In this study, we obtained scRNA-seq data from 8 blood samples of sepsis patients from the GEO database (GSE151263 and GSE167363). Firstly, we filtered out clusters with cell counts less than 3, cells with the number of genes mapped less than 300, cells with more than 15% of mitochondrial genes, cells with more than 60% of ribosomal genes and cells with more than 0.5% of erythrocytic genes. Then, data normalization was performed via “NormalizeData” package and the top 40 principal components (PCs) were extracted through principal component analysis (PCA). Subsequently, the Harmony algorithm was used to integrate the data from multiple samples and remove batch effects (Korsunsky et al., 2019). Uniform manifold approximation and projection (UMAP) was utilized for unsupervised clustering and unbiased visualization of cell subpopulations on a two-dimensional map. Cell cluster annotations were based on canonical gene markers. When comparing the differences of gene expression level between a cluster and all other clusters, differential expression analysis was conducted to compare the expression differences of gene between a cluster and all other clusters using the “FindAllMarkers” function. |log2 (fold change) | > 1.5 and P-value< 0.05 were set as the threshold of marker genes for each cluster. Finally, we annotated the cell subpopulations of the different clusters using the “SingleR” package (Aran et al., 2019). We downloaded the RNA-sequencing data and clinical information of blood samples of sepsis patients from GEO database (http://www.ncbi.nlm.nih.gov/geo/).

The WGCNA package was used to screen genes most involved in monocytes proportion (Langfelder and Horvath, 2008). The overall relevance of all samples in the dataset was ascertained via clustering samples and then outliers was removed. The optimum soft thresholding power β was determined and selected according to the lowest power for which the scale-free topology fit index had a high value. 30 was set as the minimum gene number/module and 3 modules were produced. Eventually, the correlation analysis between modules and traits was performed to explore the most relevant modules for monocytes content.

First, 90 genes related to monocytes were obtained from single-cell transcriptome analysis and 7749 antiquewhite module genes from WGCNA. Through taking the intersection of the two, we demonstrated 22 candidate monocyte-related genes for sepsis. Subsequently, in order to obtain monocyte-related genes that is involved in prognosis, least absolute shrinkage and selection operator (LASSO) regression analyses were performed.

For demonstrating genetic variants involved in gene expression, eQTL analysis was conducted, using transcriptome and genotype data from a variety of cohorts. Summary eQTL data leveraged in the current study were obtained from the GWAS Catalog website (https://gwas.mrcieu.ac.uk/). The “TwoSampleMR” package was then employed to confirm strongly associated SNPs (p value<5e − 8) as instrumental variables (IVs). The threshold of linkage disequilibrium was set at r² < 0.001 and clumping distance = 10,000 kb. According to the filter of “F-test value >10”, the week SNPs were removed.

The cathepsin S (CTSS) gene was selected based on the aforementioned analysis for further research, and we compared the expression level of signature genes in blood samples of sepsis or the normal control. Subsequently, according to the expression levels of CTSS, the genes in GSE dataset were classified into a low expression group and a high expression group. The differentially expressed genes (DEGs) were then screened via the Limma package. The parameters |log2FC|>0.5 and p < 0.05 were used as the screening criteria for DEGs. Additionally, volcano plot and heatmap of DEGs were drawn using ggplot2 and pheatmap R packages.

To reveal the potential biological functions of CTSS, we used the R package “clusterProfiler” to analyze Gene Ontology (GO) enrichment of DEGs (Yu et al., 2012). GO terms, including biological processes (BPs), with adjusted p < 0.05, were considered statistically significant.

In order to explore the clinical factors and the value of CTSS for clinical prognosis of patients with sepsis, this study used the R package rms to establish a prediction nomogram that set 1-, 2-, and 3-week OS as the endpoints. Additionally, the calibration curve was utilized to assess the accuracy and resolution of the nomogram. In addition, “survivalROC” R package was used to depict the time-dependent receiver operating characteristic (ROC) curves and the area under curve (AUC). Finally, decision curve analysis (DCA) was also performed to evaluate the net clinical benefit of nomogram on patient prognosis (Petanceska et al., 1996).

