Our scientific research was performed using the TEC-conditional homozygous Dhx9-deleted mice model. We recombined the Cre-loxp system in TECs via crossing the Dhx9 ^loxp/loxp mice to the Foxn1^Cre mice. We used the Dhx9 ^loxp/loxp littermates as WT controls. All mice were kept under specific pathogen-free conditions. Age-matched mice displayed a single representative experiment in figures. The Dhx9 ^loxp/loxp mice came from Dr. Baojun Zhang, Department of Pathogenic Microbiology and Immunology, School of Basic Medical Sciences, Xi’an Jiaotong University (Xi’an, Shaanxi, China) (24). The Foxn1^Cre mice were provided by Dr. Yu Zhang of Peking University, Health Science Center (Beijing, China). With the improvement of the Animal Ethics Committee of the Institute of Zoology (Beijing, China), all animal experiments were operated under the guidance of the ethical care and use of laboratory animals.

TECs are an extreme low frequency cell population in thymus. We enriched TECs through free sedimentation, combined with magnetic-activated cell sorting (MACS) or fluorescence-activated cell sorting (FACS). Then 4-week-old mice were euthanized to collect the thymic lobes. After being washed twice in cold PBS, the thymic lobes were immediately cut into homogeneous tiny pieces and collected with Dulbecco’s Modified Eagle’s Medium (DMEM), containing 2% FBS into 15 mL numbered centrifuge tubes. The dissociative thymocytes were released into DMEM while the tissue fragments settled freely to the bottom of the tube as thymic stroma. To acquire single cell suspension, the tissue fragments were resuspended with DMEM, containing 2% FBS, 1 mg/mL collagenase/dispase (11097113001; Sigma-Aldrich), and 20 U/mL DNAse I (D5025; Sigma-Aldrich) and then incubated at 37°C for 45 min. When the tissue pieces disappeared, we used 5 mL of PBS containing 1% FBS and 5 mM EDTA to neutralize cell mixture gently, which was blown several times and filtered with a 200-mesh filter to obtain single-cell suspension. The cells were then centrifuged, resuspended in DMEM (containing 2% FBS), and counted for further experiments. As for MACS, anti-mouse CD45 microbeads (Miltenyi Biotec, 130052301) were used to further TEC enrichment, according to the supplier’s protocols.

For flow cytometry analysis, 1×10⁶ fresh isolated cells were resuspended with 100 μL FACS buffer (PBS containing 0.1% BSA, 0.02% NaN₃) and incubated with fluorescein conjugated antibody mixture in 4 °C for 30 min. Anti-CD16/32 antibody (12-0161-82, clone 93) and fixable viability dye eFluor™ 506 (65–0866–18) are indispensable for blocking the Fc receptor and eliminating dead cells. The fixation buffer (eBioscience, 00-5123-43 and 00-5223-56) and permeabilization buffer were used to fix and permeabilize cells for intracellular staining, such as Aire, Ki67, Helios, Foxp3, RORγt, and PLZF. For the detection of thymic tuft cells, the secondary antibody Alexa Fluor 647-conjugated donkey anti-rabbit IgG (H+L) (Jackson ImmunoResearch Laboratories, 711-605-152) was used after intracellular staining of DCAMKL1 (DCLK1) (Abcam, ab31704). All operations were shielded from light at 4 °C until analyzed by flow cytometry.

