The basic two-step terminal differentiation model of the medullary thymic epithelial cell (mTEC) lineage from immature MHC class II (MHCII)lo to mature MHCIIhi mTECs has recently been extended to include a third stage, namely the post-Aire MHCIIlo subset as identified by lineage-tracing models. However, a suitable surface marker distinguishing the phenotypically overlapping pre- from the post-Aire MHCIIlo stage has been lacking. In this study, we introduce the lectin Tetragonolobus purpureas agglutinin (TPA) as a novel cell surface marker that allows for such delineation. Based on our data, we derived the following sequence of mTEC differentiation: TPAloMHCIIlo → TPAloMHCIIhi → TPAhiMHCIIhi → TPAhiMHCIIlo. Surprisingly, in the steady-state postnatal thymus TPAloMHCIIlo pre-Aire rather than terminally differentiated post-Aire TPAhiMHCIIlo mTECs were marked for apoptosis at an exceptionally high rate of ∼70%. Hence, only the minor cycling fraction of the MHCIIlo subset (<20%) potentially qualified as mTEC precursors. FoxN1 expression inversely correlated with the fraction of slow cycling and apoptotic cells within the four TPA subsets. TPA also further subdivided human mTECs, although with different subset distribution. Our revised road map emphazises close parallels of terminal mTEC development with that of skin, undergoing an alternative route of cell death, namely cornification rather than apoptosis. The high rate of apoptosis in pre-Aire MHCIIlo mTECs points to a “quality control” step during early mTEC differentiation.
This article is featured in In This Issue, p.3385
The thymus establishes and maintains the fundamental T cell arm of the adaptive immune system of vertebrates by generating a self-MHC–restricted and self-tolerant T cell repertoire. Thymic epithelial cells (TECs), encompassing a cortical (cTEC) and medullary (mTEC) lineage, provide the structural and functional framework for key steps of T cell development, that is, lineage induction, TCR repertoire generation and selection, T cell differentiation, and proliferation (1). There has recently been considerable progress in delineating the road map of TEC lineage specification and subsequent terminal differentiation. However, new TEC subsets continue to be described and their interrelationship and respective physiological roles in the embryonic versus adult thymus remain unclear (2–4).
In the fetal thymus both TEC lineages originate from a common bipotent stem cell/progenitor pool (5). The phenotype of these TEC stem cells and the immediate/early steps of TEC lineage bifurcation still remain poorly defined. It had been commonly assumed that self-renewing stem cells symmetrically bifurcate into the cTEC and mTEC lineages. However, recent data favor an asymmetric model whereby bipotent progenitors transit through a “cTEC-like” stage and express markers typically associated with terminally differentiated cTECs, namely β5t, CD205, and Ly51. In this model, cTEC commitment represents the default pathway, and entry into the mTEC lineage would entail downregulation of cTEC markers, concomitant with upregulation of markers specifying the early mTEC lineage, such as claudin 3/4, SSEA-1, Rank, and Pdpn (2). In addition to the functional identification of fetal bipotent TEC stem cells, there have been several recent reports on the identification of bipotent TEC stem cell candidates in the adult thymus (2).
Following commitment, further progression within both lineages is characterized by sequential upregulation of sets of markers that specify each lineage. Specifically, cTECs are known to upregulate CCRL1, IL-7, and MHC class II (MHCII) upon terminal differentiation (2, 6). Progressive differentiation of mTECs includes upregulation of CD80, MHCII, and autoimmune regulator (Aire) as well as tissue-restricted self-antigens (TRAs) (7–9). Aire+MHCIIhi mTECs are fully competent APCs that are instrumental in central tolerance imposition by induction of negative selection and generation of thymically derived regulatory T cells (10). At the population level, TRA expression in mature mTECs displays a stochastically generated mosaic pattern, whereas in single cells TRA expression is regulated in a coordinated fashion (11–13).
The expression of Aire was thought to mark the end stage of mTEC differentiation, with Aire inducing apoptosis in mature mTECs (7, 14). However, there has been accumulating evidence of a post-Aire stage. Post-Aire mTECs were shown to downregulate certain TRAs, along with their maturation markers, that is, Aire, CD80, and MHCII expression. At the same time, they upregulated keratin 10 (K10) and involucrin (Ivl) expression (15–19), thus portraying features similar to terminally differentiated keratinocytes in the upper layers of cornified skin (20, 21). Despite distinct transcriptional profiles, pre- and post-Aire MHCIIlo mTECs were so far phenotypically indistinguishable by available surface markers. Hence, analysis of the major MHCIIloCD80lo mTEC subset in wild-type mice reflected an averaging of two widely separated differentiation stages along mTEC development, namely early pre- and late post-Aire mTECs. This limitation confounded the definitive establishment and characterization of sequential developmental stages along the road map of mTEC differentiation.
In this study, we introduce the lectin Tetragonolobus purpureas agglutinin (TPA) as a cell surface marker that allows for such delineation. This novel tool allowed us to expand the initial two-step mTEC terminal differentiation model (7–9, 22) to four major steps. Salient features of this revised road map are the exceptionally high rate of apoptosis among MHCIIlo pre-Aire mTECs and the terminal differentiation/clearance of post-Aire mTECs via a mechansism akin to cornification of keratinocytes in the upper skin layer. Notwithstanding certain quantitative differences, this novel TPA-based mTEC staging also applies to the human thymus.
Materials and Methods
C57BL/6N adult, pregnant female mice and SD rats were purchased from Charles River WIGA (Sulzfeld, Germany) or bred at the German Cancer Research Center animal facility. The FoxN1–enhanced GFP mice (23) were provided by Thomas Boehm (Max Planck Institute for Immunobiology and Epigenetics, Freiburg, Germany). Aire−/− and R26R-EYFP mice on the C57BL/6 background were genotyped as previously described (24, 25). All animals were housed under specific pathogen-free conditions at the German Cancer Research Center in accordance with the European Convention for the Protection of Vertebrate Animals Used for Experimental and Other Scientific Purposes and the German Legislation (35-9185.81/G-273/14). Aire-CreERT2 × Ai6 (Rosa26-Stopfl-zsGreen) (18) mice were bred in a specific-pathogen free facility at the University of California, San Francisco, and the University of California, San Francisco Institutional Animal Care and Use Committee approved all experiments.
Human thymus tissue
Human thymus samples were obtained in the course of corrective cardiac surgery at the Department of Cardiac Surgery, Medical School of the University of Heidelberg. This study has been approved by the Institutional Review Board of the University of Heidelberg.
