Immune landscape reclassifies human TETs.
It is reported that lymphocyte infiltration level in tumor was correlated with the histological classification of TETs for decades 13. However, details about the immune landscape of human TETs need to be decoded. To this end, we used mass cytometry, single-cell RNA sequencing, TCR profiling, FCM, HE and IF staining to investigate the immune landscape and its underlying mechanism of human TETs (Fig. 1A). Firstly, we established a panel of more than 40 markers to uncover the immune landscape of human thymus and TETs by CyTOF (Supplementary Table S1). Three normal thymus and twelve TET tissues were performed mass spectrometric flow tests (Supplementary Table S2). Single cell analysis showed that the cell composition of TETs was not only obviously different from normal thymus tissues, but also had a significant internal difference among tumors (Fig. 1B). Importantly, we found the tumor microenvironment of TETs, defined by cellular composition, can be further classified into three types (type1, 2, 3) (Fig. 1B). Furthermore, each type of human TETs showed a unique histological morphology and lymphocytes infiltrating pattern (Fig. 1C).
Next, we annotated the composition of immune cells in TETs and normal thymus (Fig. 1D and 1E). The immune cell compartment of TET samples similar to normal thymus comprised major immune lineages (Fig. 1D and Supplementary Fig. S1A), among which CD3−CD4+CD8+ and CD3+ T cells were most abundant (Supplementary Fig. S1B). Through inter-group comparisons among normal thymus tissue and each type of TETs, we found obvious differences in the major cellular components between groups, especially in CD3−CD4+CD8+ and CD3+ T lymphocytes (Fig. 1F, 1G, and Supplementary Fig. S1C). The proportion of CD3−CD4+CD8+ lymphocytes was highest in type 1 and lowest in type 3 (Fig. 1G). Moreover, from type 1 to type 3, there was an obvious decreasing tendency of CD3− CD4+ CD8+ lymphocytes in tumor (Fig. 1G). In contrast, the proportion of CD3+ T cells, B cells, and NK cells in tumor was lowest in type 1 and highest in type 3, which had a gradual increasing tendency among three types (Fig. 1G and Supplementary Fig. S1C). Monocytes and granulocytes were more abundant in tumor of type 3, and the proportions of DCs in tissues of both type 2 and type 3 were more abundant than normal thymus and type 1 (Fig. 1G and Supplementary Fig. S1C). It is interesting that except CD3−CD4+CD8+ lymphocytes, other immune cells which were related to peripheral immune response in tumor were more abundant in type 3 TETs than that in type 1 and type 2 (Fig. 1G and Supplementary Fig. S1C). In addition, the discrepant abundance of CD3−CD4+CD8+ lymphocytes and CD3+ T cells in tumors among three types of TETs were further validated by flow cytometry (Fig. 1H). Our results showed that the immune landscape of tumor dramatically changed during the development of TETs. Moreover, these findings proposed that human TETs could be reclassified according to the change of immune landscape in tumor.
T cell developmental pattern of TETs.
The discrepant amount of CD3−CD4+CD8+ and CD3+ T lymphocytes among tumors suggested that there was a discriminatory T cell developmental pattern in each TETs type. In order to further investigate the change of T cell development in TETs, we used well accepted markers, which were highly correlated with T cell development in thymus 21, to annotate the cell subsets of CD3+ T cells in tumor and normal thymus (Fig. 2A, Supplementary Fig. S2A and S2B). Major subsets of T cells involved in thymic T cell development were represented in the dataset 12, including double negative (DN, CD4−CD8−), double positive (DP, CD4+CD8+), CD4+ single positive (CD4+ SP), CD8+ single positive (CD8+ SP), and FOXP3+ regulatory T (Treg) cells (Fig. 2A). Phenotypic analysis showed that T cell subsets in tissues of thymus and tumors which we defined were representative (Fig. 2B). We found the T cell composition in tumor of TETs was observably different from that in normal thymus (Fig. 2C and 2D). Importantly, the T cell compositions in tumor were significantly different among TETs types (Fig. 2C and 2D). Among tumor types, immature DP cells dominated in type 1, while mature SP cells dominated in the tumor of type 2 and type 3 (Fig. 2D and 2E). Further analysis showed that the proportion of DN and DP cells which were at the early stage of T cell development in tumor was the highest in the tumor of type 1 and lowest in type 3 (Fig. 2E and Supplementary Fig. S2C). It was interesting that, the proportion of DN and DP cells in tumor of TETs, especially DP cells, showed a tendency to decrease from type 1 to type 3 (Fig. 2E and Supplementary Fig. S2C). On the contrary, the proportion of CD4+ and CD8+ T cells which were at the late stage of T cell development was the lowest in tumor of type 1 and highest in type 3 (Fig. 2E). Moreover, the proportion of CD4+ and CD8+ T cells in tumor of TETs showed a gradual increasing trend from type 1 to type 3 which was contrary to DN and DP cells (Fig. 2E). Flow cytometry detection further confirmed the discriminatory amount of DN, DP and SP cells in each TETs type (Fig. 2F). Furthermore, immunofluorescence staining showed that DP and SP cells were located in the cortex or medulla like structure in tumor of type 2 TETs respectively. However, the typical cortex and medulla like structures were not observed and DP and SP cells were distributed in the stroma of tumor in type 1 and type 3 TETs disorderly (Fig. 2G).
