The implications of TMSB4X in TIM3 hypermethylation and CD8+ T cell exhaustion in diffuse large B-cell lymphoma Authors: Yiming Yang, Zujun Qin, Xin Liu, Huimin Yang, Xiaolin Wang, Chao Mai, Shunhai Jian, Qiqi Zhu Published: 2026 ABSTRACT Epigenetic modifications of immune checkpoints (IC) are crucial for immune checkpoint blockade (ICB) efficacy. TIM3 accelerates CD8+ T cell exhaustion in diffuse large B-cell lymphoma (DLBCL), yet its epigenetic regulation remains unclear. We analyzed CD8+ T cell localization and TIM3 expression using single-cell RNA sequencing (scRNA-seq) and immunofluorescence. DNA methylation analysis evaluated the epigenetic status of typical IC genes across TIM3 expression groups. Integrating genetic mutation, DNA methylation, scRNA-seq, and bulk RNA sequencing data identified a hub gene for TIM3 regulation. The biological and clinical significance of this hub gene, including impacts on immune activity, tumor microenvironment composition, clinicopathological features, prognosis, ICB response, and potential therapeutic options, was comprehensively investigated using scRNA-seq, bulk RNA sequencing data, and immunohistochemistry. Differentially methylated positions were identified in key exhaustion signature genes (TIM3, PD1, CTLA4, LAG3, STAT1) between TIM3-high and TIM3-low groups. Hypermethylation of TIM3’s 3’UTR region correlated with transcriptional upregulation, suggesting positive regulation through TMSB4X loss, frequently altered by somatic mutation. Decreased TMSB4X expression is associated with epigenetic dysregulation, impaired immune function, aggressive clinicopathological features, poor prognosis, and inferior chemotherapy response. Patients with low TMSB4X may benefit from ICB therapy and certain epigenetic-targeting agents. TMSB4X links mutation and epigenetics in TIM3-induced CD8+ T cell exhaustion, providing a prognostic biomarker and potential therapeutic strategies for DLBCL.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-54670-2. Introduction As a major subtype of aggressive lymphatic hematologic malignancies, diffuse large B-cell lymphoma (DLBCL) accounts for about 30 ~ 40% of non-Hodgkin lymphoma cases1. The heterogeneity in genomics present in DLBCL significantly affects pathological manifestation and clinical course, as exampled by the subset classification of activated B-cell (ABC) and germinal center B-cell (GCB) which correlates with variable prognosis1. Many DLBCL patients benefited from the current treatment regimens, represented by rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP), as well as polatuzumab-based regimens, and potential options such as autologous stem cell transplant or chimeric antigen receptor (CAR) T-cell therapy2,3. However, additional therapeutic approaches are needed for the remaining relapsed/refractory patients. The tumor environment (TME) in DLBCL is mainly composed of CD4+ T cells, CD8+ T cells, macrophages, natural killer (NK) cells, dendritic cells (DC), and possible benign B cells4. CD8+ T cells could kill tumor cells through direct killing and secretion of cytotoxic cytokines, and are essential in anti-tumor responses, while the exhausted CD8+ T cells induced by immune checkpoint (IC) over-expression may facilitate tumor immune evasion5,6. Over-expression of TIM3, a critical IC, has been reported in CD8+ T cells and indicated the dysfunction of CD8+ T cells, immune deficiency, and tumor progression in DLBCL7,8. Targeting TIM3, a novel IC encoded by HAVCR2, has emerged as a promising therapy to complement traditional chemotherapy, radiotherapy, and immune therapy9. However, only a limited number of patients demonstrated effective response to IC blockade (ICB) therapy including TIM3, and the underlying mechanism remains to be explored10. The regulation of gene expression that incorporates inherent transcriptional patterns with external signals, is influenced by environmental variables, and is independent of changes in DNA sequence, is known as epigenetics, mainly including DNA methylation and histone modification11. Emerging studies reveal that epigenetic determinants are central to the formation and maintenance of CD8+ T cell exhaustion12. More importantly, the overall epigenetic landscape in exhausted CD8+ T cells is nearly unaltered after ICB treatment in tumor and chronic infections, limiting persistent anti-tumor immunity and critically influencing ICB efficacy13. Meanwhile, overcoming of the epigenetic barrier may be a promising strategy to rejuvenate exhausted CD8+ T cells13. Besides, the intricate interplay between epigenetics and gene mutation could also determine the fate of the terminal transition