The following Abs were used for flow cytometry of murine mTECs: anti–CD45-PerCP or allophycocyanin-Cy7 (clone 30-F11; BD Biosciences, Heidelberg, Germany; eBioscience, San Diego, CA), anti–CDR1-Pacific Blue or PE-Cy7 (26), anti–Ly51-FITC (clone 6C3; BD Biosciences), anti–EpCAM-Alexa Fluor 647 or allophycocyanin-Cy7 (clone G8.8; BioLegend, San Diego, CA) (27) and anti–MHCII-PE or Pacific Blue or Alexa Fluor 700 (clone 2G9, BD Biosciences; clone M5/114.15.2, BioLegend), anti–CD80-PE (clone 16-10A1; BD Biosciences), CD11c-PE-Cy7 (clone N418; BioLegend), annexin V–FITC or Alexa Fluor 680 (Invitrogen, Darmstadt, Germany), anti-active caspase-3–FITC (clone C92-605; BD Biosciences); anti-K10 (clone PRB-159P; Covance, Princeton, NJ), and Ki67-PE-Cy7 (clone SolA15; eBioscience). The following Abs were used for flow cytometry of human mTECs: biotinylated anti-EpCAM (clone HEA125, German Cancer Research Center; Sav-PE, BD Biosciences), anti-CDR2–Alexa Fluor 488– (26), Alexa Fluor 647–, or Alexa Fluor 680–conjugated mAb HLA-DR (clone L243; German Cancer Research Center) and anti–CD45-PerCP (clone 21D; BD Biosciences), and annexin V–Alexa Fluor 680 (Invitrogen). The following Abs were used for immunohistochemistry: anti–keratin 14 (clone AF64; Covance), anti–keratin 5 (clone GP-CK5; Progen, Heidelberg, Germany), anti–keratin 8 (clone Troma-1; Developmental Studies Hybridoma Bank, Iowa City, IA), anti-K10 (clone EP1607Y; Epitomics, Burlingame, CA), anti-involucrin (clone SY5; Acris, Herford, Germany), and biotinylated Ulex europaeus agglutinin (Vector Laboratories, Burlingame, CA). The following Abs were used for both flow cytometry and immunohistochemistry: biotinylated TPA (Vector Laboratories) and anti-involucrin (clone PRB-140C; Covance), and anti-Aire (mouse, clone B1/02-5H12-2, provided by H. Scott, Adelaide, SA, Australia; human, clone 6.1, provided by P. Peterson, Tartu, Estonia). The indirect FACS and immunohistochemistry stainings were performed using appropriate secondary Abs from Jackson ImmunoResearch (West Grove, PA)/Dianova (Hamburg, Germany) and Molecular Probes (Invitrogen). For Western blotting, anti–α-tubulin (clone B-5-1-2; Sigma-Aldrich, St. Louis, MO), anti–full-length poly(ADP-ribose) polymerase (PARP; clone C2-10; BD Biosciences), and anti-cleaved PARP (clone E51; Abcam, Cambridge, U.K.) Abs were used along with their corresponding anti-mouse or anti-rabbit HRP detection Abs from Jackson ImmunoResearch and Santa Cruz Biotechnology (Dallas, TX), respectively.
Isolation of mTECs
Thymi were removed from embryonic, postnatal, and/or adult mice and placed into RPMI 1640 medium (containing 5% FCS) on ice. The remaining digestion steps for adult and postnatal thymi were performed as described previously (12, 28). The thymi from embryos were digested using collagenase/dispase for 20–30 min at 37°C with intermittent agitation using a 1000- or 200-μl pipette. Human TECs were purified as described previously (29) with some modifications (11). The mTECs were enriched using either mouse or human anti-CD45 MicroBeads (Miltenyi Biotec, Bergisch Gladbach, Germany). mTECs from SD rat thymi were isolated, stained, and FACS purified as previously described (30). The cells were then sorted on a BD FACSAria III cell sorter (BD Biosciences) using either propidium iodide (0.2 μg/ml) or fixable viability dye eFluor 780 or eFluor 450 (eBioscience) to exclude dead cells. Fluorescence analysis was performed using a BD LSR II flow cytometer (BD Biosciences). Murine mTECs were gated as living CD45−Ly51−EpCAM+ cells, which were further subdvided into TPAlo or TPAhi MHCIIlo and MHCIIhi mTECs (9, 28).
RNA preparation, cDNA synthesis, and gene expression profiling
The RNA from single-cell suspensions was isolated using the High Pure RNA Isolation Kit (Roche, Mannheim, Germany) and reverse transcribed into cDNA with SuperScript II reverse transcriptase (Invitrogen) for quantitative PCR. Real-time PCR reactions were performed in a final volume of 25 μl with optimal concentrations of forward and reverse primers (50–900 nM) using Power SYBR Green PCR Master Mix (Applied Biosystems, Darmstadt, Germany). Reactions were run on a sequence detection system (GeneAmp 7300; Applied Biosystems) and expression values were normalized to Ubiquitin or GAPDH expression relative to complete mouse or human thymus tissue using the comparative Ct method. Primer pairs spanning at least one intron were purchased from Eurofins MWG Operon (Ebersberg, Germany). Sequence information on primer pairs used is available upon request.
FACS-purified annexin V−TPA−/+ mTEC subsets from four independent experiments were processed using the μMACS SuperAmp protocol (Miltenyi Biotec) for RNA amplification (31). The amplified cDNA was labeled and hybridized to microarrays for gene expression profiling using Illumina’s BeadArrays.
Lineage tracing mouse model
Lineage-tracing experiments were performed as previously described (18). Briefly, Aire-CreERT2 × Ai6 mice were treated every other day with 2 mg of tamoxifen (Sigma-Aldrich) dissolved in corn oil (Sigma-Aldrich) by oral gavage for a total of three treatments before thymic isolation on day 7.
In vivo proliferation assay using EdU
In vivo cell proliferation was measured by EdU labeling. Two milligrams of EdU per mouse were injected daily i.v. for different time intervals (data not shown). Four hours after the last injection mice were killed. For each time point, the thymi of four to eight mice were pooled separately for the EdU- and PBS-injected control groups. EdU staining was performed using the EdU flow cytomerty kit 488 or coumarin azide (BaseClick, Neuried, Germany) as per the manufacturer’s protocol. Costaining for PE-conjugated surface markers and intracellular markers were performed after the EdU reaction.
