Thymic Epithelial Cell-Derived IL-15 and IL-15 Receptor α Chain Foster Local Environment for Type 1 Innate Like T Cell Development

Expression of tissue-restricted antigens (TRAs) in thymic epithelial cells (TECs) ensures negative selection of highly self-reactive T cells to establish central tolerance. Whether some of these TRAs could exert their canonical biological functions to shape thymic environment to regulate T cell development is unclear. Analyses of publicly available databases have revealed expression of transcripts at various levels of many cytokines and cytokine receptors such as IL-15, IL-15Rα, IL-13, and IL-23a in both human and mouse TECs. Ablation of either IL-15 or IL-15Rα in TECs selectively impairs type 1 innate like T cell, such as iNKT1 and γδT1 cell, development in the thymus, indicating that TECs not only serve as an important source of IL-15 but also trans-present IL-15 to ensure type 1 innate like T cell development. Because type 1 innate like T cells are proinflammatory, our data suggest the possibility that TEC may intrinsically control thymic inflammatory innate like T cells to influence thymic environment.

In addition to crucial roles of TCR signals for both αβT and γδT cell development, local environment plays important roles in these innate like T cell maturation and differentiation to effector lineages. IL-15 is critical for development of iNKT cells, especially, for the NK1.1 + CD44 + stage 3 and IFNγ-producing Tbet + iNKT1 cells (20)(21)(22)(23). Similarly, γδT cell effector lineages are also controlled by local cytokines. IFNγ-producing γδT1 cells are severely decreased in pLNs in IL-15 or IL-15Rα deficient mice. IL-15 induces γδT1 cell proliferation and survival via upregulating Bcl-xL and Mcl-1 (24,25). An important feature of IL-15 signaling is that IL-15Rα serves as a high affinity IL-15-binding protein to trans-present IL-15 to the IL-15Rβ/γc complex on neighboring cells (26)(27)(28)(29)(30). IL-15Rα mediated trans-presentation of IL-15 promotes NK cells and CD8 T cell homeostasis (26)(27)(28)(29)(30). Interestingly, IL-15Rα deficiency causes severe impairment of stage 3 iNKT1 cell development (6,7). Although it has been reported that radiation-resistant cells in the thymus provide IL-15 and trans-present IL-15 via IL-15Rα to promote iNKT cell development (6,7), the exact cellular source of IL-15 and the cell type(s) that trans-present IL-15 via IL-15Rα have been unclear as the thymus contains many cell types including radiation resistant non-hematopoietic cells and some hematopoietic cells that could also be radiation resistant.
Thymic epithelial cells (TECs) are crucial for thymopoiesis and thymus function to generate a vast repertoire of T cells that are able to perform immune defenses but are also selftolerated. Cortical TECs (cTECs) and medullary TECs (mTECs) localize in discrete regions in the thymus and perform different function (31)(32)(33). cTECs are mainly responsible for positive selection of developing thymocytes expressing functional TCRs capable of recognition of self-peptide/MHC complexes (34)(35)(36)(37). mTECs ensure highly self-reactive T cells are ablated to establish central tolerance via presentation of promiscuously expressed tissue restricted antigens (TRAs) controlled by Aire and Fezf2 (34,36,(38)(39)(40)(41). In this report, we analyzed publicly available databases and revealed that TECs indeed express a variety of cytokine and cytokine receptors at various levels. We demonstrated further that ablation of either IL-15 or IL-15Rα in TECs selectively impaired development and/or homeostasis of iNKT1 and γδT1 cells in the thymus, indicating that TECs not only serve as an important source of IL-15 but also trans-present IL-15 to ensure type 1 innate like T cell development. Our data suggest that possibility that TEC may intrinsically control thymic inflammatory innate like T cells, which may in turn influence thymic environment.

Expression of a Variety of Cytokines and Cytokine Receptors Including IL-15/IL-15Rα by mTECs
To determine the expression of cytokines and cytokine receptors in mTECs, we searched the publicly available Skyline RNAseq database from The Immunological Genome Project (Immgen.org) for mRNA levels in mTEC. mRNAs of many cytokines and their receptors could be detected in mTECs at various levels ( Figures 1A,B). For cytokines, Il7 is expressed at high levels and Il23a is expressed close to high levels ( Figure 1A); Csf1, IL12a, Il15, Il27, Tgfb2, Tgfb3, Tnf, Tnfsf9, and Tnfsf10 are expressed at intermediate levels; Many other cytokines such as Il10, Il12b, il17c, Il1b, Il4, Il33, and several Tnf superfamily members are expressed at low levels; several other cytokines such as Ifng, Il17a, Il17d and Tgfb1were expressed at very low or trace levels. For cytokine receptors, Csf2rb, Ifngr2, Il11ra1, Il13ra1, Il1rn, Il2rg, and Il4ra are expressed at high levels, whereas most cytokine receptors including Il15ra are expressed at intermediate levels and a few of cytokine receptors such as Il22ra2, Csf3r, and Il17rd were expressed between low and trace levels. Compared with different types of immune cells and other stromal cells, mTECs were among the highest expressers of mRNAs for multiple cytokines and cytokine receptors such as Il7, Il10, Il11ra1, Il13, Il15, Il15ra, Il17c, Il20rb, Il23a, Il27, Tnfsf4, Tnfsf9, and Tnfsf15 ( Figure 1C). Thus, mTECs express mRNAs of many cytokines and cytokine receptors at various levels.