of progenitor-exhausted CD8+ T cells, the target of which synergized with anti-PD-L1 treatment achieved enhanced anti-tumor ability14. However, the role of epigenetics and somatic mutation in TIM3 expression remains poorly studied in DLBCL. In this study, the expression of TIM3 in CD8+ T cells from DLBCL was analyzed through single-cell RNA sequencing (scRNA-seq) and validated by immunofluorescence. The difference in epigenetic status in overall genes and typical IC, together with the detailed CpG sites between high and low TIM3 expression groups were evaluated utilizing DNA methylation analysis. The hub gene (TMSB4X) involved in TIM3 expression regulation was further explored based on the intersection of candidates from somatic mutation and DNA methylation analysis. The expression of TMSB4X in DLBCL was examined, along with its associations with CD8+ T cell immunological activity, TME composition, clinicopathological characteristics, and prognosis. Additionally, the significance of TMSB4X in ICB response and the potential drugs were predicted. The purpose of this research is to investigate the regulatory role of epigenetics and somatic mutation in TIM3 expression and offer a biomarker for clinical assessment, which also potentially facilitates the improvement of DLBCL therapy. Materials and methods (The study design is shown in supplementary methods) Patients Eighty-eight cases of de novo DLBCL were sourced from the Affiliated Hospital of North Sichuan Medical College’s Department of Pathology from March 2017 to December 2024. The World Health Organization’s categorization criteria for DLBCL (4th edition, 2018) were used to determine the diagnoses. This study excluded patients with a history of secondary or recurrent DLBCL. Single-cell RNA sequencing The scRNA-seq genomic data of DLBCL and healthy controls originated from the heiDATA database15 and the Gene Expression Omnibus (GEO) database (GSE182434 and GSE182436)16. We used our previous approach to analyze the scRNA-seq data17. The single-cell biological activities of exhausted CD8+ T cells were assessed using the Gene Set Variation Analysis (GSVA) focused on the Reactome database. The AUCell package was applied to determine the activity of exhaustion, cytotoxicity, and inflammatory responses in exhausted CD8+ T cells. Immunofluorescence The co-expression of CD8 and TIM3 was examined by immunofluorescence using formalin fixed, paraffin embedded (FFPE) slices of DLBCL lymph node tissue. Representative cases from the same DLBCL cohort that was analyzed for immunohistochemistry (IHC) (see Sect. 2.7) were used. The primary antibodies used were anti-TIM3 (Abcam, EPR22241) and anti-CD8 alpha (Abcam, EPR22483-288). The immunofluorescence experiment was performed by Wuhan Servicebio Biotechnology (Wuhan, China). Analysis of DNA methylation For 48 DLBCL cases, matching RNA sequencing data and DNA methylation information from Illumina 450 K arrays were gathered from The Cancer Genome Atlas (TCGA). Normalized CpG beta values were used to contrast the methylation levels of the high and low TMSB4X groups based on Differential Methylation Probe (DMP) analysis. Analysis of genetic mutations The TCGAbiolinks software was used to extract the genetic mutation profile of 48 DLBCL samples from the TCGA database17. We used the Sangerbox platform (http://sangerbox.com/) to characterize somatic mutations across various TIM3 expression groups. RNA sequencing For DLBCL patients, RNA sequencing data and accompanying clinical and pathological parameters have been collected from TCGA and GEO datasets, comprising GSE181063, GSE117556, GSE11318, GSE53786, GSE32918, GSE10846, and GSE31312. We used CIBERSORT (https://cibersort.stanford.edu/) to assess the relationship between the TME constitution and TMSB4X expression. IHC IHC was used to examine TMSB4X protein expression in FFPE lymph node tissue slices of DLBCL (n = 88). Anti-TMSB4X (Proteintech, Catalog: 19850-1-AP) was the primary antibody against TMSB4X. TMSB4X expression was mainly observed in the cytoplasm of TILs from DLBCL. The average numbers of TMSB4X+ lymphocytes per high-power field (HPF) at the hot spot area were used to assess the stains manually. Predicting drug sensitivity The effectiveness of ICB administration in DLBCL was measured using the Tumor Immune Dysfunction and Exclusion (TIDE) score. The Genomics of Drug Sensitivity in Cancer (GDSC) database was used to predict the sensitivity of DLBCL patients to prevalent medications. The oncoPredict package was used to compare the half-maximal inhibitory concentration (IC50) of the associated drugs between the high and low TMSB4X groups. Analytical