Cell cycle analysis
The isolated, CD45-depleted mTECs from mouse and human thymus were first stained for various surface markers. Intracellular staining was performed using BD Cytofix/Cytoperm (BD Biosciences) reagents. Thereafter, the fixed cell suspension was incubated with DAPI for a minimum of 10 min and measured on a BD LSR II flow cytometer (BD Biosciences). Data were analyzed using FlowJo (Tree Star, Ashland, OR) and the cell cycle indices were calculated as (S + G2-M)/G1-G0 (represented as arbitrary units). Along with DAPI, mouse TPA-defined mTEC subsets were additionally stained for Ki67 using the Foxp3/transcription factor staining buffer set (eBioscience).
Two-dimensional feeder coculture
FACS-enriched mouse TECs were cocultured with human fibroblast feeders according to the method described for human epidermal keratinocytes (33). mTECs (i.e., 10,000–100,000 cells) were cultured on irradiated (50 Gy) human fibroblasts (360,000 cells) in cFAD medium (three fourths DMEM/one fourth DMEM–Ham’s F12, 5% FCS, 0.1 U/ml penicillin, 0.1 mg/ml streptomycin, 24 mg/l adenine, 400 μg/l hydrocortisone, 8.33 μg/l cholera toxin, 5 mg/l insulin, 1 μg/l epidermal growth factor) in a 10-cm2 petri plate. The cultures were terminated after 14.5 d, and TEC clones were enumerated following staining with rhodamine B or immunohistochemical staining for keratin 14 expression.
Reaggregate thymus organ cultures and three-dimensional culture on Alvetex scaffold
Reaggregate thymus organ cultures (RTOCs) were prepared from embryonic thymus stroma as described previously (34). Embryonic thymi (embryonic day [E]13–14) were digested with collagenase/dispase solution at 37°C until a single-cell suspension was obtained. These embryonic cells (5 × 105 cells per RTOC) were spiked with sorted YFP+ mTEC subsets to form RTOCs. After 2–4 d of culture in vitro, RTOCs were digested with collagenase/dispase at 37°C. The single-cell suspensions were reanalyzed by flow cytometry.
An Alvetex scaffold (Reinnervate, Glasgow, U.K.) was additionally used for three-dimensional in vitro culture of sorted TPA mTEC subsets as previously described (11). Between 50,000 and 300,000 sorted mTECs were seeded onto the scaffold in a minimal amount of RPMI 1640 media (10–15 μl) supplemented with 5% FCS and RANKL (0.1 μg/ml; R&D Systems, Wiesbaden, Germany). After 4 d, the cultures were harvested and processed for FACS analysis. The purity of the sorted TPA subsets was between 84 and 95% upon reanalysis.
Sorted mTECs and RMA cells were lysed in RIPA buffer along with complete protease inhibitor mixture (Roche) for 30 min on ice, along with vortexing every 10 min, followed by centrifugation at 12,840 × g for 15 min. The cell (nuclear) lysates were loaded on a small 10% SDS-PAGE gel (Bio-Rad, Hercules, CA). The proteins were separated on the gel, blotted onto a nitrocellulose membrane, stained using Ponceau S, blocked with milk solution, hybridized with suitable Abs, and detected by ECL. β-Tubulin was used as protein loading control. The RMA cell line without and with UV irradiation was used as negative and positive controls, respectively, for PARP detection.
Immunohistochemistry and image acquisition
Complete mouse thymus or portions of a human thymus were embedded in Tissue-Tek (Miles, Elkhart, IN) and frozen on dry ice. Cryostat sections (5 μm) were mounted on slides, air dried, and processed for immunohistochemistry. Briefly, the cryosections were either left unfixed or fixed in absolute acetone at −20°C for 10 min and blocked with 3% BSA in PBS. Incubation with the primary Abs was carried out in a humidity chamber at 37°C for 30 min followed by 30 min at room temperature. In the case of double staining, the two primary Abs were applied simultaneously. After washing, the sections were incubated with fluorochrome-conjugated secondary Abs as mentioned above. Upon completion of staining, the sections were mounted in Kaiser’s glycerol gelatin (Merck, Darmstadt, Germany) and stored in the dark.
Immunofluorescence specimens were analyzed using a Zeiss Axio Imager (Zeiss, Jena, Germany) Z1 microscope equipped with epifluorescence illumination. Micrographs were recorded with a CCD camera (AxioCam MRm) applying AxioVison 4.8 software.
Bioinformatic analysis of microarrays
Quantile normalization (35) and limma differential gene expression analysis (36) were performed to identify differentially expressed genes between the different TPA-defined mTEC subsets. The array barcode was introduced as an additional factor to account for potential batch effects between the two arrays that have been used. An adjusted p value of ≤0.01 (Benjamini–Hochberg method) and fold change of ≥2 or ≤0.5 were the criteria used to define genes as differentially expressed. All probes identified as significant in any of the comparisons between TPA-defined mTEC subsets were pooled, transformed to z scores (zero mean and unit variance), and subjected to hierarchical clustering (Euclidean distance metric, complete linkage algorithm). The resulting dendrogram was cut into four clusters based on visual inspection of the heat map. The overlap of the gene lists for each cluster, and of the entire set, with predefined lists from QuickGO (37) was determined by mapping probes to either gene symbols or ENSEMBL gene IDs. The enrichment of genes from predefined lists was determined by a Fisher exact test. Heat maps were based on values transformed to z scores (yellow, high; blue, low; black, zero), where rows were reordered by hierarchical clustering (Euclidean distance metric, complete linkage algorithm). Columns are given in the order of the putative differentiation sequence described in the main text. All calculations were performed in R version 3.2.3 (38) using the extension packages limma (version 3.26.5) (39), heatmap.plus (version 1.3) (40), and lumiMouseAll.db (version 1.22.0) (32). TRAs were defined as previously described (11, 30).
An unpaired t test and one-way ANOVA were performed using Prism (GraphPad Software).
Identification and characterization of a post-Aire mTEC stage
The terminal differentiation model of the mouse mTEC lineage proceeding from immature MHCIIlo to mature MHCIIhi has recently been extended to include an additional post-Aire stage. This post-Aire stage had been initially identified by mTEC lineage tracing in genetically altered mice (16, 18, 41). However, post-Aire cells could not be isolated in steady-state from wild-type mice, because their surface phenotype essentially overlapped with the immature MHCIIlo subset. This “contamination” of immature MHCIIlo mTECs by the more mature post-Aire cells is exemplified by their high expression levels of Ivl, a bona fide terminal differentiation marker of keratinocytes and mTECs (16, 18, 20, 21) (Fig. 1A, Supplemental Fig. 1A). A specific surface marker for the ex vivo isolation of the post-Aire subset was hitherto lacking. In the search for such a marker we resorted to a previous study (42) reporting that the lectin TPA specifically stained scattered single mTECs and mTECs forming concentric cell aggregates resembling Hassall’s corpuscles (HCs), a pattern that would comply with late-stage mTECs. Importantly, as shown by transmission electron microscopy, TPA labeled the outer membrane/surface of these mTECs.
