To investigate the role of Sirt6 in TECs, we crossed Foxn1-cre mice (Soza-Ried et al., 2008; Liu et al., 2013) and Sirt6^flox/flox mice (Kim et al., 2010; Wang et al., 2016) to generate Sirt6 conditional knock out mice (designated as Sirt6 cKO mice henceforth) and we further confirmed Sirt6 was indeed inactivated (Supplementary Figures 1A–C). First, we checked the effect of TEC-specific Sirt6 knockout on the thymus of the 4-week-old mice. Compared with wild-type mice, the specific deletion of Sirt6 in TECs led to severe thymic atrophy (Figures 1A,B). The morphological analysis showed that thymic medullary region decreased significantly while the cortical region had no obvious change after Sirt6 ablation (Figures 1C,D).

As a longevity gene, whether Sirt6 deletion leads to thymus senescence attracts our attention. Interestingly, although thymic atrophy still could be seen in 8-month-old Sirt6 cKO mice, the extent of reduction in thymus size was smaller than that in 4-week-old Sirt6 cKO mice (Supplementary Figure 2A). The rate of thymus weight loss (normalized for total body weight) and the rate of thymocytes number reduction did not increase in 8-month-old Sirt6 cKO mice compared with wild-type control littermates (Supplementary Figure 2B). The morphology of thymus in 8-month-old Sirt6 cKO mice showed no obvious signs of aging and the cortico-medullary junctions was integrated as assessed by H&E staining and Masson staining (Supplementary Figures 2C,D). Loss of naïve T cells in spleen is another indicator of age-related thymic degeneration (Aw and Palmer, 2012). However, from the 4 weeks to 8 months of age, the decline of naïve T cells in the spleen of Sirt6 cKO mice did not worsen compared with control mice (Supplementary Figures 2E,F). Thus, we focused our studies on the role of Sirt6 in the development and differentiation of TECs but not on the aging of TECs. Unless otherwise indicated, further experiments were carried out in 4-week-old Sirt6 cKO mice and their littermate controls.

To assess the endogenous role of Sirt6 in TECs, we checked the frequency and total number of TECs (CD45^–EpCAM⁺) in Sirt6 cKO mice and their littermate controls. Flow cytometry analysis revealed that Sirt6 deficiency resulted in the reduction of TECs, both in proportion and in absolute cell number (Figures 1E,F). These results suggested the thymic hypoplasia in Sirt6 cKO mice was caused by impaired TEC development. As we mentioned above, TECs are composed of cTECs and mTECs, we further scrutinized the effect of Sirt6 ablation on cTECs and mTECs. We found the percentage and cell number of mTECs reduced remarkable and absolute number of cTECs was obviously unchanged although the proportion of cTECs was relatively increased in Sirt6 cKO mice (Figures 1G,H). These results showed that Sirt6 deficiency predominately influence the development of mTECs rather than cTECs.

Due to the decreased thymus size and the cell number of mTECs in Sirt6 cKO mice, we examined the proliferation and apoptosis of mTECs. The results showed that the percentage of Ki67-positive cells in total mTECs reduced by nearly half in Sirt6 cKO mice (Figures 2A,B). Whereas, the expression of active caspase 3, an indicator of intrinsic apoptosis, did not increase in mTECs of Sirt6 cKO mice compared with littermate controls (Supplementary Figure 3).

To better understand how Sirt6 regulates cell proliferation, we performed RNA-sequencing analysis (RNA-seq) on mTECs (CD45^–EpCAM⁺UEA-1⁺ly51^–) sorted from 4-week-old wild-type or Sirt6 cKO mice, with two parallel samples in each group (named WT.1 and WT.2, Sirt6 cKO. 1 and Sirt6 cKO. 2). Gene Set Enrichment Analysis (GSEA) showed that DNA replication process was significantly reduced in Sirt6-deficient mTECs (Figure 2C), indicating that Sirt6 ablation may affect the cell cycle progression of mTECs. The expressions of many proliferation-related genes changed in Sirt6 deficient mTECs (Figure 2D). The expressions of genes promoting cell cycle and cell proliferation such as Cdc7, Ccna1, Rpa1, Pole2, Rfc2, and Rfc3 (Chen and Wold, 2014; Klajic et al., 2014; Tudzarova et al., 2016; Hu et al., 2020; Rogers et al., 2020) were down-regulated in Sirt6 deficient mTECs compared with wild-type TECs, whereas genes inhibiting cell cycle and proliferation such as E2f3, Smad3, and Gadd45b (Salvador et al., 2013; Park et al., 2016; Clijsters et al., 2019) were significantly up-regulated in Sirt6 deficient mTECs (Figure 2D). The modulated expressions of these genes would collectively contribute to the poor mTECs proliferation in Sirt6 cKO mice. 5-bromo-2’-deoxyuridine (BrdU) incorporation assay was used to detect newly synthesized DNA in S phase. The portion and the absolute number of proliferating cell (BrdU-positive cells) decreased markedly in mTECs of 2-week-old Sirt6 cKO mice in comparison with their littermate controls (Figures 2E,F). In contrast, Sirt6 deficiency did not affect the proliferation of cTECs as showed by the similar BrdU incorporation (Supplementary Figure 4A).