Expression of Discrete Cytokines in Murine TEC Subsets
Recently, murine TECs have been defined into 5 subsets based on single cell RNA sequencing analysis (42)(43)(44)(45)(46)(47)(48). To further investigate expression of cytokines and their receptors in TEC subsets, we analyzed scRNAseq data of TECs generated by the Ido Amit group, which had sequenced more TECs than other reports (42). Using the Seurat package approach (49), we could define TECs from 4 to 6 week old mice into 10 populations (Figure 2A). Populations 3, 4, and 8 are Psmb11 + and represent cTECs; populations 2 and 9 are Krt14 + and represent mTEC-I; populations 1, 6, and 7 are Aire + and Fezf2 + and represent mTEC-II; population 5 is enriched with Il25, Pou2f3, and Dclk1 and represents mTEC-IV or Tuft cells; population 0 is the most abundant population that expresses the highest levels of multiple molecules such as H2-ab1, Psmb11, Krt14, Aire, Fezf2, and Dclk1 as well as cytokines and cytokine receptors, although at low frequencies. This population may represent mTEC-III ( Figure 2B). Interestingly, Aire + /Fezf2 + populations 1, 6, and 7 (mTEC-II) also contain high levels and/or frequencies of cytokines/cytokine receptor mRNAs such as Il13, Il23a, Il27, and Tnf. In addition to Il25, mTEC-IV also is the highest Il10 expresser. Although cTECs (populations 3, 4, and 8) contain highest frequencies of Il7 + cells, populations 1, 2, and 9 (mTEC-I/III) contain cells expressing higher levels of Il7 than cTECs. Il15 is expressed at high frequencies in population 1 and its levels appear higher in mTEC populations than cTEC populations, which is consistent with the detection of IL-15 reporter expression in the medulla in the mouse thymus (50). Il15ra is expressed at higher frequencies in populations 1 and 2 of mTECs and populations 3 and 4 of cTECs. However, the expression levels in these mTECs appear higher than in cTECs.
Overall, Aire/Fezf2 + mTECs appear to express multiple cytokines at levels higher than cTECs while cTECs express higher levels of Il7 than mTECs.

Expression of Cytokines and Cytokine Receptors in Human TEC Subsets
Similar to murine TECs, a recent report has found human TECs could also be defined into multiple populations based scRNAseq transcriptomic analysis (51). Human TECs also contain TEC-I -IV populations that mimic their murine counterparts. In addition, human TECs also contain MYOD1and MYOG-expression myoid TEC-myo and NEUROD1-and NEURODG1-expressing TEC-neuro populations ( Figure 3A) (51). We searched the Human Fetal Thymic Epithelium Gene Expression Web Portal (https://developmentcellatlas.ncl.ac.uk/ datasets/HCA_thymus/human_epi/) for cytokines and cytokine receptors and revealed that human TECs also express many cytokine mRNAs at various levels ( Figure 3B). IL15, IL15RA, IL11RA, IL13RA1, IL1R1, IL23A, IL32, IL34, TGF1B1, TNF, and CSF1 are noticeably expressed at intermediate or high levels. Thus, similar to murine TECs, human TECs also expressed various cytokine/cytokine receptors at the mRNA levels.