We screened sepsis patients according to the following inclusion and exclusion criteria. The inclusion criteria were (1) a suspected or documented infection at the time of admission to the ICU; (2) compliance with the international consensus definition of sepsis (Sepsis-3.0); (3) age ≥ 18 years; (4) enrollment within 24 hours of hospital arrival. The exclusion criteria were (1) age < 18 years; (2) other infections and diseases, such as coronary heart disease, malignant diseases, aseptic inflammation, cerebral infarction, and brain injury. We collected 1ml of peripheral blood samples from healthy volunteers and sepsis patients, which was approved by the Ethics Committee of Second Affiliated Hospital of Fujian Medical University (Approval number: 2024542). After collecting blood samples via venipuncture and mixing it thoroughly with equi-voluminal whole blood and tissue diluent, the mixture was slowly added to a centrifuge tube containing equi-voluminal separation solution (Solarbio, Beijing, China). The centrifuge tube was then centrifuged at 2000 rpm for 30 min at room temperature. We then transferred the second layer of milky white lymphocytes to another centrifuge tube and added 10 ml cell washing solution. After being centrifuged at 1500 rpm for another 10 min, the supernatant was removed and 10 ml cell washing solution was added again. This procedure was repeated three times to isolate monocytes from peripheral blood samples. Once isolated, the monocytes were suspended and cultured in RPMI1640 medium with 10% FBS and 1% P/S.

The RAW264.7 and THP-1 cell lines were purchased from the Pricella Life Science & Technology Co., Ltd (Wuhan, China). The THP-1 and RAW264.7 cell lines were cultured in 1640 medium with 10% fetal bovine serum (FBS) and 1×penicillin-streptomycin. Cells were routinely cultured in a CO₂ incubator at 37°C.

We extracted RNA from cells using TRIzol (Gene Copoeia, MD, USA) and performed the reverse transcription of 1 µg total RNA from each sample to cDNA via the PrimeScript RT reagent Kit (RR047A, Takara, Japan). The reaction conditions were listed below: 37°C for 15 min and 85°C for 5 s. RT-qPCR was then performed on the Bio-Rad Real-Time PCR cycler. The reaction mixture contained TB Green Premix Ex Taq II 5 μL, upstream primer 0.4 μL, downstream primer 0.4 μL, cDNA 2 μL, and RNase Free dH₂O₂ 2 μL. The reaction conditions were as follows: 95°C for 30 s, followed by 40 cycles of 95°C for 10 s and 60°C for 30 s. Each well was set up with 3 replicates. The relative expression levels of genes were calculated using the 2^-ΔΔct method. The primer sequences were listed in Supplementary Table S1 .

After determining the concentration of protein using a BCA protein assay kit (Thermo), equal amounts of proteins (40 µg) in each sample were separated by 12% SDS-PAGE and then transferred to a PVDF membrane (Millipore, Billerica, MA, USA). Subsequently, we blocked the membrane using 5% skimmed milk at room temperature for 1 h and incubated them with the anti-CTSS (Signalway, cat. 43650; 1:800) and anti-β-actin (Affinity, cat. AF7018, 1:10000) primary antibodies. Thereafter, the membranes were incubated with the horseradish peroxidase (HRP)-conjugated secondary antibody. Finally, the images were visualized using the ChemiDoc imaging system (Bio-Rad) and blots were quantified via Image J software.

Peripheral monocytes (1×10⁵) from sepsis patients or healthy volunteers were added to the upper chamber of the Transwell plate (24-well plate, 5 μm pores; Corning, NY, USA) in 200 μL of serum-free medium. RPMI 1640 medium supplemented with 10% FBS and 1% P/S was added to the lower compartment of the plate. After incubation at 37°C for 24 h, the migrated cells on the lower surface of the membrane were fixed in 10% formalin and then stained with crystal violet. Finally, we selected five random fields of each well to be photographed, and the number of cells were obtained. The experiment was repeated three times independently.

Male C57BL/6 mice were maintained in a temperature-controlled environment with a regular 12-hour light and dark cycle. All procedures in this study were approved by the Ethics Committee of Second affiliated hospital of Fujian Medical University (ethical approval number: 2024542). Sepsis model was established via cecal-ligation puncture (CLP) by referencing a surgical procedure as previously described. In brief, the cecum was exteriorized and ligated after anesthesia with 2–4% sevoflurane (inhaled), and then punctured with a 21G needle. After 12 hours post-CLP, mice were euthanized with a pentobarbital overdose (100 mg/kg) and tissue samples were preserved at −80°C for further immunofluorescent staining.

Statistical analysis was conducted using SPSS software (version 26.0; SPSS, Chicago, IL, USA). Data are expressed as mean ± standard deviation (SD). The quantitative data between two groups were compared and analyzed by Wilcoxon test (two-tailed), while a one-way analysis of variance (ANOVA) was used to compare multiple groups, followed by Bonferroni post hoc tests. All images were drawn using R version 4.1.3 (https://www.r-project.org/) and its adequate packages. The threshold of statistical significance was set at p < 0.05.

