Skip to main content
NCBI home page
Journal of Inflammation Research logoLink to Journal of Inflammation Research
. 2026 Sep 22;19:632459. doi: 10.2147/JIR.S632459

Integrated Multi-Omics Analyses Identify TYMP as a Candidate Protective Immunoregulatory Marker in CD4⁺ T Cells During Sepsis

Yanan Wang 1,2,*, Jianglin Zhao 1,2,*, Yushi Zhang 1,*, Ji Shen 1, Yuan Shi 1, Qingxiang Liu 3,✉, Xuefeng Zhang 1
PMCID: PMC13615804  PMID: 42801117

Abstract

Background

Cellular heterogeneity in sepsis complicates the identification of cell-specific therapeutic targets. We aimed to identify genetically supported candidate immunogenetic regulators associated with sepsis through integrative multi-omics analyses.

Methods

We conducted an integrative observational multi-omics study combining single-cell expression quantitative trait locus-based Mendelian randomization (sc-eQTL MR), bulk transcriptomic survival analysis, genetic colocalization, single-cell RNA sequencing (scRNA-seq), and a prospective observational clinical cohort. Publicly available summary-level or de-identified datasets were analyzed, including FinnGen sepsis GWAS data (GWAS ID: finn-b-O15_PUERP_SEPSIS), OneK1K sc-eQTL data, bulk transcriptomic data (GSE65682), and scRNA-seq datasets (GSE167363 and GSE151263). For prospective validation, 30 consecutive adult patients admitted to the intensive care unit who fulfilled the Sepsis-3 criteria were enrolled at the Affiliated Jiangyin Hospital of Nantong University between January 2026 and April 2026. Peripheral blood was collected within 24 hours after fulfillment of the Sepsis-3 criteria. TYMP concentrations in plasma were quantified using ELISA.

Results

TYMP showed a nominal inverse association with sepsis susceptibility in CD4 effector memory/TEMRA cells (OR = 0.644, P = 0.008) in the exploratory MR screening, while genetic colocalization suggested a shared genetic signal with sepsis susceptibility (PPH4 = 0.78). In an independent cohort of 30 patients with sepsis, plasma TYMP levels were significantly elevated in survivors than in non-survivors (124.9 ± 20.7 vs 109.6 ± 15.1 pg/mL, P = 0.033). TYMP exhibited prognostic value for 28-day mortality (AUC = 0.701, P = 0.042), and elevated TYMP expression was associated with improved survival (HR = 0.259, P = 0.029). Single-cell and pseudotime analyses associated TYMP-expressing CD4⁺T cells with immunoregulatory transcriptional programs involving hypoxia and TGFB1 signaling. Pseudotime analysis demonstrated dynamic transcriptional reprogramming toward an immunoregulatory state.

Conclusion

Integrated genetic, transcriptomic, single-cell, and clinical analyses nominate TYMP as a genetically supported candidate protective immunoregulatory marker associated with CD4⁺T cells in sepsis.

Keywords: sepsis, mendelian randomization, single-cell eQTL, TYMP, CD4+T cells, inflammation regulation, single-cell transcriptomics

Introduction

Sepsis is a life-threatening syndrome characterized by a dysregulated host immune response to infection, leading to acute organ dysfunction and high mortality rates worldwide.1–3 The immunopathogenesis of sepsis is highly complex and dynamic, encompassing both excessive inflammation and profound immunosuppression that collectively disrupt immune homeostasis.4,5 Despite decades of intensive research and numerous clinical trials, effective targeted therapies for sepsis remain elusive, largely owing to its remarkable cellular and molecular heterogeneity.6,7 This complexity highlights an urgent need to identify robust cell-type-specific regulators that can serve as actionable therapeutic targets and facilitate the development of precision immunomodulatory strategies for sepsis.

In recent years, bulk transcriptomic profiling has substantially advanced our understanding of the molecular alterations associated with sepsis.8,9 However, conventional transcriptomic analyses are inherently limited in their ability to distinguish causal drivers from downstream disease-associated changes. Mendelian randomization (MR) provides a robust genetic framework for causal inference by leveraging genetic variants, typically single-nucleotide polymorphisms (SNPs), as instrumental variables for exposures of interest.10 Because genetic variants are randomly allocated at conception, MR minimizes confounding and mitigates reverse causality, thereby serving as a natural experiment analogous to a randomized controlled trial.11 More recently, the integration of MR with single-cell expression quantitative trait loci (sc-eQTL) data has enabled causal inference at cell-type-specific resolution, offering unprecedented opportunities to dissect the immunogenetic architecture of complex diseases.12,13 Such integrative approaches facilitate the systematic identification and prioritization of high-confidence, cell-type-specific therapeutic targets within the heterogeneous immune landscape of sepsis.

Phosphorylase (TYMP), also known as platelet-derived endothelial cell growth factor, is an intracellular enzyme involved in pyrimidine metabolism and has been extensively implicated in vascular remodeling,14 thrombosis,15 and tumor progression.16 Emerging evidence further suggests that TYMP may influence inflammatory signaling and immune-cell responses in a context-dependent manner. However, whether TYMP participates in sepsis-associated immune dysregulation and whether its potential effects are restricted to specific immune-cell populations remain largely unknown.

In this study, we sought to identify genetically supported, cell-type-specific candidate regulators associated with sepsis through an integrative multi-omics framework. By combining sc-eQTL-based MR, transcriptomic survival analysis, genetic colocalization, and scRNA sequencing, we systematically evaluated the immunogenetic landscape of sepsis across multiple peripheral immune-cell populations. Through this unbiased cell-type-resolved screening, TYMP was prioritized as a candidate protective regulator because its genetically predicted expression was specifically associated with reduced sepsis susceptibility in the CD4⁺ effector memory/TEMRA compartment. Adult ICU patients fulfilling the Sepsis-3 criteria were selected for prospective validation to ensure a standardized and clinically relevant case definition and to evaluate circulating TYMP during the early acute phase of illness, when short-term outcome stratification is particularly relevant. Collectively, our findings nominate TYMP as a potential cell-type-associated prognostic and immunoregulatory marker in sepsis.

Methods

Data Sources and Study Design

Genome-wide association study (GWAS) summary statistics for sepsis were obtained from the FinnGen dataset (GWAS ID: finn-b-O15_PUERP_SEPSIS). The single-cell expression quantitative trait loci (sc-eQTL) summary data were derived from the OneK1K cohort (https://onek1k.org/), a comprehensive and high-resolution resource providing precise genetic regulatory information across distinct peripheral blood mononuclear cell (PBMC) subtypes. Bulk transcriptomic data and corresponding clinical survival information from patients with sepsis were downloaded from the Gene Expression Omnibus (GEO) database (Accession numbers: GSE6568217). In addition, single-cell RNA sequencing (scRNA-seq) datasets of sepsis patients were obtained from GEO (Accession number: GSE167363, GSE15126318,19) and were used to characterize immune-cell heterogeneity and investigate the cell-type-specific expression patterns of candidate genes in sepsis.