The proportion of naïve SP cell (naïve CD4+ and naïve CD8+ cell) and memory SP cell (memory CD4+ and memory CD8+ cell) which have been fully-developed was lowest in type 1 and highest in type 2 TETs (Fig. 2E and Supplementary Fig. S2C). Interestingly, the tissue resident CD4+ T (CD4+ TRM) cells, tissue resident CD8+ T (CD8+ TRM) cells and Treg cells, which generally accumulated in peripheral tissues 22, showed a higher tendency in the tumor of type 3 TETs than the other tumor types (Supplementary Fig. S2C). Besides, there is no significant bias in the ratio of CD4+ and CD8+ T cells among tumor types (Supplementary Fig. S2C). Taken together, these findings uncovered the unique pattern of T cell development in each type of tumor which further supported our reclassification and suggested that T cell developmental dysfunction occurred in the tumors of human TETs.
Dysfunction of T cell development in TETs.
To further uncover the T cell developmental dysfunction in tumor of human TETs, we performed scRNA-seq and TCRαβ profiling on six tumor samples which had been detected the immune landscape by mass cytometry using the droplet-based 10x Genomics platform (Fig. 1A). The data of a normal thymus sample published by Jong-Eun Park et al. was also reanalyzed as normal control 21. After quality control including doublet removal, a total of 52,788 cells from tumors and 2,845 cells from normal thymus were included in our study (Supplementary Table S3). Uniform manifold approximation and projection (UMAP)-based clustering analysis of our samples resulted in 26 cell clusters (Supplementary Fig. S3A). Based on known lineage markers, we annotated cell clusters into nine major cell types (Supplementary Fig. S3B-S3E). These cell types included T cell (CD3E, CD3D), B cell (CD19, MS4A1, CD79A), epithelial cell (EPCAM, FOXN1), dendritic cell (DC) (ITGAX, CD86, CLEC9A, LILRA4), macrophage (CD14, C1QA), monocyte (CD86, CD14, S100A9), endothelial cell (CDH5, PECAM1), fibroblasts (PDGFRA) as well as vascular smooth muscle cell (VSMC) (ACTA2, RGS5) (Supplementary Fig. S3C-S3E, Supplementary Table S4). Consistent with the results of CyTOF, cellular atlas showed that T cells dominated in the discrepancy of cell types between tumor and normal tissue as well as among tumor types of TETs (Supplementary Fig. S3F and S3G).
In order to further investigate the T cell developmental dysfunction in TETs, T cells were re-clustered into 18 cell clusters (Supplementary Fig. S4A). These cell clusters were further annotated (Fig. 3A), according to the marker genes related to T cell development in thymus (Supplementary Fig. S4B and 3B). We defined the major lymphocyte subsets involved in thymic T cell development, including DN, DP, αβT (entry), SP, CD8+ resident memory T (CD8+ TRM) (ZNF683, ITGAE, XCL1, XCL2), Treg (FOXP3, IL2RA), NKT (NKG7, KLRB1, ZBTB16) and Th-like (STAT3, STAT4, RORA, AHR) cells (Fig. 3B, Supplementary Table S5). Previous CyTOF results showed there were two groups (CD3+ and CD3−) of DP cells in tumor of TETs (Fig. 1D and 2A). However, solely based on the transcriptional level of CD3 expression we were unable to define the CD3− subset of DP cells. To solve this divergence, we further annotated DP cells into proliferating (P) and quiescent (Q) subsets according to the expression level of MKI67 and CDK1 as previous study 21 (Supplementary Fig. S4C and S4D). Consistently, the marker genes of CD3− DP cells, which were obtained from the protein marker-based sorting strategy in another study 23, were obviously highly expressed in DP (P) cells (Supplementary Fig. S4E). These results suggested a close match between the DP (P) cells defined by scRNA-seq and CD3− DP cells detected by CyTOF in our study.