In order to detailedly depict the cellular constitution of blood sample at single-cell resolution and demonstrate the cell markers of monocytes, we reanalyzed the scRNA-seq data of blood samples from 8 sepsis patients. Supplementary Figure S1A displayed the depth of sequencing, the range of detected gene numbers, and the percentage of mitochondria, hemoglobin and ribosome content in each sample. After removing low-quality cells through strict quality control, 27069 cells were included in the subsequent analysis. We used the PCA method for dimensionality reduction ( Figure 1A ), and selected 40 PCs for further analysis ( Figure 1B ). Subsequently, Harmony algorithm was used to integrate the data and correct the possible batch effects. We then obtained 13 clusters, which were visualized using the UMAP algorithm ( Figure 1C ). The relative expression of marker genes in cell subpopulations were presented in the heatmap ( Figure 1D ). The “singleR” algorithm was utilized to annotate cell subpopulations and indicated that clusters 2, 5, 7 and 12 were defined as monocyte subpopulations ( Figure 1E ). Due to the special focus on the monocytes in this study, we scrupulously annotated these clusters again using canonical markers (FCN1, CD14) and determined their identity of monocyte subpopulations ( Figure 1F ; Supplementary Figure S1B ). Finally, we obtained 90 monocyte marker genes of sepsis according to |logFC| >1.5 and p value< 0.05 ( Supplementary Table S2 ). As expected, the GO enrichment analysis suggested that monocyte gene markers were enriched in biological processes including positive regulation of defense response, phagocytosis and regulation of endocytosis ( Figure 1G ).

To fully anatomize monocyte-related biomarkers, we performed WGCNA in the bulk RNA-seq dataset of blood sample in sepsis. Initially, after removing the outliers from the blood samples, six were determined as the optimal soft-threshold power and the modules were demonstrated by WGCNA algorithm ( Figure 2A , Supplementary Figure S2 ). We compared data from two immune infiltration algorithms, including MCPCOUNTER and CIBERSORT, as we expected, there was excellent congruity on the percentage of monocytes in both data (R=0.288, P=0.002; Figure 2B ). In order to precisely determine the key modules of WGCNA involved in monocytes, we separately calculated the correlation between the two types of immune infiltration data and the modules. As displayed in Figure 2C , the antiquewhite4 module showed the highest correlation to the deconvolution result. Thus, we demonstrated the gene in the antiquewhite4 module as a potential biomarker for monocytes in sepsis ( Figures 2D, E ; Supplementary Table S3 ). In addition, amazingly, the results from the GO enrichment analysis of the antiquewhite4 module genes were consistent with the enrichment analysis of the monocytes marker genes demonstrated by scRNA-seq data ( Figure 2F ). In conclusion, our analysis identified an obvious correlation between the antiquewhite4 module and monocytes in sepsis via WGCNA.

By intersecting the 90 monocyte marker genes with the 7749 antiquewhite4 module genes for fetching, 22 candidate monocyte-related genes were obtained ( Figure 3A ). Initially, LASSO analysis was performed to derive a prognostic signature composed of 4 key monocyte-related marker genes (MRMGs), including cluster of differentiation 14 (CD14), CTSS, C-X-C motif chemokine ligand 8 (CXCL8) and Thrombospondin 1 (THBS1) ( Figures 3B, C ; Supplementary Table S4 ). Subsequently, we investigated the causal effect of key MRMGs expression on sepsis risk and mortality using their e-QTL data from GWAS Catalog website (https://gwas.mrcieu.ac.uk/) via MR analysis. Figures 3D, E displayed that only CTSS exhibited protective causality with sepsis risk and mortality. Thus, we further examined the expression of CTSS in the GSE131761 dataset. We discovered that the expression of CTSS was significantly downregulated in sepsis patients than that in control healthy ( Figure 4A , p<0.001). Furthermore, we utilized the survival data of sepsis patients in GSE65682 dataset to further explore the prognostic significance of CTSS ( Figure 4B ). We found that patients in the low group of CTSS expression displayed obviously worse prognosis when compared with high group. This series of analysis demonstrated the crucially protective role of CTSS in sepsis development and prognosis, promoting it to be selected for subsequent analysis.

