Proliferative arrest and rapid turnover of thymic epithelial cells expressing Aire

The Aire gene is transcribed predominantly in the MEC^hi subset (3, 7), but it was not known how homogeneously these cells express the Aire protein. Therefore, we performed flow cytometric analysis of adult thymic stroma using an mAb specific for Aire (Fig. 2 A) (9). Intracellular staining of enriched TEC suspensions from aire^−/− mice gave no signal but revealed a subset of Aire⁺ cells composing 50–60% of the MEC^hi population in aire^+/+ mice (Fig. 2 B). A very low proportion of MEC^lo were Aire⁺, whereas no staining above background was detected in cortical epithelial cell (CECs; Fig. 2 B). These values translated to an average of 8.4 × 10⁴ (±2.5 × 10⁴) Aire⁺ MECs in the thymus of a 6-wk-old mouse.

Aire⁺ MEC^hi were not distinguished from Aire⁻ MEC^hi by expression of the MEC cell-surface markers, Ulex europeaus agglutinin 1 (UEA-1; Fig. 2 C) and CD80 (not depicted), nor by expression of a range of other markers previously associated with precursor cell types, including CD24, CD44, K5, K8, or Sca-1 (not depicted). However, both the Aire⁺ and Aire⁻ MEC^hi subsets expressed CD86 and programmed death 1 ligand (PD-L1) at variable levels (Fig. 2 C), indicating further heterogeneity among MEC^hi and the potential to provide distinct co-stimulation signals to thymocytes. Nonetheless, we found the phenotypes of Aire⁺ and Aire⁻ MEC^hi to be very similar overall.

The proliferative status of Aire⁺ MECs is important for determining whether the terminal differentiation or progressive restriction model better describes MEC differentiation. A terminally differentiated population would be postmitotic, whereas, conversely, a precursor MEC population would rely on division for maintenance and phenotypic diversification. Previous studies showed that the adult MEC^hi subset was more proliferative than CECs and MEC^lo, according to assessments of DNA content (6), Ki67 expression (7), and BrdU labeling (7); however, the Aire⁺ MEC^hi component was not specifically analyzed. On the other hand, relying on immunohistology, Hamazaki et al. found that adult Aire⁺ MECs did not incorporate BrdU 4 h after injection (5), but this technique precluded quantitative analysis, and it was not clear whether this time point labeled sufficient MECs for detection. In this study, we optimized flow cytometric analysis of TECs for concurrent detection of BrdU and Aire to assess proliferation of Aire⁺ and Aire⁻ TEC subsets, and we also followed the flow of BrdU label with time to investigate their population dynamics.

First, to assess the relative levels of cell division among TEC subsets, we injected mice with a single dose of BrdU, waited 12h, stained enriched suspensions of thymic stromal cells with antibodies to TEC markers and BrdU, and analyzed them by flow cytomety. A substantial proportion of TECs had incorporated BrdU 12 h after injection (Fig. 3 A). In accordance with previous studies (6, 7), there was considerably lower incorporation by the MEC^lo compared with the MEC^hi population (average of 7.1 ± 2.9% vs. 15 ± 2.7%, respectively). Within the MEC^hi subset, the Aire⁻ population had a surprisingly high level of BrdU incorporation, with 27% of the cells in this subset dividing during the pulse (Fig. 3 A). In contrast, minimal BrdU incorporation was detected in the Aire⁺ MEC^hi subset, indicating that few, if any, Aire⁺ cells were cycling (Fig. 3 A). Second, to follow the short-term fate of cells that divided during the BrdU pulse, we measured BrdU retention by TECs at various time points after injection. Within the whole TEC population, labeled cells diminished by one third within the first 3 d, then stabilized for the next 2 d (Fig. 3 B, far left). The pattern of decay was similar in the MEC^lo and Aire⁻ MEC^hi subsets at magnitudes reflecting their initial BrdU incorporation (Fig. 3 B, left and right). In contrast, the proportion of BrdU⁺ cells in the Aire⁺ MEC^hi population increased over the first 3 d after the pulse, then diminished (Fig. 3 B, far right). Given that Aire⁺ MEC^hi were not cycling, the entry of BrdU into this compartment must be derived from cycling Aire⁻ precursors.

