CXCL13 Correlates with Prognosis, Immune Infiltration, and T Cell Exhaustion in Ovarian Cancer


 Background: CXCL13 is an important chemotactic factor closely related to the biology of cancer cells. The presence work focused on exploring the significance of CXCL13 in prognosis prediction and analyzing the associations of CXCL13 with T cell function and immune infiltration in various cancers, especially ovarian cancer (OV).Purpose: CXCL13 is associated with prognosis, immune infiltration, and T cell failure of ovarian cancer.Methods: The Oncomine, GEPIA2 and HPA databases were utilized for analyzing CXCL13 levels within diverse cancers. The significance of CXCL13 in prognosis prediction was explored through Kaplan-Meier Plotter, TCGAportal, and GEPIA2. Meanwhile, the associations of CXCL13 with clinical stage, gene marker sets, and immune infiltration were examined through TISIDB, GEPIA2, and TIMER databases. Besides, CXCL13 was screened to analyze the biological processes (BPs) and KEGGs enriched by co-expression genes. The miRWalk database was employed for analyzing the gene-miRNA interaction network of CXCL13 within OV.Results: CXCL13 expression decreased in many cancers, which predicted the dismal survival of OV. CXCL13 upregulation was in direct proportion to the increased immune infiltration degrees of many functional T cells (like exhausted T cells) and immune cells. Additionally, some critical genes of exhausted T cells, such as TIM-3, PD-1, LAG3, TIGIT, GZMB, and CXCL13, were closely associated with CXCL13. Moreover, CXCL13 was related to immune response regulatory signaling pathway, leukocyte cell-cell adhesion, cell adhesion molecules (CAMs), and hematopoietic cell lineage. Conclusion: CXCL13 can serve as a biomarker to predict cancer prognosis, particularly OV. CXCL13 upregulation remarkably elevates the immune infiltration degrees of numerous immune cells, like mast cells, CD8+ T cells, natural killer (NK) cells, and dendritic cells (DCs). Furthermore, CXCL13 is suggested to be closely related to exhausted T cells, which may be used as a candidate regulating factor for T cell exhaustion within OV. Detecting CXCL13 levels contributes to prognosis prediction and CXCL13 regulation within exhausted T cells, which provides a new approach to maximizing the anti-OV efficacy of immunotherapy.


Introduction
Ovarian cancer (OV) presents a high mortality rate, and the death-to-incidence ratio is impressively high. At present, OV shows no obvious symptoms, and no effective screening approaches are available. As a result, the diagnosis of many OV cases is made in the later stages (FIGO stages III-IV), and the 5-year overall survival (OS) is about 40-45% [1]. Currently, OV is mainly treated with chemotherapy combined with surgery. The preferred treatment is tailored according to individual features, which usually comprises primary cytoreductive surgery combined with platinumbased chemotherapy and eventual maintenance using bevacizumab (for wild type (WT) BRCA patients or those with HRD su ciency) or PARP inhibitors (for mutant BRCA cases or those with HRD de ciency) [2]. Notably, immunotherapy has emerged as the standard treatment for different cancers, since it enhances the body immunity against tumor cells. Multiple immune checkpoints can be triggered via ligand-receptor interactions, which may be dysregulated in the tumor for reducing the antitumor response [3]. Therefore, it is urgently needed to illustrate the tumor-immune interaction immunophenotypes and to identify new immune therapeutic targets for OV.
CXCL13, also called BCA-1 (B cell-attracting chemokine 1) or BLC (Blymphocyte chemoattractant), belongs to the homeostatic chemokine. CXCL13 can be produced via stromal cells within the B-cell regions in secondary lymphoid tissues (follicles), like lymph nodes, spleen, Peyer's patches and tonsils [4][5][6][7][8]. CXCL13 level is signi cantly related The present work conducted a comprehensive analysis on the CXCL13 level and its relation with cancer survival of patients obtained from several databases, including GEPIA2, Oncomine, LOGpc, HPA, Kaplan-Meier plotter, miRWalk, and LinkedOmics. In addition, Tumor Immune Estimation Resource (TIMER) was utilized to analyze the relations between CXCL13 and tumor-in ltrating immune cells (TIICs) within a diverse tumor microenvironment (TME).
Results in the present work can help to illustrate the key function of CXCL13 in OV and elucidate the possible relation of CXCL13 with tumor-immune interactions and the underlying mechanism.