According to the identification of different cell populations, the fluorochrome-conjugated antibodies used in the experiment were listed below. For TECs, we used CD45 (Biolegend; 103132; clone 30-F11), EpCAM (Biolegend; 118215; clone G8.8), Fluorescein-labeled Ulex Europaeus Agglutinin I (UEA1) (FL-1061; Vector Laboratories), Ly51 (Biolegend; 108312; clone 6C3), Ly51 (BD Biosciences; 740882; clone BP-1), CD40 (Biolegend; 124610; clone 3/23), I-A/I-E (Biolegend; 107632; clone M5/114.15.2), CD80 (eBioscience; 12-0801-82; clone 16-10A1), and Aire (eBioscience; 50-5934-80; clone 5H12). For conventional T cells, we used TCRβ (Biolegend; 109222; clone H57-597), CD4 (Biolegend; 100414; clone GK1.5), CD8 (Biolegend; 100738; clone 53-6.7), CD19 (Biolegend; 152404; 1D3/CD19), CD11b (eBioscience; 11-0112-85; clone M1/70), F4/80 (eBioscience; 11-4801-82; clone BM8), NK1.1 (eBioscience; 11-5941-81; clone PK136), CD44 (Biolegend; 103055; clone IM7), CD279 (Biolegend; 135216; clone 29F.1A12), CD5 (Biolegend; 100625; clone 53-7.3), Ter119 (Biolegend; 116206; clone TER-119), CCR7 (Biolegend; 120106; clone 4B12), Foxp3 (eBioscience; 11-5773-82; clone FJK-16s), CD25 (eBioscience; 12-0251-82; clone PC61.5), CD24 (eBioscience; 11-0242-82; clone M1/69), CD69 (eBioscience; 48-0691-82; clone H1.2F3), CD62L (eBioscience; 12-0621-82; clone MEL-14), and CD45RB (eBioscience; 11-0455-82; clone C363.16A). For the iNKT cells, we used PLZF (eBioscience; 53-9320-80; clone Mags.21F7), RORγt (eBioscience; 562894; clone Q31-378), and CD1d tetramer (ProImmune; D001-2X). For the kit, we used the PE Active Caspase-3 Apoptosis Kit (BD Biosciences; 550914). We performed a flow cytometric detection with a Gallios Flow Cytometer (Beckman Coulter, United States) or a BD LSRFortessa X-20 Flow Cytometer (BD Biosciences, United States) and analyzed the results with FlowJo software (BD Biosciences).

TECs have the ability to proliferate when cultured in Thymic Epithelial Cell Medium (TyEpiCM, ScienCell Research Laboratories, Catalog #3911). We collected the thymic lobes from neonatal WT and Dhx9 cKO mice, respectively, and cut them into homogeneous tiny pieces, as described above. But, unlike in “Purification of thymic epithelial cells”, the small pieces were resuspended by TyEpiCM directly and incubated in a sterilized condition of 37°C with 5% CO₂ for 7 days. In the meantime, we washed the culture dishes with 37°C PBS and changed the medium every other day to remove nonadherent thymocytes and the dead cells.

Cell cycle distribution was determined by a BD LSRFortessa X-20 Flow Cytometer and analyzed by a “cell cycle” module in FlowJo software. Generally, cultured TECs were disassociated by trypsin EDTA 1× (25-053-CI, Corning) at 37°C for 5-10 min, collected into 5 mL polystyrene round-bottom tube and washed twice with cold PBS. Cells were fixed in 75% ethanol overnight at -20°C and followed by washing twice with 3-5 mL PBS. The fixed cells were then incubated with 100 μL 100 μg/mL RNaseA and 25 μL 100 μg/mL propidium iodide (PI) for 30 min at room temperature.

In fact, 4% paraformaldehyde fixed, paraffin embedded tissue sections (6 µm) were deparaffinized in xylene and rehydrated in an ethanol series. Hematoxylin and eosin (H&E) staining was directed by the reported standard protocol. Finally, these slides were captured by Leica Aperio VESA8. Inflammatory cell infiltration has five levels, as previously reported (29, 30).