Single Cell eQTL MR Analysis

To investigate the cell-type-specific causal effects of gene expression on sepsis risk, we performed a two-sample MR analysis using the TwoSampleMR package.20 (version 0.6.14) in R. Genetic variants associated with gene expression were selected as instrumental variables (IVs) based on genome-wide significance thresholds. Palindromic single-nucleotide polymorphisms (SNPs) with intermediate allele frequencies were excluded during data harmonization to ensure the robustness of the causal inference. Because the eligible IVs for the prioritized genes in our analysis were restricted to a single genetic variant (nSNP = 1), the Wald ratio method was exclusively employed to obtain MR estimates. The results were reported as odds ratios (ORs) with corresponding 95% confidence intervals (CIs). Statistical significance was defined as a two-sided P value < 0.01.

Machine Learning Feature Selection and Survival Analysis

To validate the clinical relevance of the causal genes identified through MR analyses, we further interrogated bulk transcriptomic datasets from patients with sepsis. A Random Forest (RF) model was constructed using the randomForest package in R to identify prognostic genes associated with sepsis mortality. Specifically, the normalized bulk transcriptomic expression profiles from the GSE65682 dataset were utilized as input variables, with the clinical outcome (survivor vs non-survivor) set as the binary dependent variable. The model was trained with the number of trees (ntree) set to [500], and the number of variables randomly sampled at each split (mtry) set to the default square root of the total number of input genes. To control for overfitting and perform internal validation, we utilized the Out-of-Bag (OOB) error estimation, which inherently evaluates model performance on the subset of data omitted during the bootstrap resampling of each tree. Candidate genes were then ranked according to their variable importance, determined by the [Mean Decrease Gini] index, which quantifies each variable’s contribution to node homogeneity. The overlap between the top-ranked RF-derived prognostic genes and MR-prioritized causal genes was visualized using Venn diagrams. Kaplan-Meier (KM) survival analyses were subsequently performed using the survival and survminer R packages, and survival differences between groups were assessed using the Log rank test. To determine whether TYMP served as an independent prognostic factor, a multivariable Cox proportional hazards regression model was established, adjusting for relevant clinical covariates, including age and sex. In addition, Spearman correlation analysis was conducted to evaluate the associations between TYMP expression and established immunosuppressive checkpoint molecules, thereby exploring its potential role in immune regulation during sepsis.

Genetic Colocalization Analysis

To determine whether the observed association between gene expression and sepsis was attributable to a shared causal variant rather than linkage disequilibrium (LD)-driven confounding, Bayesian genetic colocalization analyses were performed using the coloc R package (version 5.2.3).21 We integrated the GWAS summary statistics of sepsis with the sc-eQTL data of the specific gene. The posterior probability of hypothesis 4 (PPH4), representing the probability that both traits share a common causal variant, was calculated to assess colocalization. Consistent with previous studies, a PPH4 value greater than 0.75 was considered strong evidence supporting a shared causal signal between gene expression and sepsis susceptibility. Regional association and locus visualization plots were subsequently generated to identify and illustrate the most likely shared causal variant.

Single-Cell RNA-Seq Data Processing and Clustering

The scRNA-seq data analysis was conducted using the Seurat package22 (version 5.2.1). The data were normalized, scaled, and subjected to principal component analysis (PCA) to reduce dimensionality. Uniform Manifold Approximation and Projection (UMAP) was used for non-linear dimensionality reduction and visualization. The CD4+T cell subpopulation was isolated based on the high expression of canonical markers (CD3E and CD4). This cluster was further stratified into TYMP-positive (TYMP+) and TYMP-negative (TYMP−) subgroups based on the expression status of TYMP.

Pseudotime Trajectory Inference

To characterize the dynamic state transitions of CD4+T cells, pseudotime trajectory analysis was performed using the Monocle2 package23(version 2.24.0) in R. Cells were ordered along a developmental trajectory according to transcriptional variation among highly variable genes, thereby reconstructing the continuum of cellular state transitions. The inferred trajectories were visualized in a reduced-dimensional space. Genes displaying significant dynamic changes across pseudotime were identified and hierarchically clustered to characterize transcriptional programs associated with cell-state transitions. Pseudotime-ordered heatmaps were generated to visualize the temporal exprewoyssion patterns of key immunoregulatory genes.

Measurement of Plasma TYMP Levels

A prospective observational study was conducted by consecutively enrolling adult patients admitted to the intensive care unit (ICU) of the Affiliated Jiangyin Hospital of Nantong University between January 2026 and April 2026. Peripheral blood samples were obtained from patients within 24 hours of fulfilling the Sepsis-3 criteria.2 Patients were excluded if they were younger than 18 years, pregnant, had a history of malignancy, chronic viral infection (including hepatitis B, hepatitis C, or HIV infection), autoimmune disease, or were receiving corticosteroid therapy. Patients were also excluded if written informed consent could not be obtained from the patient or, when the patient lacked decision-making capacity because of impaired consciousness or critical illness, from a legally authorized representative. Blood samples were centrifuged at 1500 rpm for 10 min at 4°C, and the plasma was separated and stored at −80°C until analysis. Plasma TYMP concentrations were determined using a commercially available human TYMP ELISA kit (BYabscience, Nanjing, China) according to the manufacturer’s instructions. Absorbance was measured at 450 nm using a microplate reader, and concentrations were calculated from a standard calibration curve. All measurements were performed in duplicate, and the mean value was used for statistical analysis.

Statistical Analysis

All statistical analyses and computational modeling were performed using R software (version 4.4.3). Continuous variables were compared using Student’s t-test or the Wilcoxon rank-sum test, as appropriate. Categorical variables were compared using the chi-square test or Fisher’s exact test. Receiver operating characteristic (ROC) curve analysis was performed to evaluate the prognostic performance of plasma TYMP levels for predicting 28-day mortality in patients with sepsis. The area under the ROC curve (AUC) and corresponding 95% confidence intervals (CIs) were calculated, and the optimal cutoff value was determined using the Youden index. Survival differences were assessed using the Log rank test. A two-sided P-value < 0.05 was considered statistically significant. This study was reported in accordance with the RECORD guidelines where applicable. The Mendelian randomization component was reported in accordance with the STROBE-MR guidelines.

