A Novel CD58 Mutation–Related Signature Predicts Prognosis Risk in Diffuse Large B-Cell Lymphoma Patients

Article information

Cancer Res Treat. 2026;58(3):979-1001

Publication date (electronic) : 2025 July 28

doi : https://doi.org/10.4143/crt.2024.1143

, Wei Hua ¹, Hua Yin ¹, Yue Li ¹, Jun-Heng Liang ³, Rui Gao ⁴, Bi-Hui Pan ¹, Xin-Yu Zhang ¹, Jia-Zhu Wu ¹, Hao-Rui Shen ¹, Li Wang ¹, Jian-Yong Li ¹, Jin-Hua Liang^,¹

, Wei Xu^,¹

¹Department of Hematology, The First Affiliated Hospital with Nanjing Medical University, Jiangsu Province Hospital, Nanjing, China

²Department of Hematology, Sir Run Run Hospital, Nanjing Medical University, Nanjing, China

³Nanjing Geneseeq Technology Inc., Nanjing, China

⁴Department of Endocrinology, The First Affiliated Hospital with Nanjing Medical University, Jiangsu Province Hospital, Nanjing, China

Correspondence: Jin-Hua Liang, Department of Hematology, the First Affiliated Hospital with Nanjing Medical University, Jiangsu Province Hospital, Nanjing 210029, China Tel: 86-25-83781120 E-mail: liangjinhua1990@126.com

Co-correspondence: Wei Xu, Department of Hematology, the First Affiliated Hospital with Nanjing Medical University, Jiangsu Province Hospital, Nanjing 210029, China Tel: 86-25-83781120 E-mail: xuwei10000@hotmail.com

*Hui-Li Wang and Chun-Yu Shang contributed equally to this work

Received 2024 November 27; Accepted 2025 July 21.

Abstract

Purpose

CD58, a ligand of the CD2 receptor on T cells and natural killer cells, is abnormally expressed in diffuse large B-cell lymphoma (DLBCL). However, data on the value of CD58 mutation (CD58^mut) in DLBCL are limited. Here, we aimed to evaluate the characteristics and prognostic value of CD58^mut in DLBCL patients.

Materials and Methods

The available clinical information and corresponding mutation data of DLBCL were obtained from published articles. Ultimately, 3,025 DLBCL patients in published cohorts were enrolled in the final analysis. Among the 202 DLBCL patients in the Jiangsu Province Hospital (JSPH) cohort, all tumor tissue samples were collected to perform next-generation sequencing and gene expression was analyzed via RNA-seq.

Results

We found that 8.2% (250/3,025) of patients were CD58^mut in integrated cohort, whereas 11.4% (23/202) in the JSPH cohort. CD58^mut patients exhibit inferior progression-free survival (the integrated cohort: hazard ratio [HR], 0.96; 95% confidence interval [CI], 0.77 to 1.20; p=0.663 the JSPH cohort: HR, 1.85; 95% CI, 0.85 to 4.04; p=0.052) and overall survival (the integrated cohort: HR, 1.43; 95% CI, 1.15 to 1.77; p < 0.001; the JSPH cohort: HR, 2.40; 95% CI, 0.83 to 6.93; p=0.026). A model based on six signature genes (MRO, OXTR, RASL11A, RLN1, SIGLEC1, and PROM2) was constructed via machine learning. To optimize risk stratification and survival prediction for CD58^mut patients, biological mechanism of the poorer prognosis in high-risk group may be related to the greater abundance of immunosuppressive cells, especially M2 macrophages.

Conclusion

Our results indicated that CD58^mut could serve as a novel prognostic factor for DLBCL patients, and further exploration of personalized treatment strategies for high-risk DLBCL patients based on the risk score model is needed.

Keywords: Diffuse large B-cell lymphoma; CD58 mutation; Prognostic model; Biological mechanism; Risk factors

Introduction

Diffuse large B-cell lymphoma (DLBCL) is a highly heterogeneous and aggressive subtype of non-Hodgkin lymphoma. Despite the significant improvement in prognosis of patients with DLBCL caused by the standard first-line regimen (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone [R-CHOP]), 40% to 50% of patients still experience relapse or refractory disease [1]. The distinctions of cell of origin (COO) add prognostic value to the widely used International Prognostic Index (IPI); however, COO classification has not been satisfactorily applied in predicting clinical outcomes and targeted therapy [2]. With the development of next-generation sequencing (NGS) technology, researchers [3] have developed a lymphocyte genetic subtype classifier. The application of high-throughput sequencing technology has enabled researchers to more fully describe the molecular heterogeneity of DLBCL, but the key molecular genetic changes associated with disease progression and their pathogenic mechanisms need to be further elucidated.

The CD58 gene is a glycosylated adhesion factor widely expressed on the surface of antigen-presenting cells [4]. As a ligand of the CD2 receptor on T and natural killer cells, CD58 mediates the proliferation and activation of lymphocytes, promotes antigen recognition, and enhances the adhesion between effector cells and target cells [5]. Researchers [6] reported that leukemia patients with poor prognoses often lack CD58 expression. CD58 mutation (CD58^mut) occurs in the process of the transformation of follicular lymphoma (FL) to DLBCL [7], and CD58 is one of the recurrence targets of lymphomas, such as peripheral T-cell lymphoma (PTCL) and adult T leukemia/lymphoma (ATLL) [8,9]. Therefore, the purpose of this study was to investigate the clinical significance and prognostic value of CD58^mut in DLBCL, construct a model to optimize survival prediction, and preliminarily explain the molecular mechanisms that lead to different risk score groupings in DLBCL patients with CD58^mut.

Materials and Methods

1. Study group

1) Public database resources

The integrated cohort of 3,025 patients was derived from seven previously published DLBCL cohorts that provided available clinical and mutational data [10-16]. Among them, 277 patients had no relevant treatment data, all the other patients received R-based therapy as the first-line treatment.