Previous studies reported that thymic T cell development started from CD4−CD8− DN cells, which gradually expressed CD4 and CD8 to become CD4+CD8+ DP cells, and then transitioned through a CCR9high αβT (entry) stage to diverge into mature CD4+ or CD8+ SP cells 21. Inter-group comparative analysis in our study showed that the composition of T cell subsets involved in T cell developmental stage was significantly different among tumor types of TETs (Fig. 3C-3E). Specifically, DP cells, which were in the early developmental stage, constituted the largest T cell subset in type 1, but least in type 3 (Fig. 3D and 3E). In contrast, SP cells, which were in the late stage of T cell development, constituted the largest T cell subset in type 3, but least in type 1 (Fig. 3D and 3E). Compared with normal thymus, the proportion of DP cells increased in type 1 TETs but decreased in type 2 and almost absent in type 3 (Fig. 3D and 3E). In contrast, SP cells were significantly increased in type 2 and type 3 TETs (Fig. 3D and 3E). It was interesting that naïve T cells were more abundant in the tumor of in type 2 than type 3 TETs, whereas memory T cells were more abundant in the tumor of in type 3 than type 2 TETs (Fig. 3D and 3E). These findings indicated that T cell developmental dysfunction in the tumor of each TETs type was different.
To further decode the details of developmental dysfunction of T cells in tumors, we performed trajectory analysis of the T cell subpopulations defined previously (Fig. 3F). Consistent with previous findings, trajectory analysis showed an obvious T cell developmental dysfunction in each type of TETs compared with normal thymus (Fig. 3G). Through inter-group comparison, we found T cells in type 1 TETs were enriched in the early stage of development, while T cells in type 2 TETs were abundant in both early stage and late stage of development (Fig. 3G). Interestingly, almost all T cells were concentrated in the late stage of development in type 3 TETs (Fig. 3G). The expression levels of Notch and Wnt signaling pathways related genes, which regulated the early stages of T cell development 24 25, were higher in the T cells isolated from tumor of type 1 TETs than type 2 and type 3 TETs (Fig. 3H and 3I). In contrast, IL-7 signaling related molecules, which could drive intrathymic expansion of positively selected thymocytes prior to their export to the peripheral T cell pool 26, were expressed higher in the T cells isolated from type 2 TETs than thymus and other tumor types (Fig. 3J). These results further demonstrated that T cell development was restrained in early stage in type1, increased in type 2 and almost absent in type 3 of human TETs.
Bias of TCR repertoire and diversity in TETs.
To investigate TCR repertoire and clonotype of T cells in tumor of TETs, TCR chains detected from the TCR-enriched 5′ sequencing libraries were filtered for full length recombinants and were associated with the cell type annotation. To analyze the usage and pairing of VJ genes in T cells, we ranked VJ genes according to genomic positions as previous study 21. For TCRβ, we found an obvious bias in VJ gene usage among tumor types of TETs (Fig. 4A), which resembled VJ gene usage of DP and SP cells respectively (Supplementary Fig. S5A). Consistently, VJ pairs of T cells were also significantly different among tumor types of TETs (Fig. 4B), and consistent with DP and SP respectively (Supplementary Fig. S5B). As for TCRα locus, we found a clear association between developmental timing and V-J pairing, as previously described 21 27. During the T cell developmental stage, recombination of proximal pairs on TCRα was mainly observed in DP stage, whereas distal pair recombination was mainly observed in SP stage (Supplementary Fig. S5C). This bias in recombination in turn restricted the V-J pairing of TCRα in DP and SP cells (Supplementary Fig. S5D). Consistently, we observed that the VJ gene usage and pairing of TCRα in each tumor type were in a similar pattern as DP and SP cells respectively (Fig. 4C and 4D).
Furthermore, we found the degree of expansion observed among the clonotypes, which were the number of cells sharing the same clonotype in the dataset, was strongly associated with the T cell developmental states defined by scRNA-Seq (Fig. 4E). It was worth noting that clonotype amplification was mainly observed in mature SP cells, but less in immature DN and DP cells (Fig. 4E). This also indicated that immature T cells (DN and DP cell) had more abundant clonal diversity than mature T cells (SP cell), which was consistent with the finding of the least clonal diversity and the most SP cells in type 3 TETs (Fig. 4F). Together, these findings uncovered the consequence in TCR repertoire and diversity of T developmental dysfunction in each tumor type.
