Specific subsets of urothelial bladder carcinoma infiltrating T cells associated with poor prognosis

Comprehensive investigation of tumor-infiltrating lymphocytes in cancer is crucial to explore the effective immunotherapies, but the composition of infiltrating T cells in urothelial bladder carcinoma (UBC) remains elusive. Here, single-cell RNA sequencing (scRNA-seq) were performed on total 30,905 T cells derived from peripheral blood, adjacent normal and tumor tissues from two UBC patients. We identified 18 distinct T cell subsets based on molecular profiles and functional properties. Specifically, exhausted T (TEx) cells, exhausted NKT (NKTEx) cells, Ki67+ T cells and B cell-like T (B-T) cells were exclusively enriched in UBC. Additionally, the gene signatures of TEx, NKTEx, Ki67+ T and B-T cells were significantly associated with poor survival in patients with BC and various tumor types. Finally, IKZF3 and TRGC2 are the potential biomarkers of TEx cells. Overall, our study demonstrated an exhausted context of T cells in UBC, which layed a theoretical foundation for the development of effective tumor immunotherapies.

Sample/single-cell preparation.The PBMC were isolated from the patients using lymphocyte separation medium (Mediatech, USA) according to the manufacturer's protocol.Briefly, the blood samples were collected from the patients who were starved overnight in a purple blood tube containing EDTA anticoagulant (BD, USA).After the dilution with PBS, 6 mL of blood samples were layered carefully on top of the lymphocyte separation medium.After the centrifugation at 500g for 30 min at RT, the interlayers, between the blood plasma and lymphocyte separation medium, which were the PBMCs, were collected and added with 2-3 mL of red blood cell lysis buffer (Solarbio, China) to lyse the red blood cells at 4 °C for 8 min.After the centrifugation at 400g for 10 min at 4 °C, the cells were carefully transferred to a new tube and washed twice with 10 mL RPMI-1640 culture medium (Corning, USA).After the centrifugation, the cells sediment were collected and resuspended in the culture medium.
The UBC tissue and adjacent normal bladder tissues (NBT) were cut into small pieces using a sterilized cutter.Subsequently, both tissues were digested with 2 mL of collagenase IV (Yuanye Biotechnology, China) in a CO 2 incubator for 2 h at 37 °C until homogenized.The cells were filtered with 40 µm cell strainer (Falcon, USA) and followed by the centrifugation at 1000-2000 rpm for 10 min.After discarding the supernatant, the cells were incubated in 4 °C for 8 min with red blood cell lysis buffer to lyse the red blood cells.Then, the cells were washed twice and resuspended in the culture medium.

Magnetic-activated cell sorting of CD3 cells. The magnetic-activated cell sorting (MACS) of CD3 +
T cells was performed using pre-cooled solutions.Firstly, the cells were resuspended in 80 µL PBS buffer and followed by the addition of 20 µL CD3 Microbeads, human (Miltenyi Biotec, Germany) for every 1 × 10 7 cells.After incubation at 4 °C for 15 min, the cells were diluted with 1-2 mL of pre-cooled buffer and followed by the centrifugation at 300 g for 10 min.After resuspension in 500 µL pre-cooled buffer, the cells were transferred to the top of the MS column (Miltenyi Biotec, Germany).The flow-through was collected which contained the CD3 − cells.The MS columns were washed 3 times with 500 µL pre-cooled buffer to complete flow-through of CD3 -cells.After removing MS columns from MACS separator, the remaining flow throughs were collected to a new centrifuge tube, which was the CD3 + cells.The cell number was counted using 0.4% trypan blue (Mediatech, USA) and maintained at a concentration of 7 × 10 5 to 1.2 × 10 6 cells/mL.The cells were retained on ice before proceeding to further experiments.

Immunofluorescent staining/flow cytometry of T cell determination.
After sorting out the CD3 + cells, the determination of T cell populations was performed via flow cytometry detection.Firstly, 1 × 10 6 CD3 + cells were stained with 10 µL of FITC Anti-Human CD3 (OKT-3) Monoclonal Antibody, PE/Cy5 Anti-Human CD4 (OKT-4) Monoclonal Antibody and APC Anti-Human CD8 (HIT8a) Monoclonal Antibody (SunGene Biotech, China) together and incubated for 30 min at 4-8 °C in dark condition.After centrifugation and washing the cells with 500 µL PBS, the stained cells were resuspended with 500 µL of RPMI-1640 medium.The flow cytometry analysis of T cell populations was performed using FACSCalibur (BD, USA) and T cell populations were analyzed using FlowJo software.
ScRNA-seq.The sorted cells were counted and assessed for the viability with 0.4% trypan blue using LUNA-II™ automated cell counter (Logosbio, Korea).The cells were then maintained at a concentration of 7 × 10 5 to 1.2 × 10 6 cells/mL with final viability of > 80% for the subsequent sequencing.The cell suspension was loaded into Chromium microfluidic chips and barcoded with a 10 × Chromium Controller (10 × Genomics).The RNA from barcoded cells was reverse-transcribed.Subsequently, single-cell library preparation was constructed using the reagents from a Chromium Single Cell reagent kit (10 × Genomics) according to the manufacturer's protocols.Sequencing was performed with Illumina (Logosbio, Korea) according to the manufacturer's protocols.
ScRNA-seq data processing.The sequenced data from Illumina sequencer (Logosbio, Korea) were first processed to filter out the low-quality reads via trimmomatic software which can be summarized below: (1) average quality per base drops below 10; (2) "N" bases quality below 3; (3) containing adaptor sequence; (4) droping reads below the 26 bases long and ( 5) not forming paired reads.The remaining reads passed the filtering criteria were counted as clean reads and converted to Fastq format from BCL format for the subsequent analyses.Provided by 10 × Genomics, the Cell Ranger software pipeline (Version 3.0.2) was used to demultiplex cellular barcodes, align the reads to the reference genome and transcriptome using STAR aligner, thereby building up a matrix of the gene expressed in each cell.The unique molecular identifier (UMI) count matrix was processed using the R package Seurat (Version 2.3.0).
Since scRNA-seq implicated noise, quality-control (QC) assessments were performed after the basic data processing following the mentioned criteria.The number of genes detected in each cell must be within the range of 200-4000 based on the sample and the least number of cells needed for the gene detection was three.Besides, the ratio of UMI for mitochondrial gene number (percent.mito)and UMI for haemoglobin gene number (percent.HB) were not exceeding 0.3 and 0.05, respectively.The information on QC criteria for further processing as detailed in Supplementary Table 2.After applying these QC criteria, a total of 30,905 cells were included for further downstream analyses.