statistics R software (version 4.4.0) and SPSS software (version 26.0; SPSS Corp., Chicago, IL, USA) were used for statistical analyses. Chi-squared tests and Mann-Whitney U-tests were used to compare categorical data, whereas nonparametric tests were used to evaluate numerical data. Spearman’s correlation coefficients were computed to investigate the associations between IC gene expression and the corresponding cpg values, TMSB4X and HAVCR2 expression, and TMSB4X and the single-cell biological activity of exhausted CD8+ T cells. For survival analysis, the optimal cut-off value for TMSB4X expression was determined using receiver operating characteristic (ROC) curve analysis. Overall survival (OS) was evaluated by the Kaplan-Meier method with the log-rank test. To integrate prognostic data across multiple cohorts, we performed meta-analysis using a random-effects model. We presented the results in a forest plot, displaying hazard ratios (HRs) with 95% confidence intervals (CIs) for each dataset and the pooled estimate. A two-tailed p-value < 0.05 was considered statistically significant for all analyses. Results The overall methylation status associated with TIM3 expression in DLBCL UMAP analysis of single-cell transcriptomic data from DLBCL and controls discerned seven kinds of major clusters: Naive T cells, CD4+ T cells, CD8+ T cells, B cells, NK cells, macrophages, and DC (Fig. 1a), and the cellular composition in the respective case is shown in Fig. 1b. The distribution of HAVCR2 expression is mainly overlapped with that of CD8+ T cells both on transcription level (Fig. 1c) and protein level (Fig. 1d). The overall methylation status in DLBCL was evaluated, and certain differences were observed between high and low TIM3 expression groups based on PCA analysis (Fig. 1e). An overview of the CpG sites is shown in Supplementary Fig. 1a. Comparison of differentially methylated sites demonstrated the marked difference in the gene body, followed by distal intergenic region (IGR), transcription start sites (TSS) 1500, TSS200, 5’ untranslated region (UTR), the first exon (1stExon), and 3’UTR (Fig. 1f), and mainly occupied with the functional region of island and opensea on DNA strand (Fig. 1g). The top 35 genes mostly enriched by significant CpGs were demonstrated in Supplementary Fig. 1b. The methylation changes of exhaustion-related signatures involved in different TIM3 expression groups A detailed description of the differential methylation probe in terms of exhaustion-related signatures revealed that multiple cpg sites were enriched in PDCD1 (Fig. 2a), CTLA4 (Fig. 2b), LAG3 (Fig. 2c), STAT1 (Fig. 2d), as well as TIM3 (Fig. 2e). The corresponding beta-value of representative cpg sites demonstrated negative correlations with their gene expression for PDCD1 (r = − 0.545, p < 0.01), CTLA4 (r = − 0.515, p < 0.01), LAG3 (r = − 0.285, p = 0.049), STAT1 (r = − 0.5421, p < 0.01), while the positive correlation for HAVCR2 (r = 0.558, p < 0.01) (Fig. 2f). The information of those representative cpg sites is shown in Fig. 2f, and the feature for PD1 and STAT1 was gene body, while their counterparts for CTLA4, LAG3, and HAVCR2 were 1stExon, TSS200, and 3’UTR, respectively (Fig. 2g). The functional region of those cpg sites located in opensea for CTLA4, STAT1, and HAVCR2, island for PDCD1, and shore for LAG3 (Fig. 2g). The hub gene involved in TIM3 epigenetic and mutational regulation The somatic mutation difference between high and low TIM3 expression groups was analyzed, and the results suggested that the high TIM3 expression group had a higher mutation rate of TMSB4X compared with the low TIM3 expression group, including missense mutation, splice site, nonsense mutation, and in-frame deletion (Fig. 3a). TMSB4XP3, the variation of TMSB4X, demonstrated a negative correlation with 3’UTR methylation level of HAVCR2 at site cg18374914 (r = − 0.425, p < 0.01) (Fig. 3b), while TMSB4X showed a negative tendency (Fig. 3c). Further analysis confirmed the negative relationships between TMSB4X and HAVCR2 expression in DLBCL through larger sample sizes from GSE10846 (Fig. 3d) and GSE53786 (Fig. 3e). These results suggested the possible regulatory role of decreased TMSB4X induced by mutation in epigenetic manipulation of TIM3 expression. The impact of TMSB4X expression on immune function in DLBCL The expression of TMSB4X in DLBCL and controls was compared. The results showed that the overall expression of TMSB4X in DLBCL (Fig. 4a) and the expression of TMSB4X in CD8+ T cells (Fig. 4b) from DLBCL were both lower than those in the control group. Meanwhile, TMSB4X was widely expressed in the TME (Fig. 4c). Among CD8+ T