Results

Systematic Sc-eQTL Mendelian Randomization Screening Identifies Cell-Type-Specific Candidate Genes Associated with Sepsis

To systematically investigate genetically supported associations between immune-cell-specific gene expression and sepsis susceptibility, we performed two-sample MR analyses using sc-eQTLs as instrumental variables. Because each eligible gene was instrumented by a single valid SNP, MR estimates were obtained using the Wald ratio method. Using a nominal P < 0.01 as an exploratory screening threshold, we identified 71 gene-cell association signals corresponding to 37 unique genes across multiple immune cell populations, including CD4+T cells, CD8+T cells, B cells, natural killer cells, monocytes, and dendritic cells (Figure 1). Both protective and risk-associated effects were observed, highlighting substantial cellular heterogeneity in the genetic architecture of sepsis. Several genes exhibited associations across multiple immune cell subsets, including BIN1, DDT, PARK7, CHMP4A, and RPS25, whereas others displayed highly restricted cell-type-specific effects. Notably, TYMP exhibited a significant protective association specifically within CD4 Effector memory/TEMRA cells (OR = 0.644, 95% CI: 0.465–0.893, P = 0.008), suggesting a potentially specialized role for this immune cell subset in sepsis susceptibility. Collectively, these exploratory findings provide a candidate gene set for subsequent clinical and biological evaluation.

Figure 1.

A forest plot of gene and immune cell exposure associations with sepsis odds ratios and confidence intervals. A forest plot with a left table and a right odds ratio plot for sepsis. Table columns read, exposure, outcome, nsnp, method, pval and OR 95 percent CI. Outcome is sepsis for all rows. Nsnp is 1 for all rows. Method is Wald ratio for all rows. The right plot x axis is labeled, OR 95 percent CI, with tick labels 0, 0.5, 1, 1.5, 2, 2.5, 3. A vertical reference line is at 1. Each row shows a point estimate with a horizontal 95 percent confidence interval. Example rows include, Natural Killer Cell DZIP3, OR 1.243 with 95 percent CI 1.103 to 1.401, pval less than 0.001; CD8 Effector memory DZIP3, OR 1.235 with 95 percent CI 1.098 to 1.389, pval less than 0.001; Natural Killer Cell CROCC, OR 1.720 with 95 percent CI 1.250 to 2.367, pval less than 0.001; CD4 Effector memory TEMRA TYMP, OR 0.644 with 95 percent CI 0.465 to 0.893, pval 0.008. The plotted intervals span both sides of 1 for some rows and stay entirely above or below 1 for others.

Open in a new tab

Single-cell eQTL (sc-eQTL)-based Mendelian randomization screening identifies exploratory cell-type-specific genetic associations with sepsis susceptibility. Forest plot showing gene-cell association signals prioritized using a nominal P < 0.01 across diverse peripheral blood immune cell populations. Sc-eQTLs were used as instrumental variables, and MR estimates were obtained using the Wald ratio method. A total of 71 gene-cell type association signals corresponding to 37 unique genes met the exploratory nominal threshold. Odds ratios (ORs) and 95% confidence intervals (CIs) are shown for each association.

Integrative Machine Learning and Survival Analysis Prioritizes TYMP as a Clinically Relevant Prognostic Biomarker

The 37 genes identified by sc-eQTL MR analysis yielded four overlapping candidates: BIN1, TYMP, MIF, and CCDC82 (Figure 2A). Kaplan-Meier survival analysis showed that TYMP and MIF were significantly associated with patient outcomes, whereas CCDC82 did not reach statistical significance (P = 0.055) (Figures 2B, S1A and B). Higher TYMP expression was associated with improved survival (P = 0.0028), while elevated MIF expression predicted poorer outcomes (P = 0.015). Although BIN1 was also associated with favorable survival (P = 1×10−4) (Figure S1C), its prognostic pattern was inconsistent with the predominantly risk-associated effects observed across multiple immune cell populations in the sc-eQTL MR analysis. Among the remaining candidates, TYMP exhibited a stronger prognostic association than MIF and showed concordant directions of association between the exploratory genetic screening and clinical outcome analyses. Specifically, genetically predicted higher TYMP expression in CD4 Effector memory/TEMRA cells showed a nominal inverse association with sepsis susceptibility in the exploratory MR analysis, while higher TYMP expression was associated with improved survival in the clinical cohort. Therefore, TYMP was prioritized for subsequent investigation.

Figure 2.

A composite figure with a Venn diagram, two Kaplan Meier plots, a TYMP bar chart and a ROC curve. Image A shows a Venn diagram with circles labeled Candidates (37) and Random Forest (252), overlapping at BIN1, TYMP, MIF, CCDC82. Image B displays a Kaplan-Meier plot for TYMP with an optimal cutpoint of 4.012. The x-axis is Time (days) from 0 to 30 and the y-axis is Probability of Survival from 0 to 100. The red line (high expression) is above the blue line (low expression) with p = 0.0028. Image C is a bar chart comparing TYMP levels between 28-day survivors and non-survivors, showing higher levels in survivors. Image D presents a ROC curve with AUC = 0.701 and P = 0.042. Image E is another Kaplan-Meier plot with a risk table, showing high expression (blue line) above low expression (red line) with P = 0.029. The risk table lists numbers at risk for low and high expression groups. Across these images, TYMP is a key gene in the overlap, with plots illustrating its expression impact on survival and ROC metrics.

Open in a new tab

Integration of machine learning and clinical transcriptomic analyses identifies TYMP as a prioritized prognostic candidate in sepsis. (A) Venn diagram showing the overlap between candidate genes identified by sc-eQTL-based MR analysis and prognostic feature genes selected using a Random Forest machine-learning algorithm in bulk RNA-seq cohorts of patients with sepsis. Four overlapping genes (BIN1, TYMP, MIF, and CCDC82) were identified for further evaluation. (B) Kaplan-Meier survival analyses of the TYMP in patients with sepsis. Patients were stratified into high-expression and low-expression groups using optimal expression cutoffs. (C) Comparison of plasma TYMP levels between 28-day survivors and non-survivors. Data are presented as mean±SD. *P < 0.05. (D) ROC curve evaluating the prognostic performance of plasma TYMP levels for predicting mortality in patients with sepsis. (E) Kaplan-Meier survival analysis based on plasma TYMP expression levels.

To further validate the clinical relevance of TYMP, we analyzed an independent cohort of 30 patients with sepsis, including 17 survivors and 13 non-survivors (Figure S2). Compared with survivors, non-survivors exhibited significantly higher APACHE II scores (29 ± 10 vs 22 ± 8, P = 0.028) and a greater requirement for mechanical ventilation (92.3% vs 47.1%, P = 0.017), whereas no significant differences were observed in age, BMI, SOFA score, or source of infection (Supplementary Table 1).