Dimensionality reduction and clustering.
Based on the filtered cell and gene matrices, Seurat (Version 2.3.0) in R package was applied to normalize the gene expression matrices.The gene expression level was divided by the total gene expression in the cells, which was then converted to relative abundance and multiplied with the normalization factor (default is 10,000) before undergoing log transformation.Then, Seurat package was used to establish the mean-variance relationship of the normalized counts of each gene in every cell.Before dimensionality reduction and clustering, QC control was conducted to remove the source of variation such as background noise, cell cycle and batch effects.Clustering analysis was performed to identify the cell subsets.In dimensionality reduction, principal component analysis (PCA) using Seurat package was performed to identify the significant principal component for the further clustering.A graph-based clustering algorithm using k-nearest neighbors (KNN) algorithms integrated with louvain community detection was conducted for the clustering of cell subsets.The t-SNE maps were generated to visualize the cluster of cells in two dimensions.

Differential gene expression analysis and gene annotation.
Seurat (Version 2.3.0) was used to perform differential gene expression analysis between different cell clusters.Using the Wilcoxon Rank Sum Test, the differentially expressed genes (DEGs) were determined.The highly expressed DEGs in each cell clusters were presented in a heat map.

DEGs enrichment analysis and protein-protein interaction analysis.
Based on the t-SNE map, the cell cluster which has the highly expressed DEGs were selected for gene annotation, gene-set enrichment analysis and protein-protein interaction analysis.These analyses were performed using Metascape (https:// metas cape.org/) to produce a high-quality graphical presentation.The enrichment analysis was based on hypergeometric distribution and p-adjusted (padj) which was less than 0.01 signified significant enrichment.The overlapped gene regions were visualized using Circos software (http:// circos.ca/).

Analysis of the correlation between stage, grade, and neutrophil-lymphocyte ratio (NLR)
with T cell subpopulations.In order to better understand the heterogeneity of T cell subsets in tumor microenvironment and validate above results in more samples, immunofluorescence staining was carried out and examined by microscope in additional 30-50 UBC samples with different stages (Supplementary Table 1).According to previous reports, the UBC case with NLR ≥ 2.5 is defined as high and that NLR < 2.5 is defined as low 17 .Graphpad Prism 8.0 was applied to analyze the relationship between T cell subpopulations and stage, grade, and NLR in UBC patients.
Prognostic and survival analysis.The overall survival (OS) and disease-free survival (DFS) of the gene signatures of cell cluster were analyzed by GEPIA2 (http:// gepia2.cancer-pku.cn/# index).The prognostic value of the respective gene signatures in different cell clusters was evaluated using Kaplan-Meier curves in GEPIA2.

Statistical analysis.
The Student t test was used to compare the mean values of two groups.In the gene expression and survival analysis, the average of gene expression was first calculated.UBC samples expressing signature genes of exhausted T (T Ex ), exhausted NKT (NKT Ex ), KI67 + T and B cell-like T (B-T) cells were defined as the high signatures group and the remaining samples as the low signatures group.The OS of each group was calculated by a Kaplan-Meier analysis, and the difference between those two groups was examined using the log rank test.A value of P less than 0.05 was regarded as statistically significant (*P < 0.05, **P < 0.01 and ***P < 0.001).

Results
Single-cell isolation and sequencing of tumor-infiltrating T cells in human UBC.To better decipher the heterogeneity of tumor-infiltrating T cells in UBC in single-cell level, the CD3 + T cells were isolated from PBMC, adjacent normal and tumor tissues from two MIBC patients (UBC1 and UBC2) via magnetic bead separation method (Fig. 1A).The qualified concentration, viability and purity of isolated T cells ensured single cell library preparation (Fig. 1B,C).Additionally, multicolor immunofluorescent staining of T cell population were performed to determine the T cell subsets in UBC by the antibodies targeting CD3, CD4, CD8 and FOXP3 (Fig. 1D).Taken together, CD3 + T cells were successfully isolated from PBMC, adjacent normal and tumor tissues from UBC patients.