Analysis based on Oncomine Database
Oncomine database covers totally 715 gene expression datasets and 86,733 samples, which contributes to data mining [21]. This database was adopted in the present work for assessing the relation of CXCL13 level with cancer prognosis (https://www.oncomine.org/resource/login.html).

Analysis based on GEPIA2 Database
Gene Expression Pro ling Interactive Analysis 2 (GEPIA2) database (http://gepia2.cancer-pku.cn/), an online approach used to interactively analyze gene levels in cancer and non-carcinoma samples obtained from GTEx (Genotype Tissue Expression) and TCGA, provides the customizable functions, such as pro le plotting, differential expression, patient survival, correlation, dimensionality reduction and similar gene detection analyses [22]. The association of CXCL13 levels with TCGA-derived cancer survival was examined through "survival analysis".
Besides, the association of CXCL13 levels with TIIC gene markers was analyzed by Spearman's correlation analysis in "correlation analysis". Both cancer and non-carcinoma samples were utilized in analyses.
Analysis based on TCGA portal Database TCGAportal (http://www.tcgaportal.org) is a web-based portal used to parallelly align several cancers and comprehensively analyze individual cancers. TCGAportal was used to further verify the prognostic value of CXCL13 expression in cancer patients (http://tumorsurvival.org/) [23]. TCGA covers the pathological and sequencing data for 30 diverse cancers [24].

Analysis based on Kaplan-Meier Plotter Database
Kaplan-Meier Plotter database has been developed as a web-based approach for the rapid access to the in uence of gene expression on 21 cancer survival, as well as 4 great datasets, namely, breast cancer (BC, n = 6234), lung cancer (LC, n = 3452), ovarian cancer (n = 2190), and gastric cancer (GC, n = 1440) [25]. It was adopted in the current work for evaluating the associations of CXCL13 levels with the survival of these 4 cancers. In addition, the pan-cancer and OV datasets were utilized for analyzing CXCK13 levels within diverse OV subtypes. We determined HR (95% CIs) and log-rank P-values and plotted the survival curves (http://kmplot.com/).

Analysis based on LOGpc Database
LOGpc, a web server covering numerous datasets that can be used to analyze survival, contains 13 survival terms from 28,098 cancer cases with 26 cancers (such as OSlms, OSkirc and OSblca) and an additional 23 online prognosis prediction approaches [26][27][28][29][30][31][32][33]. Such case samples were mostly obtained from GEO and TCGA databases. LOGpc is publicly accessible and user friendly. The 26 cancer types are divided as 11 system categories based on TCGA. At present, LOGpc only allows for o cial gene symbol input. After the gene symbol is set by the user and the relevant parameters are selected, the "Kaplan-Meier plot" button is pressed and the results are displayed on the output webpage. For meeting the speci c demands of diverse investigators, the clinical confounders are de ned for subsequent subgroup analyses. (http://bioinfo.henu.edu.cn/DatabaseList.jsp).

Analysis based on TISIDB Database
The TISIDB database covers altogether 988 immune-associated anticancer genes, non-carcinoma multi-omics data, molecular pro ling data, high-throughput screening (HTS) technologies, and different immunological data resources collected in 7 publicly accessible databases [34]. It allows to analyze the associations between the screened genes and chemokines, lymphocytes, and immunomodulators. The present work employed TISIDB for assessing the association of Annexin levels with OV clinical stages and investigating the relationship of CXCL13 level with immunomodulators and lymphocytes (http://cis.hku.hk/TISIDB).