We further investigated the effect of Sirt6 deficiency on the differentiation and maturation of mTECs. Compared with immature mTECs, functional mature mTECs highly expressed CD80, MHC II and Aire, which were considered to play an critical role in negative selection (Kadouri et al., 2020). Flow cytometry analysis showed that the expression of MHC II and Aire on mTECs was similar between wild-type and Sirt6 cKO mice, whereas the proportion of CD80⁺ mTECs increased significantly in Sirt6 cKO mice (Figure 3A). Due to the overall reduction of mTECs, the absolute number of CD80⁺ mTECs, MHC II^high mTECs and Aire⁺ mTECs all decreased dramatically (Figure 3B). Specifically, Aire expressed on CD80⁺MHC II^high mTECs (mTEC^high) and then regulated the expression of thousands of TRAs (Heino et al., 2000; Gray et al., 2007; Wells et al., 2020). Sirt6 deletion accelerated the maturation of CD80^–Aire^– mTECs to the differentiation of CD80⁺Aire^–mTEC, but did not affect its further differentiation into CD80⁺Aire⁺ mTECs (Figures 3C,D). As shown by similar expression of MHC II and CD40, Sirt6 did not affect the maturation of cTECs (Supplementary Figures 4B,C).

TNFR family members CD40, RANK, and LTβR play a critical role in promoting the development of mTECs (Rossi et al., 2007; Zhu et al., 2007; Akiyama et al., 2008, 2012; Hikosaka et al., 2008; Seach et al., 2008; Cosway et al., 2017). There was no significant difference in Cd40 and Rank expression in mTECs between Sirt6 cKO mice and wild-type mice, but the expression of Ltbr in Sirt6 deficient mTECs decreased slightly (Supplementary Figure 5).

In order to figure out the regulation of mTEC maturation in molecular basis, we performed KEGG analysis for the RNA-seq results. We found that the endocytosis, NF-κB signaling pathway and focal adhesion were upregulated in Sirt6-deficient mTECs (Figure 4A). Among them, NF-κB signaling pathway has been reported to be essential in the development and maturation of mTECs (Burkly et al., 1995; Weih et al., 1995; Kajiura et al., 2004; Akiyama et al., 2005; Kinoshita et al., 2006; Zhang et al., 2006; Zhu et al., 2006; Riemann et al., 2017; Jin and Zhu, 2018). NF-κB target geneset was also enriched in Sirt6 absent mTECs, indicating that NF-κB signaling pathway was indeed activated after Sirt6 deletion (Figures 4B,C).

Next, we focused on the NF-κB signaling pathway to understand why Sirt6 specifically restricts the development of CD80⁺ mTECs. We analyzed the network of NF-κB related transcription factor and found that Spib, as an important transcription factor related to the development and differentiation of mTEC (Akiyama et al., 2014), was notably increased in Sirt6-deficient mTECs (Figure 4D). Among the known transcription factors involved in the development and differentiation of mTEC, Spib showed the greatest difference between WT and Sirt6 deficient mTECs (Supplementary Figure 6A). Furthermore, Spib was the highest expressed compared with other transcription factors that changed (p < 0.05) between Sirt6 deficient mTECs and wild-type control (Figure 4E).

In order to verify the contribution of upregulated Spib in Sirt6 deficient mTECs, we compared the genes directly regulated by Spib (Christie et al., 2015) and the genes changed in Sirt6 deficient mTECs. Our RNA-seq results showed that 920 genes was down-regulated (p < 0.05), and 1,217 genes was up-regulated (p < 0.05) in the Sirt6 deficient mTECs. Among the genes regulated by Spib, 461 genes were up-regulated (38% of 1,217 upregulated genes) and 281 genes were down-regulated (31% of 920 down-regulated genes) (Figure 4F). The genes directly regulated by Spib accounted for 35% of those with statistical difference between wild-type and Sirt6 deficient mTECs.