Plasma TYMP concentrations were significantly higher in survivors than in non-survivors (pg/mL, 124.9 ± 20.7 vs 109.6 ± 15.1, P = 0.033) (Figure 2C). Receiver operating characteristic (ROC) curve analysis demonstrated that plasma TYMP levels possessed moderate prognostic value for predicting 28-day mortality, with an area under the curve (AUC) of 0.701 (95% CI: 0.507–0.854; P = 0.042) (Figure 2D). Patients were subsequently stratified into high-expression and low-expression groups according to the optimal cutoff value. Kaplan-Meier survival analysis revealed that patients with high TYMP expression had significantly improved 28-day survival compared with those with low TYMP expression (HR, 0.259; 95% CI: 0.077–0.869; P = 0.029) (Figure 2E). These findings further support TYMP as a protective factor and clinically relevant prognostic biomarker in sepsis.

TYMP Is an Independent Prognostic Factor and Is Associated with Immunoregulatory Gene Signatures in Sepsis

To further evaluate the clinical relevance of TYMP, we performed multivariate Cox proportional hazards regression analysis adjusting for age and sex. Consistent with the Kaplan-Meier survival analysis, elevated TYMP expression remained independently associated with a reduced risk of sepsis-related mortality (HR = 0.52, 95% CI: 0.33–0.81, P = 0.004) (Figure 3A). In contrast, increasing age was associated with a higher mortality risk (HR = 1.02, 95% CI: 1.01–1.03, P = 0.005), whereas sex showed no significant association with outcome. These findings indicate that TYMP represents an independent protective prognostic factor in sepsis beyond basic demographic characteristics.

Figure 3.

Two plots showing TYMP hazard ratios and a Spearman correlation matrix for TYMP and immune genes. The image A showing a forest plot titled, Hazard Ratios for TYMP (Optimal Cutpoint). The horizontal axis label is not shown; tick labels run from 0.4 to 1.8. The vertical axis lists groupopt, age and gender. groupopt: Low (N equals 67) is marked Reference. High (N equals 412) has hazard ratio 0.52 with 0.33 to 0.81 and p value 0.004 with two asterisks. age (N equals 479) has hazard ratio 1.02 with 1.01 to 1.03 and p value 0.005 with two asterisks. gender: female (N equals 207) is marked Reference. male (N equals 272) has hazard ratio 1.15 with 0.79 to 1.68 and p value 0.456. A dotted vertical reference line is at 1. Text at bottom reads, number Events: 114; Global p value (Log Rank): 0.00076355. The image B showing a heatmap titled, Spearman correlation matrix (significant genes). The x axis label is not shown; the x axis categories are TYMP, CD274, IL1B, CD14, IDO1, CD33, TGFB1, PTGS2, HAVCR2, CSF1R. The y axis label is not shown; the y axis categories are TYMP, CD274, IL1B, CD14, IDO1, CD33, TGFB1, PTGS2, HAVCR2, CSF1R. Each cell contains a Spearman correlation coefficient. Row TYMP: TYMP 1.00, CD274 0.35, IL1B 0.34, CD14 0.33, IDO1 0.30, CD33 0.25, TGFB1 0.23, PTGS2 0.14, HAVCR2 0.13, CSF1R 0.13. Row CD274: CD274 1.00, IL1B 0.34, CD14 0.08, IDO1 0.31, CD33 minus 0.06, TGFB1 0.18, PTGS2 0.08, HAVCR2 0.10, CSF1R minus 0.18. Row IL1B: IL1B 1.00, CD14 0.34, IDO1 0.29, CD33 0.11, TGFB1 0.09, PTGS2 0.37, HAVCR2 minus 0.05, CSF1R 0.05. Row CD14: CD14 1.00, IDO1 0.01, CD33 0.47, TGFB1 0.20, PTGS2 0.03, HAVCR2 0.33, CSF1R 0.50. Row IDO1: IDO1 1.00, CD33 0.04, TGFB1 minus 0.09, PTGS2 0.11, HAVCR2 minus 0.01, CSF1R minus 0.07. Row CD33: CD33 1.00, TGFB1 minus 0.05, PTGS2 minus 0.03, HAVCR2 0.20, CSF1R 0.64. Row TGFB1: TGFB1 1.00, PTGS2 minus 0.11, HAVCR2 0.06, CSF1R minus 0.04. Row PTGS2: PTGS2 1.00, HAVCR2 minus 0.05, CSF1R minus 0.12. Row HAVCR2: HAVCR2 1.00, CSF1R 0.15. Row CSF1R: CSF1R 1.00. A vertical color scale at right is labeled from negative 1 to 1.

Open in a new tab

TYMP is an independent prognostic factor and is associated with an immunoregulatory gene signature in sepsis. (A) Forest plot showing multivariate Cox proportional hazards regression analysis of TYMP expression and sepsis survival. Hazard ratios (HRs) and 95% confidence intervals (CIs) are shown for each variable. ** P < 0.01. (B) Spearman correlation matrix illustrating the relationships between TYMP expression and selected immune-related genes in the sepsis cohort. Color intensity and numerical values represent Spearman correlation coefficients (r).

To gain insight into the biological processes associated with TYMP, we next examined its relationship with key immune-regulatory genes. Spearman correlation analysis demonstrated that TYMP expression was positively correlated with multiple immune-related molecules, including CD274 (PD-L1, r = 0.35), IL1B (r = 0.34), CD14 (r = 0.33), IDO1 (r = 0.30), CD33 (r = 0.25), and TGFB1 (r = 0.23) (Figure 3B). These results suggest that elevated TYMP expression is associated with a distinct immunoregulatory transcriptional program in sepsis.

Genetic Colocalization Suggests a Shared Genetic Signal Between TYMP Expression and Sepsis Susceptibility

To further assess whether the association signals for TYMP expression and sepsis susceptibility may originate from a shared underlying genetic variant rather than distinct variants in linkage disequilibrium, we performed Bayesian genetic colocalization analysis integrating sepsis GWAS summary statistics with TYMP eQTL data (Figure 4A). Regional association analyses revealed overlapping association signals within the TYMP locus on chromosome 22 (Figure 4B). Colocalization analysis provided evidence consistent with a shared genetic signal between TYMP expression and sepsis susceptibility, with a PPH4 of 0.78 (Figure 4C). Locus-specific visualization identified rs78007348 as a candidate variant shared by the TYMP eQTL and sepsis GWAS association signals.