Analysis based on TIMER Database
TIMER (Tumor Immune Estimation Resource; cistrome. shinyapps.io/timer), a kind of easy-to--to-use web interface, has provided a computational approach for oncology investigators to comprehensively and dynamically analyze and monitor cancer genomic and immunologic data [35]. It contains gene expression pro ling data of 10 897 samples covering 32 different kinds of TCGA-derived cancers, which can be used to estimate the 6 TIICs subpopulation abundances, including CD4+ T cells, CD8+ T cells, dendritic cells (DCs), B cells, neutrophils and macrophages. In this study, we adopted constrained least squares tting for speci c gene levels, which showed a negative correlation with the tumor purity of all cancers [36], to predict the 6 TIIC subpopulation abundances. Furthermore, "Gene module" and "Diff Exp module" were utilized for analyzing the CXCL13 level within diverse cancers and the associations of CXCL13 levels with 6 TIIC subpopulation abundances. Wilcoxon test was utilized to assess the signi cant difference in CXCL13 levels. Statistical signi cance and purity-adjusted partial Spearman's correlation were adopted to assess the association between CXCL13 levels and immune in ltration. Tumorin ltration degrees across cancers showing distinct somatic copy number alterations (SCNA) were compared for CXCL13 using "SCNA module" de ned by GISTIC 2.0. The module includes high ampli cation (2), arm-level gain (1), diploid/normal (0), arm-level deletion (−1), and deep deletion (−2). [37] In addition, we utilized "Correlation

Results
The mRNA Expression Levels of CXCL13 in Different Types of Human Cancers For evaluating CXCL13 levels within cancer and non-carcinoma samples, this study determined CXCL13 levels in diverse cancer and non-carcinoma samples based on Oncomine database. As a result, CXCL13 levels increased in diverse cancers, such as breast cancer (BC), bladder cancer, cervical cancer (CC), leukemia, head and neck cancer (HNC), lymphoma, lung cancer, and OV, in comparison with non-carcinoma samples ( Figure 1A). In addition, its expression decreased in kidney cancer, colorectal cancer (CRC), and sarcoma within certain datasets.
Supplementary Table 1 presents more details on CXCL13 levels within diverse cancers.
For evaluating CXCL13 levels within human cancers, this study determined CXCL13 levels based on RNA-seq data from TCGA-derived cancers. Figure 1B

Analysis based on Human Protein Atlas (HPA) Database
The HPA database (https://www.proteinatlas.org/) covers gene and pathological data collected from numerous studies conducted using different cell lines and tissue types [42]. This database was adopted in the present work for examining CXCL13 levels within diverse tissues together with CXCL13 mRNA localization in cells.

Analysis based on LinkedOmics Database
LinkedOmics represents an openly accessible database, which covers multi-omics data of 32 TCGA-derived cancers [43]. Pearson test was conducted for statistical analyses of CXCL13 co-expression by LinkedOmics of "LinkFinder".

Correlation between CXCL13 and gene markers of immune cells
To reveal the underlying associations of CXCL13 expression with immune in ltration degree, we analyzed the associations of CXCL13 level with TIIC gene markers in OV based on GEPIA2 and TIMER.
Based on prior studies, the current work adopted the commonly used TIIC gene markers and diverse functional T cells. Table 2 presents the tumor purity-adjusted results of correlation analysis for OV. Clearly, CXCL13 was markedly related to gene markers of monocytes, B cells, TAMs, CD8+ T cells, neutrophils, T cells, DCs, mast cells, NK cells, and many functional T cells. It was interesting that such observations veri ed that CXCL13 was related to T cells, B cells, as well as functional T cells reported previously, and illustrated the tight relationship of CXCL13 with mast cells.
For con rming the above observations, this study also examined the correlations of CXCL13 levels with TIIC gene markers within OV and non-carcinoma tissues using GEPIA2 (Table 3) (in ammation), and C4 (lymphocyte depletion) types. The highest and lowest CXCL13 levels were detected in C2 and C3 types, respectively ( Figure 5A). This study also detected CXCL13 levels within diverse OV molecular subtypes in TISIDB. There are 4 molecular subtypes detected in OV [49], including immunoreactive, mesenchymal, proliferative, and differentiated. As discovered by TISIDB, the greatest and lowest CXCL13 levels were detected in immunoreactive and proliferative subtypes, respectively ( Figure 5B), suggesting the close relationship between CXCL13 expression and tumor immune microenvironment (TIME). Similarly, the comparison of different OV stages ( -IV) was signi cant based on the GEPIA2 database (P=0.0288) ( Figure 5C). As revealed by HPA-based data analysis, intense CXCL13 staining was detected in OV samples relative to non-carcinoma ovarian tissues ( Figure  5D).

