Adult Thymic Epithelium Contains Nonsenescent Label-Retaining Cells

Thymic aging precedes that of other organs and leads to a progressive decline in thymic function with age in all jawed vertebrates (1–5). By the age of 45 y, ∼75% of epithelial cells have been replaced by fibroblasts and adipocytes in the human thymus (1, 6–8). Nonetheless, the precocious loss of thymic epithelial cells (TECs) is intriguing considering that TECs are not postmitotic cells such as mature neurons or differentiated cardiomyocytes. Indeed, based on Ki67 and BrdU labeling, Gray et al. (9) found that in 4-wk-old mice, ∼10% of TECs arose from proliferation daily. The proportion of dividing TECs decreased with age, but some Ki67⁺ TECs could still be detected in 12-mo-old mice (9). In line with this, Rode and Boehm (10) demonstrated that in preadolescent and adult mice, after acute loss, cortical TECs (cTECs) were able to proliferate vigorously, albeit in a sexually dimorphic manner. Why is the substantial regenerative potential of TECs insufficient to prevent age-related thymic involution? Our inability to answer this question illustrates our lack of understanding of TEC homeostasis and the scarcity of data on the nature and properties of TEC progenitors (4, 11–13).

Two groups directly demonstrated the existence of bipotent TEC progenitors. Rossi et al. (14) reported that a single embryonic TEC injected into a fetal thymus created both cTECs and medullary TECs (mTECs). Using in vivo cell lineage analysis, Bleul et al. (15) demonstrated the presence of bipotent TEC progenitors at birth, but also detected unipotent progenitors. Nonetheless, the proportion and functional importance of unipotent versus bipotent TEC progenitors are unknown and, alike what occurs in other tissues, might well change over time (4, 15). Thus, the mammary gland initially develops from multipotent embryonic progenitors, which create both myoepithelial cells and luminal cells, but later in life the maintenance of each lineage is ensured by unipotent progenitors able to differentiate into either myoepithelial or luminal lineages (16). Progress in our understanding of TEC renewal and homeostasis is hindered by the lack of markers or colony-forming assays for TEC progenitors.

Stem and progenitor cell populations display remarkable diversity in their proliferative behavior (17). In some tissues, stemness is associated with quiescence. Thus, hematopoietic stem cells (HSC) and hair follicle bulge divide infrequently to prevent stem-cell exhaustion (18–20). In other tissues, there are no slow-cycling or quiescent stem cells (21, 22). Thus, the esophageal epithelium is maintained by a single population of cells that divides stochastically to generate proliferating and differentiating daughters with equal probability (23). Stem-cell behavior can also change as a function of age: HSCs proliferate extensively until 3 wk of age and then switch to a quiescent adult phenotype (24, 25).

The primary goal of our study was to discover whether nonsenescent quiescent cells were present in neonatal and adult TECs. To this end, we used a transgenic label-retaining cell (LRC) assay in which a histone H2B-GFP fusion protein (H2B-GFP) is expressed under the control of the reverse tetracycline-controlled transactivator (rtTA) and the tetracycline operator minimal promoter (18, 20, 26). One advantage of this system is that H2B-GFP–retaining cells can be visualized and isolated alive, thereby enabling performance of additional analyses on LRCs. Our secondary objective was to compare the turnover of cTECs and mTECs in neonatal and adult thymus. We report that: 1) in adult but not neonatal mice, cTECs and mTECs show different turnover rates; 2) adult TECs proliferate more extensively in female than male mice, and 3) LRCs displaying putative features of progenitor cells are present in adult TECs.

B6.Cg-Gt(ROSA)26Sor^{tm1(rtTA*M2)Jae}/J and STOCK Tg(tetO-HIST1H2BJ/GFP)47Efu/J mice purchased from The Jackson Laboratory (Bar Harbor) were bred and housed under specific pathogen-free conditions in sterile ventilated racks at the Institute for Research in Immunology and Cancer. For H2B-GFP pulse-chase experiments, doxycycline was incorporated in food (2 g/kg; Harlan Laboratories) or in drinking water (2 mg/ml doxycycline supplemented with 5% sucrose) (27, 28). All procedures were in accordance with the Canadian Council on Animal Care guidelines and approved by the Comité de Déontologie et Expérimentation Animale de l’Université de Montréal.