Previous studies showed that Spib1 rather than Spib2 is highly expressed in mTECs (Akiyama et al., 2014). To verify the expression of Spib in mTECs, we isolated mTECs (CD45^–EpCAM⁺UEA-1⁺ly51^–) from wild-type or Sirt6 cKO mice. Quantitative PCR analysis showed that Spib1, not Spib2, was upregulated after Sirt6 deletion (Figure 4G). In addition, western blot result of sorted TECs (CD45^–EpCAM⁺) confirmed that SPIB was indeed up-regulated after Sirt6 conditional knockout (Figure 4H and Supplementary Figure 6B). It has been reported that Spib promotes osteoprotegerin (Opg) expression and participates in the mTEC developmental stage–specific negative feedback regulation (Akiyama et al., 2014). The development of mTECs was limited in Sirt6 cKO neonates but not in E16.5 (Supplementary Figures 7A,B), which was consistent with the previous report that Spib regulated the development of mTECs in neonates rather than embryos (Akiyama et al., 2014). mTEC^hi population (CD45^–EpCAM⁺UEA-1⁺CD80⁺MHC II^high) were sorted from wild-type or Sirt6 cKO mice (Supplementary Figure 6C), and the expression of Cd80 and Opg were evaluated by quantitative PCR analysis. After Sirt6 deletion, the expressions of Cd80 and Opg were up-regulated (Figures 4I,J).

Positive and negative selection T cells in thymus is mainly orchestrated by TECs (Anderson et al., 2000; Rodewald, 2008). We examined the development of thymocytes in Sirt6 cKO mice to determine whether the impaired mTEC development affected its function in orchestrating thymocytes development. Although the proportion of DN, DP, CD4 SP, and CD8 SP was unaffected by Sirt6 deletion, their cellularity decreased significantly in Sirt6 cKO mice compared with wild-type control (Figures 5A,B). The further maturation of post-selected thymocytes downregulates the expression of CD24, CCR7 and upregulates the expression of CD62L (Zuklys et al., 2016; Liang et al., 2018). We found the frequency of CD24^lowCD62L^highCD4⁺CD8^–TCRβ⁺ and CD24^lowCD62L^highCD4^–CD8⁺TCRβ⁺ reduced in Sirt6 cKO mice (Figure 5C and Supplementary Figure 8A). The percentage of CD24^–CCR7^loCD4⁺CD8^–TCRβ⁺CD5⁺Foxp3^– thymocytes and CD24^–CCR7^loCD4^–CD8⁺TCRβ⁺CD5⁺ thymocytes decreased in Sirt6 cKO mice (Figures 5D,E and Supplementary Figure 8B), implying the further maturation of thymocytes was restrained by dysfunctional mTECs in Sirt6 cKO mice. The number of CD4⁺ T cells or CD8⁺ T cells and the portion of naïve CD4⁺ T cells or naïve CD8⁺ T cells reduced in spleen, indicating that T cells output from thymus decreased after Sirt6 deletion in TECs (Figures 5F,G and Supplementary Figure 8C).

In addition to the elimination of self-reactive T cells, mTECs play an important role in promoting the diversion of nTreg cells lineage (Cowan et al., 2013; Kadouri et al., 2020). The results showed that the percentage and the cell number of mature nTreg cells (Foxp3⁺CD4 SP) decreased obviously in Sirt6 cKO mice compared to their littermate controls (Figure 5H). By detecting the maturation process of nTreg cells (Lio and Hsieh, 2008), we found the ratio of immature nTreg cells (Foxp3^–CD25⁺CD4 SP) to mature nTreg cells (Foxp3⁺CD25⁺CD4 SP) increased notably in Sirt6 cKO mice, implying that the maturation of nTreg cells was blocked during the differentiation of Foxp3^–CD25⁺CD4 SP to Foxp3⁺CD25⁺CD4 SP (Figures 5I,J and Supplementary Figure 8D).