Figure 4.

Three scatter plots and a regional Manhattan plot comparing sepsis GWAS and TYMP eQTL signals. The image A showing a scatter plot titled, GWAS vs eQTL minus log10 p Correlation Scatterplot. The x axis label is, GWAS minus log10 p, with ticks at 0.0, 0.5, 1.0, 1.5 and 2.0. The y axis label is, eQTL minus log10 p, with ticks at 0, 2, 4, 6 and 8. Many points cluster near y equals 0 across x from 0.0 to about 2.1. A fitted line rises slightly from near y about 0.5 at x equals 0.0 to near y about 0.8 at x equals 2.0. Several higher points appear, including near x about 2.0 with y about 6.5 and y about 8.0. The image B showing a scatter plot titled, GWAS Regional Manhattan Plot: chr22. The x axis label is, Genomic Position bp, with ticks at 50900000, 50950000, 51000000 and 51050000. The y axis label is, minus log10 p, with ticks at 0, 1, 2, 3 and 4. Points are densest below y equals 1 across the full x range. The highest point is near y about 4 around x about 50900000. Additional points reach about y 2.5 to 3.0 at several positions. The image C showing three plots with the text, PPH4 equals 0.78, centered above the two right plots. Left plot: a scatter plot with x axis label, GWAS summary statistics minus log10 p, ticks at 0.0, 0.5, 1.0, 1.5 and 2.0 and y axis label, TYMP eQTL minus log10 p, ticks at 0, 2, 4, 6 and 8. A labeled point, rs78007348, lies near x about 1.9 and y about 8.0. Most points lie below y equals 1 across x from 0.0 to about 2.1, with a smaller cluster around y about 3 to 4 near x about 0.3 to 1.0. A vertical legend labeled, r squared, shows tick labels 0.2, 0.4, 0.6, 0.8. Top right plot: a scatter plot with y axis label, GWAS summary statistics minus log10 p, ticks at 0.0, 0.5, 1.0, 1.5 and 2.0. The x axis has tick marks but no readable label. A labeled point, rs78007348, is near the top around y about 2.1. Most points lie between y equals 0.0 and y equals 1.0. Bottom right plot: a scatter plot with x axis label, chr22 Mb, with ticks around 50.9, 51.0, 51.0 and 51.0 and y axis label, TYMP eQTL minus log10 p, with ticks at 0, 2, 4, 6 and 8. A labeled point, rs78007348, is near y about 8.0 at x about 51.0. Most points lie below y equals 1 across the x range, with a small cluster around y about 3 to 4 near x about 51.0.

Open in a new tab

Genetic colocalization suggests a shared genetic signal between TYMP expression and sepsis susceptibility. (A) Scatterplot illustrating the correlation of genetic association signals between the sepsis GWAS and TYMP eQTL datasets. (B) Regional Manhattan plot displaying the landscape of sepsis GWAS signals across the specific genomic locus on chromosome 22. (C) Detailed regional association and colocalization plots. The analysis yielded a PPH4 of 0.78, with rs78007348 representing a candidate variant shared by the TYMP eQTL and sepsis GWAS association signals.

Taken together, the colocalization findings provide complementary genetic evidence that TYMP expression and sepsis susceptibility may share an underlying genetic signal. In conjunction with the exploratory MR findings, these results support the prioritization of TYMP as a candidate for further investigation.

Single-Cell Transcriptomic Analysis Reveals Immunosuppressive Signatures in TYMP+CD4+T Cells

To further investigate the cellular context and potential functional role of TYMP within the CD4+T cell compartment, we analyzed single-cell RNA sequencing data and performed dimensionality reduction and clustering analyses. A CD4+T cell population was first identified based on the expression of canonical T-cell markers, including CD3E and CD4 (Figure 5A and B). According to TYMP expression levels, this population was subsequently divided into TYMP-positive (TYMP+) and TYMP-negative (TYMP−) CD4+T cell subsets (Figure 5C and D).

Figure 5.

UMAP, violin, heatmap: CD4+ T cell markers, immunosuppression scores, pathway activity. The image analyzes CD4+ T cells through various graphs. Image A shows a UMAP plot of CD3E expression, with darker blue indicating higher levels. Image B uses a similar blue scale for CD4 expression. Image C illustrates TYMP expression, with darker blue for higher levels. Image D differentiates TYMP-negative and TYMP-positive CD4+ T cells in a grouped UMAP plot, marked by red and blue. Image E highlights TGFB1 expression, with darker blue for higher levels. Image F compares immunosuppression scores between TYMP-negative and TYMP-positive CD4+ T cells using a violin plot, showing higher scores in TYMP-positive cells, with a significant difference. The vertical axis ranges from 0.0 to 0.6. Image G is a heatmap of pathway activity scores for TYMP-negative and TYMP-positive CD4+ T cells, with red for higher activity and blue for lower. Pathways like TGFB and PI3K show notable differences. The analysis links marker expression to immunosuppression scores and pathway activity differences.

Open in a new tab

Single-cell transcriptomic profiling identifies a distinct immunoregulatory phenotype in TYMP+CD4+T cells. (A and B) UMAP plots showing the expression distribution of canonical T cell markers, CD3E and CD4. (C and D) UMAP plots illustrating the expression of TYMP and the subsequent stratification of CD4+T cells into TYMP-positive (TYMP+) and TYMP-negative (TYMP−) subpopulations. (E) UMAP plot highlighting the enriched expression of the immunosuppressive cytokine TGFB1 within the TYMP+ subset. (F) Violin plot comparing the overall Immunosuppression Score between TYMP+CD4+T cells and TYMP−CD4+T cells (****P < 0.0001, Wilcoxon rank-sum test). The immunosuppression score was calculated using the AddModuleScore function in Seurat, based on the expression of a curated panel of classical immunosuppressive and exhaustion markers, including TGFB1, IL10, IDO1, CD274 (PD-L1), CTLA4, HAVCR2 (TIM-3), LAG3, and TIGIT. (G) Heatmap displaying the differential activation scores of key metabolic and immunological signaling pathways across the two cellular subsets. Pathway activity was inferred from single-cell gene expression data using the PROGENy (Pathway RespOnsive GENes for activity inference) algorithm.

Analysis of immune regulatory gene expression revealed that TGFB1, a key mediator of immune regulation, was preferentially expressed in TYMP+CD4+T cells (Figure 5E). Consistent with this observation, immunosuppression scoring analysis demonstrated significantly higher immunosuppression scores in TYMP+CD4+T cells compared with TYMP−CD4+T cells (Figure 5F), suggesting an enhanced immunoregulatory phenotype within the TYMP expressing population.