CXCL13 co-expression network in OV
To further understand CXCL13's biological signi cance in OV, the LinkedOmics of "LinkFinder" module was adopted for checking the CXCL13 co-expression patterns. Figure 6A revealed that, altogether 8577 genes (red dots) showed positive correlation with CXCL13, whereas 11433 (green dots) presented a negative correlation (p < 0.05). Figures   6B and 6C display the heatmaps for the 50 most signi cant CXCL13-related genes (both positive and negative). According to the GSEA-annotated GO terms, CXCL13 co-expression genes were mostly associated with leukocyte differentiation, leukocyte cell-cell adhesion, response to a molecule of bacterial origin, immune response-regulating signal transduction pathway, and the regulation of immune effector process. In contrast, it is not involved in cytoskeleton-dependent intracellular transport, ribonucleoprotein complex subunit organization, and microtubule bundle formation ( Figure 6D). KEGG analysis revealed that the genes were mostly related to osteoclast differentiation, cell adhesion molecules (CAMs), cytokine-cytokine receptor interaction, chemokine signaling pathway, hematopoietic cell lineage, natural killer cell-mediated cytotoxicity, NK-kappa B signaling pathway, and phagosome ( Figure 6E). miRNA screening of regulatory CXCL13 miRWalk was applied to screen the targeted miRNAs of CXCL13. Then, miRWalk was used to draw the miRna gene network. As illustrated in Figure 6F, the interaction network consists of CXCL13 and 191 miRNAs. Moreover, the contribution level of miRNAs to CXCL13 is presented as the number of lines. Additionally, the top 20 miRNAs targeting CXCL13 are presented in Figure 6F.

