This study utilized the publicly available scRNA-seq dataset GSE167363, which includes peripheral blood mononuclear cells (PBMCs) from 10 patients with Gram-negative sepsis (survivor group: n = 6; non-survivor group: n = 4) and 2 healthy controls (HCs) (26). Additionally, PBMCs from six sepsis patients (survivor group: n = 3; non-survivor group: n = 3), recruited from the Department of Emergency Medicine, Tianjin Medical University General Hospital, were used for m6A methylation sequencing. Informed consent was obtained from all participants, and the study protocol was approved by the Ethics Committee of Tianjin Medical University General Hospital (Approval ID: IRB2023-YX-045-01).

Single-cell transcriptomic data were processed using the Seurat package (v5.0) in R. Low-quality cells were filtered out based on the following criteria: number of features < 200 or > 6000, and mitochondrial gene content > 15%. Data were normalized using the “LogNormalize” method, and the top 2000 highly variable genes were selected for downstream analysis. Principal component analysis (PCA) was performed for dimensionality reduction, with the number of components selected via elbow plot. Data integration across samples was conducted using Seurat’s canonical correlation analysis (CCA) to minimize batch effects. Louvain algorithm was employed for clustering, and visualization was performed using UMAP and t-SNE embeddings.

Cell type annotation was conducted using a combination of GPTCelltype (v1.1.1), SingleR (v1.0.0), and scCATCH (v3.1.1), with reference to established marker gene databases and published literature. Differentially expressed genes (DEGs) were identified using the wilcoxauc function from the RPresto package, with the significance threshold set at log₂ fold change > 0.25 and adjusted p < 0.05. Functional enrichment analyses, including Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and Gene Set Enrichment Analysis (GSEA), were performed using clusterProfiler (v4.6.0).

To explore developmental dynamics, Monocle2 (v2.22.0) was applied to infer pseudotime trajectories of key immune cell subsets, including B cells, T cells, regulatory T cells (Tregs), and platelets, based on their transcriptional profiles.

Cell–cell communication networks were predicted using CellphoneDB (v2.1.7) to identify significant ligand–receptor interactions, with a minimum expression threshold of 10% set for inclusion in downstream analyses. NicheNet was used to assess ligand activity and predict target gene regulatory potential. The interaction network was visualized using the CCplotR R package.

Total RNA was extracted from PBMCs using TRIzol™, followed by DNase I treatment and enzymatic purification. For m6A profiling, 20 μg of total RNA was fragmented in Mg²⁺ lysis buffer at 70°C for 6 minutes, and the reaction was terminated with EDTA. RNA fragments were purified using the Zymo RNA Clean & Concentrator-5 kit. m⁶A immunoprecipitation (MeRIP) was performed using an anti-m6A antibody and Protein A/G magnetic beads, incubated at 4°C for 2 hours. After elution and purification, sequencing libraries were constructed and subjected to paired-end sequencing (PE150) on the Illumina NovaSeq 6000 platform.

Raw sequencing reads were processed using Cutadapt (v2.5) to trim adaptors and remove low-quality reads. Clean reads were aligned to the human genome (hg38) using Hisat2. Peak calling and differential methylation analysis were performed using exomePeak2 (v2.13.2), with a significance threshold of p < 0.05. Motif discovery and annotation were conducted using HOMER (v4.10.4). Differentially methylated genes were subjected to GO and KEGG enrichment analysis. Peak distribution and m⁶A site visualization were completed using Integrative Genomics Viewer (IGV, v2.19.1).

To identify key candidate genes, we integrated three datasets: (1) differentially m6A-modified genes in PBMCs; (2) DEGs in platelets between the survivor and non-survivor groups; and (3) DEGs in platelets between HCs and the survivor group. Genes shared across all three datasets were defined as core candidate genes. The correlation between the expression of these candidate genes and the platelet ligand APP was further assessed. In addition, we evaluated the correlation between candidate gene expression and 10 major m6A regulatory factors.

All data visualizations were performed using R packages, including ggplot2, ComplexHeatmap, VennDiagram, and circlize. Statistical analyses were conducted based on data distribution and type, applying the Wilcoxon rank-sum test, Student’s t-test, or ANOVA as appropriate. A p-value < 0.05 was considered statistically significant.

To investigate the immunological landscape of PBMCs in Gram-negative sepsis patients with distinct clinical outcomes (survivors vs. non-survivors), we performed scRNA-seq analysis based on the publicly available dataset GSE167363, which includes PBMCs from 10 sepsis patients (survivor group: n = 6; non-survivor group: n = 4) and 2 HCs.