2) JSPH cohort

Our cohort included 202 newly diagnosed DLBCL patients who were diagnosed at the First Affiliated Hospital with Nanjing Medical University (Jiangsu Provincial Hospital, JSPH) from September 2020 to July 2023. Histopathological diagnosis was determined by at least two experienced pathologists separately according to the World Health Organization classification. All newly diagnosed DLBCL patients received standard R-CHOP–like regimens as first-line treatment. All tumor tissue samples from these patients underwent NGS. Furthermore, we analyzed gene expression levels in a subset of 160 DLBCL samples via RNA sequencing (RNA-seq).

2. Mutation analysis

The mutation data for the integrated cohort utilized in this article were downloaded from the Gene Expression Omnibus (GEO) database (GSE117556 dataset and GSE87371 dataset) and applied for genetic subtyping according to the LymphGen algorithm. In JSPH cohort, we included all 202 patients with DLBCL who had complete NGS data, genomic DNA extracted from tumor tissue samples of all patients was subjected to mutation analysis via a clinically validated NGS mutation panel as previously described [17].

3. RNA sequencing

To explore the potential biological mechanisms underlying CD58 mutations and establish clinically available risk models, we extracted data from 339 DLBCL patients with complete mRNA expression profiles from the GSE117556 dataset (REB cohort) as the training cohort. Additionally, excluding the missing RNA data, we utilized 202 cases from the DS cohort and 160 cases from the JSPH cohort as two independent validation cohorts. Total RNA was isolated using TRIzol RNA Isolation method (Invitrogen) and Library was sequenced by Illumina HiSeq NGS platforms (Illumina).

4. Risk score model

We employed least absolute shrinkage and selection operator (LASSO) regression, univariate and multivariate Cox regression analyses, and random forest (RF) to identify key genes, ultimately selecting six feature genes to construct the risk model. The risk score formula, based on the expression of each included gene weighted by its LASSO regression coefficient, is as follows: risk score=Σ_(i=1) coef*gene expression. Using the risk score formula, we evaluated the risk score for each patient in the REB cohort, DS cohort, and JSPH cohort.

5. Statistical analysis

All the statistical analyses were performed via SPSS ver. 27.0 software (IBM Corp.), GraphPad Prism 10.0 software (GraphPad Software Inc.), and R software ver. 4.2.3 (R Foundation for Statistical Computing). The normally distributed data were compared via an unpaired Students’ t test, and the nonnormally distributed data were compared via the Wilcoxon test. Survival analysis was performed via Kaplan-Meier method and log-rank or Breslow test. Univariate and multivariate analyses were performed via Cox hazards regression models. The receiver operating characteristic (ROC) and corresponding area under the curve (AUC) were calculated to evaluate the prognostic ability of the model.

More detailed information is available in the Supplementary Materials and Methods.

Results

1. Mutation profile of CD58 mutations in DLBCL

The workflow of this study is shown in S1 Fig. In our study, we analyzed an integrated cohort of 3,025 DLBCL patients from the seven public databases and 202 DLBCL patients from our center. Among the integrated cohort, 250 patients (8.2%) had CD58^mut. Patients with CD58^mut usually exhibited worse overall survival (OS) (hazard ratio [HR], 1.43; 95% confidence interval [CI], 1.15 to 1.77; p < 0.001) than those with wild-type CD58 (CD58^wt) (Fig. 1A and B). A total of 23 patients (11.4%) in the JSPH cohort had CD58^mut, and patients with CD58^mut had worse progression-free survival (PFS) (HR, 1.85; 95% CI, 0.85 to 4.04; p=0.052) and OS (HR, 2.40; 95% CI, 0.83 to 6.93; p=0.026) (Fig. 1C and D). The majority of the 250 patients with mutations were frame-shifting mutations (n=74, 29.6%) and nonsense mutations (n=65, 26.0%). JSPH showed the same results (37.1% for frame-shift mutations, 29.6% for nonsense mutations) (Fig. 1F).

Fig. 1.

CD58 mutations and clinical characteristics in diffuse large B-cell lymphoma (DLBCL). (A, B) Kaplan-Meier curves of overall survival and progression-free survival according to CD58 mutational status in integrated cohort. (C, D) Kaplan-Meier curves of overall survival and progression-free survival according to CD58 mutational status in Jiangsu Province Hospital (JSPH) cohort. (E) The lollipop plot illustrating specific localization of CD58 variants. (F) Bar graph demonstrating quantity of top 5 of high frequency of CD58 mutation types. (G) Grouped bar graph demonstrating proportion of CD58 mutation according to different base substitution. (H, I) Accumulative bar diagram demonstrating proportion of CD58 mutation according to variant exon and different cell-of-origin. (J) Prevalence of DLBCL subtypes classified by LymphGen subtype in CD58^mut patients. ABC, activated B-cell-like; GCB, germinal center B-cell-like; UNC, unclassified.

The distribution pattern of mutations was analyzed, exon 2 was most common in both the integrated cohort (45.8%) and the JSPH cohort (42.3%) (Fig. 1H). In addition, M1 and S87 were identified as hotspots for CD58^mut (Fig. 1E). The CD58^mut was associated mainly with C>T transitions (Fig. 1G).

Among the CD58^mut patients in the integrated cohort, the largest proportion of them were other subtype (47.1%). There was no statistical significance in distribution of CD58^mut as regard to lymphGen genetic subtype. Among the CD58^mut cohort of JSPH, most of them can be classified as MCD subtype (47.8%) (Fig. 1J). The COO subtype information in the integrated cohort and JSPH cohort was listed in Fig. 1I.