To assess the turnover of these populations, we analyzed the kinetics of BrdU incorporation over a 2-wk period of constant access to label. The proportion of cycling cells at the outset of incorporation was assessed 2 h after BrdU injection. Again, the Aire⁻ MEC^hi population incorporated the most BrdU and Aire⁺ MEC^hi only background levels (Fig. 3 C, right and far right), confirming the data in Fig. 3 A, which used a longer pulse. Interestingly, almost all Aire⁻ MEC^hi were BrdU⁺ by 5.5 d (Fig. 3 C, right), indicating that this subpopulation is composed entirely of rapidly dividing cells with a prodigious replacement rate of 20% per day. In contrast, the TEC, MEC^lo, and Aire⁺ MEC^hi subsets showed bimodal incorporation curves, with saturation of a rapidly dividing component (∼60% of each) within 6 d, followed by a reduced rate of incorporation by a component with a lower level of proliferation (Fig. 3 C, far left, left, and right). This result indicates that mixtures of cells with both high and low turnover rates maintain these subsets.

A lag period of 36 h before BrdU⁺ cells accumulated in the Aire⁺ MEC^hi population (Fig. 3 C, far right) reflected the precursor/product relationship observed in the decay kinetics and represents the minimum amount of time such a precursor takes to up-regulate Aire after cell division. Taking this delay into account, we estimate that an average of 13% (±2.1%) of this subset is replaced each day.

The slow turnover of MEC^lo and their expression of the putative TEC precursor marker, MTS-24 (10), are reminiscent of immature epithelial cells in other tissues (11). To determine whether adult MEC^lo were indeed precursors of Aire⁺ MECs, MEC^lo were purified and reaggregated with MHC-mismatched embryonic stroma and adult CD4/8 single-positive (SP) thymocytes. After 5 d of culture, a substantial proportion and number of MEC^lo had up-regulated MHC II and Aire (Fig. 3 D). Aire⁺ MECs did not incorporate BrdU in vitro (Fig. 3 D), indicating that there was no expansion of trace Aire⁺ MEC^lo and, therefore, that Aire⁻ MEC^lo were differentiating into the Aire⁺ cells observed. A proportion of these Aire⁺ cells were MHC II^lo, perhaps reflecting slower kinetics of differentiation in vitro compared with in vivo and/or differential provision of the signals driving Aire and MHC II molecule up-regulation. Collectively, the observations that Aire⁺ MEC^hi are postmitotic and are derived from cycling, Aire⁻ MEC^lo precursors establish that Aire expression is a late event in MEC differentiation.

To investigate the fate of Aire⁺ MEC^hi, we analyzed their longevity in BrdU decay experiments. Mice were exposed to BrdU for 2 wk to achieve maximal incorporation (80–100%); label was then withdrawn and retention was measured at various times thereafter. During the chase period, BrdU⁺ TECs diminished in both label intensity and proportion labeled, suggesting further division by cycling cells and loss of cells by apoptosis or some other means (Fig. 4).

Continued cycling, manifest as reduced BrdU intensity, was evident for the Aire⁻ MEC^lo and MEC^hi subsets (unpublished data). The presence of a relatively large proportion of MEC^lo with high levels of BrdU 4 wk after label withdrawal (unpublished data) indicates a low level of turnover and is in accord with low levels of division.

The label in the noncycling Aire⁺ MEC^hi subset showed a rapid decline, with 90% of labeled cells gone 2 wk after withdrawal of BrdU (Fig. 4). The rate of loss during the first 2 wk (6.5% per day) is likely to be an underestimate (compared with the 13% per day observed in the incorporation study) because of the continued entry of labeled precursors during this time. Nevertheless, these data indicate very rapid turnover of this population relative to the Aire⁻ MEC^hi and MEC^lo subpopulations.