Discussion
CXCL13, a kind of homeostasis chemokine, was originally called BCA-1 or BLC. CXCL13 participates in tumor genesis, proliferation, metastasis and survival of cancer cells [4][5][6][7][8]. However, its relations with T cell function, immune in ltration, and prognosis of diverse cancers remain unclear. Therefore, this study examined cancer samples from multiple databases for analysis. As a result, CXCL13 expression was related to the prognosis of different cancers, in particular OV. Moreover, CXCL13 co-expression genes also have signi cant prognostic signi cance in ovarian cancer. CXCL13 levels revealed a positive correlation with immune in ltration degree within OV. After examining the associations of gene levels among diverse T cells, CXCL13 was con rmed to signi cantly relate to many functional T cells within OV, especially exhausted T cells. Therefore, CXCL13 might serve as a candidate prognostic biomarker for OV, which offers a new direction to analyze the associations of CXCL13 expression with T cell function and immune in ltration degree.
The present work analyzed CXCL13 expression with systematic prognosis of diverse cancers based on separate datasets from Oncomine and 33 TCGA-derived cancers from GEPIA2. CXCL13 was differentially expressed between tumor and non-carcinoma samples in diverse cancers. According to Oncomine database-based analysis, CXCL13 expression increased in Bladder cancer, BC, CC, HNC, lymphoma, leukemia, OV, and lung cancer compared to normal tissues, whereas certain datasets indicated that CXCL13 was lowly expressed within CRC, kidney cancer and sarcoma ( Figure 1A). However, TCGA-based data analysis revealed that CXCL13 was upregulated in BRCA, CESC, COAD, DLBC, ESCA, HNSC, KIRC, LUAD, LUSC, OV, PAAD, READ, SKCM, STAD, TGCT, THYM, UCEC, and UCS, compared with normal adjacent tissues ( Figure 1B). Human Protein Atlas data also veri ed that CXCL13 expression increased within ovarian cancer, as suggested by Immunohistochemistry ( Figure 5D).
The different CXCL13 levels within diverse cancers from diverse databases might re ect different data extraction methods and biological properties. Based on Kaplan-Meier Plotter and GEPIA2 data analysis, CXCL13 downregulation predicted the dismal survival of OV, BRCA, ACC, and HNSC ( Figure 2). In addition, based on univariate analysis and multivariate analysis, CXCL13 expression was signi cantly correlated with TNM, stage of the patient, age, gender, histology, and grade, except for race. CXCL13 level was related to TNM stage, corresponding to LNM degree within OV, and TNM stage exhibited the highest HR (Table 1). Collectively, the above results indicated that CXCL13 might serve as a prognostic marker for OV.
The present work evaluated the association of CXCL13 levels with the immune system based on TISIDB database. According to our ndings, CXCL13 was the most signi cantly related to lymphocytes (including Th1, Act-B, and Act-CD8), immunoinhibitors (such as CTLA4, PDCD1LG2, and TIGIT), MHC molecules (like HLA-B, TAP1, HLA-F), and immunostimulators (such as CD27, CD48, and ICOS). Epigenetic silencing of T1-type chemokines can be a new immune escape mechanism in cancer, while selective epigenetic reprogramming promotes the anti-OV therapeutic effect [50]. Membrane-bound PD-L1 has been the most signi cant OV biomarker over the last decade, which is induced by TAMs-derived soluble in ammatory factors, resulting in immune invasion [51]. Simultaneous blocking of PD-1-PD-L1and CXCL12-CXCR4 pathways can suppress OV proliferation and avoid immunosuppression [52].
Therefore, CSF1R inhibitors may be used to promote the PD-L1 antibody e cacy together [53]. Consequently, CXCL13, which is related to the above immune molecules, offers a novel target to study immune escape in OV, which can be used to be the immunotherapeutic target for OV.
However, OV is by no means a single disorder, which is further classi ed as numerous molecular subtypes.
According to TISIDB database-based analysis, CXCL13 gene displayed the greatest expression within the immunoreactive subtype, while that within the mesenchymal type ranked the second place, and CXCL13 was lowly expressed within differentiated and proliferative types. Differential CXCL13 expression within OV of diverse immune subtypes was detected. The results suggested that the C2 displayed the greatest expression relative to that in the remaining 3 subtypes. The integrative analysis of CXCL13 gene levels across OV and diverse subtypes from various databases possibly suggests that CXCL13 is closely associated with the immune characteristics in TME.
Given that CXCL13 has an important effect on the immune system and on predicting the prognosis of OV, this study examined the association of CXCL13 with immune in ltration degree within OV ( Figure 4A). As a result, CXCL13 upregulation was closely associated with the in ltration degrees of many immune cells, like B cells, CD4+T cells and especially, CD8+ T cells, dendritic cells and neutrophils, which have a stronger correlation levels. Additionally, DC in ltration was signi cantly correlated with OV prognosis ( Figure 4B). The diverse SCNA of CXCL13 did not signi cantly affect the macrophage immune in ltration degrees within OV ( Figure 4C), and we paid attention to the relationship of CXCL13 with immune cells.
As suggested by subsequent analyses on the relationships of CXCL13 with TIIC gene markers, CXCL13 interacted with many immune cells and diverse functional T cells, including central memory T cells, effector T cells, and exhausted T cells (Tables 2 and 3). Since T cell exhaustion accounts for a leading reason for the ineffective anticancer immunity [54][55][56], the measures for preventing T cell exhaustion represent the keys to anticancer immunotherapy. Based on our results, CXCL13 upregulation showed a positive correlation with several critical genes related to exhausted T cells, such as TIM-3, PD-1, LAG3, TIGIT, and GZMB. These are therapeutic targets for immunotherapy [57,58].
According to our ndings, CXCL13 exerts dual functions, where its upregulation shows a positive correlation with favorable survival of some cancers including OV. In the meantime, it can induce T cell exhaustion that may induce ine cient anticancer immunity. Consequently, CXCL13 has important yet different functions in normal immune development and in the regulation of TME, which deserves further investigation.
This study identi ed that CXCL13 was related to mast cells within OV, which has not been reported previously. Mast cells exert the effector activity in the case of TH2-skewed autoimmune and allergic in ammation, enhance su cient in ammatory responses, and activate T cell in cooperation with DCs [59]. Some recent studies suggest that mast cells do signi cantly affect TME conformation or promote cancer development [60,61]. In our study, Availability of data and materials Data supporting our ndings are already included in the manuscript.

Competing interests
The authors declare that they have no competing interests.     Supplementary Files