2. Prognostic impacts of CD58 mutations in DLBCL

A summary of the clinical characteristics for the integrated cohort and the JSPH cohort is presented in S2 Table. We observed that poorer OS was associated with the CD58^mut in different age groups (HR, 1.70; 95% CI, 1.16 to 2.48; p=0.006; HR, 1.35; 95% CI, 1.09 to 1.68; p=0.006), normal lactate dehydrogenase (LDH) (HR, 1.68; 95% CI, 1.10 to 2.56; p=0.016), low IPI risk group (HR, 1.72; 95% CI, 1.02 to 2.91; p=0.043), female (HR, 1.71; 95% CI, 1.16 to 2.53; p=0.007), other subtype (HR, 1.56; 95% CI, 1.19 to 2.05; p=0.001), and germinal center B-cell-like (GCB) subtype (HR, 1.59; 95% CI, 1.19 to 2.11; p=0.001). In addition, the PFS of CD58^mut patients in the younger subgroup (HR, 1.49; 95% CI, 1.02 to 2.18; p=0.039), normal LDH level (HR, 1.62; 95% CI, 1.04 to 2.52; p=0.034), female (HR, 1.61; 95% CI, 1.05 to 2.49; p=0.031), and GCB subtype (HR, 1.41; 95% CI, 1.04 to 1.91; p=0.028) was also worse (Fig. 2A and B). The trend of the JSPH cohort was the same as that of the integrated cohort (Fig. 2C and D).

Fig. 2.

Univariate Cox regression analyses of clinical characteristics for overall survival (OS) and progression-free survival (PFS). (A, B) Univariate Cox regression analyses of clinical characteristics for OS and PFS in the integrated cohort by CD58^mut mutational status. (C, D) Univariate Cox regression analyses for OS and PFS between patients in the Jiangsu Province Hospital (JSPH) cohort grouped by CD58^mut mutational status. ABC, activated B-cell-like; CI, confidence interval; COO, cell-of-origin; ECOG, Eastern Cooperative Oncology Group; GCB, germinal center B-cell–like; HR, hazard ratio; IPI, International Prognostic Index; LDH, lactate dehydrogenase; NA, not available; prt-FLR, poor response to first-line treatment with R-CHOP like regimen; UNC, unclassified.

In the integrated cohort, CD58^mut patients with mutations located in exons 1 to 3 all had poor OS (S3A and S3C Fig.). Patients with CD58^mut in exon 2 have a poor prognosis, but the difference is not statistically significant (S3B Fig.). No statistically significant difference in OS was observed among patients with missense mutations (S3E Fig.). In terms of mutation type, CD58^mut patients with frame-shift and nonsense mutations presented poor OS (S3D and S3F Fig.). No statistically significant difference was found in PFS (S3J-S3O Fig.). In the JSPH cohort, patients with CD58^mut located in exon 2 had poor OS and PFS (S3I and S3R Fig.).

Among the DLBCL patients in the integrated cohort who received R-CHOP–like regimens in the integrated cohort, we performed a prognostic analysis on the overall response rate (ORR). CD58^mut patients had worse OS and PFS (S3G and S3P Fig.), which is consistent with the JSPH cohort (S3H and S3Q Fig.). Univariate and multivariate Cox analyses suggested that CD58^mut was an independent prognostic factor for OS in the integrated cohort (S4A-S4C Fig.).

Notably, four CD58^mut DLBCL patients without survival data were excluded from the integrated cohort. We further analyzed 246 CD58^mut patients in the integrated cohort and found that patients in the older, Eastern Cooperative Oncology Group (ECOG) > 1, high-risk IPI subgroup had poorer OS (S5A-S5C Fig.) while no statistically significant differences in OS were observed in other risk factors (S5D-S5I Fig.). Furthermore, PFS was significantly poorer in older patients and the high-risk IPI group subgroup (S5K and S5L Fig.) while no statistically significant differences in PFS were observed in other risk factors (S5J, S5M-S5R Fig.).

3. Potential mechanisms of DLBCL patients with different CD58 mutation statuses

RNA-seq analysis was performed on the GSE117556 database (REB cohort) to explore potential biological mechanisms, and a volcano plot of the significantly upregulated and downregulated genes were shown in Fig. 3A. The heatmap and Gene Ontology (GO) analysis results were presented in Fig. 3B and C. Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis further elucidated the activation of crucial pathways such as the NOD signaling pathway, cytokine-cytokine receptor interaction and cell adhesion signaling pathway in CD58^mut DLBCL patients (Fig. 3D). Similar to the KEGG results, gene set enrichment analysis (GSEA) revealed that key pathways, such as the mitogen-activated protein kinase (MAPK) pathway, Ras signaling pathway, and apoptosis pathway, were further enriched (Fig. 3E).

Fig. 3.

Bioinformatics analysis of the differentially expressed genes (DEGs) underlying the different prognosis in the REB cohort. Volcano plot (A) and heat map (B) between CD58^mut and CD58^wt groups. Gene Ontology (GO) plots (C), Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis (D), and gene set enrichment analysis (GSEA) results (E) of the DEGs. AGE, advanced glycation end product; CD58^mut, mutated CD58; CD58^wt, wild-type CD58; IL, interleukin; MAPK, mitogen-activated protein kinase; RAGE, receptor for AGE.

4. Risk score model construction and model validation by screening key genes based on the CD58 mutation status

We used three machine algorithms to identify signature genes from the candidate differentially expressed genes (DEGs) of DLBCL patients. After univariate analysis (S6B Fig.), a second screening of LASSO analysis revealed 23 signature genes (S6C and S6D Fig.), whereas 117 signature genes were identified via RF analysis (S6A Fig.). The intersection of the 23 genes obtained from univariate and LASSO analyses and the 117 genes from RF analysis was performed to obtain 19 genes (S6E Fig.). Multivariate Cox analysis was performed on these 19 genes. Through the interaction of these three algorithms, six signature genes were ultimately identified as independent prognostic factors, including MRO, OXTR, RASL11A, RLN1, SIGLEC1, and PROM2 (S6F Fig.).