Landscape of tumor-infiltrating T cells in UBC1.
The acquired qualified CD3 + T cells were sequentially analyzed by scRNA-seq (Fig. 1A).The single-cell sequencing data and filteration parameters were summarized in Supplementary Fig. 1 and Supplementary Tables 2, 3.After filteration, a total of 30,905 T cells were selected for T cell subsets analysis according to gene expression signature (Supplementary Figs. 2, 3).
A total of 10,895 T cells were identified in PBMC, NBT, and UBC tissue from UBC1.Through clustering analysis, a graph based clustering algorithm is used, which constructs a KNN graph through euclidean distance.By default, louvain algorithm is used to group cells and optimize modules.Based on the clustering results, 13 clusters of T cell populations were displayed using the t-SNE dimensionality reduction algorithm (Fig. 2A).Based on this, the expression of CD3D, CD4, CD8A, and FOXP3 in each cluster was further analyzed (Supplementary Fig. 4A-C).Seven, eight and six T subsets were identified in PBMC, NBT and UBC sample, respectively (Fig. 2B-D).The composition of the T cell population in PBMC was similar to that of NBT (Fig. 2E).Notably, cluster 0 and cluster 2 from PBMC were effector T cells and named as CD8-C0-T Eff -1 and CD8-C2-T Eff -2, respectively (Fig. 2B).Similarly, cluster 0 and cluster 5 from NBT were effector T cells and named as CD8-C0-T Eff -1 and CD8-C5-T Eff -1, respectively (Fig. 2C).They highly expressed various effector molecules-related genes, such as CCL4, GZMA, IFNG, NKG7 and KLRG1 18 (Fig. 2A, Supplementary Fig. 2A and Supplementary Tables 4-6).
In addition, cluster 1 from PBMC and cluster 2 from NBT were naïve T cells and named as CD4-C1-T N and CD4-C2-T N , respectively, which highly expressed various naïve T cell markers, such as CCR7 and SELL 19 , not effector molecules (Fig. 2A-C, Supplementary Fig. 2A and Supplementary Tables 4-6).Furthermore, cluster 3 from PBMC was transitional naïve T cells, named as CD4-C3-T TN , which highly expressed GPR171, GPR183, LTB and RGS1, not effector molecules (Fig. 2A-B, Supplementary Fig. 2A and Supplementary Tables 4, 5).Compared with T N , the expression of CCR7, LEF1, MYC, SELL and TCF7 decreased in T TN , which represented a transition state from T N cells into T CM cells.
The composition of the T cell populations in UBC was remarkably different compared to those of PMBC or NBT (Fig. 2D-E).Cluster 0 (CD8-C0-T Ex -1) and cluster 3 (CD8-C3-NKT Ex ) were dominate in UBC and highly expressed exhaustion-related genes LAG3, HAVCR2, PDCD1 and CTLA4 (Fig. 2A and Supplementary Table 7), suggesting the exhausted phenotype in T cells in UBC microenvironment.Specifically, CD8-C3-NKT Ex highly expressed NK cell surface receptor coding genes KLRB1, KLRC2 and KLRD1.Interestingly, cluster 1 (CD4-C1-T CM ) and cluster 4 (CD4-C4-T CM ) from UBC were T CM cells, which highly expressed T CM cell markers, such as IL7R and CCR7 23 (Fig. 2A,D, Supplementary Fig. 2C, Supplementary Table 7).Intriguingly, a small subset of CD4-C1-T CM in UBC, representing 2.2% of the unique cell clusters, highly expressed immunoglobulin-related genes, such as IGHA1, IGHG1-4 and IGLC2-3 (Supplementary Table 4), which were responsible for antibody synthesis of B cells.More importantly, cluster 5 (CD3-C5-Ki67 + T) was distinctively found in UBC and divided into two groups (Fig. 2D,E).This cell cluster expressed high levels of proliferation factors MKi67, KIAA0101 and STMN1 (Supplementary Table 7), suggesting the rapid proliferation state.In short, there is a high similarity between the T cell population in PBMC and NBT of UBC1, and T Ex -1, NKT Ex , B-T and Ki67 + T cells were exclusively enriched in UBC, which potentialy participated in the formation of the immunosuppressive microenvironment of UBC.

Profile of tumor-infiltrating T cells in UBC2.
A total of 20,010 T cells were characterized in PBMC, NBT and UBC from UBC2.Through clustering analysis, a graph based clustering algorithm is used, which constructs a KNN graph through euclidean distance.By default, louvain algorithm is used to group cells and optimize modules.Based on the clustering results, 17 clusters of T cell populations were displayed using the t-SNE dimensionality reduction algorithm (Fig. 3A and Supplementary Fig. 3).Based on this, the expression of CD3D, CD4, CD8A, and FOXP3 of each cluster were analyzed (Supplementary Fig. 4D,F).