cells, TMSB4X expression level was relatively low in exhausted subgroups, while high in cells with moderately enhanced cytotoxic function and inflammatory responses (Fig. 4d). The results of Reactome analysis showed that TMSB4X was involved in the variable processes of epigenetic regulation, such as methylation of DNA and protein, and acetylation of histones, and high TMSB4X was positively correlated with the immune response, including antigen processing cross-presentation and cytokine production in exhausted CD8+ T cells (Fig. 4e). The significance of TMSB4X in TME composition, clinicopathological features, and prognosis evaluation CIBERSORT was used to compare the immune cell infiltration of the high and low TMSB4X groups (Fig. 5a). The results showed that T regulatory cells, activated mast cells, activated NK cells, resting NK cells, T follicular helper cells, and naive B cells were reduced in the high TMSB4X group, while gamma delta T cells, activated memory CD4+ T cells, naive CD4+ T cells, and resting mast cells were more infiltrated (Fig. 5b). The clinicopathological analysis’s findings revealed that patients older than 60, with high PS scores (2–5) and ABC-subtype, had lower TMSB4X expression than their counterparts, and also demonstrated a poorer response to treatment (Fig. 5c). The prognostic significance of TMSB4X was first evaluated in seven independent GEO datasets (GSE181063, GSE117556, GSE11318, GSE53786, GSE32918, GSE10846, and GSE31312). The analysis revealed that low TMSB4X expression was associated with poorer OS in these cohorts (Fig. 6a-g). A forest plot integrating all datasets confirmed that low TMSB4X expression consistently represented a higher risk to patients (Fig. 6h). Subsequently, TMSB4X protein expression was examined by IHC in our in-house DLBCL cohort (patient profiles in Table 1), and patients were stratified into high and low groups according to the median count of TMSB4X+ lymphocytes (4/HPF; Fig. 6i). In line with the previous findings, the results from our cohort further confirmed that patients with low TMSB4X expression had inferior OS (Fig. 6j). Table 1Baseline profiles of the enrolled DLBCL patients.CharacteristicNoIn group(%)Age, mean (range)61.1(25–88)Age > 60 years50/8856.8 ≤ 60 years38/8843.2Gender Male46/8852.3 Female42/8847.7COO GCB20/8822.7 Non-GCB68/8877.3PS 0–156/8466.7 2–528/8433.3Stage I/II25/8529.4 III/IV60/8570.6IPI 0–127/8531.8 2–558/8568.2B-symptom Yes48/8258.5 No34/8241.5Primary site Nodal49/8756.3 Extranodal38/8743.7LDH > 220 IU/L Yes35/6851.5 No33/6848.5Treatment CHOP- based therapy67/7391.8 Other therapy6/738.2Response to CHOP CR + PR37/5863.8 SD + PD21/5836.2COO, cell of origin; GCB, germinal center B-cell-like; PS, performance status; LDH, lactate dehydrogenase; CHOP, cyclophosphamide, doxorubicin, vincristine, prednisone; CR, complete remission; PR, partial remission; SD, stable disease; PD, progressive disease. Table 1Baseline profiles of the enrolled DLBCL patients.CharacteristicNoIn group(%)Age, mean (range)61.1(25–88)Age > 60 years50/8856.8 ≤ 60 years38/8843.2Gender Male46/8852.3 Female42/8847.7COO GCB20/8822.7 Non-GCB68/8877.3PS 0–156/8466.7 2–528/8433.3Stage I/II25/8529.4 III/IV60/8570.6IPI 0–127/8531.8 2–558/8568.2B-symptom Yes48/8258.5 No34/8241.5Primary site Nodal49/8756.3 Extranodal38/8743.7LDH > 220 IU/L Yes35/6851.5 No33/6848.5Treatment CHOP- based therapy67/7391.8 Other therapy6/738.2Response to CHOP CR + PR37/5863.8 SD + PD21/5836.2COO, cell of origin; GCB, germinal center B-cell-like; PS, performance status; LDH, lactate dehydrogenase; CHOP, cyclophosphamide, doxorubicin, vincristine, prednisone; CR, complete remission; PR, partial remission; SD, stable disease; PD, progressive disease. Baseline profiles of the enrolled DLBCL patients. COO, cell of origin; GCB, germinal center B-cell-like; PS, performance status; LDH, lactate dehydrogenase; CHOP, cyclophosphamide, doxorubicin, vincristine, prednisone; CR, complete remission; PR, partial remission; SD, stable disease; PD, progressive disease. The significance of TMSB4X in ICB therapy and potential drugs screening The TIDE algorithm was used to assess the relationship between TMSB4X expression and the effectiveness of ICB therapy. The findings indicated that lower TIDE scores were associated with lower TMSB4X expression (Fig. 7a), indicating that DLBCL patients with decreased TMSB4X levels may be susceptible to ICB therapy. The IC50 values of several medications were examined in order to give an optional regimen for individuals with low TMSB4X expression (Fig. 7b). The findings showed that several epigenetic regulators, including GSK343, Fludarabine, Zoledronate, PCI-34,051, and