Enrichment of thymic stromal cells was performed as previously described (9, 29). Thymic stromal cells were stained with biotinylated Ulex europaeus lectin 1 (UEA1; Vector Laboratories) and PE-Cy7 or PE-Texas Red–conjugated streptavidin (BD Biosciences) and the following Abs: Alexa Fluor 700 anti-CD45 and PE anti–I-A^b from BD Biosciences and Alexa Fluor 647 anti-Ly51 and allophycocyanin-Cy7 anti-EpCAM from Biolegend. Viability of cells was assessed using 7-aminoactinomycin D (7-AAD; BD Biosciences). TECs were selected as CD45⁻EpCAM⁺. In neonatal mice, mTECs and cTECs were defined as UEA-1⁺Ly51⁻ and UEA-1⁻Ly51⁺, respectively. In adult mice, mTECs and cTECs were defined as Ly51⁻ and Ly51⁺, respectively (see Supplemental Fig. 1 for gating strategies and isotype controls). Adult female H2B-GFP^+/− rtTA⁺ mice were treated with doxycycline, and thymic stromal cells from two to three mice were sorted and pooled for quantitative RT-PCR (qRT-PCR) analysis. Live TECs (7-AAD⁻ CD45⁻ EpCAM⁺) were sorted as cTECs (Ly51⁺ I-A^{b low}) and mTECs (Ly51⁻ I-A^{b low}), and further separated in GFP^hi (LRCs) and GFP⁻ (non-LRCs) populations (see Supplemental Fig. 2 for gating strategy). TECs were sorted on a three-laser FACSAria (BD Biosciences) or analyzed on a three-laser LSR II (BD Biosciences) using FACSDiva and FlowJo softwares (BD Biosciences) (30).

Gene expression was compared between LRCs and non-LRCs from the same TEC preparations. Total RNA was isolated using TRIzol as recommended by the manufacturer (Invitrogen) and then further purified using RNeasy Micro columns (Qiagen). Sample quality was assessed using Bioanalyzer RNA Pico chips (Agilent). Reverse transcription was performed using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems) as described by the manufacturer. Samples were preamplified for 12 cycles using TaqMan PreAmp Master Mix (Applied Biosystems). Gene expression levels were determined using prevalidated TaqMan Gene Expression Assays (Applied Biosystems) and Universal Probe Library Assays (Roche). Actb (Mm00607939), Gapdh (Mm4352932), and Hprt (Mm01545399) were used as endogenous controls. Detection of qRT-PCR was done using an ABI PRISM 7900HT Sequence Detection System (Applied Biosystems) as described previously (31).

Unless stated otherwise, results are expressed as means ± SD, statistical significance was tested using Student unpaired two-tailed t test, and differences with a p value <0.05 were considered significant. The half-life of GFP⁺ cells was calculated with GraphPad Prism software V5.01 using a one-phase decay nonlinear regression analysis. Comparison between GFP dilution kinetics has been determined using extra sum-of-squares F test.

Doxycycline leads to expression of H2B-GFP in all tissues of mice bearing the rtTA and the H2B-GFP genes. When doxycycline is withdrawn, because of the dynamic interchange of H2B among nucleosomes, H2B-GFP dilutes equally between two daughter cells during cell division (18, 20, 23, 32). Thus, the H2B-GFP content can be analyzed in a manner similar to that used for CFSE staining (32). In quiescent cells, H2B-GFP is detectable for at least 6 mo (33–35), and at least five cell divisions are required for a H2B-GFP⁺ cell to be rendered undetectable by flow cytometry (GFP⁻) (19, 32). For pulse-chase experiments in adult mice, doxycycline administration was initiated at 4–5 wk of age and was pursued for 6 wk. For “neonatal” mice, doxycycline was administered throughout the gestation period and was terminated at birth. During the chase period, beginning at T0, we analyzed two parameters. First, we studied the percentage of GFP⁺ cells: cells whose GFP fluorescence intensity was greater than that of negative controls (mice with the H2B-GFP but not the rtTA gene). Second, we assessed the frequency of cells that had undergone 0 or 1 division only during the chase period (GFP^hi cells); cells that remained GFP^hi at the end of the chase period were considered to be LRCs. When attempting to identify GFP^hi cells, we took into account that, in the absence of cell division, the GFP content of labeled cells decreases by 50% every 24 d because of H2B-GFP degradation (32); the gate defining GFP^hi cells was shifted accordingly during the chase period.