Subsequently, on the basis of the identified feature genes, we constructed a scoring formula to calculate the risk score and stratified the patients into low-risk and high-risk subgroups according to the median risk score. The OS (HR, 5.09; 95% CI, 2.81 to 9.20; p < 0.001) and PFS (HR, 3.28; 95% CI, 2.07 to 5.22; p < 0.001) of low-risk patients were significantly better than those of high-risk patients. The AUCs of time-dependent receiver operating characteristic curve (time-ROC) suggested that the risk score had satisfactory sensitivity and specificity (Fig. 4A-D). Similar results were obtained for the external validation cohort DS cohort (Fig. 4E-H). In the JSPH validation cohort, low-risk patients had better OS (HR, 3.19; 95% CI, 1.40 to 7.24; p=0.010) and PFS (HR, 1.45; 95% CI, 0.81 to 2.58; p=0.210) than high-risk patients. The prognosis of the high-risk PFS group showed a clear trend, but this difference was not statistically significant. The time-ROC curve of the JSPH cohort has satisfactory sensitivity and specificity, but it is worse than that of the REB training cohort and the DS validation cohort (Fig. 4I-L), which may be related to the shorter follow-up time of the JSPH cohort.

Fig. 4.

Assessment of the prognostic value of the 6-gene model in diffuse large B-cell lymphoma (DLBCL) patients. Kaplan-Meier curves (A) and time-dependent receiver operating characteristic (ROC) (B) of overall survival in the training cohort (REB) grouped by the predicted risk scores. Kaplan-Meier curves (C) and time-dependent ROC (D) of progression-free survival in the training cohort (REB) grouped by the predicted risk scores. Kaplan-Meier curves (E) and time-dependent ROC (F) of overall survival in DS cohort by the predicted risk scores. Kaplan-Meier curves (G) and time-dependent ROC (H) of progression-free survival in DS cohort by the predicted risk scores. Kaplan-Meier curves (I) and time-dependent ROC (J) of overall survival in Jiangsu Province Hospital (JSPH) cohort by the predicted risk scores. Kaplan-Meier curves (K) and time-dependent ROC (L) of progression-free survival in JSPH cohort by the predicted risk scores. The distributions of risk score (M), the distributions of survival status (O), and expression profile of signature genes between the high- and low-risk groups of REB cohort (Q). The distributions of risk score (N), the distributions of survival status (P), and expression profile of signature genes between the high- and low-risk groups of DS cohort (R). AUC, area under the curve.

The distributions of the risk score and survival status of the two cohorts, as shown in Fig. 4M and O (REB cohort) and Fig. 4N and P (DS cohort), both suggested that patients in the high-risk group had higher mortality than those in the low-risk group. The six genes in the risk score model were expressed similarly in both cohorts (Fig. 4Q and R). These findings indicate that the risk score model we constructed had good stability.

CD58^mut tended to reduce mRNA expression levels (Fig. 5A and B). Therefore, we used the surv-cutoff function to calculate the best cutoff point and divided our patients into a high-expression group and a low-expression group. Notably, compared with the high-expression group, the low-expression group had a lower abundance of stromal cells and immune cells and greater tumor purity (Fig. 5C), and the difference was statistically significant. In the REB and DS cohorts, high levels of CD58 expression were significantly associated with favorable PFS (HR, 1.96; 95% CI, 1.17 to 3.31; p=0.004 and HR, 1.57; 95% CI, 0.93 to 2.65; p=0.071) and OS (HR, 2.62; 95% CI, 1.36 to 5.08; p=0.001 and HR, 2.35; 95% CI, 1.29 to 4.28; p=0.003) (Fig. 5D-G).

Fig. 5.

CD58 genetic alterations, expression level, and prognostic impact in diffuse large B-cell lymphoma. (A) CD58 gene expression in tumor samples in REB cohort revealed by RNA-seq. (B) CD58 gene expression in tumor samples in REB and DS cohort revealed by RNA-seq. (C) ESTIMATE algorithm revealing the stromal scores, immune scores, ESTIMATE scores, and tumor purity in the CD58 high-expression group vs. the CD58 low-expression group in the REB cohort. Overall survival (D) and progression-free survival (F) curves in REB cohort according to CD58 expression level. Overall survival (E) and progression-free survival (G) curves in DS cohort according to CD58 expression level. (H-K) p-values and hazard ratios (HRs) with 95% confidence intervals (CIs) for risk score, International Prognostic Index (IPI), and clinical features in REB cohort by univariate and multivariate Cox regression analysis. COO, cell-of-origin; ECOG, Eastern Cooperative Oncology Group; LDH, lactate dehydrogenase.

^****p < 0.0001.

5. The clinical significance and independent prognostic value of the risk score

To elucidate the clinical significance of the risk score, clinical variables such as age, sex, stage, ECOG, and IPI were included in the correlation analysis. The box plot revealed that the CD58 low-expression group had a higher risk score (Fig. 6A). Among the different subtypes of COO, the ABC subtype had the highest risk score, followed by the UNC (unclassiﬁed) and GCB subtype (Fig. 6B). The IPI group, age, sex, stage, ECOG, extranodal locations, and LDH level were not significantly related to the risk score (Fig. 6C-H). Finally, univariate (Fig. 5H and I) and multivariate (Fig. 5J and K) Cox analyses of OS and PFS were performed on clinical features such as IPI, age, sex, and risk score models in the REB cohort, suggesting that risk score was an independent prognostic factor for DLBCL.

Fig. 6.

Correlation analysis of risk score with clinical features and assessment of the prognostic value of the 6-gene model in diffuse large B-cell lymphoma (DLBCL). (A-H) Risk score in different DLBCL subgroups. (I, J) Kaplan-Meier curves of overall survival and progression-free survival in REB cohort according to CD58 expression level and predicted risk scores. (K, L) Kaplan-Meier curves of overall survival and progression-free survival in DS cohort according to CD58 expression level and predicted risk scores. ABC, activated B-cell-like; CD58exp, CD58 expression level; GCB, germinal center B-cell-like; LDH, lactate dehydrogenase; UNC, unclassified.