Thymus, embedded with the optimal cutting temperature compound, was cut into 6 μm slides. The slides or cells were fixed for 20 min with 4% polyoxymethylene and blocked in PBS containing 1% BSA. Immunofluorescence staining was performed by the reported standard protocol (31). For analysis of the thymic medulla and cortex region, the thymic slides were strained via primary antibodies [rabbit anti-KRT5 (Covance; PRB-160P; clone AF 138) and rat anti-KRT8 (DSHB; ab531826; Troma-I)] and followed by the secondary antibodies Alexa Fluor 594-conjugated donkey anti-rabbit IgG (H+L) (Jackson ImmunoResearch Laboratories; 711-586-152) and Alexa Fluor 488-conjugated donkey anti-rat IgG (H+L) (Jackson Immuno Research Laboratories; 712-546-150). For the detection of antinuclear antibodies in 8-month-old WT and Dhx9 cKO mice, Sera were diluted by 1:30 as primary antibodies. The combination was detected via Alexa Fluor 488-conjugated donkey anti-mice IgG (H + L) antibodies (Jackson Immuno Research Laboratories; 715-546-150). Nuclei were stained with 4’6-diamidino-2-phenylindole (DAPI; Sigma-Aldrich; D9542). Images were taken by a laser-scanning N-SIM super-resolution confocal microscope (Nikon, Tokyo, Japan).

The MicroElute Total RNA Kit (Omega Bio-Tek, R6831) or TRIzol Reagent (Invitrogen, 15596-018) were used to extract total RNA from the sorted TECs, according to the manufacturer’s instructions. Next, unstable mRNA was reversely transcribed into cDNA by the SuperScript III Reverse Transcriptase Kit (Invitrogen, 18080-093). Using SYBR Premix Ex TaqTM (TaKaRa, RR420), a quantitative PCR of target genes was performed on a CFX96 apparatus (Bio-Rad Laboratories). For data analysis, expressions of target genes were normalized by Hprt using the 2-ΔΔCt method. All primers used in this study were Hprt-forward 5′-TGAAGAGCTACTGT-AATGATCAGTCAA-3′, Hprt-reverse 5′-AGCAAGCTTGCAACCTTAACCA-3′, Dhx9-forward 5′-CCACCCATACTTAGCGACAC-3′, and Dhx9-reverse 5′-CCATAGCCAGAAGACTCAACC-3′.

Single-cell suspensions of mTEC^hi and mTEC^lo from WT and Dhx9 cKO mice were sorted using a Fusion cell sorter (BD Biosciences), with two independent parallel samples in each group ( Figure S2 ). The operations are consistent with the previously published method (32). We used DEGseq to identify the differential expressed genes (q-value < 0.05, |log₂FC| > 0) of mTEC^hi and mTEC^lo from the different genotype mice, respectively (33). According to the differential expressed genes in mTECs of WT and Dhx9 cKO mice, GSVA was performed to calculate the individual gene set enrichment scores, which were visualized by R3.6.0. The analysis of the KEGG pathway was performed by KOBAS 3.0 (34). GSEA was carried out by searching the KEGG database (34). All analyses were carried out with P < 0.05 as the cutoff criterion. The raw data of RNA-seq sequencing used in this article could be found in the National Genomics Data Center (NGDC), part of the China National Center for Bioinformation (CNCB) under number PRJCA008466.

We used a student’s unpaired t-test for the statistical analysis of the WT and Dhx9 cKO mice. Sample sizes for the statistical analysis were at least n = 3, and the statistical significant P-value was taken as P < 0.05. Errors were shown as standard deviations (SD) throughout. There were no limitations on the repeatability of the experiments, and no samples were excluded from this analysis.