Carmustine, had reduced IC50 values within the low TMSB4X expression group. Additionally, the IFN pathway inhibitor JAK1_8709 shown increased sensitivity in the group with decreased TMSB4X expression. Discussion ICB therapy aims to rewire exhausted CD8+ T cells to enhance immune function and suppress tumor proliferation mainly through targeting up-regulated IC, and has improved the treatment of many cancer types18. TIM3 is a crucial member of IC apart from PD-1 and CTLA4, and could induce CD8+ T cells exhaustion through the interaction with its ligands represented by galectin-9, accompanied with diminished cytotoxic function and inferior outcome19. This phenomenon is frequently observed in solid tumors and lymphohematopoietic neoplasms, such as lung cancer20, liver cancer21, colorectal cancer22, follicular lymphoma23, as well as DLBCL8. Consistently, our study confirmed the over-expression of TIM3 in CD8+T cells from DLBCL, suggesting its potential role in the formation of exhausted status and possible therapeutic value of ICB7. However, the therapeutic effect of ICB in DLBCL is unsatisfactory24, which is similar to the circumstance in other malignant neoplasms and needs further investigation to expand the proportion of benefited patients. Recent studies stress the importance of epigenetics alterations in the persistent aberrant expression of signature markers related to CD8+ T cells exhaustion, including typical IC, transcription factors (TF), and effector function related genes5,25. Meanwhile, ICB fails to change the epigenetic profile of exhausted phenotype, indicating the transient potentiated anti-tumor responses and correlates ICB resistance and tumor relapse26. These highlight the pivotal role of epigenetic alterations in restoring and sustaining the tumor cell-killing ability of CD8+ T cells27. DNA methylation, a primary form of epigenetic profile, has been implicated in shaping CD8+ T cells exhaustion28. Our study identified the obvious difference in overall methylated CpG sites between high and low TIM3 expression groups, implying that the DNA methylation plays a possible role in TIM3 induced CD8+ T cells exhaustion. Interestingly, focus of exhaustion-related signatures revealed that TIM3 was hypermethylated while the remaining IC, including PD1, CTLA4, LAG3, and STAT1, the critical TF of TIM37, were all hypomethylated in high TIM3 group. Consistently, the methylation beta-value of TIM3 showed a positive correlation with gene expression, while the methylation level of PD1, CTLA4, LAG3, and STAT1 showed a negative correlation with their respective gene expressions. Further analysis demonstrates that the 3’UTR was the crucial TIM3 hypermethylated region, while the hypomethylated site of remaining IC and TF located in the gene body, TSS, and the first exon. Generally, hypermethylation indicates regressed gene expression, while hypomethylation induces active gene transcription. However, the influence depends on context by the exact differentially methylated sites, including 5’UTR, enhancer, promoter, TSS, the first exon, gene body, and 3’UTR. For example, promoter methylation usually decreases while gene body methylation frequently increases but could also decrease gene expression, suggesting the DNA methylated region possibly determines the specific function of gene regulation29. In fact, Sood AK elucidated that increased 3’UTR methylation of TIM3 indicated its enhanced expression, which could be reversed by methyltransferases inhibitor30. Moreover, epigenetic reprogramming of MiR-379-5p forced expression could decrease the TIM3 protein expression in CD8+ T cells by directly targeting 3’UTR, and exhibited improved capability to eradicate cancer31. Collectively, our study demonstrated the significant difference in DNA methylation especially for typical IC in high and low TIM3 expression groups, and suggested that the enrichment of 3’UTR methylation in TIM3 may play a critical role in upregulating TIM3 expression, highlighting the significance of epigenetic changes in TIM3 induced CD8+ T cells exhaustion from DLBCL. Genetic mutation has profound effects on epigenetic modifications. For instance, the mutating bases32 and motifs33 might impede gene transcription by inducing minor changes in DNA organization and interfering with the interplay between enhancer and promoter. Meanwhile, specific gene mutation, especially for histone methyl transferases and other epigenetic regulaters34, may cause immense impact on DNA methylation35 and result in widespread epigenetic disorder36. Interestingly, epigenetic alterations could also contribute to DNA mutation37, and the chromatin structure