R programming was applied for integrating, analyzing, and visualizing the data. The transcriptome profile and clinical data of DLBCL were collected from the cancer genome atlas (TCGA, https://portal.gdc.cancer.gov/) and the Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo/), including GSE83632 (29) (n = 163), GSE56315 (30) (n = 88), GSE87371 (31) (n = 223), NCICCR_DLBCL (1) (n = 481), and GSE31312 (32) (n = 498). Differential expression of S100A8 mRNA in DLBCL tissues and normal controls (NCs) was compared using limma package (33), and the performance of S100A8 in DLBCL was analyzed based on the receiver operating characteristic (ROC) curve. The protein levels of S100A8 were investigated through The Human Protein Altas (HPA, https://www.proteinatlas.org). The value of S100A8 in predicting DLBCL patient survival was analyzed utilizing the log-rank test and Cox regression model using survival and survminer packages.

The differentially expressed genes (DEGs) of high and low S100A8 expressions divided by median were screened using limma package (33) in training and validation datasets and defined as significantly DEGs according to |logFC| > 1 and adjusted P value < 0.05. The overlapping DEGs between training and validation datasets were identified using ggvenn package, and inputted into the Database for Annotation, Visualization, and Integrated Discovery (DAVID) (34) where Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genome (KEGG) analysis (35) were performed. Gene set enrichment analysis (GSEA) was applied to analyze signaling pathways enriched for the overlapping DEGs in both training and validation datasets through the WEB-based Gene SeT AnaLysis Toolkit (WebGestalt, http://www.webgestalt.org/#) (36). The DEGs were inputted to the Search Tool for the Retrieval of Interacting Genes (STRING, http://string-db.org) (37), and protein-protein interaction (PPI) analysis was further investigated.

In this study, the formalin-fixed paraffin-embedded (FFPE) lymph node tissue blocks of 39 patients, including 25 DLBCL patients and 14 reactive hyperplasia (RHL) patients, were retrospectively enrolled from the Department of Pathology in our institution; the clinical characteristics were provided in Supplementary Table S1 and stored at room temperature. This study was authorized by the Affiliated Hospital of Putian University’s Ethics Committee (ID: 2023075).

Immunofluorescence (IF), as an important immunochemical technique, can allow the detection and localization of a wide variety of antigens in different types of tissues. The FFPE blocks collected were sliced (4 μm sections) and deparaffinized in xylene, rehydrated in alcohol, and then washed in distilled water. For antigen retrieval, the sections were microwaved in citrate buffer and blocked using a protein block. The sections were incubated with rabbit anti-S100A8 polyclonal (Beyotime, China) and mouse anti-IL17F monoclonal (Santa Cruz, USA). Subsequently, secondary antibodies FITC-labeled goat anti-rabbit IgG (Beyotime, China) and cy3-labeled goat anti-mouse IgG (Beyotime, China) were added and incubated at room temperature for visualization. The sections were counter-stained with DAPI (Beyotime, China), and then images were captured by an automatic quantitative pathology imaging system (Vectra Polaris, PerkinElmer, Waltham, MA). For performing immunohistochemistry (IHC), the sections were incubated with the mouse anti-S100A8 monoclonal (Immunoway, China) overnight at 4°C, followed by HRP-labeled goat anti-mouse IgG(H+L) (Beyotime, China) treatment at room temperature for 2 hours. Third, 3,3-diaminobenzidine (DAB, Beyotime, China) was used for visualization, and the sections were imaged by a DM1L inverted microscope (Leica, Germany).

The three cell lines were enrolled in this study, including one human normal B-cell line (GM12878, BeNa, China) and the DLBCL cell lines SUDHL2 (a gift from Southeast University) and SUDHL4 (Meisen, China). These cell lines were cultured in RPMI-1640 (Biosharp, China) supplemented with 10% fetal bovine serum (FBS, Gibco, US), 1% streptomycin, and penicillin (Gibco, US), and incubated at 37°C in a 5% CO₂ incubator (1300 series A2, Thermo Fisher, US). Differential concentrations of Paquinimod (ABR-215757, MedChemExpress, US) were applied to DLBCL cell lines for 24 h.

Total RNA of cell lines was isolated using TRIzol reagent (Invitrogen, US), and then HiScript Q RT SuperMix for qPCR (Vazyme, China) was used to transcribe the RNA into cDNA. The quantitative real-time polymerase chain reaction (qRT-PCR) was conducted on the CFX Connect Real-Time PCR Detection System (BioRad, US) with an HQ SYBR qPCR Mix (without ROX) (Zomen, China). The S100A8 expressions were quantified with the 2^−ΔΔCt method, and β-actin expression was used as an endogenous control. The primer sequences are shown in Supplementary Table S1 .