IL-15Rα Expressed in TECs Selectively
Promoted iNKT1 Cell Development IL-15Rα can trans-present IL-15 to IL-15R to trigger IL-15R signaling (26,27). It has been reported that radiationresistant thymic stromal cells may trans-present IL-15 to promote stage 3 and iNKT1 cell development via enhancing Bcl-2 mediated survival. The data were generated in lethally irradiated IL15Rα −/− mice reconstituted with WT bone marrow cells (6,7). However, these studies did not distinguish the role of TECs, other stromal cells, and radiation-resistant tissue resident macrophages or lymphoid tissue inducer cells. To investigate whether IL-15Rα expressed on TECs has biological consequences, we analyzed TEC-specific IL-15Rα deficient, Il15ra f /f -Foxn1Cre mice. Thymocyte development was not grossly affected in Il15ra f /f -Foxn1Cre mice (Figure 5A).
Similarly, IL-15Rα deficiency in TECs in Il15ra f /f -Foxn1Cre mice did not obviously affect total γδT cell percentages or numbers (Figures 6E,F). However, γδT1 but not γδT17 cells in the thymus were decreased 69.1% in percentages and 70.4% numbers (Figures 6G,H). Thus, IL-15Rα on TECs also selectively promoted γδT1 cell differentiation but appeared dispensable for γδT17 differentiation.

DISCUSSION
It has been long appreciated that TECs control local environment to shape both conventional and innate like T cell development. We analyzed publicly available RNAseq and scRNAseq data and found that TECs, especially mTECs, express mRNAs for numerous cytokines and cytokine receptors such as Il13, Il23a, Il15, and Il27 as well as Il15ra in mouse and/or human. Some cytokines and cytokine receptors including IL-15 and IL-15Rα are single chain molecules. It is conceivable that these molecules could be expressed as biologically functional molecules in TECs if they are properly processed inside these cells. While multiple previous studies have found radioresistant cell derived IL-15 and/or IL-15Rα or have suggested that mTEC-derived IL-15 and/or IL-15Rα are important for iNKT cell, especially iNKT1 cell, development, no TEC-specific ablation of these molecules have been reported (6,7,62). We examined how TEC-specific IL-15 or IL-15Rα deficiency affects T cell, especially innate like T cell, development. We found that ablation of either IL-15 or IL-15Rα in TECs causes significant impairment of iNKT1 and γδT1 cell development in the thymus. Our data reveal that TECs not only serve as an indispensable source of IL-15 but also trans-present IL-15 for proper type 1 innate T cell development. At present, we do not known whether expression of various cytokine and cytokine receptors in TECs is dependent on Aire or Fezf2 and whether they function in TECs as TRAs to ensure T cell central tolerance. Nevertheless, our observations, together with those that mTEC-IV-derived IL-25 promotes iNKT2 development in the thymus (42,43), suggest the possibility that some cytokines and cytokine receptors expressed in TECs may function both as TRAs and biologically active molecules that can exert their canonical biological functions in the thymus to shape local thymic environment to regulate T cell, particularly innate like T cell, development. Further studies are needed to examine whether TEC-specific ablation of IL-15 and IL-15Rα leads to escape the negative selection of T cells reactive to these molecules.
Of note, TEC-deficiency of IL-15 or IL-15Rα does not completely abolish type 1 innate like T cell development. It is possible other cell types such as dendritic cells and macrophages in the thymus may play partially redundant roles with TECs. Interestingly, TEC-specific IL-15 deficiency weakly reduced iNKT17 numbers in the thymus. This observation is consistent with previous reports that injection of IL-15/IL-15Rα complex induced expansion of both thymic iNKT1 and iNKT17 cells in mice (62,63). Thus, TECderived IL-15 also plays an important role for iNKT17 cell development. Of note, our study does not distinguish the role of mTEC and cTEC derived IL-15/IL-15Rα for iNKT1 and γdT1 development as Foxn1Cre ablates genes in both mTECs and cTECs. However, IL-15 appears to be expressed mainly in mTECs and IL-15Rα is expressed at higher levels in mTECs than cTECs (Figure 2). Additionally, it has been found that mTECs are critical for iNKT1 cell development and induction of IL15R signaling by injecting IL-15/IL-15Rα complex into micer is able to overcome mTEC deficiency to promote iNKT1 development (62,63). Similarly, γdT cells differentiate into effector lineages in the medulla (64). Together, these observations support that mTECs provide critical source of IL-15 for iNKT1 and γδT1 cell development.
Although mRNAs encoding many cytokines and cytokine receptors are expressed in TECs, some of them are biologically active only after complex with other molecules. For example, IL-12 and IL-23 that are heterodimers of an IL-12B (IL-12p40) subunit and the IL-12A (IL-12p35) subunit or the IL-23A (IL-23p19) subunit, respectively. Simultaneous expression of both subunits in the same cells would be required for formation of a functional protein. It is intriguing that expression levels among cytokines and cytokine receptors varies drastically in TECs. Il23a is expressed at the highest levels in mTECs. Whether such high levels of expression ensure full deletion of IL-23A reactive T cells, increase the chance of coexpression with IL-12B in some TECs, or IL-23A itself has biological activity in TECs remain to be explored.
The ability of TECs to produce cytokines and trans-presentation of cytokine(s) to shape thymic environment to control innate like T cell effector lineage differentiation/homeostasis in the thymus could have important implications for thymus biology. Despite the importance of the thymus for T cell generation, it undergoes involution or atrophy with advancing age. Thymic involution may contribute to the decline of immune functions, increased infection-induced mortality and morbidity, and autoimmune diseases in the elderly population (65)(66)(67). Although many extrinsic factors that can modulate the course of thymic involution have been identified, none is able to prevent or stop thymic involution. It has been noted that age-associated thymic involution is associated with accumulation of fatty tissue and inhibition of adipogenesis delays thymic involution. Interestingly, adipogenesis is promoted by local inflammation that is negatively controlled by iNKT2 and M2 macrophages but positively controlled by IFNγ and M1 macrophages (68)(69)(70). Given the ability of TEC sublineages to control type 1 and type 2 innate like T cell differentiation and iNKT cells can in turn regulate mTECs and thymic dendritic cells (63,71), it is possible that thymic involution is an intrinsically programmed process encarved in and triggered by TECs (particularly mTECs) via shaping local thymic environment and presence of innate like T cell effector lineages in the thymus. A hypothesis warrants further investigation.
Mice were maintained in a pathogen free facility. All mouse experiments were performed following a protocol approved by the Institutional Animal Care and Use Committee of Duke University.