According to the constructed risk score model and the high and low expression of CD58, the risk stratification was optimized and the patients were divided into four groups. The high-risk group was divided into a high-risk group with high CD58 expression and a high-risk group with low CD58 expression. The low-risk group was also divided into a low-risk group with high CD58 expression and a low-risk group with low CD58 expression. In the REB and DS cohorts, patients with high CD58 expression in the low-risk group had the best OS (Fig. 6I: HR, 2.19; 95% CI, 1.60 to 2.99; p < 0.0001; Fig. 6K: HR, 1.76; 95% CI, 1.34 to 2.35; p < 0.001) and PFS (Fig. 6J: HR, 1.78; 95% CI, 1.42 to 2.22; p < 0.001; Fig. 6L: HR, 1.53; 95% CI, 1.21 to 1.92; p < 0.001). In contrast, high-risk patients with low CD58 expression had the worst OS and PFS in both cohorts.

6. Difference analysis based on the risk score between the high- and low-risk groups

RNA-seq analysis of the high-risk group and low-risk group in GSE117556 (REB cohort) revealed that the genes significantly upregulated and downregulated according to the volcano map are shown in S7A Fig., and the heatmap and GO analysis results are shown in S7B-S7C Fig. Compared with patients in the low-risk group, those in the high-risk group were enriched in the dysregulation of cell apoptosis, cytokine and cytokine receptor interaction, the Nod signaling pathway, cell adhesion, the MAPK signaling pathway, the phosphoinositide 3-kinase (PI3K)-Akt signaling pathway, and the Ras signaling pathway (S7D Fig.).

Considering the different prognostic outcomes in high-risk group and low-risk group based on the risk score model, tumor immunophenotype analysis was performed using RNA-seq data to compare immune cell infiltration. In low-risk DLBCL patients, the infiltration of M2-type polarized macrophages and (p < 0.05) neutrophils (p < 0.001) was lower (Fig. 7A). In addition, compared with the high-risk score group DLBCL patients, the low-risk score group had significantly increased relevant immune checkpoints that promote the activation, proliferation and survival of CD4+ T and CD8+ T cells, including CD48 (p < 0.01), ICOS (p < 0.001), TNFRSF8 (p < 0.01) and TNFRSF9 (p < 0.001), while the expression of negative regulatory immune checkpoints, including LAG3 (p < 0.05), IDO1 (p < 0.01), TNFSF9 (p < 0.01), etc., was significantly increased in high-risk patients (Fig. 7C).

Fig. 7.

Biological mechanism of difference in prognosis between high-risk and low-risk diffuse large B-cell lymphoma (DLBCL) patients based on risk score model; A nomogram based on the risk score, International Prognostic Index (IPI), and clinical features. (A) The proportion of 22 immune cells between low-risk and high-risk groups. (B) Estimated half-maximal inhibitory concentration (IC₅₀) value of each DLBCL patient of training set with low-risk and high-risk scores for anticancer drugs. (C) The distribution of immune checkpoints between the high-risk and low-risk groups. (D) A nomogram for the prediction of DLBCL patients’ 1-, 2-, 3-, and 4-year survival probability according to clinical factors. (E) Nomogram-predicted percentages and the observed probabilities of 1-, 2-, 3-, and 4-year survival. (F) Time-dependent receiver operating characteristic analysis of the nomogram and International Prognostic Index (IPI) at 4 years. ABC, activated B-cell-like; AUC, area under the curve; CI, confidence interval; COO, cell-of-origin; ECOG, Eastern Cooperative Oncology Group; GCB, germinal center B-cell-like; LDH, lactate dehydrogenase; OS, overall survival; UNC, unclassified. ^*p < 0.05, ^**p < 0.01, ^***p < 0.001; ns, not significant.

7. Drug sensitivity analysis

According to the Genomics of Drug Sensitivity in Cancer (GDSC) database, patients in the high-risk group were more sensitive to some drugs than those in the low-risk group were, including the tyrosine kinase inhibitor dasatinib (p < 0.001) and the PI3K inhibitor ZM44739 (p < 0.001). The drugs to which the low-risk group was more sensitive included the purine analogue fludarabine (p < 0.001), the alkylating agent temozolomide (p < 0.001), and the anaplastic lymphoma kinase inhibitor crizotinib (p < 0.001) (Fig. 7B).

8. Nomogram structure

For better application in clinical practice, a prognostic nomogram was constructed based on risk score and clinical characteristics (Fig. 7D). The 1-4 years OS predicted by the nomogram showed good consistency with the observed OS, indicating a favorable match of the calibration curves (Fig. 7E). In addition, we performed pairwise comparisons between the nomogram and risk score and the widely used IPI. The ROC curves revealed that the AUC of the risk score (AUC, 0.771; 95% CI, 0.693 to 0.848) was greater than that of the IPI score (AUC, 0.630; 95% CI, 0.545 to 0.715), and there was a significant difference between the AUC value of the risk score and that of the IPI. The ROC curve showed that the AUC of the nomogram (AUC, 0.843; 95% CI, 0.778 to 0.908) was higher than that of the risk score (AUC, 0.771; 95% CI, 0.693 to 0.848). Meanwhile, the AUC of nomogram presented the significant difference compared with that of the risk score (Fig. 7F), which demonstrated that the nomogram had a greater ability to predict the prognosis of DLBCL patients than the IPI and risk score did.

Discussion

Genetic alterations in CD58 have been demonstrated to occur in various subtypes of lymphoma, including primary central nervous system lymphoma [18], primary mediastinal large B-cell lymphoma [19], FL [7], and classical Hodgkin lymphoma (cHL) [20]. Numerous studies have linked CD58 to adverse prognoses in various lymphoma subtypes, including PTCL [9], anaplastic large cell lymphoma [21], and cHL [22]. All of which provide basis for the clinical value of CD58 in lymphoma. As for the biological mechanism, a previous study reported that mechanism by which CD58 alterations regulated the antitumor immune responses was through inducing PDL1 and IDO in DLBCL [4]. In other lymphoma subtypes, research primarily focuses on the role of CD58 in signaling pathways and the immune microenvironment, suggesting the potential to exert a substantial impact on the therapeutic response to immunochemotherapy.