The protein expression of S100A8 in cell lines was validated through WB analysis. First, the total protein was extracted using RIPA buffer (Beyotime, China) with 1% PMSF (Beyotime, China), and the concentration of protein was measured using a BCA protein assay kit (Beyotime, China). Then, the extracted protein was loaded onto a 12.5% SDS-PAGE gel (Meilunbio, China) and transferred onto polyvinylidene fluoride (PVDF) membranes. The membranes were incubated with primary antibodies [anti-S100A8 polyclonal (Immunoway, China), anti-BCL2 polyclonal (Beyotime, China), anti-Bax polyclonal (Beyotime, China), anti-Caspase-3 polyclonal (Beyotime, China), and anti-β-actin polyclonal (Beyotime, China)] at 4°C overnight. After the membranes were incubated with HRP-labeled goat anti-rabbit IgG (Beyotime, China), At room temperature, the level of protein was detected using BeyoECL Plus (Beyotime, China) and quantified using Fiji (version 2.9, fiji.sc).

A density of 5 × 10⁴ DLBCL cell lines were seeded in the 96-well plates in 200 μL culture medium. After exposure to Paquinimod for 24 h, 20 μL of CCK-8 reagent (Biosharp, China) was added into each well, and the absorbance (OD) of every well was measured by a Thermo Multiskan FC enzyme reader (Thermo Fisher, USA) after incubation for 2 h. Each concentration accounted for in the control was presented as cell inhibition, and the half maximal inhibitory concentration (IC50) was calculated using R programming. Meanwhile, cell apoptosis was conducted according to the manufacturer’s protocols using an Annexin V–FITC/propidium iodide (PI) Apoptosis Detection Kit (Zomanbio, China), and then the cells were detected by a FacsCalib flow cytometer (BD Biosciences, USA).

The single-cell RNA sequencing (scRNA-seq) was extracted from the GSE182434 (38), the Seurat package (39) to normalize and scale the expression matrices. The top 2000 most highly variable genes were selected for principal component analysis. An Uniform Manifold Approximation and Projection (UMAP) algorithm was used to visualize cells. Then, the expression of S100A8 in differential cell types was annotated, and the enrichment was explored.

The connection between immune infiltrates and S100A8 expression in TCGA_DLBC was explored through the TIMER2 website (40) (http://timer.cistrome.org). The S100A8-enriched immune cells were selected, and CIBERSORT, CIBERSORT-ABS, EPIC, MCPcounter, quanTIseq, TIDE, TIMER, and xCell algorithms were utilized to estimate immune infiltration. The Spearman’s rank correlation coefficient (rho) values were provided. Next, for identifying the potential connection between S100A8 and TME features, the estimate package (41) was applied to calculate the immune and stromal scores on both training and validation datasets using single sample gene set enrichment analysis (ssGSEA). Next, the correlation between the status of S100A8 and immune cell infiltration was explored through the Cell-type Identification by Estimating Relative Subsets of RNA Transcripts (CIBERSORTx, http://cibersort.stanford.edu/). Also, immunedeconv package (42) was applied to analyze the immune cell fractions. It seemed that the interaction between angiogenesis and TME is an important promoter of tumor growth, invasion, and metastases (8, 43). The angiogenesis-related genes (ARG) were obtained from GeneCards (https://www.genecards.org), and the correlation between the expression of S100A8 and ARGs was investigated.

Checkpoint targets play a key role in tumor immunotherapy (44). We identified the correlation between S100A8 and checkpoint targets, which was obtained from the checkpoint therapeutic target database (CKTTD) (44). To evaluate the correlation between S100A8 expression and anticancer drug activity, the RNA expression data (RNA: RNA-seq) and drug data (compound activity: DTP NCI-60) were obtained from the CellMiner database (www.discover.nci.nih.gov) (45). Next, for predicting the drug sensitivity based on the heterogenicity of tumor samples, the oncoPredict package (46) was used to estimate the half-maximal inhibitory concentration (IC50) of cancer cell lines to build a predictive model to determine drug sensitivity in DLBCL. The potential anti-tumor compounds were screened based on the drug response data in the Genomics of Drug Sensitivity in Cancer (GDSC) database (https://www.cancerrxgene.org/), which had been analyzed to determine the relationship between S100A8 expression and sensitivity to chemotherapy and targeted drugs (P < 0.01).

All data analysis was performed in R programming (version 4.3.1) and visualized using the ggplot2 package. P < 0.05 was considered statistically significant.