TEC subsets change in number and proportion as a function of age, and these changes are particularly dramatic during the first month of life in mice (9). Therefore, as a prelude to our study of TEC turnover in young mice, we analyzed the numbers of cTECs (CD45⁻EpCAM⁺Ly51⁺UEA1⁻) and mTECs (CD45⁻EpCAM⁺Ly51⁻UEA1⁺) at several time points during the first 12 wk of life (Fig. 1A, 1B). In accordance with previous results (9), cTECs represent the major TEC population at birth. Numbers of mTECs increased very rapidly during the first 3 wk, more slowly thereafter, and reached a zenith at around 7 wk. By contrast, the number of cTECs decreases sharply between weeks 1 and 3 (Fig. 1A, 1B). Hence, cTECs represent only 1.99 ± 0.66% of TECs by 3 wk. Notably, a similar decrease of cTECs was previously observed in 4-wk-old C57BL6 mice (9). Also, the loss of cTECs could not be ascribed to a downregulation of the cTEC marker Ly51 because similar results were observed with another cTEC marker, CD205 (Supplemental Fig. 3). Following a nadir at 3 wk, cTEC numbers increased progressively and adult cTEC/mTEC ratios were reached at about 7 wk of age (Fig. 1A, 1B).

Mice bearing the rtTA and the H2B-GFP genes were pulsed during gestation as their mothers were given doxycycline from conception to birth. Thereafter, GFP expression was chased from birth to 12 wk of age. H2B-GFP dilution occurred at same rate in cTECs and mTECs: they exhibited similar proportions of GFP⁺ and GFP^hi cells (Fig. 2A–C), and the half-life of GFP⁺ cells was similar for both TEC subsets (Table I). The transient increase in the percentage of GFP⁺ cTECs at 3 wk (Fig. 2B) was not significant and probably resulted from a loss of GFP⁻ cTECs because the absolute number of GFP⁺ cTECs in the thymus did not increase during this time frame (data not shown). Notably, all TECs were GFP⁻ after 12 wk of chase (Fig. 2A, 2B). Hence, none of the TECs containing GFP at birth became an LRC. Our data are consistent with previous observations based on BrdU and Ki67 labeling showing that TECs proliferate extensively during the neonatal period (9). In addition, our data demonstrate that, in young mice, cTECs and mTECs have similar turnover kinetics.

These observations raise the question: if neonatal mTECs and cTECs display similar proliferation rates (Fig. 2A–C, Table I), how can we explain the sharp decrease in cTEC numbers observed between weeks 1 and 3 (Fig. 1B)? During this period, the proportion of apoptotic cTECs was inferior or equal to that of apoptotic mTECs (Fig. 1C). Hence, the cTEC loss cannot be explained by preferential apoptosis of cTECs. We found a plausible explanation to this conundrum by using a metric that we called the GFP content index, which is the sum of the GFP fluorescence intensity of all GFP⁺ cells. In a closed system, without any input of new GFP⁺ cells, the total GFP content cannot increase; it can only decrease as a function of cell division and GFP catabolism. Any increase in the total GFP content implies the arrival of new GFP⁺ cells during the chase period. At T0, the GFP content of cTECs was superior to that of mTECs and, as expected for cells in a closed system (no new input), the GFP content of cTECs decreased progressively during the chase period (Fig. 2D). The salient finding was that the GFP content of mTECs showed a significant increase during week 1 (p = 0.026) and thereafter remained superior to that of cTECs (Fig. 2D). The increase in the GFP content of mTECs means that, at least during week 1, mTECs received new cell input. In other words, GFP⁺ cells that did not possess mTEC markers at birth acquired these markers after birth.