To further characterize the biological features of TYMP+CD4+T cells, pathway activity analysis was performed. TYMP+CD4+T cells exhibited increased activation of multiple pathways related to immune regulation and cellular stress adaptation, including Hypoxia signaling, TGF-β signaling, NFκB signaling, and TNF-α signaling (Figure 5G). These findings indicate that TYMP expression is associated with a distinct transcriptional state characterized by enhanced immune regulatory activity and adaptation to inflammatory and hypoxic microenvironments.

Taken together, these single-cell analyses suggest that TYMP expression marks a CD4⁺T cell population with prominent immunoregulatory features. These observational findings provide additional biological context for the exploratory genetic and clinical associations and motivate subsequent functional investigation.

Pseudotime Trajectory Analysis Reveals Transcriptional States Associated with TYMP+CD4+T Cells

To further delineate the continuous state transitions of CD4+T cells within the septic microenvironment, we constructed a single-cell pseudotime developmental trajectory. The trajectory plots revealed a distinct continuum of cellular differentiation: TYMP-negative (TYMP−) cells were predominantly localized at the root of the trajectory, whereas TYMP-positive (TYMP+) cells were highly concentrated towards the terminal branches (Figure 6A and B). This distribution indicates that TYMP expression is preferentially associated with later pseudotime states of CD4+T cells.

Figure 6.

Three plots of CD4 T cell pseudotime: two branching scatter maps and a gene expression heatmap. Image A displays a scatter plot of CD4 T cells along a branching trajectory with two branch points. The x-axis is labeled Component 1, ranging from -15 to 15 and the y-axis is labeled Component 2, ranging from -5 to 10. A legend indicates two groups: TYMP-CD4+ T and TYMP+CD4+ T. The plot forms a three-arm structure converging near Component 1 at -5 and Component 2 at 2, with markers 1 and 2 at the junction. Image B shows the same trajectory colored by pseudotime, with a color scale from 0 to 30. The same markers 1 and 2 are present. Image C presents a heatmap of gene expression across pseudotime, with clustered gene modules. The x-axis is labeled Pseudotime, with ticks at 0, 10, 20, 30. The y-axis lists genes such as TRIB2, GNB1, IL6R and others. A vertical color scale ranges from -3 to 3. A right-side legend labeled Cluster shows groups 1 to 4.

Open in a new tab

Pseudotime analysis reveals dynamic transcriptional programs associated with TYMP+CD4+T cells. (A) Distribution of TYMP+CD4+T cells and TYMP−CD4+T cells along the inferred pseudotime trajectory. Numbers 1 and 2 indicate the two branch points identified in the pseudotime trajectory. (B) Pseudotime projection showing the progression of CD4+T-cell states during differentiation. Numbers 1 and 2 indicate the two branch points identified in the pseudotime trajectory. (C) Heatmap of genes dynamically expressed across pseudotime. Distinct gene modules displayed temporal expression patterns along the trajectory.

To characterize transcriptional programs associated with this trajectory, we performed hierarchical clustering of genes dynamically expressed along the pseudotime axis (Figure 6C). The heatmap demonstrated that as pseudotime progressed, the cells progressively downregulated canonical genes required for maintaining naïve or resting T cell states (eg, TCF7, SELL, and IL7R). More importantly, at the terminal stages of the trajectory, dominated by TYMP+cells—a specific cluster of genes intimately associated with immune suppression, inflammation resolution, and cellular tolerance was robustly activated and reached peak expression.

Specifically, the highly expressed genes during this terminal phase included: (1) Amphiregulin (AREG), a well-characterized tissue-repair factor frequently secreted by regulatory T cells to mitigate immune-mediated tissue damage; (2) Nuclear Receptor Subfamily 4 Group A Member 2 (NR4A2), a critical transcription factor widely reported to induce T cell exhaustion and enforce immune tolerance; (3) Dual Specificity Phosphatases1 (DUSP1) and DUSP2, classical negative regulators of the pro-inflammatory MAPK signaling cascades that effectively “brake” excessive cytokine production; (4) Regulator of G Protein Signaling 1 (RGS1), whose upregulation is a hallmark of diminished T cell chemotaxis and advanced immune exhaustion. Taken together, these pseudotime analyses show that TYMP+CD4+T cells are enriched at later trajectory states characterized by increased expression of genes associated with immune regulation, tissue repair, and reduced inflammatory signaling.

Discussion

In this study, we integrated sc-eQTL-based Mendelian randomization, genetic colocalization, clinical transcriptomic analyses, and single-cell transcriptomics to identify genetically supported cell-type-associated candidate regulators in sepsis. We prioritized TYMP as a candidate marker associated with sepsis susceptibility and clinical outcome. Single-cell analyses further showed that TYMP was enriched in a distinct immunoregulatory CD4+T-cell subset characterized by specific transcriptional programs and state distributions. Collectively, these findings nominate TYMP as a candidate CD4+T-cell-associated immunoregulatory and prognostic marker in sepsis.

TYMP, also known as platelet-derived endothelial cell growth factor, is a thymidine phosphorylase that has been extensively implicated in vascular remodeling,24 thrombosis,25 and tumor progression.26 However, its role in immune regulation remains incompletely understood. In the present study, TYMP emerged as one of the candidate genes showing convergent evidence across exploratory genetic screening, prognostic analysis, and cell-type-resolved transcriptomic analyses. The nominal inverse association observed in the exploratory MR screening was complemented by genetic colocalization, which suggested that TYMP expression and sepsis susceptibility may share an underlying genetic signal. Higher TYMP expression was also associated with improved survival in patients with sepsis and remained independently associated with mortality after adjustment for clinical covariates. Interestingly, TYMP+CD4+T cells displayed increased expression of TGFB1 and significantly higher immunosuppression scores compared with TYMP-counterparts. Although excessive immunosuppression has traditionally been considered detrimental in sepsis,27,28 accumulating evidence suggests that immunoregulatory pathways may also serve as essential counterbalances to uncontrolled systemic inflammation. Therefore, the association between higher TYMP expression and favorable outcomes may be consistent with a counter-regulatory immune state that limits excessive inflammation and supports immune homeostasis.