The flow of this study is depicted in Figure 1A . Based on the TIMER2 website (40), the S100A8 expression in all TCGA tumors was explored ( Supplementary Figure S1A ). Next, the GSE83632 (29) (training dataset) and GSE56315 (30) (validation dataset) were used to identify DEGs between DLBCL and NCs using limma package (33) ( Figures 1B, D ), and the dramatically upregulated S100A8 expression was shown (P<0.05) in Figures 1C, E . The area under the curve (AUC) of the ROC curve was 1 ( Supplementary Figure S1B ) and 0.928 (95% confidence interval (CI): 0.889-0.967, Supplementary Figure S1C ), suggesting that S100A8 had a good ability of distinguishing DLBCL from NCs. The protein levels of S100A8 in malignant lymphomas were higher than in normal lymph nodes ( Supplementary Figure S1D ). We further investigated the prognostic impact of S100A8 on DLBCL patients. The elevated S100A8 expression was closely related to poor overall survival (OS) in the GSE87371 (31) (HR = 1.505, 95% CI: 1.108–2.044, P = 0.007, Figure 1F ), NCICCR_DLBCL (1) (HR = 1.544, 95% CI: 1.025–2.326, P = 0.038, Figure 1G ), and GSE31312 (32) (HR = 1.362, 95% CI: 1.082–1.714, P = 0.006, Figure 1H ). Poor progression-free survival (PFS) was linked to increased S100A8 in the GSE87371 (31) (HR = 1.407, 95% CI: 1.018–1.944, P = 0.033, Supplementary Figure S1E ) and the GSE31312 (32) (HR = 1.338, 95% CI: 1.061–1.688, P = 0.011, Supplementary Figure S1F ). Meanwhile, high expression of S100A8 was seemed to be associated to poor OS in DLBCL patients with high international prognostic index (IPI, score > 2) between GSE87371 (HR = 1.764, 95% CI: 1.058–2.944, P = 0.021, Supplementary Figure S1G ) and NCICCR_DLBCL (HR = 1.768, 95% CI: 0.873–3.580, P = 0.082, Supplementary Figure S1H ).

The DEGs of high and low S100A8 expression based on the median expression value were identified in the training (GSE87371, 232 upregulated and 21 downregulated genes, Figure 2A ) and validation (NCICCR_DLBCL, 236 upregulated and 92 downregulated genes, Figure 2B ) datasets. The 123 overlapping genes were classified through a venn plot ( Figure 2C ). The biological processes (BP) of these genes were enriched by inflammatory response, immune response, and chemotaxis. The main enrichment cellular components (CC) were plasma membranes, extracelluar regions, and extracelluar space. Meanwhile, molecular function (MF) was associated with transmembrane signaling receptor activity, signaling receptor activity, and serine-type endopeptidase activity ( Supplementary Figure S2A ). The KEGG pathway analysis indicated that DEGs were mainly involved in the IL-17 signaling pathway, viral protein interaction with cytokine and cytokine receptor, Pertussis, and Cytokine-cytokine receptor interaction ( Supplementary Figure S2B ). GSEA analysis of the training and validation datasets by WebGestalt found that DEGs could activate the IL-17 signaling pathway, complement and coagulation cascades, and inhibit the TNF signaling pathway ( Figures 2D, E ). As shown in Figure 2F , the connection between the top 50 DEGs was explored through Cytoscape software (version 3.9.1) using the CytoHubba plugin based on the MCC score. The PPI plot of these DEGs was drawn through STRING ( Supplementary Figure S2C ).

S100A8 might be mainly involved in the IL-17 signaling pathway ( Figure 3A ), which was identified by GSEA analysis. In order to verify the role of the IL-17 signaling pathway in DLBCL patients, we collected 25 DLBCL patients and 14 RHL patients in our institution. The clinical characteristics are provided in Supplementary Table S1 . The IF assay was performed and shown in Figure 3B ; the green fluorescence represented the expression of S100A8, and red indicated the expression of IL17F. The results showed S100A8 expression in DLBCL was higher than that in RHL (P < 0.05, Figure 3D ). Next, the IHC assay also suggested overexpression of S100A8 in DLBCL (P < 0.05, Figures 3C, E ).