We first confirmed the in situ distribution of TPA-positive single cells and HC-like epithelial cell aggregates at low frequencies in the murine thymic medulla with no cells stained in the thymic cortex (Fig. 1B, Supplemental Fig. 1B). Next, we analyzed the surface expression of TPA in immature MHCIIlo and mature MHCIIhi subsets by flow cytometry. TPA expression in both subsets displayed a broad disttibution without clearly delineable clusters. To determine whether the level of TPA expression could serve to discriminate early from late mTEC stages, we costained TPA with Ivl and K10, two markers that are known to be progressively upregulated during terminal differentiation of mTECs. We set eight gates to slice across the TPA expression profile (Fig. 1C, left panel) and assessed the relative expression level of Ivl and K10. A significant upregulation of both markers was only observed within the 4.7% expressing the highest level of TPA among MHCIIhi (slices 7 and 8) and within the 18.8% of MHCIIlo (slices 4.5–8) (Fig. 1C, right panel). Based on this differential expression of Ivl and K10, we operationally subdivided MHCIIlo and MHCIIhi mTECs into TPAlo (slices 1–6 in MHCIIhi and slices 1–4.5 in MHCIIlo) and TPAhi (slices 7–8 in MHCIIhi and slices 4.5–8 in MHCIIlo) subsets (Fig. 1C, left panel, red lines). The TPA gating strategy and subset designation in subsequent experiments was strictly based on this operational definition. TPAhi mTECs displayed a 2.4-fold enrichment in the immature mTECs (18.1%) as compared with the mature compartment (7.7%) in C57BL/6 mice (Fig. 1C, 1D). Note that this frequency was strain-dependent, as in BALB/c mice the two subsets amounted to 12.9 and 11.6% TPAhi, respectively (data not shown).
To corroborate whether the TPAhi fractions in both subsets indeed represented more mature stages, we analyzed the mRNA expression of a panel of genes denoting mTEC differentiation and promiscuous gene expression (pGE). For Ivl and K10, which are markers for terminal differentiation of keratinocytes and mTECs, we observed steadily increasing expression levels in the following sequence: TPAloMHCIIlo → TPAloMHCIIhi → TPAhiMHCIIhi → TPAhiMHCIIlo. Of note, MHCIIlo cells separated on the basis of TPA expression levels showed a >10-fold difference in Ivl and K10 expression levels. In contrast, Aire, FoxN1, and CD80 reached their maximal expression in TPAloMHCIIhi and then decreased in TPAhi mTECs. Next, we analyzed prototypic TRAs representing the four delineable gene pools of pGE (9). Male-enhanced Ag 1 (Mea1) and phosphorylase kinase, γ2 (Phkg2), expressed both in cTECs and mTECs (pool I), and the extracellular proteinase inhibitor (Expi) and carbonic anhydrase 8 (Car8), expressed in mTECs irrespective of their maturation (pool II), were equally transcribed in all four populations. In contrast, casein-β (Csnb) and glutamic acid decarboxylase (GAD67), only expressed in mature mTECs in an Aire-independent manner (pool III), and insulin 2 (Ins2) and chemokine (C motif) ligand 1 (Xcl1), only expressed in mature mTECs in an Aire-dependent manner (pool IV), showed maximal expression in MHCIIhi, irrespective of TPA expression. Interestingly, the transcription level of these two gene pools decreased in TPAhiMHCIIlo mTECs. Expression of the chemokine (C-C motif) ligand 21 (Ccl21) was consistently highest in TPAloMHCIIlo (Fig. 1E). Thus, the gene expression patterns, as far as analyzed, were in accordance with the TPAhi cells being more mature among the respective MHCII subsets, insinuating the following mTEC differentiation sequence: TPAloMHCIIlo → TPAloMHCIIhi → TPAhiMHCIIhi → TPAhiMHCIIlo mTECs.
Next, we performed a comparative transcriptome analysis using whole-genome arrays. The gene expression signatures of the four TPA subsets (excluding dead propidium iodide [PI]+ cells and dying annexin V+ cells) could be grouped into four main patterns (Fig. 2): cluster 1, genes that were upregulated solely in MHCIIhi mTECs, independent of TPA expression; cluster 2, genes that were highly expressed in pre-Aire TPAloMHCIIlo mTECs, transiently downregulated in mature mTECs, and again partially upregulated in post-Aire TPAhiMHCIIlo mTECs; cluster 3, genes that gradually increased in expression along the differentiation sequence; and cluster 4, genes that were downregulated in post-Aire TPAhiMHCIIlo mTECs. All clusters were enriched for epidermal differentiation genes to various degrees (p = 2.4 × 10−5 to 7.6 × 10−7) except cluster 4 and TRAs (p < 2.2 × 10−16; cluster 4 = 0.02). Cluster 1 contained TRAs of gene pools III and IV and was also significantly enriched for genes involved in Ag processing and presentation (p = 0.02). Cluster 2 was enriched for cell adhesion and transcriptional regulator genes. Cluster 3 showed enrichment for terminal differentiation genes of epidermal origin, namely, keratin 1 (K1), K10, Ivl, loricrin (Lor), kallikrein related-peptidase 8 (Klk8), corneodesmosin (Cdsn), cornifelin (Cnfn), transmembrane protease, serine 13 (Tmprss13), and caspase 14 (Casp14) akin to patterns observed in skin cornification. Cluster 4, the smallest gene cluster, showed no enrichment for discernible cell- or pathway-specific signatures.
The subset-specific gene expression patterns observed at the mRNA level were confirmed for Aire, FoxN1, K10, and Ivl at the protein level by flow cytometry and immunohistochemistry (Supplemental Figs. 1C–I, 2). We followed FoxN1 expression in the various mTEC subsets using FoxN1-GFP reporter expression as a surrogate marker. At the protein level, up to 69.7% of TPAhiMHCIIhi mTECs were Aire+ and 91.6% were FoxN1+ (Supplemental Figs. 1C, 1F, 1G, 2). Of note, none of the marker combinations analyzed in this study completely segregated with any of the four subsets, likely indicating further subset specification.