Gene expression signature of exhausted CD8 + T cells. Based on the global characterization of tumor-
infiltrating T cells from two MIBC patients, T Ex cells dominated the microenvironment of UBC (Fig. 4A,B).Specifically, 273 and 201 differentially expressed genes were identified in T Ex cells from UBC1 and UBC2, respectively (Supplementary Table 11).Obviously, T Ex cells expressed high levels of immune checkpoints genes TIM3, PDCD1, LAG3 and CTLA4 (Fig. 4C,D), and the majority of the differentially expressed genes were overlapped between T Ex cells from UBC1 and UBC2 (Fig. 4E).By gene annotation (GO) and pathway enrichment analysis, the differential expression genes of T Ex cells significantly enriched in T cell activation, adaptive immune system and oxidative phosphorylation, etc. (Fig. 4F and Supplementary Fig. 5).Furthermore, IKZF3 and TRGC2 are novel biomarkers of T Ex (Fig. 4G,H) and CD8 + PD1 + TIM3 + TRGC2 + T Ex significantly enriched in UBC (3.47%) compared to NBT (0.66%) (Fig. 4I,J).Specifically, the proportion of T Ex cells enhanced with the increase of the stage (P < 0.001) (Fig. 4K).Similarly, UBC patients with high grade (P < 0.05) and NLR (P < 0.01) had a increased proportion of T Ex (Fig. 4L,M).
Survival analyses via GEPIA2 database indicated that UBC patients acquiring high expression of T Ex signature genes (Supplementary Table 12) had a decreased OS as compared to those of UBC patients acquiring low expression of these signature genes (Fig. 4N).Additionally, lung adenocarcinoma (LUAD), lung squamous cell carcinoma (LUSC), liver hepatocellular carcinoma (LIHC), brain lower grade glioma (LGG) and adrenocortical carcinoma (ACC) patients acquiring high expression of T Ex signature genes also had a decreased OS and DFS as compared to those of tumor patients acquiring low expression of these signature genes (Supplementary Fig. 6).Taken together, T Ex was exclusively identified in UBC, which highly expressed IKZF3 and TRGC2.The gene expression signature of T Ex was associated with poor prognosis of patients with UBC and multiple other tumor types.

Gene expression profile of NKT Ex cell population.
Besides T Ex , NKT Ex was another major subpopulation in UBC (Fig. 5A,B).Specifically, 165 and 152 differentially expressed genes were identified in NKT Ex cells from UBC1 and UBC2, respectively (Supplementary Table 13).Similar to T Ex , NKT Ex expressed high level of immune checkpoints such as TIM3, PDCD1, LAG3 and CTLA4 (Fig. 5C,D), and there was a high overlapping rate of specific genes among them (Fig. 5E).Moreover, the differential expression genes of NKT Ex significantly enriched in the response to adaptive immune system, NK cell mediated immunity, T cell activation, etc. (Fig. 5F, Supplementary Fig. 7).Furthermore, ATF3 and NR4A1 were the important biomarkers in identification of NKT Ex (Fig. 5G,H) and CD8 + CD56 + PD1 + TIM3 + NKT Ex significantly infiltrated into UBC (16.92%) rather than NBT (0.92%) (Fig. 5I,J).Compared with T1 stage, the proportion of NKT Ex cells in T2 and T3/T4 stages  www.nature.com/scientificreports/ was significantly increased (P < 0.001) (Fig. 5K).Similarly, UBC patients with higher grading (P < 0.05) and NLR (P < 0.01) had a higher proportion of NKT Ex cells (Fig. 5L,M).UBC patients acquiring high expression of NKT Ex signature genes (Supplementary Table 14) had a decreased OS of UBC patients compared to those of UBC patients acquiring low expression of these signature genes (Fig. 5N).Additionally, LUAD, LUSC, LIHC, LGG and ACC patients acquiring high expression of NKT Ex signature genes had a decreased OS and DFS as compared to those of tumor patients acquiring low expression of these signature genes.Taken together, NKT Ex was significantly enriched in UBC (Supplementary Fig. 8), which highly expressed ATF3 and NR4A1.The gene expression signature of NKT Ex correlated with poor prognosis of patients with UBC and other tumor types.
Ki67 + T cells are enriched in UBC.The Ki67 + T cells were existed in both UBC1 and UBC2 (Fig. 6A,B).Specifically, 742 and 556 differentially expressed genes were identified in Ki67 + T cells from UBC1 and UBC2, respectively (Supplementary Table 15).Ki67 + T cells expressed high level of immune checkpoints such as MKI67, AURKA, TOP2A and UBE2C (Fig. 6C,D), and there was a high overlapping rate of specific genes among them (Fig. 6E).Moreover, the differential expression genes significantly enriched in the cell cycle, cell proliferation and DNA replication, suggesting an active proliferation status of Ki67 + T cell clusters (Fig. 6F, Supplementary Fig. 9).Furthermore, MKI67 and TOP2A were the important biomarkers in identification of Ki67 + T cells (Fig. 6G,H).Additionally, immunofluorescence assay demonstrated that CD3 + Ki67 + T cells significantly enriched in UBC (11.85%) compared to NBT (0.46%) (Fig. 6I,J).
Specifically, the proportion of Ki67 + T cells increased with the increment of stages (P < 0.001) (Fig. 6K).Similarly, as the grading (P < 0.01) and NLR (P < 0.01) of UBC patients increasd, the proportion of Ki67 + T cells augmented (Fig. 6L,M).The UBC patients with high expression of Ki67 + T signature genes (Supplementary Table 16) had a lower OS as well as DFS as compared to those of UBC patients acquiring low expression of these signature genes (Fig. 6N,O).Additionally, LUAD, LUSC, LIHC, LGG and ACC patients acquiring high expression of Ki67 + T signature genes had a decreased OS and DFS as compared to those of tumor patients acquiring low expression of these signature genes (Supplementary Fig. 10).Hence, Ki67 + T cells was remarkably enriched in UBC, which highly expressed MKI67, AURKA, TOP2A and UBE2C.The gene expression signature of Ki67 + T cells correlated with poor prognosis of patients with UBC and other tumor types.