Patients with sepsis in GSE65682 dataset was classified into a low-expression group and a high-expression cohort according to the median value of CTSS, with p value < 0.05 and |log2FC| >0.5; 161 low-expressed genes and 354 up-expressed genes were obtained ( Figure 4C ). The heat map only showed the top 50 DEGs in |logFC| order ( Figure 4D ). In order to further study the functional effects of different CTSS expressions, we first analyzed the 515 DEGs of GSE65682 by GO enrichment analysis. The top five groups functional effects of BP included immune response-regulating signaling pathway, cell killing, positive regulation of cytokine production, positive regulation of defense response, and leukocyte activation involved in immune response ( Figure 4E ). The outcomes of the GO enrichment analysis of DEGs were also similar with the enrichment analysis of the monocyte marker genes and the antiquewhite4 module genes.

Besides the obvious survival outcomes found between populations with high/low-expression of CTSS, distinct variances were also seen in their clinicopathologic characteristics. An incremental decrease in the expression level of CTSS was observed with declining endotype class, indicating a potential association between the expression level and the progression of sepsis ( Figure 5A ). However, the expression level of CTSS did not display significant difference in different gender and ages ( Figures 5B, C ). Subsequently, we established a nomogram by integrating clinicopathologic factors and CTSS to predict 1-, 2-, and 3-week survival probabilities of sepsis patients, respectively ( Figure 5D ). Calibration plots demonstrated the high consistence between the observed values and the predicted values ( Figures 5E–G ). Additionally, The AUCs showed the nomogram provided more clinical net benefit in forecasting prognosis at 1-, 2-, and 3-weeks ( Figures 5H–J ). DCA showed that the nomogram also held the optimum clinical net benefit for 1-, 2-, and 3-weeks survival ( Figures 5K–M ). These findings indicated that the nomogram based on the expression level of CTSS could be served as an effective way to forecast prognosis of sepsis patients in clinical practice.

Aforementioned analysis demonstrated the important role and prognostic significance of monocyte-related marker CTSS in sepsis. Interestingly, activation of monocytes is one of main characteristics of sepsis. Thus, we questioned whether sepsis-induced CTSS downregulation is involved in activating monocytes. Monocytes were firstly isolated from the peripheral blood of healthy individuals and septic patients using commercial kit. qRT-PCR analysis demonstrated obviously lower expression levels of activated monocyte-related markers (MCP-1, TNF-α, IL-6, IL-1β) mRNA in the peripheral monocytes of septic patients, compared to monocytes from healthy individuals ( Figures 6A–D ) (p < 0.05). The chemotactic ability of monocytes was obviously stronger in septic patients than that in healthy control ( Figures 6E, F ) (p < 0.05). qRT-PCR, western blot and immunofluorescence analysis verified the lower expression levels of the CTSS mRNA and protein in the peripheral monocytes of septic patients ( Figures 6G–K ) (p < 0.05). The expression level of CTSS was subsequently investigated in THP-1 and RAW264.7 cells following LPS stimulation. The findings indicated a significant decrease in the mRNA and protein levels of CTSS in both cell lines upon LPS treatment ( Figures 7A–F ) (p < 0.05).

Following the administration of LPS, we investigated the role of CTSS in monocyte activation during sepsis by treating monocytes obtained from healthy individuals with a CTSS agonist. The qRT-PCR results ( Figures 7G–J ) showed a significant increase in the expression levels of activated monocyte-related markers (MCP-1, TNF-α, IL-6, IL-1β) mRNA in the LPS group compared to the control group. In contrast, the CTSS agonist group exhibited a significant decrease in the expression levels of these marker genes compared to the LPS group (p < 0.05). Migration analysis ( Figure 7K ) revealed an increase in the number of invaded cells in the LPS group compared to the control group. Conversely, the CTSS agonist group exhibited a decrease in the number of invaded cells compared to the LPS group.

Tissue inflammation is one of major characteristic of sepsis. Thus, we mined the GSE207651 dataset profiling the single-cell immune landscape throughout lung sepsis to understand the impact of CTSS on the tissue inflammatory responses of septic mice. After data filtration, integration and removal of potential batch effects, we annotated each cell subpopulation as epithelial cells, T cells, B cells, endothelial cells, fibroblasts, granulocytes, macrophages, monocytes, and NK cells ( Figure 8A ). Thereafter, with a special focus on the CTSS, we observed monocytes and macrophages showed significant decrease of this gene in septic lung injury, which was believed to play a role in the function of monocytes/macrophages ( Figures 8B, C ). Subsequently, to further elucidate the possible role of CTSS in monocytes/macrophages during sepsis, we assessed the infiltration of macrophages into the septic and normal lungs of mice. The results showed that the number of infiltrated macrophages in the lungs of septic mice were higher than that in the lungs of control mice, as evidenced by the higher F4/80 fluorescence intensity in septic lungs. However, CTSS agonist effectively decreased the number of infiltrated macrophages in the lungs of septic mice ( Figure 8D ).