repertoire. Key early players produced by thymic stromal cells are stem cell factor (SCF)/Kit-ligand and IL-7, which provide survival/proliferation signals. DN1 and DN2 cells express high levels of cell surface Kit. In mice with naturally occurring dominant loss-of-function mutation of Kit or Scf, the numbers of pre–β-selection early thymocytes decreased by 40-fold, although total thymic cellularity at birth decreased only by half, because of compensative proliferation (transit-amplifying cell rescue) during later stages of thymocyte development (21). The reduction in early thymocyte numbers was mediated at least partially by reduced cell proliferation as determined by BrdU incorporation.

Elegant transplant experiments showed that engraftment of an Scf mutant thymus under the kidney capsule of a wild-type (WT) recipient led to impaired thymocyte proliferation, demonstrating that it was primarily the loss of intrathymic SCF that affected the cellularity (21). Intrathymic SCF is primarily produced by thymic endothelial cells and cortical TECs (cTECs), with additional contribution from some mesenchymal/fibroblast cells (22). SCF-expressing endothelial cells were found almost exclusively in the cortex and coexpress high-levels of mRNA of Dll4 and Cxcl12 but not Ccl25 (22). There are two forms of SCF, membrane bound and secreted/soluble. DN1 cells but not DN2 cells were found preferentially associated with endothelial cells expressing membrane-bound SCF. Knocking out membrane-bound SCF specifically in endothelial cells led to a 30-fold reduction in the earliest DN1 cells, a 16-fold reduction in late DN1 cells, and an 8-fold reduction in DN2 cells, whereas a knockout of membrane-bound SCF specifically in TEC led to a more significant loss of DN2 cells (15-fold) than DN1 cells (4-fold), consistent with the proximity of DN1 cells to membrane-bound SCF-expressing endothelial cells (22). It should be noted that, although lymphoid progenitor cells in bone marrow express Kit and endothelial cells in bone marrow express SCF, the authors did not observe any decrease in lymphoid progenitors in bone marrow from these mice (22). Furthermore, it was shown that there was increased apoptosis but no difference in cell division when purified DN1 cells were cocultured in vitro with primary thymic stromal cells deficient in membrane-bound SCF compared with WT stromal cells, which led the authors to conclude that membrane-bound SCF mainly provides survival, rather than proliferative, signals to early DN cells (22). However, the in vitro culture conditions might not be optimal for cell proliferation. Given the significant difference in BrdU incorporation in early thymocytes from naturally occurring Scf mutants compared with that from WT mice thymus (21), it is likely that SCF provides both survival and proliferative signals to early developing thymocytes.

IL-7 is a key lymphopoietic cytokine produced solely by TECs in the thymus (23, 24). The IL-7R is composed of two subunits, the IL-7R unique α-chain (IL7Rα/CD127) and the common γ-chain (γc/CD132), shared with IL-2R, IL-4R, IL-9R, IL-15R, and IL-21R. Cell surface IL-7R expression could be detected in all but DP stages of thymocytes, with DN2 and DN3 cells showing the most responsiveness to IL-7 in an in vitro assay (25, 26). Germline deletion of γc, Il7ra, or Il7, as well as TEC-specific deletion of Il7, all led to >90% reduction in thymic cellularity in adult mice (27–31). The more severe phenotype than that seen in mice with a loss of SCF/Kit signaling is consistent with the fact that IL-7 signaling impacts all but the DP stages of thymocyte development, in contrast to the confined effect of SCF/Kit signaling on DN1/2 cells.