Gene expression landscape of B-T cells. A specific cluster named B-T cells existed in both UBC1 and
UBC2 patients (Fig. 7A,B).Specifically, 86 and 107 differentially expressed genes were identified in B-T cells from UBC1 and UBC2, respectively (Supplementary Table 17).B-T cells expressed high level of immune checkpoints such as GHG2, IGHG1, JCHAIN and CD79A (Fig. 7C,D), and there was a high overlapping rate of specific genes among them (Fig. 7E).Moreover, the differentially expressed genes were involved in activation of B cell, B cell proliferation, induction of immune response, etc. (Fig. 7F), suggesting that B-T cells was actively involved in proliferation and antibodies production.However, the functional relations between these genes of B-T cells from UBC1 and UBC2 were rather diminutive (Supplementary Fig. 11), indicating the high heterogeneity of B-T cells.Furthermore, this cell cluster had increased expression of IGHG2, IGHG1, JCHAIN and CD79A (Fig. 7C,D,G,H) and CD3 + CD19 + IGHG1 + T cells significantly infiltrated in UBC (30.88%) compared to NBT (0.98%) (Fig. 7I,J).
Compared with T1 stage, the proportion of B-T cells in T2 (P < 0.01) and T3/T4 (P < 0.001) stages significantly increased (Fig. 7K).Similarly, the proportion of B-T cells significantly increased in UBC patients with higher grades (P < 0.05) and NLR (P < 0.01) (Fig. 7L,M).More importantly, UBC patients acquiring high expression of B-T cells signature genes (Supplementary Table 18) had a decreased OS as well as DFS compared to those of UBC patients acquiring low expression of these signature genes (Fig. 7N,O).Additionally, LIHC, uveal melanoma (UVM), LGG and ACC patients acquiring high expression of B-T cells signature genes had a decreased OS and DFS as compared to those of tumor patients acquiring low expression of these signature genes (Supplementary Fig. 12).In short, B-T cells were exclusively identified in UBC, which highly expressed IGHG2, IGHG1, JCHAIN and CD79A.The gene expression signature of B-T cells was associated with poor prognosis of patients with UBC and other tumor types.