Epithelial origin decodes T cell developmental dysfunction in TETs.
Thymic T cell development was orchestrated by heterogeneous TECs 10. To further investigate the epithelial origin of TETs and its impact on T cell development in TETs, the subsets of epithelial cells (tumor cells) in thymus and tumors were further analyzed (Fig. 5A and Supplementary Fig. S6A). UMAP-based clustering analysis showed a different composition of epithelial cells in tumor of human TETs, reflected the heterogeneity of tumor cells among different types (Fig. 5A). The epithelial cells were most abundant in type 3 TETs among all tumor types, while similar in type 1 and type 2 (Fig. 5A and 5B). Gene expression analysis showed the epithelial cell in type 3 expressed high level of cancer stem cell (CSC) related markers (NOTCH2, CD24, ALDH1B1 and PROM1) (Fig. 5C). At the protein level, the epithelial cell in type 3 TETs expressed higher level of Ki67 further suggesting its malignancy (Supplementary Fig. S6B). In addition, trajectory analysis showed the epithelial cells in type 1 and type 2 TETs were at different developmental stages compared with type 3 TETs (Fig. 5D).
In order to further uncover the heterogeneity of tumor cells among TETs subtypes, epithelial cells in tumors of TETs were re-clustered into 17 cell clusters (Supplementary Fig. S6C). The cellular clusters were annotated into seven subsets based on a combination of known TEC markers, including cTEC-like cells (PRSS16, CCL25, ZBED2), KRT14+ mTEC-like cells (KRT14), CHI3L1+ TEC-like cells (CHI3L1), GNB3+ mTEC-like cells (POU2F3, GNAT3 and GNB3), MYOG+ TEC-like cells (MYOG), CHGA+ TEC-like cells (CHGA) and mcTEC-like cells (PRSS16, CCL25, KRT14 and DLK2) (Fig. 5E and 5F, Supplementary Fig. S6D, Supplementary Table S6). Based on this definition, we found that CTSL, encoding the proteases cathepsin L1, and PRSS16, encoding thymus-specific serine protease which was reported to play an important role in thymic T cell positive selection 28, were mainly expressed in cTEC-like cells and mcTEC-like cells in TETs (Supplementary Fig. S6E). The genes, FEZF and AIRE, responsible for thymic T cell negative selection were mainly expressed in CHI3L1+ mTEC-like cells, MYOG+ TEC-like cells, GNB3+ mTEC-like cells and CHGA+ TEC-like cells (Supplementary Fig. S6E). Inter-group comparison showed the composition of tumor cell subsets in TETs subtypes was significantly different (Fig. 5E and 5G). Besides, our results revealed the major tumor cell subsets in each type of TETs, such as KRT14/GNB3+ mTEC-like cells in type 1, CCL25+ cTEC-like cells in type 2 and CHI3L1+ mTEC-like cells in type 3 (Fig. 5E and 5G). The main tumor cell subsets in each TETs type were further confirmed by immunofluorescence (Fig. 5H). Interestingly, we found the mcTEC-like cells presented in both type 1 and type 2 tumors had more molecular characteristics of thymic epithelial progenitor cells (TEPCs) compared to KRT14+ mTEC-like and CCL25+ cTEC-like cells 28 (Supplementary Fig. S6F), indicating KRT14+ mTEC-like and CCL25+ cTEC-like cells might be differentiated from mcTEC-like cells. Through pathway analysis, we found that NF-κB signaling pathway of mcTEC-like cells in type 1 was obviously activated, while that in type 2 was inhibited (Supplementary Fig. S6G). This meant the differentiation of mcTEC-like cells in type 1 was more inclined to be mTEC-like cells, while the differentiation of mcTEC-like cells in type 2 was more inclined to be cTEC-like cells, which was consistent with the result that mTEC-like cells increased in type 1 and cTEC-like cells increased in type 228 (Fig. 5G). Moreover, immunofluorescence staining result showed the AIRE expression in type 2 TETs was significantly less than the other two types (Supplementary Fig. S6H), which was correlated with the incidence of autoimmune disease in TETs 13.