First, canonical immune cell marker genes were integrated to generate a bubble plot depicting marker gene expression patterns across different immune cell subsets ( Figure 1A ). Based on these expression profiles, unsupervised clustering was performed, followed by dimensionality reduction using t-SNE1 and UMAP1, leading to the successful identification of multiple PBMCs subpopulations, including B cells, CD14+ monocytes, FCGR3A+ monocytes, dendritic cells (DCs), T cells, regulatory T cells (Tregs), natural killer (NK) cells, platelets, erythroid progenitors, neutrophils, macrophages, plasmacytoid dendritic cells (pDCs), and hematopoietic stem cells ( Figure 1B ). Notably, the distribution of these cell subsets varied significantly among HCs, survivors, and non-survivors, indicating pronounced immune heterogeneity across groups ( Supplementary Figures 1A–D ).

At the cellular level, a comparison of total PBMCs counts among the HC, survivor, and non-survivor groups revealed distinct patterns of cellular abundance ( Figure 1C , Supplementary Table S1 ). Further analysis of immune cell composition showed a progressive increase in the relative proportion of platelets from HCs to survivors and non-survivors ( Figure 1D ). This trend suggests that platelets may not only participate in the immune response during sepsis but also be closely associated with disease severity and clinical outcomes. Taken together, these findings indicate that the immune cell composition of PBMCs in Gram-negative sepsis patients undergoes significant alterations. In particular, the increasing proportion of platelets during disease progression suggests their potential key role in the pathogenesis and adverse outcomes of sepsis.

To further investigate the immunological potential of platelets in sepsis, we performed pseudotime trajectory analysis on PBMCs. The results revealed that B cells, T cells, and Tregs exhibited continuous and well-defined developmental trajectories along the pseudotime axis, ultimately converging toward a platelet-like phenotype ( Figure 2A , Supplementary Figures 2A–C ). These findings suggest that, under septic conditions, certain immune cells may undergo phenotypic reprogramming or functional transdifferentiation, acquiring platelet-like features. This differentiation process was accompanied by dynamic changes in gene expression profiles within these cell subsets, further supporting the presence of a platelet-directed reprogramming trajectory ( Figure 2B ).

To assess potential intercellular communication between immune cells and platelets, we constructed and analyzed the PBMCs cell–cell interaction network. Global interaction strength analysis indicated a progressive weakening of immune cell communication with advancing disease severity, with the non-survivor group exhibiting the sparsest interaction network ( Figures 2C–E ). Focusing on the ligand–receptor interactions between B cells, T cells, Tregs, and platelets, we observed that in healthy individuals, communication between these subsets and platelets was relatively active ( Figure 2F ). However, in the sepsis survivor group ( Figure 2G ) and the non-survivor group ( Figure 2H ), the network became increasingly impaired. Notably, the signaling intensity of the APP → CD74 axis, a key communication route between platelets and B cells, decreased progressively across groups: 10.88 in the healthy control group, 8.864 in the survivor group, and 6.636 in the non-survivor group. This trend indicates that platelet–B cell signaling is markedly suppressed during sepsis progression. Given the critical role of CD74 as both a chaperone of MHC class II molecules and a regulator of inflammatory signaling, attenuation of this axis may reflect impaired immune activation or defects in antigen presentation capacity. Taken together, these findings reveal that the immune communication network between platelets and immune cells undergoes substantial remodeling during the course of sepsis. The pronounced weakening of this network, particularly in non-survivors, suggests a functional impairment of platelets in maintaining immune homeostasis, which may contribute to immune dysregulation and poor clinical outcomes in sepsis.

To further explore the functional alterations of PBMCs under different clinical outcomes in sepsis, we first performed differential gene expression analysis across immune cell subsets between HCs and the sepsis survivor group ( Figure 3A , Supplementary Table S2 ). The results revealed extensive transcriptomic differences across multiple immune cell populations, suggesting that sepsis induces widespread immune activation and functional reprogramming in specific cell subsets. Subsequent comparison of PBMCs between the survivor and non-survivor groups revealed more pronounced transcriptional alterations in the non-survivors across several immune cell types ( Figure 3B , Supplementary Table S3 ), particularly in B cells, T cells, and platelets. These findings indicate that these key immune populations may undergo functional transitions that are closely associated with disease progression.