Discussion
UBC is characterized by a high rate of recurrence and has limited therapeutic options.Tumor-infiltrating T cells are of high heterogeneity in the tumor tissues, whose composition and numbers are closely related to the prognosis and therapeutic effects of UBC treatment.However, the composition and function of tumor-infiltrating T cells in UBC remain unclear.Here, scRNA-seq of the T cell populations in PBMC, NBT and tumor tissue from two UBC patients were performed.The transcriptional profiles of 30, 905 individual T cells provided a deep interrogation of phenotype and functional heterogeneity among tumor resident immune cell populations.We also identified four clusters, T Ex , NKT Ex , Ki67 + and B-T subtypes, were significantly enriched in UBC, which correlated with the poor prognosis of UBC.
In our study, T N , T TN , T Eff , NKT, Treg and DC like T cell subtypes were mainly enriched in the PBMC and NBT of UBC patients.T Eff highly expressed effector molecules-related genes such as CMC1, and NKT cells expressed NK cell surface markers and effector molecules-related genes, such as IFNG, GZMB and GZMA, suggesting the potential of effector functions.There were T N , T TN , T CM in PBMC, indicating the different developmental stages of T cell population, suggesting the differentiation process and functional activation in T cell population of PBMC.
T Ex is a exhausted subset of T cells that specifically overexpressing IKZF3, LUC7L3, TRGC2, LAG3, and TIM3, which were mainly involved in T cell development, antigen recognition, and expression of heterogeneous receptors.It is particularly noteworthy that LAG3 and TIM3 are genes that express T cell inhibitory receptors and negatively regulate the killing ability of T cells.By blocking the signal reception of inhibitory receptors, the exhaustion state of T cells could be partially alleviated, thereby enhancing their ability to kill tumors.Compared with T cells in PBMC and NBT, T Ex cells were significantly enriched in UBC and highly expressed immune checkpoints encoding genes, which was similar to the T Ex cells in liver cancer 15 , breast cancer 14 and non-small-cell lung cancer 13 .Additionally, cytokine granulocyte-macrophage colony stimulating factor (GM-SCF) encoding gene CSF2 and chemokine CCL5 encoding gene CCL5 were specifically expressed in T Ex cells, which recruted Treg cells into tumor microenvironment and promoted tumor growth 27 .Furthermore, IKZF3 and TRGC2 were also highly expressed in T Ex and CD8 + PD1 + TIM3 + TRGC2 + T Ex accounts for 3.47% of CD3 + T cells in UBC, signifying that IKZF3 and TRGC2 are potential biomarkers of T Ex .More importantly, the gene expression signature of T Ex cells remarkably associated with poor prognosis of UBC, LUSC, LIHC, LGG and ACC.In recent years, using the autoimmune system to fight against tumor cells has become an important approach in tumor immunotherapy.A variety of treatment methods have been explored for different targets of reversing the process of T cell exhausted, such as immunocheckpoint inhibitor treatment, epigenetics treatment, CAR-T adoptive cellular immunotherapy and other programs.Some studies have shown that the combination of inhibitory receptors, such as the use of anti PD-1/PD-L1 and anti CTLA-4 antibodies, can significantly reverse the exhaustion of T cells and enhance the immune response of various tumor patients 28 .In the CT26 colon adenocarcinoma model 29 , the combination of anti TIM-3 and anti PD-1 drugs is more effective than anti PD-1 therapy alone.Guo et al. 30 pointed out that the combination of anti TIM-3 and anti CD137/4-1BB antibodies can also significantly inhibit the growth of mouse ovarian tumors.Thus, T Ex cells and its biomarkers were the promising targets for cancer immunotherapy and render the development of robust and effective immunotherapy strategies towards UBC.
The killing ability of T cells mainly depends on antigen presentation.Unlike T cells, NKT cell can kill without contacting antigens.However, if NKT cell are exposed to antigens for a long time, they will also be in a state of exhaustion like T cells.NKT Ex cells specifically overexpresses ATF3, CSF2, NR4A1, LAG3, and TIM3, which are mainly involved in transcriptional activation, regulation of lymphocyte production, and expression of heterogeneous receptors.We also observed a higher distribution of NKT Ex (CD8 + CD56 + PD1 + TIM3) in UBC, which accounts for 16.92% of CD3 + T cells in UBC.NKT Ex cells expressed high level of immune checkpoint encoding genes and exhausted-related genes, which was prone to be an exhaustive state and similar to the NKT Ex cells in pancreatic cancer 12 .Previous studies indicated that NKG2A targeting with monalizumab is a novel checkpoint inhibitory mechanism promoting anti-lymphoma immunity by enhancing the activity of both T and NK cells 31 .Here, we also found that KLRC1 (encoding NKG2A) was highly expressed in both T Ex and NKT Ex cells of UBC, which could be a promising target.Furthermore, ATF3 and NR4A1 are potential biomarkers of NKT Ex cells.More importantly, not only UBC patients, but also LUSC, LIHC, LGG and ACC ones acquiring high expression of NKT Ex cells signature genes had a decreased OS as compared to those of tumor patients acquiring low expression of these signature genes.As is well known, immune checkpoints play an important role in the production of immunosuppressive tumor microenvironment, leading to NK cell failure and tumor immune escape.Therefore, NK cells must reverse their dysfunctional state and increase their effector effect to improve the efficiency of cancer immunotherapy.Blocking immune checkpoints not only saves NK cells from failure, but also enhances their powerful anti-tumor activity 32 .In melanoma, fibrosarcoma, colon cancer and leukemia models, the co-blocking of TIM-3 and PD-1 showed higher anti-tumor efficacy, more complete tumor regression and longer survival period 29,33,34 .Besides, the combination of NKG2A monoclonal antibody or TIGIT monoclonal antibody plus PD-1 monoclonal antibody or cetuximab showed encouraging results in the treatment of patients with advanced solid tumors 31,35 .Thus, NKT Ex cells and its biomarkers were the promising candidates for UBC immunotherapy.
Ki67 + T cells were identified in the UBC patients for the first time and accounted for 11.85% of CD3 + T cells in UBC.MKI67, AURKA, TOP2A and UBE2C were highly expressed in Ki67 + T cells, indicating an active proliferation status and potential contribution to the occurrence, development, and metastasis of UBC.Ki67 + T cell subsets were previously identified in colorectal cancer 36 and multiple myeloma 37 .Moreover, the percentage of Ki67 + lymphocytes was significantly higher in patients with multiple myeloma and monoclonal gammopathy compared with the normal controls, which was associated with disease stage.Here, we found that the gene expression signature of Ki67 + T cells remarkably associated with poor prognosis of UBC, LUSC, LIHC, LGG and ACC.Therefore, Ki67 + T cells are potential candidate cells for UBC immunotherapy, and further research is needed on their application in immunotherapy in the future.
Finally, we identified another novel and specific subpopulation, namely B-T cells, in UBC.Previous studies suggested that the proportion of B-T cells was high in cancer compared with that of controls, such as esophageal cancer, non-Hodgkin's lymphoma, lung cancer, breast cancer, liver cancer, etc. 38 .Gene expression profiling revealed that this cell cluster had increased expression of IGHG2, IGHG1, JCHAIN and CD79A, These genes are mainly involved in the coding of immunoglobulins and the formation of proteins, indicating the B cell properties in T cell populations.Besides, it also had the expression of CD3, suggesting the existence of matured T cell population.More importantly, not only UBC patients, but also LIHC, UVM, LGG and ACC ones acquiring high expression of signature genes of B-T cells had a decreased OS and DFS as compared to those of tumor patients acquiring low expression of these signature genes.Thus, B-T cells were a potential candidate for UBC immunotherapy, whose phenotype and role need further investigation.
In UBC, T cell exhaustion might result from long-duration antigen exposures and continuous inflammation 39 .The cell-to-cell signals including prolonged T cell receptor (TCR) engagement and co-stimulatory and/or coinhibitory signals, inflammatory cytokines and suppressive cytokines, and tissue and microenvironmental influences are the potential reasons of T cell exhaustion 40 .In this study, exhausted T or NKT cells may be converted by effector CD8 + T cells 41,42 .Mechanically, persistent antigen exposure, the elevated expression of inhibitory receptors (TIM3, PDCD1, LAG3, CTLA4, IKZF3 and TRGC2), surface markers (CD3, CD4, CD8, CD19 and CD56), transcrition factors (NR4A1, ATF3), immunosuppressive cytokines (VEGF, IL10 and IL4) are the potential reasons of T cell exhaustion in UBC.Although Ki67 + T cells and B cell like T cells were previously reported in colorectal cancer and multiple myeloma, esophageal cancer, non-Hodgkin's lymphoma, lung cancer, breast cancer, liver cancer, etc..However, the causes of these increase have not been clearly described, which need further investigation.
In short, this article conducts scRNA-seq from T cells derived from tumor tissue, normal tissue adjacent to tumor and PBMC of two UBC patients, and more UBC cases with different pathological characteristics were needed to better understand the heterogeneity of T cell subsets in UBC microenvironment.In addition, a large number of literatures have shown that T cell function exhaustion is an important mechanism of tumor immune escape 40 .The functional assay to investigate immune-suppressive activity of T Ex cells, NKT Ex cells, Ki67 + T cells and B-T cells could be carried out in the future to reveal the relationship between the major T cell subtypes and immune escape.