CXCL12 and CCL25, the key cytokines promoting homing of blood-borne lymphoid progenitor cells into thymus 28, were almost not expressed in CHI3L1+ mTEC-like cells (Fig. 5I), the main component of the tumor cells of type 3 TETs. This observation suggested the tumor cells of type 3 TETs lost the function for recruitment of thymus progenitor cells, resulting in the T cell developmental dysfunction in initial period. Having identified the major TEC-like tumor cells in TETs subtypes, we used CellPhoneDB analysis to investigate the interactions between TECs-like tumor cells and T cells as previous study 29. T cell development in thymus was a complex process involving TEC-lymphocyte cell interactions, lymphocyte cell migration and lymphocyte cell localization 30. Here, we focused on the tumor cells and lymphocyte cell interactions mediated by chemokines, which enabled lymphoblast migration and anatomical localization 30. Our results demonstrated that mTEC-like and mcTEC-like tumor cells in TETs could induce DP cells migration for positive selection through CCL25:CCR9 (Fig. 5J, Supplementary Table S7) and CXCL12:CXCR3 interaction (Supplementary Fig. S6I, Supplementary Table S7). Interestingly, cTEC-like tumor cells also had a crucial role in inducing SP cell migration through CCL19:CCR7 interaction (Fig. 5J), which differed from the cTEC in normal thymus 31.
To further illustrate the differences of T cell development among TETs subtypes in molecular level, we used Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis to discover the difference of signaling pathway enrichment in DP cells among tumor types. Compared with type 1 TETs, the genes high-expressed in DP cells of type 2 were significantly enriched in signaling pathways related to T cell expansion and positive selection, including Oxidative phosphorylation, NF-κB signaling pathway, Th17 cell differentiation, Th1 and Th2 cell differentiation and TNF signaling pathway 28 32 (Fig. 5K). In the process of positive selection, most cells would be eliminated by apoptosis, so the enrichment of highly-expressed genes in apoptosis-related pathways also indicated that more DP cells in type 2 were undergoing positive selection (Fig. 5K). Compared with normal thymus tissue, genes high-expressed in DP cells of type 2 were also enriched in the relevant signal pathways for positive selection (Supplementary Fig. S6J). Moreover, we analyzed the expression of marker genes which indicated the pre- or ongoing state for positive selection of DP cells 32. We found that DP cells in type 1 TETs mainly expressed pre-selection related genes, while DP cells in type 2 TETs highly expressed genes related to ongoing state of positive selection (Fig. 5L). These results indicated the positive selection of T cell development in type 2 TETs was promoted by tumor cells, while blocked in type 1. This was further supported by the observation that the expression level of HLA-DR molecule of tumor cells in type 1 TETs was lower than other types (Supplementary Fig. S6K). Taken together, these findings demonstrated that unique epithelial origin of each type of TETs led to a different dysfunction of epithelium and T cell interaction and consequence in the discrepancy T cell developmental dysfunction among tumor types.
Tumor cells induce CD8 + TRM cell-mediated immune response in type 3 TETs.
Among three types of TETs, it was interesting that the thymic T cell development seemed absent in type 3 TETs, whereas mature T cells were still abundant in the mesenchyme of tumor tissue (Fig. 2C-2E). A possible reason for this phenomenon was the infiltration of immune cells induced by malignant cells of cancer, which could be commonly observed 33. We confirmed that the CD8+ TRM (CD103+CD69+) cells, playing a central role in immune sensing network of peripheral tissues especially in the recruitment of various types of immune cells when they are activated 34, were significantly enriched in type 3 TETs (Fig. 6A and 6B). The unique abundance of CD8+ TRM cells in type 3 TETs was further confirmed by flow cytometric analysis (Supplementary Fig. S7A). Compared with other epithelial subpopulations, CHI3L1+ mTEC-like cells, which were most highly enriched in type 3 TETs, were predicted to have a largest number of interactions with CD8+ TRM cells (Fig. 6C and Supplementary Table S7). Some specific ligand-receptor interactions between CHI3L1+ mTEC-like tumor cells and CD8+ TRM cells were identified, including TNFSF9:HLA-DPA1, CCL20:CXCR3, CEACAM5:CD8A (Fig. 6D and Supplementary Table S7). Consistently, CD8+ TRM cells in type 3 TETs expressed higher level of activation and proliferation related molecules, including CD38, CD39, Ki67 and CD137, while the inhibitory molecule was decreased (Fig. 6E and Supplementary Fig. S7B). Furthermore, genes related to activation and TCR signaling at the transcriptional level were also significantly upregulated in CD8+ TRM cells compared with other types of T cells in the tumor of type 3 TETs (Fig. 6F). Pathway analyses revealed the upregulated gene of CD8+ TRM cells significantly enriched in TCR signaling pathway, antigen processing and presentation, cytokine-cytokine receptor interaction and chemokine signaling pathway (Fig. 6G). Further analysis demonstrated CD8+ TRM in tumor of type 3 TETs could potentially recruit T cells, B cells, DCs and macrophage via CXCL12:CXCR4, CCL21:CCR7, CXCL13:CXCR5, XCL1/2:XCR1 and CX3CL1:CX3CR1 interactions, respectively (Fig. 6H and Supplementary Table S7). IF staining further validated the co-localization of epithelial cells and CD8+ TRM cells, as well as CD8+ TRM cells and DCs, B cells in tumor of type 3 TETs (Fig. 6I and 6J). Our results proved that CHI3L1+ mTEC-like tumor cell induced CD8+ TRM cell activation playing a central role in the immune response of type 3 TETs.