More DN cells from IL-7–deficient thymus than WT thymus were found in the G₀/G₁ phase of the cell cycle and were undergoing apoptosis (32). In addition, reduced cell proliferation was also observed in positively selected IL-7R–deficient thymocytes (33, 34), suggesting that both increased cell death and reduced cell proliferation are responsible for a severe reduction in thymic cellularity in the absence of IL-7 signaling. Mechanistically, IL-7 signaling has been found to regulate the expression of trophic receptors and cell cycle regulators in pre- and post–β-selection DN cells (35). In addition, IL-7 signaling has also been shown to upregulate the expression of anti-apoptotic Bcl2 and Mcl1 mRNA (32, 36). Consequently, forced expression of BCL-2 or knocking out proapoptotic Bim could partially restore thymic cellularity in mice deficient of IL-7 signaling (37–39). Besides the survival and proliferative activities, IL-7 signaling has also been proposed to coordinate the proliferation, differentiation, and Tcra recombination during β-selection and to participate in the lineage decision of CD8SP cells (35, 40).

In addition to SCF and IL-7, thymic stromal cells have also been proposed to regulate thymocyte survival and proliferation via Notch, CXCL12/CXCR4, Wnt, Hedgehog and BMP signaling (41–46). The role of Notch signaling will be discussed in detail in the following section.

Notch signaling provides the absolutely essential intrathymic cue that dictates lymphoid progenitors with multilineage potential to differentiate exclusively toward the T cell lineage. In the absence of Notch signaling, the thymus becomes a place where the development of B cells instead of T cells is supported (47–49). Although lymphoid progenitors and developing thymocytes express NOTCH1, NOTCH2, and NOTCH3 receptors, whereas thymic stromal cells express Notch ligands JAGGED1, JAGGED2, DLL1, and DLL4, only interactions between NOTCH1/DLL4 are physiologically required for T cell development in mice (50–53). DLL4 is expressed at high levels in most cTECs but not mTECs as well as thymic endothelial cells (22, 50). Although the functional significance of DLL4 on cTECs has been clearly demonstrated, the role of DLL4 expressed on thymic endothelial cells has not been directly investigated. It should be noted that knocking out Dll4 in all endothelial cells using Cdh5-CreERT2 led to an 80% reduction in the number of ETP and an even more severe (∼90%) decrease in total thymic cellularity (54). The impaired T cell lymphopoiesis was attributed to disruption of the DLL4-mediated bone marrow niche, which led to an ∼2-fold reduction in the frequency of common lymphoid progenitors, a TSP candidate, in the bone marrow. This interpretation is consistent with the recent finding from our laboratory that prethymic Notch signaling is required for generation of TSP in bone marrow (55). Nonetheless, it does not exclude the possibility that loss of DLL4 on thymic endothelial cells could partially contribute to the severely impaired thymopoiesis observed in these pan-endothelial Dll4 knockout mice (54).

A comprehensive recent study compared the distribution of the above mentioned four Notch ligands between postnatal human and mouse thymus and further examined their temporal changes in human thymus using immunofluorescence on thymic sections (56). The authors observed a significant reduction in DLL4 levels in human postnatal (<18 mo) cTEC compared with fetal (11–19 wk of embryonic development) cTECs. Of note, only low levels of DLL4 were observed on rare cTEC at subcapsular regions as well as on some mTEC; although, thymic endothelial cells, perivascular mesenchymal (CK19⁻CD34⁺) cells, and medullary myeloid (CD11c⁺) cells still expressed high levels of DLL4. This is consistent with a previous report showing that only 10–20% of cTECs obtained from the thymus of early postnatal humans expressed detectable levels of cell surface DLL4 as determined by flow cytometry (57). Nonetheless, the high levels and frequencies of DLL4 expression seen in fetal cTECs when thymocyte development is highly active, as well as its high expression levels on endothelial cells and perivascular cells at the CMJ, where TSPs enter the thymus, suggest an important role of DLL4 in human T lymphopoiesis (56).