Conclusions
In this study, scRNA-seq was applied for the first time to provide a thorough description of T cells in PBMC, NBT and tumor tissues of UBC patients, and describing gene expression characteristics and potential functions.The heterogeneity of T cell subsets in tumor microenvironment was revealed, which contributed to the immunosuppressive microenvironment.Notably, T Ex , NKT Ex , Ki67 + T and B-T cells were exclusively enriched in UBC, which were significantly associated with poor survival in patients with UBC and various tumor types.The identified T cells subsets and novel biomarkers provided a theoretical foundation for the development of effective tumor immunotherapies.

Figure 1 .
Figure 1.Dissociation of tumor-infiltrating T Cells from UBC patients (A) Overview of the study design.(B) The UBC patient information, and the viability and numbers of T cells isolated.(C) Confirmation of the percentage of CD3 positive T cells of magnetic beads isolated single-cells by FACS Calibur.(D) Representative immunofluorescent staining with anti-CD3, CD4, CD8, and FOXP3 antibodies in UBC.Bar = 125 μm.

Figure 2 .
Figure 2. Single-cell profiling of tumor-infiltrating T cells in UBC1 (A) Heat map from single-cell analysis via expression recovery (SAVER) imputed data with cells grouped into clusters (indicated by colored bars at the top).The top ten genes differentially expressed for each cluster are shown on the y axis, and key genes are also shown for each cluster.(B) The t-SNE projection of T cells of PBMC from UBC1, showing the formation of seven main clusters shown in different colors.(C) The t-SNE projection of T cells of NBT from UBC1, showing the formation of eight main clusters shown in different colors.(D) The t-SNE projection of T cells of UBC tissue from UBC1, showing the formation of six main clusters shown in different colors.(E) The frequencies of each cluster within the PBMC, NBT, and UBC tissue samples are depicted by balloon plots.

Figure 3 .
Figure 3. Single-cell analysis of tumor-infiltrating infiltrating T cells in UBC2 (A) Heat map from SAVER imputed data with cells grouped into clusters (indicated by colored bars at the top).The top ten genes differentially expressed for each cluster are shown on the y axis, and key genes are also shown for each cluster.(B) The t-SNE projection of T cells of PBMC from UBC2, showing the formation of seven main clusters shown in different colors.(C) The t-SNE projection of T cells of NBT from UBC2, showing the formation of 11 main clusters shown in different colors.(D) The t-SNE projection of T cells of UBC tissue from UBC2, showing the formation of 11 main clusters shown in different colors.(E) The frequencies of each phenograph cluster within the PBMC, NBT, and UBC tissue samples are depicted by balloon plots.

Figure 4 .
Figure 4. Characterization of exhausted CD8 + T Cells and its clinical implication in UBC (A, B) The t-SNE projection of exhausted CD8 + T cells (red) in tumor tissues of UBC1 and UBC2.(C, D) Expression levels of TIM3, PDCD1, LAG3 and CTLA4, across 7939 (UBC1 2446 and UBC2 5493) single T cells illustrated in t-SNE plots.(E) The gene expression status between tumor tissues of UBC1 and UBC2 were visualized in a circular layout via the Circos analysis tool.The red line of the outer circle manifests UBC1, and the blue one stands for UBC2.The orange line of the inner circle manifests the common genes shared by both UBC1 and UBC2.The light orange line in the circle indicates the functional correlation between genes.(F) Pathway enrichment of 487 (UBC1 273, UBC2 214) signature genes in T Ex .(G) UMI counts of IKZF3 and TRGC2 in C0-T Ex -1, C1-T CM , C2-Treg, C3-NKT Ex and C4-T CM clusters in UBC1.(H) UMI counts of IKZF3 and TRGC2 in C0-NKT Ex , C1-T Ex -1, C2-Treg, C6-T Ex -2, C7-DC like T, C8-CK19 + T, C9-B cell like T and C10-SPARC + T clusters in UBC2.(I) Representative immunofluorescent staining with anti-CD8, PD1, TIM3 and TRGC2 antibodies in NBT and UBC tissues.Bar = 125 μm.(J) The parentage of CD8 + PD1 + TIM3 + TRGC2 + T Ex in NBT and UBC tissues.(K) Correlation analysis between staging and T Ex cells in UBC patients.(L) Correlation analysis between grading and T Ex cells in UBC patients.(M) Correlation analysis between NLR and T Ex cells in UBC patients.(N) Kaplan-Meier curves comparing the OS between UBC patients expressing high or low levels of gene signature in T Ex , log-rank test.n, patient number.Data are presented as mean ± SD. *P < 0.05, **P < 0.01.