Based on the above bioinformatics results, S100A8 is mainly involved into IL-17 signaling pathway, which can intersect with apoptosis (47, 48). To validate the potential process of cell apoptosis S100A8 mediated, We demonstrated the role of S100A8 in DLBCL cells. First, the protein level of S100A8 was decreased in SUDHL2 and SUDHL4 cell lines comparison with GM12878 cell line using qPR-PCR analysis ( Figure 4A ) and WB analysis ( Figures 4B, C ). Paquinimod, a specific inhibitor of S100A8, could participate in regulation of inflammation and immune responses (49, 50). While S100A8 inhibition with differential concentration of Paquinimod in a dose-dependent manner, the IC50 of Paquinimod treated to SUDHL2 and SUDHL4 was 409.51 nM and 181.78 nM respectively using drc package (51) ( Figure 4D ). Meanwhile, the protein levels of S100A8 could inhibit by differential concentration of Paquinimod ( Figures 4E, F ). After the treatment of 100 nM Paquinimod, cell apoptosis rates of SUDHL2 and SUDHL4 increased ( Figure 4G, H ), the protein levels of S100A8, BCL2 were significantly decreased, while Bax and Caspase-3 levels were increased ( Figure 4I, J ). These findings indicated that inhibition of S100A8 could augment cell apoptosis and induce cell death of DLBCL cell lines.

The single-cell analysis revealed that S100A8 was a highly variable gene ( Figure 5A ). We found that S100A8 was mainly expressed at different levels in macrophages, CD8 T cells, and B cells, specifically macrophages ( Figures 5B–D ), and S100A8 expression was highest in macrophages.

The dynamic interaction between tumor cells and the surrounding TME is of paramount importance in the processes of tumor initiation, development, metastasis, and therapeutic response (9, 10, 52). Based on the result of scRNA-seq analysis, we found a statistically significant negative association between S100A8 expression and B cells, also S100A8 was a positive correlation with macrophages and CD8 T cells ( Figure 6A ) through the TIMER2 website (40). As shown in Supplementary Figure S3 , Figures 5B, C , the high stromal, immune, and estimate scores were associated with DLBCL patients with a high level of S100A8 in both GSE87371 (31) and NCICCR_DLBCL (1) datasets. Meanwhile, there was not a significant difference (P>0.05) between high- and low- estimate scores both two datasets ( Supplementary Figures S3G, H ). On the other hand, the CIBERSORT algorithm (53), xCell algorithm (54), quanTIseq algorithm (55), and MCPcounter algorithm (41) were used to estimate the distribution and proportion of macrophages, CD8 T cells, and B cells. When combined with all selected algorithms, the significantly decreased abundance of macrophages M1, CD8 T cells were associated with high S100A8 expression, and B cells expressed the opposite trend ( Figures 6B, C ). There were 79 ARG enrolled ( Supplementary Table S3 ), and the positive correlation between S100A8 expression and the expression levels of IL1RN, IL1B, CXCL8, TYMP, PTGS2, CCL2, HGF, SOD2, NRP2, THBS1, ITGAV, and TIMEP2 ( Supplementary Figures S3I, J ) were investigated, indicating S100A8 might involve into the regulation of angiogenesis in TME.

Immune checkpoint inhibitors can block checkpoint proteins or their partner proteins so that immune cells called T cells recognize and kill tumors (44, 56). We downloaded a list of the immune checkpoint targets from CKTTD (44), The correlation between the S100A8 expression and the immune checkpoint targets was calculated ( Figure 7A ). The result showed that S100A8 were positively correlated with CCR5, CD274(PD-1), IDO1, IL15, LAG3, LILRB2, TDO2, TLR4, and TLR8, while there was a negative correlation with WEE1 in both GSE87371 (31) and NCICCR_DLBCL (1) datasets. Based on the CellMiner database (45), the relationship between S100A8 expression and drug response was explored ( Figure 7B ). We observed that S100A8 expression was negatively associated with AZD-2014, GDC-0349, AZD-3147, AZD-8055, and INK-128. While a positive correlation with Artemether, Imexon, GSK-2194069, ABT-199, Nandrolone phenpropionate, Cyclophosphamide, Hydroxyurea, JNJ-54302833, Chelerythrine, AZD-5991, SW-044248, BMS-536924, Carboplatin and S100A8 expression was identified. On the other hand, we assessed the drug therapeutic responses between high- and low-S100A8 levels in both GSE87371 (31) and NCICCR_DLBCL (1) datasets using oncoPredict packages (46). Fourteen drugs and twenty-three drugs were explored in GSE87371 (31) and NCICCR_DLBCL (1) datasets, respectively ( Supplementary Figure S4 ). There were eleven drugs that Selumetinib, IAP_5620, SB216763, ERK_6604, KU.55933, JAK_8517, SCH772984, Ribociclib, Trametinib, PD0325901, and AZD8055 were considered potential sensitivity small molecule inhibitors for both of them ( Figure 7C ).