The Dhx9 knock-out mouse model exhibited an embryonic lethal feature (35). Focusing on the specific regulation of Dhx9 in TECs, we generated mice with conditional Dhx9 deletion in TECs (Dhx9 ^flox/flox and Foxn1^Cre Dhx9 ^flox/flox, designated as WT and Dhx9 cKO mice, respectively) ( Figure S1A ). We detected the Cre-mediated specific deletion of Dhx9 in TECs via real time PCR and confirmed the effective knockout of Dhx9 in TECs rather than in thymocytes from Dhx9 cKO mice ( Figure S1B ). Dhx9 cKO mice were born without obvious abnormalities compared to WT littermates. We first evaluated the effect of Dhx9 deficiency in the thymus morphologically and found that Dhx9 cKO mice possessed a more shrunken thymus, embodied in a significant decreased thymus weight and the total cell number of thymocytes ( Figures 1A–C ). We next examined the cytokeratin 8 (KRT8) and cytokeratin 5 (KRT5)- immunofluorescence-stained sections of the thymus from 4-week-old WT littermates and Dhx9 cKO mice. We observed that the medullary regions, highlighted by KRT8 (red), showed a remarkable decreased in Dhx9 cKO mice, whereas the cortical region had an increased tendency in turn ( Figure 1D ). The similar phenotype was shown by the hematoxylin and eosin strain (H&E strain), with a shrunken medulla region in Dhx9 cKO mice, compared with WT littermates ( Figure 1E ). These results showed that Dhx9 might play an important role in the formation of thymic medullary regions. Subsequently, we appraised the endogenous effect of Dhx9 on TEC subsets via flow cytometry analysis. TEC-specific inactivation of Dhx9 heavily reduced the percentage and cell number of CD45^-EpCAM⁺ TECs in Dhx9 cKO mice compared with WT littermates ( Figure 1F ). TECs are subdivided into cTECs and mTECs based on the disparate expression spectrum of specific markers, including Ly51 and UEA1 (2). Flow cytometry analysis unveiled that Dhx9 cKO mice represented a measurable reduced percentage of mTECs but increased the percentage of cTECs ( Figure 1G ). More conspicuously, the cell number of mTECs in Dhx9 cKO mice was cut down to one-tenth of those in WT littermates, whereas the cell number of cTECs in Dhx9 cKO mice was similar to that of WT littermates ( Figure 1G ). Furthermore, we examined the percentage and cell number of TECs, mTECs and cTECs in 1-week-old WT and Dhx9 cKO mice ( Figures S4A–D ). The results exhibited a significant reduced percentage and cell number of TECs ( Figures S4A, B ) and mTECs ( Figures S4C, D ), but not of cTECs, which was similar with the altered phenotypes detected in 4-week-old Dhx9 cKO mice. These results collectively indicated that Dhx9 mignt regulate TEC development intrinsically and the obvious thymic atrophy in Dhx9 cKO mice was mainly attributed by the hindered development of mTECs.

We were also interested in the effects of Dhx9 on the differentiation and maturation of mTECs. The competent mature mTECs are capable of expressing a variety of TRAs randomly and facilitating the establishment of the central immune tolerance by eliminating self-reactive thymocytes (5, 36). In contrast with immature mTECs, the maturation process of mTECs is accompanied by elevated expressions of several key genes, such as CD40, CD80, and MHCII (37, 38). Pursuantly, we scrutinized the expression of CD40, CD80, and MHCII on mTECs. As shown in Figure 1H , TEC-specific Dhx9 deletion significantly accelerated the maturation process of mTECs, as illustrated by the remarkable increasing proportion of CD40⁺, CD80⁺, and MHCII^hi mTEC populations ( Figures 1H and S4E ). According to MHCII and CD80 expressed on mTECs, two diacritical cell subsets were generally defined as MHCII^loCD80^- (mTEC^lo) and MHCII^hiCD80⁺ (mTEC^hi) ( Figure S2 ) (3). Flow cytometry analysis exhibited an apparent increased proportion of mTEC^hi in Dhx9 cKO mice compared to WT littermates ( Figures 1H and S4F ). Previous studies have proved that Aire, as an indispensable transcriptional modulator for promiscuous gene expression of TRAs, is primarily expressed in the highly maturated mTEC population (39–41). Consistently, the percentage of intermediate mature Aire^-CD80⁺ and mature Aire⁺CD80⁺ mTECs from Dhx9 cKO mice increased significantly compared with WT littermates, whereas the change of unmatured Aire^-CD80^- mTECs exhibited a contrary tendency when Dhx9 was deleted in TECs ( Figures 1I and S4G ). However, the cell number of mature mTECs marked by CD40, CD80, MHCII, and Aire reduced significantly in Dhx9 cKO mice ( Figure S3 ) because of the decreased total TEC cell number in these mice. Thus, Dhx9 deficiency in TECs accelerated the maturation process of mTECs noticeably.