To assess the prognostic value of CD58^mut in detail, we identified six signature genes via multivariate Cox analysis and the intersection of the LASSO and RF results in the GEO cohort and used them to develop a prognostic risk model. The risk model includes MRO, OXTR, RASL11A, RLN1, SIGLEC1, and PROM2. MRO, also known as B29 or C18orf3, primarily functions in tumors through its association with M2 macrophages and its role in the tumor microenvironment [23]. OXTR gene have been associated with the occurrence and development of various cancers, such as pancreatic cancer, colorectal cancer, and breast cancer [24]. RASL11A is a small GTPase gene belonging to the Ras superfamily, and it exhibits tumor-suppressive effects in various types of cancer. For instance, in prostate cancer, the expression of RASL11A is downregulated, suggesting that it may play an inhibitory role in tumorigenesis and progression [25]. RLN1 activates multiple signaling pathways, such as cAMP, nitric oxide (NO), cGMP, and the MAPK pathway, via its receptor RXFP1, thereby promoting the growth, invasion, and metastasis of tumor cells [26]. SIGLEC1 (also known as CD169) plays a crucial role in the tumor microenvironment, primarily by modulating the functions of immune cells to influence tumor progression. In research on melanoma, high expression of SIGLEC1 is associated with immune therapy responses [27]. Mechanistic studies suggest that PROM2 may enhance the migratory capacity of tumor cells by affecting cell adhesion and cytoskeletal reorganization [28].

This risk score model effectively predicted prognosis in GEO training cohort and was subsequently validated in both the external GEO validation cohort and the real world. On this basis, a nomogram was drawn combining the CD58 RNA expression level and clinical characteristics, which was more convenient and improved the predictive ability to a greater extent.

Building on previous studies, we performed immune analysis (CIBERSORT) to investigate differences of immune landscapes based on risk score model. As expected, the prognosis of high-risk group was poor. The high-risk group had an immunosuppressive microenvironment with high infiltration abundance of immunosuppressive cells M2 macrophages and neutrophils, high expression of negative regulatory immune checkpoint and low expression of immune checkpoint promoting T cell activation and proliferation [29]. In further detailed grouping of this study, the highest-risk patients in the REB and DS cohorts were those with high-risk scores and low CD58 expression. Not only did CIBERSORT analysis suggest that there was an immunosuppressive microenvironment, but the ESTIMATE algorithm also suggested low immune cell and stromal cell infiltration and an immune microenvironment with high tumor purity. An increased risk score suggests a “cold tumor” [30], a weakened effect of immunotherapy, and an immunosuppressive tumor microenvironment caused by metabolic reprogramming, which are consistent with a poor prognosis.

Macrophages are key immune infiltrating cells in both solid tumors and hematological tumors and can polarize into an inflammatory (M1) or immunosuppressive (M2) phenotype under external stimuli [31]. M1 macrophages are involved in antitumor immunity and antiangiogenic effects, and M2 macrophages are considered immunosuppressive cells that promote the immunosuppressive microenvironment and tumor immune evasion [32]. In our study, we observed that in the high-risk group, the abundance of immunosuppressive cells, especially M2 macrophages, was greater. Previous studies have shown that M2 macrophages contribute to the progression of lymphomas and cause chemotherapy resistance and a poor prognosis in patients with DLBCL [33], which corroborated our findings. Interestingly, our study suggested that the abundance of CD8+ T cells in the high-risk group was greater than that in the low-risk group (p < 0.01). We speculated that although the abundance of CD8+ T cells was greater in the high-risk group than in the low-risk group, their function was inhibited. The expression of the immune checkpoint gene LAG-3 was increased, confirming our speculation. LAG-3 functions as a negative regulator of T-cell expansion and homeostasis [34], and LAG-3 expression can be described as a marker of T-cell exhaustion in the context of the invasiveness of solid tumors and lymphomas [35]. High expression of LAG-3 can be detected in T cells and is associated with an invasive tumor phenotype and overall poor prognosis [36].

On the basis of the GDSC database, targeted chemotherapeutic drugs such as dasatinib, crizotinib, fludarabine, and temozolomide can be used to predict DLBCL patients in the high-risk and low-risk group [37]. Among these drugs, the tyrosine kinase inhibitor dasatinib had the lowest IC₅₀ in the high-risk group of DLBCL patients. Dasatinib has been studied in preclinical DLBCL models and has been shown to have cytotoxic effects on cell lines both in vitro and in vivo, mainly through the inhibition of FYN [38]. Other experimental results have shown that, in addition to its cytotoxic activity, dasatinib can inhibit BCAR1-RAC1–mediated tumor dissemination [39]. Dasatinib can be used to prevent extranodal recurrence of DLBCL patients with BTG1 mutations [40]. Scuoppo et al. [38] reported that dasatinib, a drug commonly used to treat B-cell acute lymphocytic leukemia and chronic myeloid leukemia, could be repurposed for DLBCL and combined with mTORC2 inhibitors for drug-resistant cases.

However, our study also has several limitations. First, this was a retrospective study, and further validation of our risk signature in an independent prospective cohort may be more convincing. Second, our survival analysis was mainly based on data extracted from public databases cited in the literature. However, the absence of complete clinical and survival information in some databases and the short follow-up time of the JSPH cohort in our center could potentially introduce discrepancies between our findings and the actual clinical scenario.