To probe the lineage relationship between the TPA-defined mTEC subsets we availed ourselves of a previously described lineage-tracing model (18). In this model system Aire-expressing cells were inducibly and indelibly labeled with a fluorescent reporter (zsGreen). These mice harbor both an inducible Aire-CreERT2 transgene and a Rosa26-Stopfl-reporter–targeted allele. Therefore, in mice treated with a single dose of tamoxifen, cells actively expressing Aire are permanently labeled with the fluorescent reporter. The lineage relationship between Aire− mTEC precusors, Aire+, and post-Aire mTECs was deduced using this lineage-tracing mouse model.
Four distinct mTEC subsets were observed based on MHCII and zsGreen reporter expression. Notably, the model confirmed the existence of a post-Aire (MHCIIlozsGreen+Aire−) population within the MHCIIlo mTEC subset (Fig. 3A, left panel, black box), which was distinct from the pre-Aire (MHCIIlozsGreen−Aire−) subset (Fig. 3A, left panel, blue box). When staining for TPA in this model, we observed a clear demarcation between pre- (MHCIIlozsGreen−) and post-Aire (MHCIIlozsGreen+) mTECs (Fig. 3A, right panel). TPA was strongly upregulated in most (67.2%) post-Aire mTECs (Fig. 3A, 3B, right panel), thus validating that TPA marks post-Aire mTECs and confirming the pre- (TPAloMHCIIlo) to post-Aire (TPAhiMHCIIlo) directionality.
To further probe the proposed mTEC lineage sequence based on TPA expression, we FACS purified YFP+TPAlo mTECs using a pan-YFP transgenic mouse strain and cultured them in RTOCs along with an unlabeled BL6 embryonic scaffold for either 2 or 4 d. After 2 and 4 d, the RTOCs were harvested and the cells were rephenotyped. The YFP+TPAlo subsets progressed along the following developmental sequence: TPAloMHCIIlo → TPAloMHCIIhi → TPAhiMHCIIhi mTECs (Fig. 3C). We further dissected the relatedness between the TPAlo/hiMHCIIhi subsets in a simplified Alvetex scaffold three-dimensional culture on FACS-purified TPAneg (slices 2 and 3) and TPAlo (slices 2.5–4) MHCIIhi mTECs (Fig. 1C, left panel) from BL6 mice. After a 4-d culture period, an increase in the level of TPA was observed in the sorted TPA subsets. They progressed along a TPAneg → TPAlo → TPAhi sequence within the MHCIIhi mTEC subset (Fig. 3D). Owing to poor survival in both models, we could not trace the downstream fate of TPAhi mTECs. Thus, collectively the lineage-tracing model and the three-dimensional culture systems supported the proposed mTEC lineage sequence.
Ontogeny and turnover of TPA-defined mTEC subsets
Next, we asked whether the emergence of mTEC subsets during ontogeny would also follow the proposed sequence. During embryonic development, the MHCIIlo subset clearly preceded the emergence of the MHCIIhi subset, initially making up most of the population (93.7%), and gradually declining in favor of mature mTECs (Fig. 4A). The first Aire+ mTECs appeared around E14.5 (16, 43). The TPAhi subsets also first emerged at E14.5 and expanded around E15.5–16.5, and the TPAhiMHCIIhi clearly preceded the TPAhiMHCIIlo subset. Thus, most TPAhi cells within the mTEC compartment at E15.5–16.5 were MHCIIhi. TPAhiMHCIIlo cells leveled around E18.5 (15.3%), and the relative enrichment of the TPAhi fraction within MHCIIlo as compared with MHCIIhi mTECs only became apparent in the early postnatal phase (Fig. 4B, Supplemental Fig. 3).
Next, we assessed the in vivo steady-state cell division of the TPA-defined mTEC subsets. To this end, we measured proliferation by in vivo EdU labeling. A cohort of young adult mice was pulsed daily i.v. with 2 mg of EdU for 12 d. At 4 h and days 1, 3, 6, 9, and 12, mice were sacrificed and mTEC subsets were analyzed for EdU incorporation. As previously reported, MHCIIhi mTECs proliferated at a higher rate than did MHCIIlo mTECs (8, 43). When MHCIIhi mTECs were further separated on the basis of TPA expression, the TPAlo initially preceded the TPAhi ones. Alternatively, independent of TPA expression MHCIIlo mTECs lagged behind their MHCIIhi counterparts by 2–12 d. Although the TPAhiMHCIIlo mTEC subsets continued to accumulate EdU label beyond day 6, the TPAloMHCIIlo subset reached a plateau after 9 d of treatment (i.e., ∼15%; Fig. 4C, lower panel). In both mice and humans, TPAhiMHCIIhi mTECs had the highest cell cycle index and the lowest percentage of cells in G0 of all four subsets. The post-Aire stage TPAhiMHCIIlo mTECs had a larger G0 and a reduced S-G2-M fraction, indicating its entry into a postmitotic stage (Supplemental Fig. 4A, 4B).
Given the controversy as to whether Aire+ MHCIIhi cells are cycling or postmitotic (7), we also followed EdU incorporation in Aire− vs Aire+ mTECs. Both subsets vigorously proliferated, with Aire+ cells displaying higher levels of EdU incorporation as compared with Aire− ones already after 1 day of EdU pulse (data not shown). The high turnover rate of the Aire+MHCIIhi subset was further supported by cell cycle analysis. Both in mice and humans, the Aire+ subset displayed a higher cell cycle index than the Aire− cells (Supplemental Fig. 4C). Despite their cell cycle activity, Aire+ mTECs have a larger fraction of cells in G0 compared with their Aire− counterparts, compatible with a slowdown of cell division (Supplemental Fig. 4D). These data are also in line with the low annexin V staining (a marker for apoptosis) of the MHCIIhi mTECs, especially within the Aire+ subset (see below, Supplemental Fig. 5B). Taken together, MHCIIhi mTECs represent a highly proliferative compartment irrespective of further subdivision according to Aire or TPA expression, compatible with an intermediate rather than a terminal position in the aforementioned four-stage model. Next, we asked whether the putative terminal post-Aire stage is prone to apoptosis.