In contrast, active Notch signaling, as indicated by staining for the cleaved NOTCH1 intracellular domain, was seen throughout the cortex, albeit more at the CMJ and subcapsular regions in postnatal human thymus, suggesting that perhaps ligand(s) other that DLL4 may provide Notch signaling at the inner cortex in the postnatal human thymus (56). JAGGED2 was found to be the most prominent Notch ligand in the cortex of postnatal and adult (6- to 11-y-old) human thymus (56), consistent with a previous report that more than 90% of cTEC expressed JAGGED2 in postnatal samples as determined by flow cytometry (57). Ectopic JAGGED2 expression in bone marrow–derived stromal OP9 cells has been shown capable of activating Notch and promoting human T cell development from cord blood–derived hematopoietic progenitor cells (57). Therefore, JAGGED2 might play a more significant role in human T lymphopoiesis as compared with mouse. Furthermore, JAGGED2 expression exhibited a gradient decreasing from CMJ toward the subcapsular region in postnatal but not adult human thymus, opposite of the gradient increasing from CMJ toward subcapsular region in the postnatal mouse thymus. The physiological significance of this gradient and its difference between human and mouse remains to be fully elucidated. In addition, whereas JAGGED1 was found primarily expressed on mTECs in human fetal and postnatal thymus, consistent with its distribution in mouse postnatal thymus, more cTECs in the subcapsular region expressed JAGGED1 in the human adult thymus. The biological relevance of this temporospatial regulation is unknown. Lastly, no significant difference in DLL1 distribution between human and mouse and its temporal regulation in human thymus was observed, suggesting a dispensable role for DLL1 in vivo, as demonstrated by earlier work using TEC-specific deletion of Dll1 in mice (58). However, further validation regarding the sensitivity and specificity of the primary Abs used was missing, raising a concern as to whether some observations might need to be re-examined.

Thymic stromal cells, TECs in particular, educate thymocytes so that newly generated T cells can exit the thymus and recognize foreign Ags in the context of self-MHC while being tolerant to self-antigen recognition in the periphery (59). In brief, cTECs express a unique β5t-containing thymoproteasome that provides a distinctive set of peptides for MHC class I loading that supports the positive selection of CD8SP while expressing CD83 to stabilize MHC class II (MHCII) expression and expressing TSSP and cathepsin L to generate MHCII-associated self-peptides in promoting positive selection of CD4SP, respectively. In contrast, mTECs, together with DCs and B cells, mediate self-tolerance and negative selection by expressing and presenting tissue-restricted self-antigens.

A special type of mTEC, named thymic tuft cells, that has a similar transcriptional profile to intestinal tuft cells were recently identified (60, 61). Its exact biological role in regulating thymocyte development still needs to be further investigated, as the two research groups observed different immunological phenotypes in the absence of thymic tuft cells. In addition to TECs, endothelial cells, DCs, and B cells, there are other types of thymic stromal cells that play various roles in supporting thymocyte development and/or maintaining homeostasis of the thymic stromal environment itself. Mesenchymal/fibroblast cells have been shown to support the organized thymic architecture and thymic vascular network as well as to promote the proliferation and differentiation of TECs (62–66). Recently, it has been reported that a subtype of thymic fibroblasts that is FSP1⁺ is essential for the homeostasis of mature mTECs (67). More significantly and unexpectedly, a recent study revealed that the generation of medullary gp38⁺ and CD26⁻ fibroblasts, and their expression of tissue-restricted Ags, relied on lymphotoxin-β signaling, as knocking out lymphotoxin-β receptor specifically in fibroblasts led to a loss of central tolerance and autoimmune phenotype (68). Furthermore, neutrophils that are recruited to the medulla by CXCL5 secreted by Hassall corpuscles–mTEC were shown to produce IL-23 to activate plasmacytoid DCs, which in turn produced IFN-α to support SP maturation (69). Moreover, the thymus is well innervated, and neurotransmitters/neuropeptides have been reported to modulate thymopoiesis (70, 71).