In summary, our study revealed the clinical relevance of CD58^mut in DLBCL. CD58 is an adverse prognostic biomarker that is associated with genomic alterations and the immunosuppressive microenvironment in DLBCL. A better understanding of the clinical and biological significance of CD58 may help in the selection of potential immunotherapy targets for future mechanism-based treatment of DLBCL.

Electronic Supplementary Material

Supplementary materials are available at Cancer Research and Treatment website (https://www.e-crt.org).

crt-2024-1143_S1_Fig.pdf

crt-2024-1143_S2_Table.pdf

crt-2024-1143_S3_Fig.pdf

crt-2024-1143_S4_Fig.pdf

crt-2024-1143_S5_Fig.pdf

crt-2024-1143_S6_Fig.pdf

crt-2024-1143_S7_Fig.pdf

crt-2024-1143_Supplementary_Materials_and_Methods.pdf

Notes

Ethical Statement

This study was approved by the Ethics Committee of the First Affiliated Hospital with Nanjing Medical University.

Author Contributions

Conceived and designed the analysis: Wang HL, Li JY, Liang JH (Jin-Hua Liang), Xu W.

Collected the data: Pan BH, Zhang XY, Shen HR.

Contributed data or analysis tools: Hua W, Yin H, Li Y, Liang JH (Jun-Heng Liang), Gao R.

Performed the analysis: Wang HL, Shang CY.

Wrote the paper: Wang HL, Shang CY, Liang JH (Jin-Hua Liang), Xu W.

Conflicts of Interest

Conflict of interest relevant to this article was not reported.

Funding

This research was funded by the National Natural Science Foundation of China (grant number 82200887 and 82370194), Jiangsu Science and Technology Department (grant number BK20220716 and BE2023780), China Postdoctoral Science Foundation (grant number 2022M7114034 and 2023M741463), Nanjing Municipal Returned Overseas Scholars Foundation (grant number BSHNJ2023011), Jiangsu Province Hospital Talent Specialized Foundation (grant number MXJL202107), Jiangsu Province “333” Project (grant number BRA2018085), Jiangsu Province Hospital Clinical Diagnosis and Treatment Technologies Innovation Project Foundation (grant number JBGS202405).

References

1. Sehn LH, Salles G. Diffuse large B-cell lymphoma. N Engl J Med 2021;384:842–58.

2. Gutierrez-Garcia G, Cardesa-Salzmann T, Climent F, Gonzalez-Barca E, Mercadal S, Mate JL, et al. Gene-expression profiling and not immunophenotypic algorithms predicts prognosis in patients with diffuse large B-cell lymphoma treated with immunochemotherapy. Blood 2011;117:4836–43.

3. Wright GW, Huang DW, Phelan JD, Coulibaly ZA, Roulland S, Young RM, et al. A probabilistic classification tool for genetic subtypes of diffuse large B cell lymphoma with therapeutic implications. Cancer Cell 2020;37:551–68.

4. Xu X, Zhang Y, Lu Y, Zhang X, Zhao C, Wang J, et al. CD58 alterations govern antitumor immune responses by inducing PDL1 and IDO in diffuse large B-cell lymphoma. Cancer Res 2024;84:2123–40.

5. Zhang Y, Liu Q, Yang S, Liao Q. CD58 immunobiology at a glance. Front Immunol 2021;12:705260.

6. Archimbaud E, Campos L, Vila L, Thomas X. Potential role for non-HLA-restricted cytotoxic cells in the immune surveillance of acute leukemia. Immunol Lett 1993;39:13–5.

7. Pasqualucci L, Khiabanian H, Fangazio M, Vasishtha M, Messina M, Holmes AB, et al. Genetics of follicular lymphoma transformation. Cell Rep 2014;6:130–40.

8. Palomero T, Couronne L, Khiabanian H, Kim MY, Ambesi-Impiombato A, Perez-Garcia A, et al. Recurrent mutations in epigenetic regulators, RHOA and FYN kinase in peripheral T cell lymphomas. Nat Genet 2014;46:166–70.

9. Kataoka K, Nagata Y, Kitanaka A, Shiraishi Y, Shimamura T, Yasunaga J, et al. Integrated molecular analysis of adult T cell leukemia/lymphoma. Nat Genet 2015;47:1304–15.

10. Lacy SE, Barrans SL, Beer PA, Painter D, Smith AG, Roman E, et al. Targeted sequencing in DLBCL, molecular subtypes, and outcomes: a Haematological Malignancy Research Network report. Blood 2020;135:1759–71.

11. Reddy A, Zhang J, Davis NS, Moffitt AB, Love CL, Waldrop A, et al. Genetic and functional drivers of diffuse large B cell lymphoma. Cell 2017;171:481–94.

12. Pedrosa L, Fernandez-Miranda I, Perez-Callejo D, Quero C, Rodriguez M, Martin-Acosta P, et al. Proposal and validation of a method to classify genetic subtypes of diffuse large B cell lymphoma. Sci Rep 2021;11:1886.

13. Chapuy B, Stewart C, Dunford AJ, Kim J, Kamburov A, Redd RA, et al. Molecular subtypes of diffuse large B cell lymphoma are associated with distinct pathogenic mechanisms and outcomes. Nat Med 2018;24:679–90.

14. Sha C, Barrans S, Cucco F, Bentley MA, Care MA, Cummin T, et al. Molecular high-grade B-cell lymphoma: defining a poor-risk group that requires different approaches to therapy. J Clin Oncol 2019;37:202–12.

15. Dubois S, Tesson B, Mareschal S, Viailly PJ, Bohers E, Ruminy P, et al. Refining diffuse large B-cell lymphoma subgroups using integrated analysis of molecular profiles. EBioMedicine 2019;48:58–69.

16. Ennishi D, Jiang A, Boyle M, Collinge B, Grande BM, Ben-Neriah S, et al. Double-hit gene expression signature defines a distinct subgroup of germinal center B-cell-like diffuse large B-cell lymphoma. J Clin Oncol 2019;37:190–201.