Pre- rather than post-Aire MHCIIlo mTECs were marked for cell death
In viable cells, phosphatidylserine is located on the cytoplasmic surface of the cell membrane. However, in apoptotic cells phosphatidylserine flips from the inner to the outer leaflet of the plasma membrane, thus getting exposed to the external environment. Annexin V strongly binds to phosphatidylserine in a calcium-dependent manner. We assessed apoptosis of ex vivo–isolated mTEC subsets by costaining for TPA and annexin V. Surprisingly, irrespective of TPA, annexin V stained ∼70% of MHCIIlo mTECs and ∼20% of the MHCIIhi mTECs (Fig. 5C, upper left panel, Supplemental Fig. 5A; p < 0.001). The highest expression of annexin V was found among TPAloMHCIIlo mTECs (75.4%) and 38% less among the TPAhiMHCIIlo mTECs (46.7%) (Fig. 5A, 5B).
These results were further corroborated using a second death marker, namely active (cleaved) caspase-3, an effector caspase that is sequentially activated through cleavage during late apoptosis. Comparable results to the ones obtained with annexin V were observed with active caspase-3 (Fig. 5C, upper right panel, Supplemental Fig. 5A; p < 0.001). Both markers costained most of the MHCIIlo mTEC population and ∼20% of the MHCIIhi mTEC subset (Fig. 5C, lower panels). Thus, the pre-Aire TPAloMHCIIlo mTECs rather than the terminally differentiated post-Aire mTECs are most prone to steady-state apoptosis. This finding contrasts with the notion that mature MHCIIhi mTECs rather than immature, pre-Aire mTECs were more sensitive to cytoablative treatment in vivo and cell culture in vitro (8, 44–46) (C. Michel and B. Kyewski, personal communication).
Because cleaved caspase-3 has been also implicated in epithelial cell differentiation (47–50), we performed two additional assays to validate the nonproliferative and apoptotic status of the annexin V+TPAloMHCIIlo mTEC subset. First, we analyzed PARP cleavage. PARP is mainly involved in DNA repair and programmed cell death (51) and is known as one of the several cellular substrates of caspases. Active caspase-3 cleaves PARP, resulting in the formation of two specific fragments: an 89-kDa catalytic fragment and a 24-kDa DNA binding domain. The 24-kDa fragment irreversibly binds to DNA strand breaks and attenuates DNA repair (52, 53). Cleavage of PARP-1 by caspases is considered to be the hallmark of apoptosis. Western blot analysis on ex vivo–isolated annexin V+/−MHCIIlo/hi mTECs showed annexin V+ mTECs to have higher amounts of cleaved PARP compared with annexin V− cell fractions. The RMA cell line with and without induction of apoptosis by UV irradiation was used as a positive and negative control, respectively, for cleaved PARP activity (Fig. 5D). Second, we applied a single EdU pulse in vivo and measured its incorporation after 12 h in mTECs costained with annexin V. Whereas annexin V− showed perceptible EdU incorporation, the annexin V+ cells did not (Fig. 5E). Thus, independent of their stage of maturation, annexin V+ mTECs in steady-state adult thymus were essentially nonproliferative.
The “dead-end” annexin V+MHCIIlo subset was not prevalent during the fetal phase, but emerged after birth (postnatal days 4–5, 44.2%) and reached maximal levels (70.1%) in young adults (Fig. 5F). By implication, the minor pre-Aire annexin V−MHCIIlo subset constitutes a likely pool of mTEC precursor cells. Accordingly, when assayed for colony formation in a two-dimensional coculture model (33), only this subset contained clonogenic activity (data not shown). No clones developed from their annexin V+ counterparts. Collectively, the data show that: 1) steady-state apoptosis levels were most prominently observed in pre-Aire MHCIIlo mTECs rather than the terminally differentiated post-Aire stage; and 2) only the annexin V−MHCIIlo mTECs displayed clonogenic/progenitor potency.
Dynamic FoxN1 expression in annexin V/MHCII–defined mTEC subsets
The FoxN1 transcription factor is essential for intact TEC biology and thus thymus development and function. Although its precise function in TEC specification and mTEC versus cTEC commitment still needs to be defined, recent data implicate an essential role of FoxN1 in terminal differentiation, maintenance, and function in both TEC lineages (54–57). Given its proposed role in TEC maintenance, we asked how FoxN1 expression relates to the apoptosis rate among these developmental subsets. During ontogeny, the percentage of FoxN1+ cells significantly decreased from embryonic life (E15.5, 69.5%) and birth (P4–5, 49.8%) to adulthood (7.5%) in the MHCIIlo mTECs, whereas it remained relatively constant in MHCIIhi mTECs (Fig. 6A). FoxN1+MHCIIlo mTECs had a slightly higher survival preference over the FoxN1− cells until E18.5 (19 versus 35% annexin V+) (Fig. 6B, upper panel). However, postnatally, when the apoptosis rate in MHCIIlo mTECs increased, FoxN1 protein expression equally partitioned into the annexin V+ and annexin V− fractions among MHCIIlo cells (compare Fig. 5F with Fig. 6B, 6C, upper panel). In contrast, in MHCIIhi mTECs, all through ontogeny until adulthood, FoxN1+ segregated into annexin V− cells, compatible with a prosurvival role of FoxN1 in MHCIIhi mTECs (Fig. 6B, 6C, lower panel, Supplemental Fig. 5C). Thus, during embryonic development, FoxN1 inversely correlated with the apoptosis rate in MHCIIlo and MHCIIhi mTECs, whereas in adulthood this correlation was only seen in the MHCIIhi subset.
Human versus mouse mTECs: differences and commonalities
To assess whether the four mTEC stages apply to human mTECs, we costained purified human mTECs for both MHCII and TPA expression. Surprisingly, the relative expression of TPA in relationship to MHCII was reversed in humans compared with mice, with 15.0% of MHCIIhi but only 6.3% of MHCIIlo expressing high levels of TPA (Fig. 7A, 7B). This distribution suggested that the presumptive post-Aire TPAhiMHCIIlo stage was less prevalent in human mTECs. This notion is consistent with the higher expression levels of IVL and K10 in MHCIIhi versus MHCIIlo mTECs, which was again the reverse pattern compared with murine mTECs (Fig. 7C, Supplemental Fig. 1A).
When stained in situ, TPA expression was largely confined to HCs that were costained with IVL and K10. Infrequent TPAhi single cells in the vicinity of HCs also costained with AIRE (Fig. 7D). Indeed, 65.9% of isolated single TPAhiMHCIIhi mTECs were AIRE+ by flow cytometry (Fig. 7E). Thus, in the human thymus the bulk of K10+ and/or IVL+ TPAhi mTECs already form part of the HCs and thus will unlikely be recovered in single-cell suspensions (Fig. 8). Nevertheless, in human mTECs, TPA serves as an additional marker to delineate a subfraction expressing maximal levels of MHCII and a high proportion of AIRE+ cells (Fig. 7A, 7E). This allows for a considerable enrichment of intact AIRE+ human mTECs (which were 5.6-fold more frequent among TPAhi vs TPAlo MHCIIhi mTECs) with a novel surface marker combination.