Due to the developmental and functional defects of mTECs in Sirt6 cKO mice, we next determined whether such defects affected the establishment of central immune tolerance. We compared the body weight of Sirt6 cKO mice with their age-matched littermate controls and found the body weight of Sirt6 cKO mice was lower obviously than their age-matched littermate controls 20 weeks after birth (Figure 6A). Lymphocytic infiltrates in multiple organs is an important index to judge the occurrence of autoimmune disease (Chuprin et al., 2015). H&E staining results showed that many organs in the 8-month-old Sirt6 cKO mice had more severe lymphocytes infiltration than wild-type control mice, including the salivary gland, kidney, lung, and liver (Figure 6B). We further examined the presence of autoantibodies in serum of 8-month-old Sirt6 cKO. Indeed, there were high levels of antinuclear antibodies in the serum of Sirt6 cKO mice (Figure 6C). What’s more, compared with wild-type control mice, Sirt6 cKO mice had more autoantibodies against many organs of Rag2 KO mice (Shinkai et al., 1992), such as liver, colon, and salivary gland (Figure 6D).

The expression of TRAs on the mature mTECs is essential for eliminating auto-reactive T cells and promoting the development of nTreg cells, which is essential for the establishment of central immune tolerance (Klein et al., 2000). We analyzed the expression of Aire-dependent and Aire-independent TRAs (Sansom et al., 2014; Rodrigues et al., 2017) and found that the expression of both Aire-dependent and Aire-independent TRAs was affected in the mTECs after Sirt6 deletion (Figure 6E). Compared with Sirt1 which only specifically regulates Aire-dependent TRAs (Chuprin et al., 2015), Sirt6 affect a wider range of TRAs expression. Most of the down-regulated TRAs in Sirt6 deficient mTECs were shown in Figure 6F. The decreased TRAs expression indicated that the function of mTECs in establishing central immune tolerance was impaired after Sirt6 deletion.

Sirt6^loxp/loxp mice (Kim et al., 2010; Wang et al., 2016) were crossed with Foxn1-Cre mice (Soza-Ried et al., 2008; Liu et al., 2013) to generate Foxn1-Cre Sirt6^loxp/loxp mice. Littermates or age-matched wild-type mice (Foxn1-Cre negative) were used as controls. We obtained Sirt6^loxp/loxp mice from Dr. Zhenyu Ju of Hangzhou Normal University, Hangzhou, China, and obtained Foxn1-Cre mice from Dr. Yu Zhang of Peking University Health Science Center, Beijing, China. Rag2 KO mice (Shinkai et al., 1992) was purchased from HFK Bioscience, Beijing, China. All mice were maintained under specific pathogen–free conditions and treated in accordance with Animal Experiments Guidelines of the animal Ethics Committee of Institute of Zoology, Beijing, China.

Thymic stromal cells was isolated from whole thymus by using the previous thymic stromal cell separation method (Sun et al., 2013; Liang et al., 2018). In brief, fresh thymus tissue was cut into pieces and suspended the thymus fragments in DMEM (Hyclone Laboratories, SH30022.01B) medium with 2% fetal bovine serum (FBS; Gibco, 16000-044). Most thymocytes in the supernatant were removed, and the remaining thymus fragments were incubated at 37°C for 15 min in 2ml solution of 1 mg/ml collagenase/dispase (Sigma-Aldrich, 11097113001) with 20 U/ml DNAse I (Sigma-Aldrich, D5025). Repeat the above digestion three times, gently shaking to make digestion more thorough. The digested cell suspension was terminated with PBS containing 1% FBS and 5 mM EDTA until all fragments disappeared. After centrifugation, cells were suspended in DMEM (containing 2% FBS). Cell suspension was gently blown and filtered with a 200 mesh filter to remove clumps and form a single cell suspension.

Fc-receptor was blocked by 2.4G2 before staining with indicated antibodies. All type of cell were stained with the appropriate fluorophore-labeled antibodies at 4°C for 30 min. For intracellular staining, the fixation buffer (eBioscience, 00-5123-43 and 00-5223-56) and permeabilization buffer (eBioscience, 00-8333-56) was used according to the supplier’s protocol. The flow cytometry was performed with Gallios Flow Cytometer (BeckMan Coulter, United States) or BD LSRFortessa X-20 Flow Cytometer (BD Biosciences, United States).