The finding that Aire⁺ MEC^hi are postmitotic and have a high turnover rate raises the question of whether Aire itself causes proliferative arrest and/or apoptosis of MEC. This notion was supported by the finding that aire^−/− mice have an increased proportion and number of MEC^hi (Fig. 5 A, dot plots) (2) at the expense of the MEC^lo subset (Fig. 5 A, right). To determine whether this increase was caused by greater proliferation of MEC^hi in the absence of Aire, we injected BrdU into aire^+/+ and aire^−/− mice and analyzed incorporation by MEC^hi 12 h later. Surprisingly, the proportion of BrdU⁺ MEC^hi in aire^−/− mice was significantly lower than in aire^+/+ controls, suggesting that the higher numbers of MEC^hi in aire^−/− mice are not derived from increased proliferation but stem from expansion of postmitotic MEC^hi (i.e., analogous to Aire⁺ MEC^hi; Fig. 5 B). Reduced Aire levels in aire^+/− mice (9) did not impinge on the MEC^hi proportion, numbers, or BrdU incorporation (unpublished data). Therefore, Aire does not directly inhibit MEC^hi proliferation; rather, it modulates the overall number of the postmitotic population in some other way, perhaps by inducing apoptosis.

Studies of apoptosis in TECs ex vivo were not feasible because of the induction of high levels of Annexin V labeling and caspase activation by the thymus digestion procedure (unpublished data). Staining of thymus sections with the activated caspase marker VAD-FMK demonstrated in situ apoptosis of thymocytes but did not label Aire⁺ or Aire⁻ K5⁺ TECs. This presumably reflects the fact that thymic macrophages and dendritic cells promptly phagocytose dying cells, as they do the vast majority of thymocytes that perish daily but escape detection by apoptosis assays (12). Therefore, we used the MEC cell line 1C6 to investigate a potential link between Aire expression and apoptosis.

To assess whether Aire overexpression could directly inhibit MEC division, we transfected 1C6 cells with vectors encoding either GFP or an Aire-GFP fusion protein and analyzed proliferation by BrdU incorporation. Comparable transfection efficiencies were observed in the two cultures (∼30% GFP⁺). Correct folding, localization, and activity of the fusion protein was confirmed by fluorescence microscopy, Western blotting, and gene array analysis of transcription by cells transfected with Aire-GFP, with up-regulation of many PTAs (unpublished data). The proportion of GFP⁺ cells incorporating BrdU was not substantially different between 2 and 3 d after transfection, indicating that Aire-GFP did not affect proliferation (Fig. 5 C). However, both the proportion and number of Aire-GFP⁺ MECs were decreased compared with GFP⁺ MECs from 3 d after transfection, declining until no Aire-GFP⁺ cells were detected after 5 d, whereas about one half of the GFP⁺ MECs remained (Fig. 5 D). Importantly, the specific loss of Aire-GFP–expressing MECs was caused by increased apoptosis at all time points (Fig. 5 E). Similar results were obtained with an Aire-GFP construct containing an internal ribosomal entry site before the GFP sequence, indicating that apoptosis was caused specifically by Aire (unpublished data). These data establish that overexpression of Aire does not directly inhibit MEC division but rather induces apoptosis. This finding offers a likely explanation for the paucity of MEC lines that express Aire.

This study revealed several important aspects of TEC differentiation and population dynamics. A surprisingly high rate of turnover for the entire TEC population of 10% per day (±0.7%) was in accordance with previous estimates (7) and was driven by proliferation, predominantly within the Aire⁻ MEC^hi subset. The rapid cycling of all cells in this subset strongly suggests that it constitutes an MEC transit-amplifying population that may regulate parameters such as overall numbers of MECs and the tempo of their differentiation. The finding that certain TECs are highly proliferative also raises questions regarding the long-held assumption that TECs are relatively radioresistant because of low mitotic activity: does radiation in fact cause loss of TEC subsets and subsequent defects in thymocyte differentiation and tolerance?