MTECs and cTECs collectively provide a specialized thymus microenvironment for positive selection and negative selection to support the development of an immunocompetent and self-protective T cell pool (31, 32, 42). We next analyzed the thymopoiesis in the thymus from 4-week-old WT and Dhx9 cKO mice. Distinguished by the expression of CD4 and CD8, thymocytes in Dhx9 cKO mice showed a significant decreased percentage of CD4⁺CD8^- (CD4SP) and CD4^-CD8⁺ (CD8SP) T cells, coupled with an increased percentage of CD4⁺CD8⁺ T cells (DP) compared to WT littermates ( Figures 2A, B ). Together with an overall decreased cell number of DN, DP, CD4SP, and CD8SP T cells in Dhx9 cKO mice ( Figure 2C ), these results demonstrated that TEC-specific Dhx9 inactivation hindered the intra-thymic T cell development. The expression patterns of CD69 and TCRβ describe a differentiation process from DP into SP thymocytes (31, 43, 44). This process was impaired in Dhx9 cKO mice, with a decreased percentage of CD69⁺TCRβ^hi thymocytes, in line with the reduced percentage and cell number of SP thymocytes ( Figures 2D, E ). As for the clonal deletion of Helios⁺PD-1⁺CCR7^-Foxp3^- self-reactive thymocytes during negative selection (45), there was no significant difference between WT and Dhx9 cKO mice ( Figures 2F, G ).

Before the export of the negative-selected SP thymocytes to peripheral T cell pool as recent thymic emigrants (RTEs), the non-fully matured SP thymocytes further undergo post-selection maturation, including the downregulation of both CD24 and CCR7, and the upregulation of CD62L (30, 46, 47). Further analysis demonstrated a blocked CD4SP maturation process, with a significant decreased percentage of mature TCRβ⁺CD62L⁺CD24^-CD4SP thymocytes in Dhx9 cKO mice and an increased proportion of the immature subset, TCRβ⁺CD62L^-CD24⁺CD4SP ( Figures 2H, I ). Similarly, the downregulation of CD24 and CCR7 during the CD8SP maturation process was also blocked in Dhx9 cKO mice, in contrast with WT littermates ( Figures 2J, K ). Moreover, TECs also critically support the development of nTregs from self-reactive CD4⁺ SP thymocytes whose TCR signal strength is not enough to be clonal deleted and drives the development of CD25⁺Foxp3^- nTreg precursors into mature CD25⁺Foxp3⁺ nTregs (48). Dhx9-deleted TECs lead to a decreased frequency of CD4⁺Foxp3⁺ nTregs ( Figures 2L, M ) and the lower ratio of CD25⁺Foxp3⁺ nTreg-to-CD25⁺Foxp3^- nTreg precursors ( Figures 2N, O ), indicating that the maturation of nTregs was hindered in the thymus of Dhx9 cKO mice.