After doxycycline administration for 6 wk, adult male and female mice possessing the rtTA and H2B-GFP genes were analyzed during a chase period of 16 wk. Mice were 26–27 wk old at the end of the chase period. Consistent with previous studies based on BrdU and Ki67 labeling (9), GFP dilution curves revealed that TECs proliferated less actively in older (Figs. 3, 4, Table I) than in younger (Fig. 2, Table I) mice. In this study, three notable findings emerged. First, in both male and female adult mice, GFP dilution was much faster in mTECs than in cTECs (Figs. 3A, 3B, 4A, 4B, Table I). Moreover, MHC class II^hi mTECs (mTEC^hi) lost their fluorescence faster than MHC class II^lo mTECs (mTEC^lo; Supplemental Fig. 4). Second, both cTECs and mTECs proliferated more actively in females (Fig. 3) than in males (Fig. 4). Indeed, based on one-phase decay nonlinear regression analysis, the half-life of TECs was about twice as short in females than in males (Table I). Third, in contrast with young mice, adult mice contained LRCs. In females, ∼10% of cTECs, but virtually no mTECs, remained GFP^hi at the term of a 16-wk chase (Fig. 3C). The proportion of GFP^hi was even greater in male mice (Fig. 4C).

The LRC phenotype is typically found in two cell types: 1) postmitotic (terminally differentiated or senescent) cells that can no longer undergo cell division, and 2) quiescent somatic progenitor/stem cells with clonogenic potential (18, 19, 33, 36). To gain insights into the nature of thymic epithelial LRCs, we analyzed the cell size and gene expression profile of TECs from adult mice chased for 16 wk. We focused on female LRCs because they represent a small, discrete cell population in a rapidly proliferating compartment (Fig. 3C). By contrast, we surmised that LRCs in males might be functionally more heterogeneous (Fig. 4C). Indeed, because the proliferation rate of male TECs is relatively modest, we hypothesized that their large LRC population might contain a mixture of progenitor cells and postmitotic cells. One hallmark of senescent cells is that they increase in size (37). In accordance with this concept, we found that the size of mTEC^hi cells was larger than that of other TEC subsets (Fig. 5A). This makes sense because Aire⁺ mTECs are uniformly mTEC^hi and postmitotic (38). Notably, we observed that GFP^hi cTECs (LRCs) were significantly smaller than GFP⁻ cTECs. We then sorted GFP^hi and GFP⁻ cTECs and GFP⁻ mTEC^lo cells (there were practically no GFP^hi mTECs in females at the end of a 16-wk chase; Figs. 3C, 6) to assess the expression of three senescence-associated transcripts: p16INK4, p19ARF, and Serpine1 (37, 39–41). We found that expression of p16INK4a and Serpine1 transcripts was lower in LRCs than in GFP⁻ cTECs and GFP⁻ mTEC^lo cells (Fig. 5B). Levels of p19ARF were similar in LRCs and GFP⁻ cTECs but were upregulated in the mTEC^lo population. To gain insight into a possible progenitor cell functionality of adult cTEC LRCs, we assessed the expression of four genes that promote TEC proliferation and maintenance, and whose absence leads to thymic atrophy: Bmi1, Foxn1, Trp63, and Wnt4 (42–47). Expression of Foxn1, Trp63, and Wnt4 adopted the following hierarchy: LRCs ≥ GFP⁻ cTECs ≥ mTEC^lo population (Fig. 5C). For Bmi1, the hierarchy was slightly different: LRCs ≥ mTEC^lo population ≥ GFP⁻ cTECs. In addition, we evaluated the expression of transcripts overexpressed in cTECs (Psmb11, Prss16) or mTECs (Cd40, Cd80, and Tnfrsf11a) (29, 48). cTEC LRCs did not show evidence of bipotentiality: their gene expression profile was typical of cTECs with no upregulation of mTEC markers (Fig. 5D). In fact, LRCs expressed even higher levels of cTEC markers (Psmb11 and Prss16) than GFP⁻ cTECs (Fig. 5D). Finally, we observed no difference in the expression of MHC class II on GFP^hi and GFP⁻ cTECs (data not shown). Collectively, these data strongly suggest that, at least in females, cortical LRCs are not senescent cells. Their transcriptomic profile rather suggests that they express genes promoting TEC replication potential and that they might therefore be TEC progenitor/stem cells.