To better understand the functional changes in platelets during sepsis, we conducted GO enrichment analysis of DEGs between HCs and survivors ( Figure 3C ), as well as between survivors and non-survivors ( Figure 3D ). The results showed that platelet DEGs in sepsis were significantly enriched in biological processes such as blood coagulation, platelet activation, and hemostasis, indicating an enhanced activation of their classical hemostatic functions. Notably, these functional enrichments were further intensified in the non-survivor group, suggesting that in addition to their conventional roles, platelets may be actively involved in inflammatory responses and immune dysregulation in advanced stages of sepsis. In summary, our data reveal substantial transcriptomic reprogramming of monocytes and platelets within the PBMCs compartment of sepsis patients. Specifically, platelets exhibit a progressively enhanced functional state across groups with worsening clinical outcomes, highlighting their potential key role in the immunopathogenesis of sepsis.

Previous analyses revealed a progressive weakening of immune communication between platelets and B cells during sepsis; however, the underlying molecular mechanisms remain unclear. To investigate the role of m6A in this process, we first assessed the expression patterns of canonical m6A regulatory enzymes across different PBMCs subsets in HCs, sepsis survivors, and non-survivors ( Figures 4A–E , Supplementary Figures 3A–C ). We observed substantial heterogeneity in the expression of demethylases (ALKBH5, FTO) and methyltransferase complex components (WTAP, METTL3, METTL14), suggesting their potential involvement in the epigenetic regulation of immune function during sepsis.

Next, we performed MeRIP-seq on PBMCs from survivors and non-survivors. In both groups, m6A peaks were predominantly enriched in coding sequence (CDS) regions (46.9%), followed by 3′ untranslated regions (3′UTR) (30.4%) ( Figure 5A ). Notably, a slight reduction in CDS-localized m6A peaks was observed in the non-survivor group compared to survivors ( Figure 5B ), implying that disease progression may alter m6A distribution and affect mRNA stability or translational efficiency. Differential expression analysis revealed that several m6A regulators were significantly dysregulated in the non-survivor group ( Figure 5C ), further supporting the dynamic involvement of m6A modification in sepsis pathogenesis. Functional enrichment analysis of differentially methylated genes revealed significant associations with pathways related to metabolic regulation, inflammatory responses, apoptosis, and immune signaling ( Figures 5D, E ), indicating that m6A modification may influence sepsis through multilayered regulatory mechanisms. In summary, sepsis patients exhibit dynamic remodeling of m6A methylation in PBMCs, characterized by altered regulator expression, redistribution of modification sites, and functional pathway enrichment. These results highlight the potential contribution of m6A-dependent regulation to disease progression and outcome in sepsis.

To identify key genes potentially involved in m6A-mediated immune remodeling in sepsis, we integrated three sets of differentially expressed or modified genes for intersection analysis: (1) Differentially methylated genes between sepsis survivors and non-survivors in PBMCs; (2) DEGs in platelets between survivors and non-survivors; (3) Differentially expressed platelet genes between HCs and survivors. This integrative analysis yielded 56 overlapping candidate genes ( Figure 6A ). To evaluate their potential role in platelet–immune cell interaction, particularly in relation to B cell communication, we examined the correlation between candidate genes and the platelet-derived ligand APP. RPA1 exhibited the strongest correlation with APP expression ( Figure 6B , Supplementary Table S4 ), suggesting a possible role in the APP–CD74 signaling axis and platelet–B cell interaction.

Further correlation analysis revealed that RPA1 expression in platelets was significantly associated with multiple m6A regulators, including demethylases (ALKBH5, FTO), methyltransferase complex components (WTAP, METTL14), and reader proteins (IGF2BP2/3, YTHDC1/2, YTHDF1/2) ( Figures 6C–L ), indicating that RPA1 may be epigenetically regulated by m6A modification. Consistently, IGV browser visualization demonstrated reduced m6A modification levels and decreased RPA1 expression in PBMCs from the non-survivor group compared to survivors ( Figure 6M ), further supporting RPA1 as a putative m6A-regulated target gene in sepsis. Taken together, RPA1 emerged as a key candidate gene identified through cross-dataset integration. It is closely correlated with the platelet ligand APP and subject to regulation by multiple m6A factors. These findings suggest that RPA1 may contribute to the regulation of immune cell communication in sepsis and holds promise as a mechanistic mediator and potential therapeutic target.