through histone modifications may have affected the rate of mutation38. Emerging evidence uncovers the complicated crosstalk between genetic mutation and epigenetic changes exerts a prominent influence on the course of the disease. The mutation of the chromatin remodeling complex subset in pancreatic cancer, such as PBRM1, fails to bind to the promoter and hinders the expression of vimentin, leading to poorly differentiated tumors and metastasis39. In head and neck squamous cell carcinomas, the NSD1 mutation defines vulnerability to metabolic inhibitors in an epigenetic manner, suggesting the value in cancer treatment40. Moreover, the EZH2 mutations in DLBCL profoundly affect the epigenetic alterations and tumor heterogeneity41, and inhibitors could enhance the immunogenicity of tumor cells, alleviate the exhaustion level of CD8+ T cells, and potentiate the therapeutic effect of CAR-T42. Our study further excavated the hub gene implicated both in gene mutation and epigenetic changes and identified TMSB4X mutation only presented in the high TIM3 expression group and the negative correlations among TMSB4X, TIM3 methylation level, and TIM3 expression. TMSB4X is a member of the thymosin family and participates in variable biological processes43. TMSB4X could stimulate immunological responses by promoting the differentiation of DCs44, prevent the acquisition of cancer-associated fibroblasts-like phenotype of fibroblasts, and disrupt the migratory capacity of tumor cells45. It is worth noting that TMSB4X mutation is frequently discovered in DLBCL and may be important in the pathogenesis and development of tumors46. Our results showed that TMSB4X was widely expressed in DLBCL, while exhibiting down-regulated expression compared with healthy controls, and demonstrated positive correlations with immune functions in exhausted CD8+ T cells and actively involved in the methylation and acetylation, indicating the beneficial role of TMSB4X-related epigenetic modification in anti-tumor immunity. According to our results, we speculate that the TMSB4X mutation may facilitate TIM3 expression in CD8+ T cells through epigenetic modifications. Notably, TMSB4X has been demonstrated to suppress MeCP247, a key regulator involved in DNA methylation48, and could influence cellular sensitivity to epigenetic therapy49. Loss of TMSB4X may therefore lead to MeCP2 derepression, which could drive aberrant hypermethylation of the TIM3 3’UTR. This hypermethylation may further be associated with increased TIM3 expression30, potentially contributing to the exhausted phenotype, impaired effector function of CD8+ T cells, and weakened immune response in DLBCL. The significance of TMSB4X in TME composition was evaluated, and our results revealed that high TMSB4X implied increased infiltration of gamma delta T cells, activated memory CD4+ T cells, naive CD4+ T cells, and resting mast cells, while decreased T regulatory cells, activated mast cells, activated NK cells, resting NK cells, T follicular helper cells, and naive B cells. Among these changes, a high proportion of gamma delta T cells and activated T cells suggested the immune activation and enhanced tumor control capacity50 associated with high TMSB4X expression, and the reduced T regulatory cells indicated an anti-tumor TME51. These alterations possibly induced by TMSB4X mutation and epigenetic modifications were consistent with the report that the formation of a suppressive TME in colorectal cancer is enabled by initial driver gene mutations and epigenetic changes52. Concurrently, the TMSB4X expression was preferentially decreased in DLBCL patients within ABC-subtype, older age, higher PS score, and lower response to treatment, all of which emphasized the significance of down-regulated TMSB4X in DLBCL progression. This was further supported by the finding that high TMSB4X expression was protective for the survival of DLBCL patients and underpinning the clinical value of TMSB4X in DLBCL. Based on the potential role of TMSB4X in TIM3-induced CD8+ T cell exhaustion and broad effects on epigenetic changes, our study evaluated the relationship between TMSB4X expression and ICB efficacy. Our results showed that DLBCL patients with low TMSB4X expression were sensitive to ICB treatment. Additional possible drugs were also predicted for those patients, and GSK343, Fludarabine, Zoledronate, PCI-34,051, Carmustine, and JAK1_8709 were extracted. Interestingly, many of those are epigenetic regulators, suggesting alternative avenues of modifying epigenetic changes to complement TMSB4X deficiency. For example, GSK343 is a methyltransferase inhibitor and could induce the hypomethylation of