Pseudotime analysis provided additional information regarding the transcriptional states associated with TYMP-expressing CD4⁺T cells. Along the inferred trajectory, TYMP expression was associated with increased expression of genes involved in immune regulation and cellular adaptation, including NR4A2, AREG, and DUSP1. NR4A family members have been implicated in T-cell tolerance and exhaustion programs,29 whereas DUSP1 negatively regulates inflammatory signaling through suppression of MAPK activation.30 AREG has also been associated with tissue repair and inflammation resolution.31 Together, these coordinated transcriptional changes suggest that TYMP expression is linked to a CD4⁺T-cell state characterized by enhanced immunoregulatory and tissue-adaptive features, providing a biologically plausible basis for its association with favorable outcomes in sepsis. Although pseudotime analysis cannot independently determine the precise temporal sequence or causal direction of these cellular changes, it provides a useful framework for identifying potential state transitions and generating mechanistic hypotheses. Future longitudinal and functional studies will be important to clarify whether TYMP actively contributes to the establishment or maintenance of these transcriptional programs.

Several limitations should be considered when interpreting our findings. First, although TYMP was validated in an independent prospective clinical cohort, the cohort was small and single-center; thus, the clinical findings require confirmation in larger multicenter cohorts. In addition, validation was limited to circulating plasma TYMP, and CD4+T-cell-specific TYMP expression was not directly assessed. Second, some analyses used public summary-level or de-identified datasets not originally designed for this question, despite complementary validation analyses, the direction and magnitude of these potential biases remain uncertain. Third, the function of TYMP in CD4+T cells was inferred from transcriptomic data and remains experimentally unvalidated. Cellular and animal studies are needed to determine whether TYMP has a functional role in T-cell regulation and immune homeostasis during sepsis. Despite these limitations, our study highlights TYMP as a genetically supported candidate cell type-related regulator and prognostic biomarker in sepsis, warranting further experimental and clinical validation.

Conclusions

This study nominates TYMP as a genetically supported candidate cell-type-associated regulator and prognostic biomarker in sepsis through integrated genetic, transcriptomic, single-cell, and clinical analyses. Our findings suggest that TYMP expression is associated with an immunoregulatory CD4+T-cell state and may have potential value for prognostic stratification. Larger multicenter prospective studies are warranted to validate its prognostic utility and to determine whether TYMP-based immune stratification may inform the design of future clinical trials.

Acknowledgments

We thank the patients enrolled in this study, physicians, and nurses working in the ICU at the Affiliated Jiangyin Hospital of Nantong University.

Funding Statement

This work was supported by the Program of Wuxi Health Commission (grant no. M202454 to Xuefeng Zhang), and Young Scholars Fostering Fund of the First Affiliated Hospital of Nanjing Medical University (grant no. PY2025016 to Qingxiang Liu).

Abbreviations

AREG, Amphiregulin; CI, Confidence interval; DUSP1, Dual Specificity Phosphatases1; GEO, Gene Expression Omnibus; GWAS, Genome-wide association study; HR, Hazard ratios; IVs, Instrumental variables; KM, Kaplan-Meier; MR, Mendelian randomization; NR4A2, Nuclear Receptor Subfamily 4 Group A Member 2; PBMC, Peripheral blood mononuclear cell; PCA, Principal component analysis; PPH4, Posterior Probability of Hypothesis 4; RNA-seq, RNA sequencing; RF, Random forest; RGS1, Regulator of G Protein Signaling 1; Sc-eQTL, Single-cell expression quantitative trait loci; SNPs, Single-nucleotide polymorphisms; TYMP, Thymidine phosphorylase; UMAP, Uniform manifold approximation and projection.

Data Sharing Statement

The GWAS summary statistics for sepsis were obtained from the FinnGen dataset (GWAS ID: finn-b-O15_PUERP_SEPSIS), sc-eQTL summary statistics were downloaded from the OneK1K cohort (https://onek1k.org/). Bulk transcriptomic data and corresponding clinical information were retrieved from the Gene GEO database under accession number GSE65682. ScRNA-seq datasets were also obtained from GEO under accession numbers GSE167363 and GSE151263. All data generated or analyzed during this study are included in this published article.

Ethics Approval and Consent to Participate

The study was approved by the research ethic committee of the Affiliated Jiangyin Hospital of Nantong University (approval ID: 2026-KY005-01) and was in full compliance with the Declaration of Helsinki. Informed consent was obtained from each patient or their legal representative prior to enrollment in the study.

Author Contributions

Yanan Wang: Conceptualization, Methodology, Investigation, Writing - original draft. Jianglin Zhao: Conceptualization, Methodology, Investigation, Writing - original draft. Yushi Zhang: Methodology, Formal analysis, Writing - original draft. Ji Shen: Investigation, Data curation, Writing - original draft. Yuan Shi: Investigation, Writing - original draft. Qingxiang Liu: Formal analysis, Funding acquisition, Writing - review and editing. Xuefeng Zhang: Conceptualization, Supervision, Funding acquisition, Writing - review and editing. All authors gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

Dr Qingxiang Liu reports support for the work from Young Scholars Fostering Fund of the First Affiliated Hospital of Nanjing Medical University (grant no. PY2025016), during the conduct of the study. Dr Xuefeng Zhang reports support for the work from the Program of Wuxi Health Commission (grant no. M202454), during the conduct of the study. The other authors have declared that no conflict of interest exists.