Gray et al. (9) previously demonstrated that the turnover of TECs diminished with age. We confirm this observation using a different mouse model and report supplemental differences between neonates and adult mice: 1) whereas cTECs and mTECs had similar proliferation rates in young mice, mTECs proliferated more actively than cTECs in adults; and 2) LRCs were detectable after a 16-wk chase in adults (Figs. 3, 4), but no LRCs were found when the chase is initiated at birth (Fig. 2). Why did we detect LRCs in adults but not young mice? H2B-GFP expression at T0 was lower in neonates than in adults: the percentage of TECs that were GFP^hi at the beginning of the chase period was ∼70% in adults but 35% in neonates (Figs. 2–4). However, this discrepancy in GFP fluorescence intensity could not be held solely responsible for the absence of LRCs in neonates. Indeed, we calculated that the clearance rate of GFP^hi and GFP⁺ TECs was significantly more rapid in young than in adult mice (Fig. 6, Table I). Accordingly, a plausible explanation for the absence of LRCs in neonates would be that the mitotic behavior of TEC progenitor/stem cells is similar to that of HSCs. Thus, mouse HSCs are actively proliferating until 3 wk after birth, at which time they abruptly switch to become a predominantly quiescent population (24, 25). Hence, by contrast with adult HSCs, neonatal HSCs are not LRCs.

One notable point with neonatal TECs is that the cTEC/mTEC ratio changed dramatically during the first weeks of life: the number of mTECs steadily increased from birth to 7 wk, whereas the number of cTECs showed a dramatic decrease between weeks 1 and 3 (Fig. 1A, 1B). The discrepancies in the population dynamics of cTECs and mTECs cannot be explained by differences in apoptosis (Fig. 1C) or proliferation rates (Fig. 2A–C, Table I). One salient observation is that the total GFP content of mTECs increased from birth to week 1. This means that during this period, mTECs were not in a closed compartment. They received some input of GFP⁺ cells that did not express mTEC markers at the beginning of the chase period. In theory, GFP⁺ cells that join the mTEC compartment during this period could be thymic or extrathymic cells. However, recent reports have revealed that the majority of mTECs are derived from β5t-expressing precursor cells (49) and that embryonic TECs expressing the cTEC marker CD205 give rise to both cTECs and mTECs in reaggregate thymus organ cultures (48). Hence our favorite hypothesis to explain the increase in mTEC total GFP content at week 1 (Fig. 2D) and the decrease of cTECs at 3 wk (Fig. 1B) is that cells with a cTEC phenotype (Ly51⁺UEA1⁻) at T0 acquire an mTEC phenotype (Ly51⁻UEA1⁺) during the neonatal period.

Two major points can be made from our observations in adult mice. First, both cTECs and mTECs proliferate more actively in females than in males. This observation is consistent with the fact that after cTEC depletion in adult Ccx-ckr1:hDTR transgenic mice, restoration of thymopoietic activity (and presumably cTEC regeneration) was observed in females but not males (10). Moreover, thymic involution in mice and humans is more precocious in males than in females (50–52). Further studies are warranted to decipher the mechanisms responsible for this sexual dimorphism. Androgens may have a direct effect on the thymus by affecting TEC metabolism or an indirect effect mediated, for instance, by changes in the microbiota that can enhance production of cytokines (such as IFN-γ) (10, 53). Second, LRCs are present in adult TECs. In females, LRCs are found exclusively in the cTEC compartment. Based on their relatively small cell size, low expression of p16INK4a, p19ARF, and Serpine1, and high expression of Bmi1, Foxn1, Trp63, and Wnt4, we conclude that female cTEC LRCs are not senescent postmitotic cells and may well be TEC progenitors. They also express higher levels of β5T than other TECs, which further support the hypothesis that LRCs would be progenitor TECs (49). To address this issue, our plan is to perform reaggregation and transplantation experiments along with transcriptome sequencing of female and male TEC LRCs. Assuming that TEC LRCs may represent adult TEC progenitor/stem cells, their transcriptomic analysis could provide us with novel insights into their functional potential and allow us to discover surface markers specific for this cell subset.

We are grateful to Juan Ruiz Vanegas, Danièle Gagné, and Gaël Dulude (flow cytometry and cell sorting), Nadine Fradet and Jennifer Huber (qRT-PCR experiments), and the staff of Institute for Research in Immunology and Cancer animal care facility for assistance. Special thanks to Sébastien Lemieux for advice on statistical analyses.

The authors have no financial conflicts of interest.