Direct analysis of BrdU incorporation by Aire⁺ MECs at the single-cell level revealed that, in the adult, these cells are postmitotic and derive from a cycling precursor. While this paper was under review, Rossi et al. (13) showed maturation of E15 CD80⁻ to CD80⁺ MECs but also that CD80 was not a surrogate marker for Aire expression. Transcription of Aire in the ex vivo CD80⁻ MEC subset left open the question of the precise identity of Aire⁺ MEC precursors in the fetal thymus. High levels of CD80 and MHC II were better correlates for Aire expression in adult MECs, and reaggregate thymus organ culture of purified MEC^lo gave rise to Aire⁺ MEC^hi, without expansion of low-level (∼1%) Aire⁺ MEC^lo. The low efficiency of MEC^lo transition to Aire⁺ MEC suggests that only a minor subpopulation of MEC^lo exhibits this potential. The influence of apoptosis on the emerging Aire⁺ population does need to be taken into account, however, as does the possibility that not all MEC^lo are exposed to the necessary inductive signals in this system. Whether the differentiation of Aire⁻ MEC^lo to Aire⁺ MEC^hi occurs via an Aire⁻ MEC^hi intermediate awaits the means to purify this latter population. Furthermore, the dependency of this adult differentiation program upon signals important for fetal Aire⁺ MECs (i.e., the requirement of RANK signals) (13) remains to be determined.

The Aire⁺ MEC^hi population has a high turnover, with loss of cells most likely through apoptosis, based on the in vivo kinetics of decay and in vitro overexpression studies. The possibility that some in vivo loss of BrdU⁺ Aire⁺ MEC^hi occurs by cells down-regulating Aire and joining another subset cannot be ruled out. Nevertheless, the findings of this study are consistent with the terminal differentiation model and not the progressive restriction model of TEC differentiation.

The induction of apoptosis by Aire further supports the finality of its activity in MEC differentiation. Interestingly, the highest levels of apoptosis were seen 4 d after transfection, suggesting that Aire may initiate a sequence of events that culminates in MEC death. This scenario raises the question of whether apoptosis occurs via direct induction of an apoptotic program or indirectly through up-regulation of PTA that overload the protein synthesis machinery to such an extent as to provoke death through ER stress. Regardless, the delay in apoptosis should allow sufficient time for the promiscuous expression of genes by MECs, followed by cell death to deliver a battery of PTAs to dendritic cells for cross-presentation and induction of central tolerance to peripheral self (14). Thus, to affect central tolerance, Aire may not only induce transcription of PTA but also promote cellular changes that ensure their efficient presentation.

Aire-deficient mice were derived and genotyped as previously described (2) and were analyzed on the C57BL/6 (B6) × 129 genetic background. Mice were housed at the Center for Animal Resources and Comparative Medicine at the Harvard Medical School, and the Institutional Animal Care and Use Committee approved procedures. BrdU labeling experiments were initially performed with 6–8-wk-old NOD × B6 mice, then with B6 mice for final replicates. B6.H-2^g7 mice (15) were time mated, and the day of vaginal plug detection was designated E0.5.

BrdU was administered as previously described (16). In brief, mice were injected intraperitoneally with 1 mg BrdU (Sigma-Aldrich) in PBS to provide a labeling pulse. Continuous labeling was maintained by incorporation of 0.8 mg/ml BrdU in sterile drinking water with artificial sweetener. 8-wk-old mice were used for this experiment because thymic stromal cell numbers remain essentially at steady state during the labeling period (7).