The significance of the transcription factor FoxN1 to thymopoiesis has been appreciated since the identification of its mutation as the genetic cause of the athymic nude (nu/nu) phenotype (72). It was further reinforced by the observation that forced expression of FoxN1 alone was enough to induce mouse embryonic fibroblasts to become functional TECs that support T cell development both in vitro and in vivo when engrafted under the kidney capsule (73). It is interesting to note that the thymus was converted to a bipotent lymphoid organ to produce both T cells and B cells when Foxn1 was replace by Foxn4, which closely resembles the metazoan ancestor gene for Foxn1, through transgenic expression of Foxn4 under the Foxn1 promoter in nude mice (74). The generation of a significant number of immature B cells in this transgenic strain was mainly due to lower and perhaps patchy levels of DLL4 and consequently a higher IL7:DLL4 ratio. However, whether the reduction in Dll4 mRNA resulted from a decrease in Dll4 expression in all cells and/or a decrease in percentage of TECs expressing Dll4 mRNA was not investigated; but it might be biologically relevant, as the immature B cells were found predominantly in the perivascular space, spatially separated from thymocytes.

Chromatin immunoprecipitation sequencing analyses of genome-wide FoxN1-binding sites confirmed that many thymopoietic genes discussed above, such as Dll4, Cxcl12, Ccl25, and Psmb11, are direct transcriptional targets of FoxN1 (75, 76). Consistent with a previous observation (77), Il7 was not found to be a FoxN1 target gene. However, it is surprising that Scf was not found to be a direct FoxN1 transcription target, given that it was not expressed in the TECs from nude thymic anlagen (75) but was upregulated during FoxN1-induced transformation of mouse embryonic fibroblasts to TEC (73). In addition, several other genes, whose expression were previously shown to be downregulated following a decrease in FoxN1 protein level (78), such as MHCII, Cathepsin L, CD40, and Aire, were not found to be high-confidence FoxN1 direct transcriptional targets (76). Therefore, transcriptional regulation of these genes by FoxN1 must be mediated by a yet-to-be-identified transcription factor(s) that itself could be a direct FoxN1 transcriptional target. In contrast, Psmb11/β5t, CD83, and Tssp/Prss16, which are required for efficient positive selection of CD8⁺ and CD4⁺ SPs, respectively, were identified as high-confidence FoxN1 direct transcriptional targets, although their regulation by FoxN1 was not previously recognized. Direct transcriptional activation of Psmb11/β5t by FoxN1 was confirmed by another independent study (79). Furthermore, many mitochondria- and Golgi-related genes were found to be direct FoxN1 transcriptional targets, although their physiological activities related to thymopoiesis have not been investigated and remain to be revealed.

Despite the critical role of FoxN1 in creating and maintaining the thymopoietic microenvironment, how FoxN1 expression within the initial thymic anlage becomes turned on at E11 in a tissue-specific manner is not completely understood. Defects in Pax1/Pax9, Eya1, Hox3a, and Six1 resulted in impaired thymus development and therefore may regulate FoxN1 expression (80–84). In addition, BMP4 and Wnt signaling have been reported to modulate FoxN1 expression levels in TECs (85–87). However, evidence that these transcription factors bind to the FoxN1 promoter and/or enhancer region to initiate and maintain FoxN1 expression in a tissue-specific manner is still lacking. Recently an enhancer region located within its first intron was found to be required for thymus-specific expression of FoxN1 (88). In addition, consistent with previous observations, putative PAX1, SIX1, and SMAD binding sites were identified in this enhancer region. However, the physiological relevance of these putative binding sites remains to be further investigated and may hold the key as to what factor(s) turn on FoxN1 expression and thus bring about thymopoiesis.

The thymus undergoes postpuberty age-associated involution in both mouse and human, and changes in thymic stromal cells, especially a decrease in the numbers of TECs, has been proposed to be the primary cause underlying age-associated involution or atrophy (89–92). In addition, a decrease in TEC numbers has also been reported to occur during pregnancy-induced thymic atrophy and in male mice undergoing chemotherapy (92, 93). In contrast, no reduction in TEC numbers were observed in two recent studies examining age-associated or pregnancy-induced thymic involution, respectively (94, 95). This discrepancy resulted mostly from how TEC numbers were measured. All studies that observed a reduction in TEC numbers used flow cytometry following enzymatic digestion to estimate TEC cellularity, which has recently been shown by two independent research groups to dramatically underestimate TEC numbers by at least 10-fold (96, 97). More importantly, the loss of different types of TECs during this procedure was not proportional (96). Therefore, the recovery rate for individual subtypes of TECs might be inconsistent under different biological conditions, and consequently, it is technically unreliable to compare total TEC numbers and composition using flow cytometry–based approaches. In the two recent studies that did not observe a decrease in TEC numbers during thymic involution/atrophy, one elegantly employed immunofluorescence microscopy to count TECs; the other still used flow cytometry but with a new enzymatic digestion protocol, which the authors reported to recover ∼five-times more TECs compared with the enzymatic digestion protocol used in previous studies (95).