The Fluorescein labeled Ulex Europaeus Agglutinin I (UEA I) (Vector Laboratories, FL-1061) was obtained from Vector Laboratories. The CD45-PerCP/Cy5.5 (Biolegend, clone 30-F11, 103132), CD326-PE/Cy7 (Biolegend, clone G8.8, 188216), CD326-FITC (Biolegend, clone G8.8, 118207), Ly51-AF647 (Biolegend, clone 6C3, 108312), CD40-PE (Biolegend, clone 3/23, 124610), I-A/I-E-Brilliant Violet 421 (Biolegend, clone M5/114.15.2, 107632), CD80-PE (Biolegend, clone 16-10A1, 104708), CD80-BV650 (Biolegend, clone 16-10A1, 104732), CD4-FITC (Biolegend, clone GK1.5, 100405), CD4-PE (Biolegend, clone GK1.5, 100408), CD4-PE/Cy5 (Biolegend, clone GK1.5, 100410), CD4-APC (Biolegend, clone GK1.5, 100412), CD4-APC/Cy7 (Biolegend, clone GK1.5, 100412), CD8a-PE (Biolegend, clone 53-6.7, 100725), CD8a-PE/Cy5 (Biolegend, clone 53-6.7, 100710), CD8a-Brilliant Violet 421 (Biolegend, clone 53-6.7, 100738), TCR-β-PE-Cy7 (Biolegend, clone H57-597, 109222), CD24-FITC (Biolegend, clone M1/69, 101805), CD24-PE/Cy5 (Biolegend, clone M1/69, 101812), CD44-FITC (Biolegend, clone IM7, 103006), CCR7-PE (Biolegend, clone 4B12, 120106), CD5-APC (Biolegend, clone 53-7.3, 100626) were purchased from Biolegend. The Aire (eBioscience, clone 5H12, 50-5934-80), Fixable Viability Dye eFluor^TM 506 (eBioscience, 65-0866-18), CD62L-PE (eBioscience, clone MEL-14, 12-0621-82), CD25-PE-Cy5 (eBioscience, clone PC61.5, 12-0251-82), Foxp3-FITC (eBioscience, clone FJK-16s, 11-5773-82), were purchased from eBioscience. CD45-BUV395 (BD Biosciences, clone 30-F11, 564279), Ly51-BV786 (BD Biosciences, clone BP-1, 740882), PE Mice Anti-Ki-67 Set (BD Biosciences, 556027), APC BrdU Flow Kit (BD Biosciences, 552598), and PE Active Caspase-3 Apoptosis Kit (BD Biosciences, 550914) were purchased from BD Biosciences.

For Brdu incorporation, 2-week-old Sirt6 cKO and littermate control mice were intraperitoneally injected with BrdU (BD Biosciences), 1 mg per mice. Twenty-four hours after injection, thymic stromal cells were isolated according to the above-mentioned method before flow cytometric analysis. After surface antigens staining, according to the supplier’s protocol (BD Pharmingen^TM BrdU Flow Kits Instruction Manual), BrdU was detected by using APC-BrdU Flow Kit (BD Biosciences, 552598) and was analyzed by flow cytometry.

Tissues were embedded in optimum cutting temperature compound (Sakura, 4583) and frozen in liquid nitrogen. Sections (6 μm in thickness) and/or cells were fixed for 20 min with 4% polyoxymethylene (Solarbio, P1110) and blocked in PSB containing 1% BSA. Then, sections and/or cells were incubated with primary and secondary antibodies for 1 h at room temperature. Samples were stained with DAPI (1:1,000) after secondary staining. The following antibodies were used for staining: rabbit anti-KRT5 (Covance, PRB-160P; clone AF 138) diluted by 1:400 and rat anti-KRT8 (DSHB, AB 531826; Troma-I) diluted by 1:200. Sera of 8-month-old wild-type and Sirt6 cKO mice (diluted by 1:30) were used as primary antibodies for the detection of the autoantibodies. The secondary antibodies were used for staining: Alexa Fluor 594-conjugated donkey anti-rabbit IgG (H + L) (Jackson ImmunoResearch Laboratories, 711-586-152) diluted by 1:400, Alexa Fluor 488-conjugated donkey anti-rat IgG (H + L) (Jackson ImmunoResearch Laboratories, 712-546-150) diluted by 1:400, and Alexa Fluor 488-conjugated donkey anti-mice IgG (H + L) antibodies (Jackson ImmunoResearch Laboratories, 715-546-150) diluted by 1:300. All antibodies were diluted in 0.5% BSA in PBS. Laser scanning confocal microscope (Zeiss LSM710, Oberkochen, Germany) were used to acquire images.

As mentioned above, thymic stromal cells were isolated from 4-week-old wild-type and Sirt6 cKO mice, with two independent parallel samples in each group. The next RNA sequencing step of sorted mTECs (CD45^–EPCAM⁺UEA-1⁺Ly51^–) was performed according to the previous method (Liang et al., 2020).