17. Hua W, Li Y, Yin H, Du KX, Zhang XY, Wu JZ, et al. Analysis of CCND3 mutations in diffuse large B-cell lymphoma. Ann Hematol 2024;103:5729–39.

18. Braggio E, Van Wier S, Ojha J, McPhail E, Asmann YW, Egan J, et al. Genome-wide analysis uncovers novel recurrent alterations in primary central nervous system lymphomas. Clin Cancer Res 2015;21:3986–94.

19. Mottok A, Hung SS, Chavez EA, Woolcock B, Telenius A, Chong LC, et al. Integrative genomic analysis identifies key pathogenic mechanisms in primary mediastinal large B-cell lymphoma. Blood 2019;134:802–13.

20. Schneider M, Schneider S, Zuhlke-Jenisch R, Klapper W, Sundstrom C, Hartmann S, et al. Alterations of the CD58 gene in classical Hodgkin lymphoma. Genes Chromosomes Cancer 2015;54:638–45.

21. Luchtel RA, Dasari S, Oishi N, Pedersen MB, Hu G, Rech KL, et al. Molecular profiling reveals immunogenic cues in anaplastic large cell lymphomas with DUSP22 rearrangements. Blood 2018;132:1386–98.

22. Liu Y, Abdul Razak FR, Terpstra M, Chan FC, Saber A, Nijland M, et al. The mutational landscape of Hodgkin lymphoma cell lines determined by whole-exome sequencing. Leukemia 2014;28:2248–51.

23. Gu Y, Zheng M, Xie J. The role of MRO as an M2 macrophage-associated gene in non-small cell lung cancer: insights into immune infiltration, prognostic significance, and therapeutic implications. Discov Oncol 2025;16:74.

24. Pierzynowska K, Gaffke L, Zabinska M, Cyske Z, Rintz E, Wisniewska K, et al. Roles of the oxytocin receptor (OXTR) in human diseases. Int J Mol Sci 2023;24:3887.

25. Louro R, Nakaya HI, Paquola AC, Martins EA, da Silva AM, Verjovski-Almeida S, et al. RASL11A, member of a novel small monomeric GTPase gene family, is down-regulated in prostate tumors. Biochem Biophys Res Commun 2004;316:618–27.

26. Kotwal A, Goldner WS, Bennett RG. Role of relaxin signaling in cancer: a review. Biochem Pharmacol 2024;230:116634.

27. Mei S, Huang Y, Zhao Y, Zhang X, Zhang P. A pan-cancer blueprint of genomics alterations and transcriptional regulation of Siglecs, and implications in prognosis and immunotherapy responsiveness. Clin Transl Med 2023;13e1262.

28. Wu KL, Chang CY, Tsai YM, Lai JC, Hung JY, Lin YS, et al. PROM2 upregulation promotes cancer cell migration and confers a poor prognosis in lung squamous cell carcinoma. Am J Cancer Res 2024;14:1561–76.

29. Lei X, Lei Y, Li JK, Du WX, Li RG, Yang J, et al. Immune cells within the tumor microenvironment: biological functions and roles in cancer immunotherapy. Cancer Lett 2020;470:126–33.

30. Galon J, Bruni D. Approaches to treat immune hot, altered and cold tumours with combination immunotherapies. Nat Rev Drug Discov 2019;18:197–218.

31. Locati M, Curtale G, Mantovani A. Diversity, mechanisms, and significance of macrophage plasticity. Annu Rev Pathol 2020;15:123–47.

32. Jones LM, Broz ML, Ranger JJ, Ozcelik J, Ahn R, Zuo D, et al. STAT3 establishes an immunosuppressive microenvironment during the early stages of breast carcinogenesis to promote tumor growth and metastasis. Cancer Res 2016;76:1416–28.

33. Marchesi F, Cirillo M, Bianchi A, Gately M, Olimpieri OM, Cerchiara E, et al. High density of CD68+/CD163+ tumour-associated macrophages (M2-TAM) at diagnosis is significantly correlated to unfavorable prognostic factors and to poor clinical outcomes in patients with diffuse large B-cell lymphoma. Hematol Oncol 2015;33:110–2.

34. Andrews LP, Marciscano AE, Drake CG, Vignali DA. LAG3 (CD223) as a cancer immunotherapy target. Immunol Rev 2017;276:80–96.

35. Lythgoe MP, Liu DS, Annels NE, Krell J, Frampton AE. Gene of the month: lymphocyte-activation gene 3 (LAG-3). J Clin Pathol 2021;74:543–7.

36. Matsuzaki J, Gnjatic S, Mhawech-Fauceglia P, Beck A, Miller A, Tsuji T, et al. Tumor-infiltrating NY-ESO-1-specific CD8+ T cells are negatively regulated by LAG-3 and PD-1 in human ovarian cancer. Proc Natl Acad Sci U S A 2010;107:7875–80.

37. Yang W, Soares J, Greninger P, Edelman EJ, Lightfoot H, Forbes S, et al. Genomics of Drug Sensitivity in Cancer (GDSC): a resource for therapeutic biomarker discovery in cancer cells. Nucleic Acids Res 2013;41:D955–61.

38. Scuoppo C, Wang J, Persaud M, Mittan SK, Basso K, Pasqualucci L, et al. Repurposing dasatinib for diffuse large B cell lymphoma. Proc Natl Acad Sci U S A 2019;116:16981–6.

39. Umakanthan JM, Iqbal J, Batlevi CL, Bouska A, Smith LM, Shostrom V, et al. Phase I/II study of dasatinib and exploratory genomic analysis in relapsed or refractory non-Hodgkin lymphoma. Br J Haematol 2019;184:744–52.

40. Delage L, Lambert M, Bardel E, Kundlacz C, Chartoire D, Conchon A, et al. BTG1 inactivation drives lymphomagenesis and promotes lymphoma dissemination through activation of BCAR1. Blood 2023;141:1209–20.

Article information Continued

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.