Figure 5 .
Figure 5. Characterization of NKT Ex cells and its clinical implication in UBC (A, B) The t-SNE projection of NKT Ex cells in tumor tissues of UBC1 and UBC2, showing in red.(C, D) Expression levels of TIM3, PDCD1, LAG3 and CTLA4, across 7,939 (UBC1 2,446 and UBC2 5493) single T cells illustrated in t-SNE plots.(E) The gene expression status between tumor tissues of UBC1 and UBC2 were visualized in a circular layout via the Circos analysis tool.The red line of the outer circle manifests UBC1, and the blue one stands for UBC2.The orange line of the inner circle manifests the common genes shared by both UBC1 and UBC2.The light orange line in the circle indicates the functional correlation between genes.(F) Pathway enrichment of 315 (UBC1 164, UBC2 151) signature genes in NKT Ex .(G) UMI counts of ATF3 and NR4A1 in C0-T Ex , C1-T CM , C2-Treg, C3-NKT Ex and C5-Ki67 + T clusters in UBC1.(H) UMI counts of ATF3 and NR4A1 in C0-NKT Ex , C1-T Ex -1, C2-Treg, C4-T FH , C6-T Ex -2, C7-DC like T and C9-B cell like T clusters in UBC2.(I) Representative immunofluorescent staining with anti-CD8, CD56, PD1 and TIM3 antibodies in NBT and UBC tissues.Bar = 125 μm.(J) The parentage of CD8 + CD56 + PD1 + TIM3 + NKT Ex in NBT and UBC tissues.(K) Correlation analysis between staging and NKT Ex cells in UBC patients.(L) Correlation analysis between grading and NKT cells in UBC patients.(M) Correlation analysis between NLR and NKT Ex cells in UBC patients.(N) Kaplan-Meier curves comparing the OS between UBC patients expressing high or low levels of gene signature in NKT Ex , log-rank test.n, patient number.Data are presented as mean ± SD. *P < 0.05, **P < 0.01.

Figure 6 .
Figure 6.Characterization of Ki67 + T Cells and its clinical implication in UBC (A, B) The t-SNE projection of Ki67 + T cells in tumor tissues of UBC1 and UBC2, showing in red.(C, D) Expression levels of MKI67, AURKA, TOP2A and UBE2C, across 7939 (UBC1 2446 and UBC2 5493) single T cells illustrated in t-SNE plots.(E) The gene expression status between tumor tissues of UBC1 and UBC2 were visualized in a circular layout via the Circos analysis tool.The red line of the outer circle manifests UBC1, and the blue one stands for UBC2.The orange line of the inner circle manifests the common genes shared by both UBC1 and UBC2.The light orange line in the circle indicates the functional correlation between genes.(F) Pathway enrichment of 1247 (UBC1 736, UBC2 551) signature genes in Ki67 + T cells.(G) UMI counts of KI67 and TOP2A in C0-T Ex , C1-T CM , C2-Treg, C3-NKT Ex , C4-T CM and C5-Ki67 + T clusters in UBC1.(H) UMI counts of KI67 and TOP2A in C0-NKT Ex , C1-T Ex -1, C2-Treg, C3-Ki67 + T, C4-T FH , C5-Treg, C6-T Ex -2, C7-DC like T, C8-CK19 + T, C9-B cell like T and C10-SPARC + T clusters in UBC2.(I) Representative immunofluorescent staining with anti-CD3, CD4 and Ki67 antibodies in NBT and UBC tissues.Bar = 125 μm.(J) The parentage of CD3 + Ki67 + Ki67 + T cells in NBT and UBC tissues.(K) Correlation analysis between staging and Ki67 + T cells in UBC patients.(L) Correlation analysis between grading and Ki67 + T cells in UBC patients.(M) Correlation analysis between NLR and Ki67 + T cells in UBC patients.(N, O) Kaplan-Meier curves comparing the OS and DFS between UBC patients expressing high or low levels of gene signature in Ki67 + T cells, log-rank test.n, patient number.Data are presented as mean ± SD. *P < 0.05, **P < 0.01.

Figure 7 .
Figure 7. Characterization of B-T Cells and its clinical implication in UBC (A, B) The t-SNE projection of B-T cells in tumor tissues of UBC1 and UBC2, showing in red.(C-D) Expression levels of IGHG1, IGHG2, JCHAIN and CD79A, across 7939 (UBC1 2,446 and UBC2 5493) single T cells illustrated in t-SNE plots.(E) The gene expression status between tumor tissues of UBC1 and UBC2 were visualized in a circular layout via the Circos analysis tool.The red line of the outer circle manifests UBC1, and the blue one stands for UBC2.The orange line of the inner circle manifests the common genes shared by both UBC1 and UBC2.The light orange line in the circle indicates the functional correlation between genes.(F) Pathway enrichment of 193 (UBC1 86, UBC2 107) signature genes in B-T cells.(G) UMI counts of IGHG2 and JCHAIN in C0-T Ex , C1-T CM , C2-Treg, C3-NKT Ex , C4-T CM and C5-Ki67 + T clusters in UBC1.(H) UMI counts of IGHG2 and JCHAIN in C0-NKT Ex , C1-T Ex -1, C2-Treg, C3-Ki67 + T, C4-T FH , C5-Treg, C6-T Ex -2, C7-DC like T, C8-CK19 + T, C9-B-T and C10-SPARC + T clusters in UBC2.(I) Representative immunofluorescent staining with anti-CD3, CD19 and IGHG1 antibodies in NBT and UBC tissues.Bar = 125 μm.(J) The parentage of CD3 + CD18 + IGHG1 + B-T cells in NBT and UBC tissues.(K) Correlation analysis between staging and B-T cells in UBC patients.(L) Correlation analysis between grading and B-T cells in UBC patients.(M) Correlation analysis between NLR and B-T cells in UBC patients.(N-O) Kaplan-Meier curves comparing the OS and DFS between UBC patients expressing high or low levels of gene signature in B cell like T cells, log-rank test.n, patient number.Data are presented as mean ± SD. *P < 0.05, **P < 0.01.