Next, we asked whether most MHCIIlo human mTECs were also marked for cell death. This was clearly not the case, as a similar proportion of MHCIIhi and MHCIIlo mTECs stained for annexin V (∼15 and ∼20%, respectively; Fig. 7F). Thus, the human mTEC road map lacks a sizeable TPAhiMHCIIlo post-Aire stage and a major death-prone MHCIIlo annexin V+ stage. Noteworthy, the human mTEC staging more closely resembled that of rats rather than mice, as based on the relative expression levels of IVL and K10 in MHCIIlo versus MHCIIhi mTECs, as well as the occurrence of well-developed HCs (Supplemental Fig. 1A, data not shown).
Charting the road map of mTEC differentiation has recently attracted considerable interest, especially in view of the paramount role of this cell type in central tolerance by virtue of pGE (2, 4). The initial two-stage terminal differentiation model held that Aire−MHCIIloCD80lo mTECs continuously differentiate into the Aire+MHCIIhiCD80hi terminal stage, which eventually undergoes apoptosis (7–9, 22). Recently, this basic model has been extended by a distinct mTEC-committed precursor pool (2) and a late post-Aire stage, now considered the terminal stage of the mTEC lineage (15–19). However, the delineation of the post-Aire stage relied on lineage-tracing models in postnatal animals, whereas the routine separation and analysis of this mTEC subset in wild-type mice had been hampered by the lack of suitable discriminatory cell surface markers.
In this study, we introduce the lectin TPA as a novel cell surface marker that allows for the phenotypic and functional delineation of the MHCIIlo and MHCIIhi mTEC compartments into distinct developmental stages. TPA has been closely considered based on its surface expression on rare, single mTECs and mTECs forming HC-like cell aggregates in the medulla of the mouse thymus (42). Costaining of TPA and MHCII allowed for further subdivision of MHCIIlo and MHCIIhi mTECs into TPAhi and TPAlo subsets. Based on increasing expression levels of Ivl and K10 and further epithelial terminal differentiation markers among TPA/MHCII-defined mTEC subsets (in analogy to keratinocyte development), we deduced the following developmental sequence: TPAloMHCIIlo → TPAloMHCIIhi → TPAhiMHCIIhi → TPAhiMHCIIlo. Importantly, mTEC subsets emerged in this sequence also during ontogeny and EdU in vivo labeling kinetics. Additionally, RTOCs and Alvetex scaffold three-dimensional in vitro cultures reinforced the proposed lineage directionality. Furthermore, TPA was strongly upregulated in the post-Aire stage as identified in a lineage-tracing model (18). In this model we observed a small percentage (28%) of TPA+ cells in pre-Aire mTECs, which can be explained by the presence of post-Aire cells in MHCIIlo before tamoxifen induction. Alternatively, the presence of TPAlo cells (32.8%) in post-Aire mTECs is possibly due to the asynchronous upregulation of TPA in relationship to downregulating MHCII. Moreover, the composition of a single transient wave of post-Aire cells in the lineage-tracing model (observed 7 d after a single pulse of tamoxifen) is less complex than the post-Aire compartment in steady-state (C. Michel and C.N. Miller, unpublished observations).
The TPAhiMHCIIlo subset displayed features previously ascribed to post-Aire mTECs: high levels of the terminal epithelial differentiation markers Ivl and K10; and reduced levels of CD80, FoxN1, and Aire, when compared with MHCIIhi cells. Notably, the decline of these markers already started among TPAhiMHCIIhi cells at the mRNA but not the protein level (see Fig. 1E versus Supplemental Fig. 1F, 1G). The same dynamic regulation pertained to pGE as far as selected genes of pools III and IV were concerned, whereas this was not the case for pools I and II (9). Applying this new scheme, we show that Ccl21 expression, although enriched in MHCIIlo mTECs, clearly segregated into the pre-Aire mTEC pool rather than the post-Aire stage as has been previously claimed (58). When the transcriptome of the four subsets was compared at the whole-genome level by array analysis, four distinct patterns emerged. These distinct expression patterns further validated the TPA-based subset division and illustrated the considerable dynamics and extent of differential gene regulation during mTEC differentiation. Thus, TPAhiMHCIIlo cells qualify as post-Aire mTECs by all analyzed criteria.
When comparing mouse and human mTECs, TPA subdivided MHCIIhi cells to similar proportions, whereby TPA expression clearly segregated with Aire, more so in humans than mice. This allows for the first time for an efficient enrichment of intact AIRE+ human mTECs, the overall frequency of which is 3-fold lower in humans than in mice. The most striking difference between the two species was the paucity of the post-Aire TPAhiMHCIIlo subset in humans. This rather low proportion of post-Aire mTECs could be either due to a longer half-life of the AIRE+MHCIIhi cells or their more rapid demise following terminal differentiation and incorporation into HCs. Note that HCs are much more prominent in the human than the mouse thymus. Interestingly, with regard to the prevalence of HCs, rats more closely resembled humans than did their mouse counterparts.
Considering the population dynamics of the four subsets based on in vivo EdU labeling, cell cycle status, and steady-state expression of death markers, it was striking that the early TPAloMHCIIlo subset showed the lowest steady-state cell turnover with ON-EdU labeling reaching a plateau at 15%, thus substantiating results from a previous study applying BrdU (8). At the same time, up to 70% of the MHCIIlo subset was marked for apoptosis, and these early death-prone mTECs segregated into the noncycling fraction. This death-prone subset only emerged postnatally and inversely correlated with FoxN1 expression. In contrast, the post-Aire stage showed steadily increasing labeling kinetics, which was delayed by ∼2–9 d in relationship to MHCIIhi mTECs. This is in line with their presumptive terminal differentiation status and estimated half-life of 6–10 d for the Aire+MHCIIhi cells (18, 41). However, the percentage of annexin V+ cells was lower and, inversely, the percentage of FoxN1+ cells was higher in post-Aire than in pre-Aire MHCIIlo mTECs. In contrast, the two TPAlo/hiMHCIIhi subsets showed the lowest apoptosis and the highest FoxN1 expression. Thus, FoxN1 expression in all four TPA subsets inversely correlated with survival and turnover, compatible with a role of FoxN1 in the development and maintenance of the MHCIIhi TEC compartment, as previously argued (46, 55, 56). Additionally, the dynamic regulation of FoxN1 during mTEC differentiation paralleled that of skin epithelium, where FoxN1 triggers off early stages of skin epithelial differentiation while repressing late differentiation markers such as Ivl and K10 (59, 60).