With the application of single-cell genomic omics and fate mapping technologies, the conception of mTEC heterogeneity expands in terms of molecular function and development (7). IL25⁺ thymic tuft cells are a newly defined terminally differentiated mTECs regulated by Pou2f3, which are similar to the molecular characteristics of the gut tuft cells (10, 11). It has been reported that thymic tuft cells express MHCII at a lower level than mTEC^hi (9). In line with this report, our RNA-seq analysis showed that tuft cell-associated gene set mainly expressed in mTEC^lo ( Figures 3A, B ). We found that the deletion of Dhx9 in mTEC^lo led to a general decreased tuft cells-associated gene set, and the downregulated representative genes included Lrmp, Avil, Trpm5, Dclk1, L1cam, Il25, and more ( Figure 3A ), among which Il25 played an important role in regulating the development of iNKT2 in the thymus (10, 49). We also noticed that mTEC^lo in Dhx9 cKO mice exhibited a decreased TPM (transcripts per million) value of Pou2f3 in RNA-seq data ( Figure 3C ). In addition, the genes involved in the taste transduction pathway were strikingly downregulated in mTEC^lo from Dhx9 cKO compared with WT littermates ( Figure 3D ). Based on this relevant analysis, we further investigated the effect on the development of thymic tuft cells when Dhx9 was specifically knocked out in TECs. The percentage of DCLK1⁺ thymic tuft cells exerted a significant reduction in Dhx9 cKO mice, as expected ( Figure 3E ). It is known that thymic tuft cells are essential for iNKT cell development (50). We analyzed CD1d-restricted iNKTs and its sub-lineages, distinguished by expression PLZF and RORγt. We found that the percentage of CD1d-restricted iNKT cells was significantly decreased in Dhx9 cKO mice compared with WT littermates, leading to more dramatically decreased cell numbers ( Figures 3F, G ). Consistent with earlier studies in Pou2f3^-/-mice (9, 10), further analysis showed reduction frequency and cell numbers in iNKT2 and iNKT17 in the thymus from Dhx9 cKO mice, whereas iNKT1 merely displayed a decreased cell number ( Figures 3H, I ). Furthermore, it was recently reported that CD104⁺CCL21⁺ mTEC^lo were important for the development of iNKTs, especially iNKT1 and iNKT17, in the thymus via IL15 (49). The RNA-seq data showed that the expressions of Ccl21a, CD104, and Il15 in mTEC^lo were comparable between WT and Dhx9 cKO mice ( Figure S5 ), which indicated that Dhx9 deficiency in TECs might not significantly influence the development of CD104⁺CCL21⁺ mTEC^lo. In conclusion, Dhx9 was indispensable for the differentiation of thymic tuft cells by regulating the expression of Pou2f3 and then the impaired thymic tuft cell differentiation in Dhx9 cKO mice led to the retardation of iNKT cell development.

Mature SP thymocytes migrate from the thymus across the thymic blood endothelium to the peripheral T cell pool, where naive T cells retain cell number and diverse repertoire via antigen-independent homeostatic proliferation (51). We next investigated whether the severe thymus atrophy in Dhx9 cKO mice influenced peripheral T cell homeostasis. The percentage and cell number of CD4⁺ and CD8⁺ lymphocytes were significantly reduced in the spleens of 4-week-old Dhx9 cKO mice compared with WT littermates ( Figures 4A–C ). The decreased T cells in spleens can be partially caused by the reduced percentage of the recent thymic emigrants (RTEs), defined as CD4⁺CD62L^hiCD45RB^int splenic lymphocytes ( Figures 4D, E ). Furthermore, according to the expression of CD44, we compared the activated state of T cells between WT and Dhx9 cKO mice. We found that Dhx9 cKO mice showed a significantly decreased percentage of CD44^-CD62L⁺ naïve CD4⁺ and CD8⁺ T cells, whereas the percentage of CD44⁺ activation phenotype in both CD4⁺ and CD8⁺ T cells increased in Dhx9 cKO mice ( Figures 4F–I ). Treg cells showed an increased percentage in Dhx9 cKO mice, but the cell number decreased significantly ( Figure 4J ). Based on these data, we concluded that Dhx9 deficiency in TECs disturbed the homeostasis of peripheral T cell pool.