promoters to regulate gene transcription53,54. Likewise, Zoledronate may alter the degree of histone methylation to govern the production and release of interleukin 1β from macrophages55. Fludarabine was also implicated in DNA methylation regulation and gene expression in chronic lymphocytic leukemia56. These regimens suggested the therapeutic value of DNA methylation modification in TIM3 expression, which may not only provide novel treatment choices but also may increase the response rate to ICB therapy. Besides, JAK1 is the inhibitor of the JAK-STAT pathway, which also plays a key role in TIM3 expression7, and may provide possible candidate therapeutic options for DLBCL patients with low TMSB4X expression. Aside from prognostic assessment, TMSB4X expression can be used to select the appropriate medications to modify the TIM3 DNA methylation and enhance ICB responses in DLBCL. Conclusions This study deciphered the DNA methylation landscape in DLBCL and discerned the differentially methylated positions of exhaustion signature genes including TIM3, PD1, CTLA4, LAG3, and STAT1 between the high and low TIM3 expression groups. Among those epigenetic alterations, hypermethylation of the 3’UTR site in TIM3 indicates its up-regulated transcription and may be positively regulated by the loss of TMSB4X induced by somatic mutation in DLBCL. The progression of DLBCL, which was characterized by aggressive clinicopathological characteristics and a dismal prognosis, was caused by decreased TMSB4X expression, which was crucial in triggering compromised immune function and shaping an immunosuppressive TME. Moreover, patients with low TMSB4X expression were predicted to be sensitive to ICB therapy and certain epigenetic reprogramming such as methyltransferase inhibitor, which also could possibly improve ICB responses (Fig. 8). Above all, our study identified TMSB4X as the hub gene in TIM3-induced CD8+ T cell exhaustion from the epigenetic and mutational perspective and provides a potential therapeutic approach to complement traditional therapy for DLBCL patients. Further experimental research is needed to reinforce our findings. Fig. 1The overall methylation status associated with TIM3 expression in DLBCL. (a–c) UMAP analysis showing (a) seven major cell clusters, (b) their cellular composition, and (c) HAVCR2 (TIM3) expression based on scRNA-seq data. CD8+ T cells are highlighted by the blue box. (d) The co-expression of CD8 and TIM3 in DLBCL on protein level. (e–g) (e) PCA analysis of the overall methylation profile, and the (f) genomic features and (g) functional regions of differentially methylated sites between high and low TIM3 expression groups. The overall methylation status associated with TIM3 expression in DLBCL. (a–c) UMAP analysis showing (a) seven major cell clusters, (b) their cellular composition, and (c) HAVCR2 (TIM3) expression based on scRNA-seq data. CD8+ T cells are highlighted by the blue box. (d) The co-expression of CD8 and TIM3 in DLBCL on protein level. (e–g) (e) PCA analysis of the overall methylation profile, and the (f) genomic features and (g) functional regions of differentially methylated sites between high and low TIM3 expression groups. Fig. 2Methylation alterations of exhaustion-related signatures across TIM3 expression groups. (a–e) Methylation levels of (a) PDCD1, (b) CTLA4, (c) LAG3, (d) STAT1, and (e) HAVCR2 in high versus low TIM3 expression groups. (f) Correlations between the beta-values of representative CpG sites and the expression of their corresponding genes. (g) Detailed characteristics of the representative differentially methylated CpG sites. Methylation alterations of exhaustion-related signatures across TIM3 expression groups. (a–e) Methylation levels of (a) PDCD1, (b) CTLA4, (c) LAG3, (d) STAT1, and (e) HAVCR2 in high versus low TIM3 expression groups. (f) Correlations between the beta-values of representative CpG sites and the expression of their corresponding genes. (g) Detailed characteristics of the representative differentially methylated CpG sites. Fig. 3Identification of a hub gene linking TIM3 regulation to mutation and epigenetics. (a) Somatic mutation profiles in high versus low TIM3 expression groups. Numbers in parentheses indicate the p‑value for the comparison of mutation frequency between the two groups. (b–c) Correlation of the TIM3 methylation level (cg18374914) with (b) TMSB4XP3 and (c) TMSB4X expression (TCGA data). (d–e) Correlation between HAVCR2 and TMSB4X expression in (d) GSE10846 and (e) GSE53786 datasets. Identification of a hub gene linking TIM3 regulation to mutation and epigenetics. (a) Somatic mutation profiles in high versus low