Epithelial origin reclassifies human TETs.
There were several kinds of histological classification and clinical stage of human TETs for decades 17. However, due to the limitation of current classification systems, novel classification which could perform more easily and have better prognostic predictive effect was urgent needed 18. To this end, we tried to establish a new classification system of human TETs based on our above findings, especially the unique epithelial origin of tumor cells in each type of TETs which played pivotal role in T cell developmental dysfunction and immune response in tumor microenvironment. Firstly, we selected representative genes based on scRNA-seq results which represented the unique epithelial cell type in each TETs types (Fig. 5F) and found that GNB3 and CHI3L1 were expressed highly in tumor of type 1 and type 3 TETs respectively (Fig. 7A and 7B). Consistently, the expression of GNB3 and CHI3L1 in tumor was further validated by IF staining (Fig. 7C). Therefore, we reclassified 119 TETs patients of TCGA cohort into three groups according to the expression level of GNB3 and CHI3L1 (Type 1, GNBhigh; Type 2, GNBlowCHI3L1low; Type 3, GNBlowCHI3L1high) (Supplementary Fig. S8A). Kaplan-Meier survival analysis demonstrated that the classification we established closely correlated with survival of TETs patients (Fig. 7D; P = 0.014). Moreover, it was surprising that our classification based on GNB3 and CHI3L1 expression showed an advantage in the prognostic prediction of TETs patients compared with both Masaoka stage and WHO classification (Supplementary Fig. S8B and S8C).
Finally, based on above findings we proposed a tripartite frame to explain the origin of tumor cells and its impact on T cell development in tumor of human TETs as below (Fig. 7E): 1) For the type 1 TETs: risk factors induced the mutation of oncogenes in original TECs, which resulted in the activation of NF-κB signaling pathway in mcTECs-like cells and expansion of KRT14/GNB3+ mTECs-like tumor cells. The expanded KRT14/GNB3+ mTECs-like tumor cells led to a bias of cellar composition and defect of cortex like structure in tumor. The lack of CCL25+ cTECs-like cells and cortex like structure resulted in restraining T cell development, which was mainly in positive selection phase. Consequently, CD8+CD4+ T cells (DP) accumulated, but CD8+ T cells and CD4+ T cells (SP) decreased in tumor. 2) For type 2 TETs: risk factors induced the mutation of oncogenes in original TECs, which resulted in the inhibition of NF-κB signaling pathway in mcTECs-like cells and expansion of CCL25+ cTECs-like cells. The expanded CCL25+ cTECs-like tumor cells led to a bias of cellar composition and defect of medulla like structure in tumor. The accumulation of CCL25+ cTECs-like cells accelerated the positive selection phase of T cell development in tumor. However, lack of KRT14/GNB3+ mTECs-like cells and medulla like structure resulted in suppression of T cell development in negative selection phase. Consequently, CD8+CD4+ T cells (DP) decreased, but CD8+ T cells and CD4+ T cells (SP) increased in tumor. 3) For type 3 TETs: risk factors induced the mutation of oncogenes in original TECs, which resulted in the transforming and expanding of CHI3L1+ mTECs-like cells in tumor. The malignant transformation of original TECs also led to the deficiency of functional mTECs, cTECs, cortex-like structure and medulla-like structure in tumor which resulted in deficiency of T cell development. However, CHI3L1+ mTECs-like tumor cells of type 3 TETs could activate the CD8+ TRM cells and recruit other immune cells to form a unique immune microenvironment, such as effector T cells, B cells, DC cells and macrophages.