Severe atrophic thymus and disrupted T cell homeostasis in Dhx9 cKO mice reminded us to observe whether Dhx9 cKO mice developed an autoimmune phenotype spontaneously. As an important indicator of autoimmune disease, lymphocytic infiltration in multiple organs from 8-month-old WT, and Dhx9 cKO mice were measured by hematoxylin and eosin (H&E) straining. The conjecture was confirmed by our observation that lymphocytic infiltration in the liver, kidney, salivary, and colon from Dhx9 cKO mice were more obvious than in WT littermates ( Figure 5A ). Furthermore, according to the published lymphocytic infiltration scoring criteria, we found the majority of tissue in Dhx9 cKO mice possessed a higher inflammatory infiltration score ( Figure 5B ). We also compared the contents of antinuclear antibodies in sera from WT littermates and Dhx9 cKO mice. Brighter green fluorescence in Dhx9 cKO indicated high levels of antinuclear antibodies in the sera of Dhx9 cKO mice ( Figure 5C ).

It is known that the mature mTECs function as special antigen-presenting cells to eliminate self-reactive T cells by expressing a range of TRA profiles (52, 53). The expression of TRAs is mainly controlled by Aire and Fezf2 (54, 55). Our RNA-seq analyses showed that both Aire and Fezf2-dependent or independent TRAs were affected after Dhx9 deletion in TECs ( Figure 5D ). Compared with WT, the RNA-seq data revealed that most TRAs were downregulated significantly in Dhx9 cKO mice ( Figure 5D ). These results suggested that Dhx9 deletion in TECs obviously decreased TRA expression, so it influenced the immune tolerance establishment and finally caused the spontaneous occurrence of autoimmune diseases.

Given the significant reduced cell number of mTECs in Dhx9 cKO mice, we wondered whether Dhx9 inactivation impaired the survival and proliferation ability of mTECs in the adult thymus. To examine this hypothesis, we analyzed TECs from 4-week-old WT and Dhx9 cKO mice with intracellular staining with antibodies against the proliferation marker Ki67 and the apoptosis indicator active caspase 3 ( Figures 6A and S6 ). The results showed a significant reduced percentage of Ki67⁺ mTECs in Dhx9 cKO mice compared with WT littermates ( Figure 6A ), whereas there was no difference in caspase 3⁺ apoptotic mTECs and the expression of cell death-associated genes between WT and Dhx9 cKO mice ( Figures S6A, B ). Therefore, Dhx9 deficiency mainly impaired the maintenance of mTEC proliferation in mice.

To further investigate the regulation of Dhx9 in mTEC proliferation, we performed a RNA-seq analysis of mTEC^lo and mTEC^hi of WT and Dhx9 cKO mice ( Figure S2 ). In consideration of Dhx9 basic biological function, we noticed that the DNA damage signaling pathway was significantly upregulated in both mTEC^lo and mTEC^hi from Dhx9 cKO mice, accompanied by upregulated cell cycle checkpoint pathways ( Figures 6B–D ). The extensive upregulated expression of many DNA damage-associated genes in mTEC^hi of Dhx9 cKO mice was observed, compared with WT littermates ( Figure 6E ), which were involved in the overall process of DNA damage process, as illustrated in Figure 6F . It has been well demonstrated that DNA damage can be detected by damage sensor proteins, such as ATM, ATR, and the Rad17-RFC complex. Additionally, downstream Chk1 and Chk2 Ser/Thr kinases initiate signal transduction cascades, which induce cell cycle arrest during G₁ to S, DNA replication, or G₂ to mitosis phase (56). The upregulated p53 signaling pathway in Dhx9 cKO mice transcriptionally induced the expression of the cyclin-dependent kinase inhibitor cdkn1a, which facilitated DNA repair by cell-cycle exit at the G1 phase and also participated in G₂/M cell cycle arrest (57–59) ( Figures 6B, E ). Moreover, cell cycle regulating kinases Wee1 and Myt1, responsible for inhibitory Cdk1 phosphorylation during DNA damage-caused G₂ cell cycle arrest (60, 61), were also upregulated in Dhx9 cKO mice compared with WT littermates ( Figure 6E ). Modulated expressions of these genes likely contributed to the poor mTEC proliferation in Dhx9 cKO mice. The effect of Dhx9 deficiency on cell-cycle distribution was examined by flow cytometric analysis with propidium iodide (PI) staining. The results showed the decreased G0/G1 phase and accumulated S and G2/M phases in Dhx9-deleted TECs compared with WT TECs. These results suggested that Dhx9 deficiency led to cell cycle arrest, mostly at the S and G2/M phases ( Figure 6G ).