The postnatal emergence of an exceptionally high fraction of death-prone TPAloMHCIIlo mTECs begs the question as to their derivation and raison d’être. Why would these MHCIIlo mTECs be generated in the first place if most undergo apoptosis? Several possible explanations come to mind: 1) dying mTECs could provide TRAs to be cross-presented by medullary dendritic cells, athough this would be confined to the pool of Aire-independent TRAs expressed in MHCIIlo mTECs (61–63); 2) maturation and survival of MHCIIlo mTECs is contingent on a stringent selection process comparable to lymphocyte development (64), for example, the ability to enact pGE; and 3) dying mTECs release TGF-β, an essential factor to induce Foxp3 expression and drive regulatory T cell generation (65). Note that the emergence of this death-prone mTEC fraction coincides with the first postnatal burst of Treg production (66, 67).
Of note, only the minor fraction of annexin V−MHCIIlo mTECs showed clonogenic activity in a simplified in vitro two-dimensional assay. Hence, this subset is likely to harbor those MHCIIlo mTECs, which have been shown to turn into MHCIIhi mTECs (7, 8, 22). Likewise, the recently described mTEC progenitor pool, which apparently is able to sustain the mTEC compartment life-long, should be included in this subset (4, 6, 68, 69). These cells may also be recruited during thymic regeneration following various thymus insults, for example, stress, sex hormones, immunosuppressive drugs, or radiation (44–46). The relative abundance of this subset during the exponential growth phase of the TEC compartment in the fetal period and its relative decline in the young adult steady-state phase is in keeping with this scenario.
The initial assumption that Aire+MHCIIhi mTECs represent a postmitotic terminal stage of the mTEC lineage with Aire possibly actively inducing apoptosis (7, 14) has been called into question with the identification of a distinct post-Aire stage. In line with this notion the Aire+MHCIIhi cells showed the highest cell turnover preceding Aire−MHCIIhi mTECs already after 1 d of the EdU pulse (data not shown). Moreover, among MHCIIhi the Aire+ fraction had a higher cycling index than did the Aire− MHCIIhi subset both in mice and humans. Continuous cell division of Aire+ mTECs is also more readily reconciled with recent concepts on the regulation of pGE at the population and single-cell levels. Thus, recurrent patterns of gene coexpression groups (both Aire-dependent and -independent) in a minority of mTECs (<1%) have been reported in the human and mouse mTEC compartments (11–13). To explain these observations it has been proposed that clusters of TRA genes, initially targeted stochastically, are bookmarked for transmission through mitosis and thus clonally amplified, a scenario that implies cell division once Aire has been turned on (13). The concept of sliding gene coexpression groups extends this idea in that coexpression groups are not strictly faithfully transmitted during mitosis but gradually shift. This would result in continuous changes in gene expression patterns along with scanning of the genome during the limited lifespan of mTECs. Again, cell division is implicit to this concept (11, 12).
There are striking commonalities between the skin and thymus epithelial compartments. The transcription factor FoxN1 is indispensable for the development and the functional integrity of both tissues (54, 59, 70, 71). Keratinocytes and TECs, and in particular mTEC lineage progression, are characterized by expression of similar sets of cytokeratins and differentiation Ags as reflected by an enrichment of an epidermal gene signature in three of the four TPA-defined clusters (19, 72–74). In this regard, TPA marked suprabasal keratinocytes, which coexpress the terminal marker K10 (data not shown). Likewise, progenitor/stem cells of both lineages share molecular markers such as Plet-1, Rac1, and Smad7 (75–77). This close relationship between both lineages may even extend to a common progenitor stage, as TEC progenitors can be reprogrammed into multipotent skin stem cells (78). Our data underscore and extend these parallels between both epithelial compartments by the notion that terminally differentiated TPAhi-MHCIIlo mTECs such as keratinocytes do not primarily die by apoptosis but rather take an alternative route of cell death, that is, cornification (79). In this regard, upregulation of the NF-κB pathway, which is instrumental for mTEC differentiation, is known to shut off proapoptotic molecules and induces antiapoptotic factors (members of the Bcl-2 family and cellular inhibitors of apoptosis) that promote cell cycle progression and differentiation (80, 81). Conventional caspases involved in apoptosis are turned off and caspases not involved in apoptosis, such as caspase 14, are expressed upon cornification of epithelial cells of skin, choroid plexus, retina, and thymus, that is, HCs (82).
Including TPA as an additional differentiation marker led to a refined road map of the mTEC lineage (Fig. 8). Salient features of this new scheme are: 1) the relative heterogeneity of the pre-Aire MHCIIlo stage, its dynamic composition during ontogeny, and its high death rate in postnatal thymus; 2) the fairly stable and homogeneous composition of the MHCIIhi compartment in line with its essential role in tolerance induction; and 3) species-specific variations in terminal mTEC differentiation. Further detailed analysis of these newly identified mTEC subsets should contribute to ongoing efforts to draw an ever-finer map of the mTEC lineage.
We are greatly indebted to Prof. C. Sebening and Prof. T. Loukanov (Department of Cardiac Surgery, Medical School of the University of Heidelberg) for making human thymus tissue available. We thank the German Cancer Research Center Flow Cytometry Core Facility for help with cell sorting and the University of California, San Francisco Parnassus Flow Cytometry Core.
This work was supported by the German Cancer Research Center (to S.P.), the Tolerage Consortium (7th Framework Program of the European Union) (to C.M.), the European Research Council (Grant ERC-2012-AdG to B.K.), and the National Institutes of Health (Grant R01 AI097457 to M.S.A.). The University of California, San Francisco Parnassus Flow Cytometry Core is supported by Grant P30 DK063720 to the Diabetes Research Center from the National Institutes of Health.
The microarray data in this article have been submitted to the Gene Expression Omnibus database (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE84956) under accession number GSE84956.
The online version of this article contains supplemental material.
Abbreviations used in this article:
cortical thymic epithelial cell
MHC class II
medullary thymic epithelial cell
promiscuous gene expression
reaggregate thymus organ culture
thymic epithelial cell
Tetragonolobus purpureas agglutinin
The authors have no financial conflicts of interest.