Expression of Cytokines and Cytokine Receptors From the Immunological Genome Project
Skyline RNAseq database from the Immunological Genome Project (Immgen.org) was searched for mRNA levels of indicated cytokines and cytokine receptors. In the Immunological Genome Project, 34 immune cell types from male and female mice were profiled by RNA-seq. Expression of mRNA was normalized for each cell types with the Z-score method. To visualize the different values among different cell types, the data for each cell were plotted as a heatmap using the pheatmap program (75).

Analyses of Murine TEC scRNAseq Data
Raw counts of scRNAseq data of TECs from 4 to 6 weeks old mice reported by Bornstein et al. (42) were downloaded from GEO Database under the accession number GSE103967. scRNAseq data were pre-processed using the Seurat package (version 3.1.1) (49) in R (version 3.5.3). Genes expressed in fewer than 3 cells and cells with no more than 50 detected genes were filtered out. Filtered datasets were normalized the gene expression measurements for each cell by the total expression multiplied with a scale factor of 10,000 by default, followed by log-transformation of the results using the globalscaling normalization method, LogNormalize. The technical noise and/or biological sources of variation were mitigated via ScaleData function to improve downstream dimensionality reduction and clustering. Highly variable genes were screened with Find Variable Features function for downstream analysis. Principle component analysis (PCA) were performed on the scaled data using the RunPCA function. Significant PCs were identified as those with a strong enrichment of low p-value genes based on the Jackstraw algorithm. For cell clustering, k-nearest neighbors were calculated and the SNN graphs were constructed using Find Neighbors. Top 20 PCs were selected for analysis using Find Clusters. Cells within the graph-based clusters determined above were co-localized for visualization on the tSNE plot via RunTSNE and TSNEPlot. Find All Markers were applied to find markers that define clusters via differential expression. Feature Plot was applied to visualize individual gene expression on a tSNE plot. VlnPlot was applied to show expression probability distributions across clusters.

Analyses of Human TEC scRNAseq Data
Expression of cytokines and cytokine receptors in human TECs was searched online based on scRNAseq analyses (https://developmentcellatlas.ncl.ac.uk/datasets/HCA_ thymus/human_epi/) (51). Data were presented as a bubble plot with bubble size representing percentages of TECs expressing individual molecules and bubble color representing expression levels.

Statistical Analysis
Data shown represent means ± SEMs and were analyzed with the two-tailed pairwise Student t-test using the Prism 5/GraphPad software for statistical differences. Each pair of mice represents sex-matched littermates and is indicated by a connecting line between test and control mice. Pvalues < 0.05 were considered significant ( * p < 0.05, * * p < 0.01).

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The animal study was reviewed and approved by Institutional Animal Care and Use Committee of Duke University.

AUTHOR CONTRIBUTIONS
HT and LL designed and performed experiments, analyzed data, and participated manuscript preparation. N-SL, KS, and SL provided critical reagents and participated in manuscript preparation. X-PZ conceived the project, designed experiments, and wrote the manuscript. JS participated in manuscript preparation. All authors contributed to the article and approved the submitted version.

FUNDING
This study was supported by the National Institutes of Health (R56AG060984 and R56AI079088 for X-PZ and R01AI131609 for SL) and by the Translating Duke Health Initiative.