The following antibodies, purchased from BD Biosciences unless otherwise stated, were used: rabbit anti–mouse K5 (Covance), rat anti-Aire (rat IgG2c, clone 5H12; a gift from H. Scott, Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia), anti–I-A/E–Pacific blue and –allophycocyanin (APC; clone M5/114.15.2; BioLegend), biotinylated anti-Ly51 (clone 6C3), anti-CD45–PerCP-Cy5.5 (clone 30-F11), anti-BrdU–Alexa Fluor 488 (clone 3D4; Caltag), UEA-1–FITC (Vector Laboratories), anti-CD86–PE (clone GL1), anti–PD-L1–PE (clone M1H5), anti–rat IgG2c-PE and -FITC (clone 2C-8F1; Southern Biotechnology Associates, Inc.), Cy3-conjugated donkey anti–rabbit IgG (Jackson ImmunoResearch Laboratories), Annexin-V–PE, and streptavidin-APC.Cy7 and -APC. Suspensions of thymic stromal cells were prepared as previously described (17). All digests were counted and analyzed for TEC enumeration, and totals were compared using the Student's t test. Intracellular Aire staining was performed using fixation/permeabilization buffers (eBioscience) according to the manufacturer's instructions. For BrdU experiments, the second of two enzymatic digestions of individual thymi was used for analysis. Stromal cells were fixed/permeabilized using BD Bioscience buffers, followed by paraformaldehyde/Tween 20, and were treated with DNase (50 Kunitz units) before staining with anti-BrdU. Data were acquired on a flow cytometer (LSR II; BD Biosciences) and analyzed using FloJo software (TreeStar, Inc.).

MEC^lo and SP thymocytes from B6 mice were sorted to >97% purity on a cytometer (MoFlo; DakoCytomation), as previously described (7). MEC^lo and SP thymocytes were reaggregated with suspensions of fresh E15.5 embryonic stroma from B6.H-2^g7 mice, as previously described (13), at ratios of 1:1:8. Reaggregates were dispersed in dispase/DNase for flow cytometric analysis. For analysis of proliferation, 10 μM BrdU was added 5 h before termination of the culture.

8-μm air-dried cryosections of thymus were stained with primary antibodies for 30 min and then subjected to three 5-min washes in PBS. Secondary antibodies were incubated for 30 min, followed by washes. Sections were then incubated in DAPI (Invitrogen) for 5 min, washed, and mounted with coverslips in aqueous mounting medium (Biomeda). Images were acquired with a confocal microscope (Axiovert 200M; Carl Zeiss MicroImaging, Inc.) using a xenon-arc lamp in a wavelength switcher (Lambda DG-4; Sutter Instrument Co.) and processed with Slidebook imaging software (Intelligent Imaging Inc.).

The 1C6 cell line, originally derived from mouse MECs (18), was a gift of M. Kasai (National Institute of Infectious Diseases, Tokyo, Japan). Cells were seeded onto 12-well plates (at 10–20% confluence) and transfected with Effectene (QIAGEN) and 300 ng of either aire cloned into pEGFP-c1 expression vectors (Aire-GFP) or 300 ng of pEGFP-c1 (GFP) 16 h later. Transfected 1C6 cells were harvested at a various time points, and the relative intensity of GFP was analyzed by flow cytometry. Apoptosis was measured with Annexin V and DAPI staining, and GFP⁺ MECs were analyzed by flow cytometry.

We thank Dr. Hamish Scott, Dr. Michiyuki Kasai, and Mr. Andrew Koh for reagents.

This work was supported by grant RO1 DK60027 from the National Institutes of Health and Young Chair funds to D. Mathis and C. Benoist, and by Joslin's National Institutes of Diabetes and Digestive and Kidney Diseases–funded Diabetes and Endocrinology Research Center core facilities. D. Gray received support from an Australian National Health and Medical Research C.J. Martin Overseas Biomedical Fellowship, and J. Abramson received support from the Juvenile Diabetes Research Foundation.

The authors have no conflicting financial interests.