TIM3 expression groups. Numbers in parentheses indicate the p‑value for the comparison of mutation frequency between the two groups. (b–c) Correlation of the TIM3 methylation level (cg18374914) with (b) TMSB4XP3 and (c) TMSB4X expression (TCGA data). (d–e) Correlation between HAVCR2 and TMSB4X expression in (d) GSE10846 and (e) GSE53786 datasets. Fig. 4Impact of TMSB4X expression on immune function in DLBCL. (a) Overall TMSB4X expression in DLBCL tissues versus controls. (b) TMSB4X expression in CD8+ T cells from DLBCL versus controls. (c) TMSB4X expression across all cell clusters in the tumor microenvironment. (d) TMSB4X expression levels and the corresponding activity scores of exhaustion, cytotoxicity, and inflammatory responses in CD8+ T cell subgroups. (e) Correlations between TMSB4X expression and the activity of epigenetic regulation or immune responses in exhausted CD8⁺ T cells. Square size represents the absolute Spearman correlation coefficient (|r|). Impact of TMSB4X expression on immune function in DLBCL. (a) Overall TMSB4X expression in DLBCL tissues versus controls. (b) TMSB4X expression in CD8+ T cells from DLBCL versus controls. (c) TMSB4X expression across all cell clusters in the tumor microenvironment. (d) TMSB4X expression levels and the corresponding activity scores of exhaustion, cytotoxicity, and inflammatory responses in CD8+ T cell subgroups. (e) Correlations between TMSB4X expression and the activity of epigenetic regulation or immune responses in exhausted CD8⁺ T cells. Square size represents the absolute Spearman correlation coefficient (|r|). Fig. 5Significance of TMSB4X in tumor microenvironment composition and clinicopathological features. (a) Landscape of immune cell composition in each case. (b) Differences in immune cell infiltration between high and low TMSB4X expression groups. (c) Associations between TMSB4X expression and cell-of-origin (COO), age, PS score, and treatment response. COO: cell of origin; CR: complete response; PR: partial response; SD: stable disease; PD: progressive disease. Significance of TMSB4X in tumor microenvironment composition and clinicopathological features. (a) Landscape of immune cell composition in each case. (b) Differences in immune cell infiltration between high and low TMSB4X expression groups. (c) Associations between TMSB4X expression and cell-of-origin (COO), age, PS score, and treatment response. COO: cell of origin; CR: complete response; PR: partial response; SD: stable disease; PD: progressive disease. Fig. 6Prognostic significance of TMSB4X in DLBCL. (a–g) Kaplan–Meier analysis of OS based on TMSB4X gene expression in seven independent GEO datasets (a: GSE181063, b: GSE117556, c: GSE11318, d: GSE53786, e: GSE32918, f: GSE10846, g: GSE31312). (h) Forest plot displaying HRs with 95% CI for the association between TMSB4X gene expression and OS across the seven independent GEO datasets. (i) The protein expression of TMSB4X detected by IHC. (j) Kaplan–Meier analysis of OS based on TMSB4X protein expression in our cohort. Prognostic significance of TMSB4X in DLBCL. (a–g) Kaplan–Meier analysis of OS based on TMSB4X gene expression in seven independent GEO datasets (a: GSE181063, b: GSE117556, c: GSE11318, d: GSE53786, e: GSE32918, f: GSE10846, g: GSE31312). (h) Forest plot displaying HRs with 95% CI for the association between TMSB4X gene expression and OS across the seven independent GEO datasets. (i) The protein expression of TMSB4X detected by IHC. (j) Kaplan–Meier analysis of OS based on TMSB4X protein expression in our cohort. Fig. 7Significance of TMSB4X in ICB therapy and drug screening. (a) TIDE scores between high and low TMSB4X expression groups. Lower scores suggest potential vulnerability to ICB therapy. (b) IC50 values of selected drugs between high and low TMSB4X expression groups. Lower IC50 values indicate higher drug sensitivity. Significance of TMSB4X in ICB therapy and drug screening. (a) TIDE scores between high and low TMSB4X expression groups. Lower scores suggest potential vulnerability to ICB therapy. (b) IC50 values of selected drugs between high and low TMSB4X expression groups. Lower IC50 values indicate higher drug sensitivity. Fig. 8The predictive model linking TMSB4X deficiency-induced TIM3 hypermethylation to CD8+ T cell exhaustion, immunosuppression, therapy resistance, and aggressive DLBCL. The predictive model linking TMSB4X deficiency-induced TIM3 hypermethylation to CD8+ T cell exhaustion, immunosuppression, therapy resistance, and aggressive DLBCL. Electronic Supplementary Material Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2 Supplementary Material 3 Supplementary Material 3