When qualitative change in thymic stromal cells during pregnancy-induced involution and postpartum regeneration was examined using bulk RNA-sequencing, the expression levels of FoxN1 and its target genes, such as Dll4, Ccl25, and Cxcl12, were found to be lower in cTEC during involution and rebound significantly during early postpartum regeneration (95). This is consistent with previous observations showing that FoxN1 expression levels gradually decreased during aging (98–100). In contrast, forced expression of FoxN1 appeared to rejuvenate an aged thymus (101), suggesting that changes in the expression levels of FoxN1 and its target genes is one of the driving forces behind thymic involution and regeneration. In addition, a change in Il7 mRNA levels was also observed during pregnancy-induced involution and postpartum regeneration as well as age-associated thymic involution (95, 98), implying a role for a reduction in IL7 levels in driving thymic involution. However, what led to the reduction in FoxN1 and Il7 mRNA levels remains to be investigated.

A profound change in cTEC cell size and morphology was recently reported to occur during age-associated involution and transient regeneration of aged thymus after surgical castration (94). Using an elegant and powerful multicolor confetti expression approach under the control of FoxN1-Cre, which labeled almost all TECs with one of the four fluorescence proteins individually or in diverse combinations, enabled the visualization and measurement of cell morphology and size. In young (∼5-wk-old) thymus, long and looping cell projections from individual cTECs form an extensive intracellular labyrinth with 20–30 holes, with each hole filled with 2–10 thymocytes. As a result, each cTEC can support the development of 100–150 thymocytes, consistent with the estimated thymocyte/cTEC ratio of ∼100 for 10-d-old thymus (96). In contrast, cell projections were difficult to detect in aged (12-mo-old) thymus, especially at the subcapsular region, because of shortening and thinning of cell projections. Consequently, the total cell surface area of individual cTEC decreased by ∼50%, whereas cell volume reduced by ∼30% in aged compared with young thymuses. During regeneration after castration, the extended cell projections and complex cell morphology were restored in some but not all cTECs. In contrast, no change in mTEC morphology was observed during involution and regeneration. Of note, no active cell proliferation was observed for cTECs during regeneration of aged thymus, although it was readily detectable for mTECs. Therefore, it was the change in cell morphology and consequently the thymopoietic activity but not the quantity of cTECs that drives thymic involution and regeneration. Given that castration has been shown to increase FoxN1 protein levels in, but not to increase the percentage of, FoxN1-expressing TEC (99), it is tempting to speculate that only FoxN1-expressing cTECs restored their morphology after castration as a result of elevated FoxN1 protein levels through the activities of high-confident FoxN1 target genes involved in mitochondria and Golgi functions.

The thymus is a sophisticated primary lymphoid organ, with its stromal cell components providing all the necessary support to promote thymocyte development. An appreciation of one key stromal cell–dependent event mediated by Notch receptor–ligand interactions enabled us to develop the OP9-DL coculture system to produce T lineage cells in vitro and open the field for further studies of the molecular mechanisms involved in this process (102, 103). We hope that, with further investigation on, and better understanding of, signals provided by thymic stromal cells during the development of distinct subtypes of T cells, we will be able to create more advanced cell culture system(s) capable of generating all types of T cells as desired and enable new avenues for their clinical and therapeutic applications.

We are grateful to Dr. Michele K. Anderson for advice and comments during the preparation of this review.

The authors have no financial conflicts of interest.