Low quality reads (Q < 20) were assessment by FastQC and the adaptor sequence was filtered by Trimgalore. Processed reads were aligned to mice genome (mm10) via HISAT2 (Pertea et al., 2016). StringTie was applied to assemble and quantify the transcripts in each sample to obtain the number of exon, transcription initiation/stop site, count and TPM (Transcripts Per Kilobase of exon model per Million mapped reads) values. The identification of differential expression genes (DEGs) was performed by using the DEseq2 (Love et al., 2014) R-packages with count value. The threshold is under the condition of adjusted p < 0.05 and | Foldchange| > 2. The normalized gene expression only retained the gene with TPM > 0.1 in all samples.

The KEGG pathway analysis of the differentially expressed genes were performed by KOBAS 3.0 on the web¹ (Ai and Kong, 2018). In addition, GSEA was carried out by searching KEGG Database (Subramanian et al., 2007). All analyses were selected with p < 0.05 as the cut-off criterion.

Functional protein association network was created by STRING (Szklarczyk et al., 2015) (Available online: https://string-db.org/) which was constructed by the transcriptional factors obtain by website (Available online: http://bioinfo.life.hust.edu.cn/AnimalTFDB/). Networks were all visualized by Cytoscape (Demchak et al., 2014).

Thymi of wild-type and Sirt6 cKO neonatal mice were cut into small pieces and suspended with DMEM containing 2% FBS. After the small pieces settled at the bottom of the tube, the cell suspension was discarded. The remaining small pieces were suspended in TyEpiCM (ScienCell Research Laboratories, Catalog #3911) and incubator at 37°C with 5% CO₂ for 7 days, exchanging the medium every other day.

Cultured wild-type and Sirt6 cKO TECs were used to detect SIRT6 expression. TECs (CD45^–EPCAM⁺) isolated from wild-type and Sirt6 cKO mice were used to detect the expression of SPIB. After being washed with cold PBS, TEC were lysed in RIPA buffer (140 mM NaCl, 10 mM Tris-Cl (pH 8.0), 1 mM EDTA, 0.5 mM EGTA, 1% Triton X-100, 0.1% sodium deoxycholate and 0.1% SDS) complemented with a proteinase inhibitor cocktail (Sigma-Aldrich, P8340). Protein concentration was detected with Bradford assay. Proteins were analyzed by 10% SDS-PAGE and transferred onto PVDF membranes (Merck Millipore, IPFL00010). Each PVDF membrane was blocked with 5% non-fat dried milk (OXOID, LP0031) for 60–90 min at room temperature and incubated with each primary antibodies overnight on the shaking table at 4°C. After cleaning PVDF membrane with TBST solution for four times, the corresponding secondary antibodies were added for 45–60 min at room temperature. Protein bands were detected by chemiluminescence (Merck Millipore, WBKLS0500). ACTIN is used as internal reference for protein standardization. The primary antibodies used for western blot are as follows: Anti-SIRT6 (Cell Signaling Technology, 12486) diluted by 1:1,000; anti-SPIB (Cell Signaling Technology, 14337S) diluted by 1:1,000; anti-ACTIN (Sigma-Aldrich, A5441) diluted by 1:20,000.

Thymic stromal cell was isolated from 4-week-old wild-type or Sirt6 cKO mice and then sorted with MoFlo XDP cell sorter (Beckman Coulter, Brea, CA, United States), characterized as TEC (CD45^–EPCAM⁺), mTEC (CD45^–EPCAM⁺UEA-1⁺Ly51^–), cTEC (CD45^–EPCAM⁺UEA-1^–Ly51⁺), mTEC^low (CD45^–EPCAM⁺UEA-1⁺CD80^–MHCII^low) and mTEC^high (CD45^–EPCAM⁺UEA-1⁺CD80⁺MHCII^high). Total RNA was isolated with MicroElute Total RNA Kits (Omega Bio-tek, R6831) and reverse transcription was performed with Super-Script III Reverse Transcription (Invitrogen, 18080–093) according to manufacturer’s instructions.

Real-time PCR was performed with multiple kits (SYBR Premix Ex Taq, Takara Bio, DRR041A) on CFX96 (Bio-Rad Laboratories, Hercules, CA, United States). All primers used in this article were listed in Table 1.

All data are presented as the means ± SD. The statistical significance of differences between two experimental groups was tested by Student’s t-test. A p-value < 0.05 was considered statistically significant.