References

  • 1.Martin-Loeches I, Singer M, Leone M. Sepsis: key insights, future directions, and immediate goals. A review and expert opinion. Intensive Care Med. 2024;50(12):2043–15. doi: 10.1007/s00134-024-07694-z [DOI] [PubMed] [Google Scholar]
  • 2.Singer M, Deutschman CS, Seymour CW, et al. The third international consensus definitions for sepsis and septic shock (Sepsis-3). JAMA. 2016;315(8):801–810. doi: 10.1001/jama.2016.0287 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Rudd KE, Johnson SC, Agesa KM, et al. Global, regional, and national sepsis incidence and mortality, 1990-2017: analysis for the global burden of disease study. Lancet. 2020;395(10219):200–211. doi: 10.1016/S0140-6736(19)32989-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Kox M, Bauer M, Bos L, et al. The immunology of sepsis: translating new insights into clinical practice. Nat Rev Nephrol. 2026;22(1):30–49. doi: 10.1038/s41581-025-01004-6 [DOI] [PubMed] [Google Scholar]
  • 5.van der Poll T. Immunotherapy of sepsis. Lancet Infect Dis. 2001;1(3):165–174. doi: 10.1016/S1473-3099(01)00093-7 [DOI] [PubMed] [Google Scholar]
  • 6.Sinha P, Meyer NJ, Calfee CS. Biological phenotyping in sepsis and acute respiratory distress syndrome. Annu Rev Med. 2023;74:457–471. doi: 10.1146/annurev-med-043021-014005 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Antcliffe DB, Burrell A, Boyle AJ, et al. Sepsis subphenotypes, theragnostics and personalized sepsis care. Intensive Care Med. 2025;51(4):756–768. doi: 10.1007/s00134-025-07873-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Yang S, Guo J, Kong Z, et al. Causal effects of gut microbiota on sepsis and sepsis-related death: insights from genome-wide mendelian randomization, single-cell RNA, bulk RNA sequencing, and network pharmacology. J Transl Med. 2024;22(1):10. doi: 10.1186/s12967-023-04835-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Liu C, Liu J, Yang Y. Bulk and single-cell transcriptomic reveals shared key genes and patterns of immune dysregulation in both intestinal inflammatory disease and sepsis. J Cell Mol Med. 2025;29(4):e70415. doi: 10.1111/jcmm.70415 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Bowden J, Holmes MV. Meta-analysis and Mendelian randomization: a review. Res Synth Methods. 2019;10(4):486–496. doi: 10.1002/jrsm.1346 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Birney E. Mendelian Randomization. Cold Spring Harb Perspect Med. 2022;12(4). doi: 10.1101/cshperspect.a041302 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Yazar S, Alquicira-Hernandez J, Wing K, et al. Single-cell eQTL mapping identifies cell type-specific genetic control of autoimmune disease. Science. 2022;376(6589):abf3041. doi: 10.1126/science.abf3041 [DOI] [PubMed] [Google Scholar]
  • 13.Nathan A, Asgari S, Ishigaki K, et al. Single-cell eQTL models reveal dynamic T cell state dependence of disease loci. Nature. 2022;606(7912):120–128. doi: 10.1038/s41586-022-04713-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Baris A, Fraile-Bethencourt E, Eubanks J, et al. Thymidine phosphorylase facilitates retinoic acid inducible gene-I induced endothelial dysfunction. Cell Death Dis. 2023;14(4):294. doi: 10.1038/s41419-023-05821-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Belcher A, Zulfiker A, Li OQ, et al. Targeting thymidine phosphorylase with tipiracil hydrochloride attenuates thrombosis without increasing risk of bleeding in mice. Arterioscler Thromb Vasc Biol. 2021;41(2):668–682. doi: 10.1161/ATVBAHA.120.315109 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Warfield BM, Reigan P. Multifunctional role of thymidine phosphorylase in cancer. Trends Cancer. 2022;8(6):482–493. doi: 10.1016/j.trecan.2022.01.018 [DOI] [PubMed] [Google Scholar]
  • 17.Scicluna BP, Klein KP, van Vught LA, et al. A molecular biomarker to diagnose community-acquired pneumonia on intensive care unit admission. Am J Respir Crit Care Med. 2015;192(7):826–835. doi: 10.1164/rccm.201502-0355OC [DOI] [PubMed] [Google Scholar]
  • 18.Qiu X, Li J, Bonenfant J, et al. Dynamic changes in human single-cell transcriptional signatures during fatal sepsis. J Leukoc Biol. 2021;110(6):1253–1268. doi: 10.1002/JLB.5MA0721-825R [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Jiang Y, Rosborough BR, Chen J, et al. Single cell RNA sequencing identifies an early monocyte gene signature in acute respiratory distress syndrome. JCI Insight. 2020;5(13). doi: 10.1172/jci.insight.135678 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Hemani G, Zheng J, Elsworth B, et al. The MR-base platform supports systematic causal inference across the human phenome. Elife. 2018;7. doi: 10.7554/eLife.34408 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Giambartolomei C, Vukcevic D, Schadt EE, et al. Bayesian test for colocalisation between pairs of genetic association studies using summary statistics. Plos Genet. 2014;10(5):e1004383. doi: 10.1371/journal.pgen.1004383 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Hao Y, Hao S, Andersen-Nissen E, et al. Integrated analysis of multimodal single-cell data. Cell. 2021;184(13):3573–3587. doi: 10.1016/j.cell.2021.04.048 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Qiu X, Mao Q, Tang Y, et al. Reversed graph embedding resolves complex single-cell trajectories. Nat Methods. 2017;14(10):979–982. doi: 10.1038/nmeth.4402 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Hong L, Yue H, Cai D, et al. Thymidine phosphorylase promotes abdominal aortic aneurysm via VSMC modulation and matrix remodeling in mice and humans. Cardiovasc Ther. 2024;2024:1129181. doi: 10.1155/cdr/1129181 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Li W, Yue H. Thymidine phosphorylase: a potential new target for treating cardiovascular disease. Trends Cardiovasc Med. 2018;28(3):157–171. doi: 10.1016/j.tcm.2017.10.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Tian S, Wang C, Zhao X, et al. A CEBPB/TYMP/GDF15 signaling axis mediates tumor growth and cisplatin resistance in bladder cancer. Transl Oncol. 2025;61:102537. doi: 10.1016/j.tranon.2025.102537 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Ovali MA, Percin S. Sepsis-associated immunosuppression: mechanistic Insights, Biomarkers, and therapeutic perspectives. Mol Biol Rep. 2025;53(1):148. doi: 10.1007/s11033-025-11312-6 [DOI] [PubMed] [Google Scholar]
  • 28.Pei F, Gu B, Miao SM, et al. Clinical practice of sepsis-induced immunosuppression: current immunotherapy and future options. Chin J Traumatol. 2024;27(2):63–70. doi: 10.1016/j.cjtee.2023.11.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Liu J, Wu M, Yang Y, et al. BTN3A1 expressed in cervical cancer cells promotes Vgamma9Vdelta2 T cells exhaustion through upregulating transcription factors NR4A2/3 downstream of TCR signaling. Cell Commun Signal. 2024;22(1):459. doi: 10.1186/s12964-024-01834-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Razazian M, Bahiraii S, Sohail A, et al. Fisetin ameliorates vascular smooth muscle cell calcification via DUSP1-dependent p38 MAPK inhibition. Aging. 2025;17(4):885–907. doi: 10.18632/aging.206233 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Wang Y, Li J, Zhang Y, et al. AREG(+) regulatory T cells mediating myocardial repair and neovascularization after myocardial infarction. Mol Med. 2025;31(1):229. doi: 10.1186/s10020-025-01281-8 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

The GWAS summary statistics for sepsis were obtained from the FinnGen dataset (GWAS ID: finn-b-O15_PUERP_SEPSIS), sc-eQTL summary statistics were downloaded from the OneK1K cohort (https://onek1k.org/). Bulk transcriptomic data and corresponding clinical information were retrieved from the Gene GEO database under accession number GSE65682. ScRNA-seq datasets were also obtained from GEO under accession numbers GSE167363 and GSE151263. All data generated or analyzed during this study are included in this published article.

Articles from Journal of Inflammation Research are provided here courtesy of Dove Press

RESOURCES