The related studies indicated that cell cycle regulators had a crosstalk with glycolysis (62, 63). It was reported that cyclin D1 could negatively regulate the hexokinase II abundance (64). Xiao et al. found that glucose transporter 1 (GLUT1) acted as a link between glycolysis and proliferation during the progression of prostate cancer (65). The gene set variation analysis (GSVA) showed an extensive downregulation of glycometabolism-related signaling pathways in the mTEC^lo and mTEC^hi of Dhx9 cKO mice compared with corresponding WT TECs ( Figures 6B and S7 , S8 ). It were proved that TECs proliferated extensively, with a high turnover rate during thymus growth (66). Cell proliferation requires many substrates, such as elevated metabolic intermediates (e.g., glucose-6-phosphate, acetyl-CoA, and fructose-6-phosphate) and continued regeneration of cofactors to provide free energy or reduce equivalents for reactions (e.g., ATP, NADPH, and NADH) (67). The retarded glycolysis and oxidative phosphorylation in Dhx9-deleted mTECs might contribute to the poor TEC proliferation.

As illustrated in Figure 1 , we noticed significant accelerated mTEC maturation kinetics in Dhx9 cKO mice. To determine the mechanisms that facilitated this maturation process, we performed a network analysis based on the differential expressed genes between WT and Dhx9 cKO mTECs. The network uncovered the central regulatory role of the p53 signaling pathway, and the enrichment of p53 phosphorylation-related genes in both mTEC^lo and mTEC^hi from Dhx9 cKO mice increased the stability of P53 protein through posttranslational modifications ( Figures 7A, B ) (68, 69). Except for upregulation of cell cycle arrest, the network analysis showed genes included in glycolysis and OXPHOS were significantly downregulated in Dhx9-deficient mTECs, and it was consistent with the existing opinion that p53 negatively regulated cellular metabolism ( Figures 6B , 7A ) (70, 71). The upregulated genes involved in the NF-κB pathway in Dhx9 cKO mice caught our attention ( Figure 7A ). It is well demonstrated that the activation of the NF-κB pathway plays an important role in the maturation of mTECs, guiding us to focus on the NF-κB pathway in Dhx9-inactivated mTECs. GSEA revealed that the NF-κB pathway was enriched in both mTEC^lo and mTEC^hi of Dhx9 cKO mice ( Figure 7C ). The expressions of many NF-κB pathway-related genes increased in Dhx9-deficient mTEC^hi and to a lesser degree in Dhx9-deficient mTEC^lo ( Figure 7D ). Subsequently, we wondered how p53 regulated the NF-κB pathway in Dhx9-deleted mTECs. The associated gene network reminded us that the regulation might indirectly upregulate the TNFRSF ( Figure 7E ). Previous studies showed that p53 promoted the expression of Tnfrsf11a (RANK) at the transcriptional level (18). We found the expression of CD40 and Tnfrsf11a increased notably in Dhx9-deficient mTECs in comparison with the WT control ( Figure 7F ). We next evaluated the expression of RANK in mTECs from WT and Dhx9 cKO mice via flow cytometry. Consistent with the upregulation of CD40 ( Figure 1H ), the percentage of RANK⁺ mTECs were increased significantly in 4-week-old Dhx9 cKO mice ( Figure 7G ). However, the expression of LtβR had no significant difference between WT and Dhx9 cKO TECs ( Figure 7H ). These data collectively indicated that Dhx9 deficiency activated the positive regulatory loop of the activated NF-κB pathway and enhanced the expression of CD40 and RANK in mTECs.