The transcriptome characteristics of the postmortem parietal cortex were compared between 12 sepsis deaths and 12 non-septic critically sick individuals in the GEO database GSE135838. The GEO database GSE135838 is available on the National Center for Biotechnology Information (NCBI) website at https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi.acc=GSE135838. The entire RNA data from the chosen samples were downloaded for subsequent analysis, but information on each sample’s age, etiology, medication treatment, or prognosis was unavailable. The “Deseq2” R package was then applied to the raw data to identify the genes differentially expressed in human brains between septic patients and non-septic critically ill patients. In each sample, genes that met the following requirements remained (1): |log2 (fold change) | > 1 and (2) p< 0.05. With the help of the “ggplot2” R package, a volcanic map was made to illustrate the results. Additionally, we used the “pheatmap” R package to conduct a bi-directional hierarchical clustering analysis of the collected DEGs, using TBtools to visualize the result as a heat map. Then, the Human Protein Atlas (HPA) protein annotation database and the Uniprot database were used to download the extracellular protein gene list GO:0005576 and the extracellular protein gene list, respectively. To identify the most commonly occurring EP-DEGs, these two lists were intersected with the DEGs from the GSE135838 dataset using a Venn diagram. The significantly highly expressed EP-DEGs were then displayed on a heat map.

The HPA Human brain dataset was used to evaluate the mRNA expression levels of the 14 top EP-DEGs, which include the cerebral cortex and the hippocampal formation in the healthy brain in 12 different regions. Expression stereograms were created using the nTPM values from the database. The HPA dataset’s protein concentration values for the 14 highly expressed EP-DEGs in normal plasma, which has been reported in multiple literature sources, were also used to construct the scatter figure.

DAVID Bioinformatics Resources v6.8 was used to do the Gene Ontology (GO) mapping and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways analysis to annotate the potential functions of EP-DEGs. The projected biological process (BP), cell composition (CC), and molecular function (MF) of each EP-DEG were evaluated, and all the EP-DEGs were posted to the DAVID website (version 6.8). The enriched GO terms were displayed using the R programming language’s GOplot package, which was also used to examine the intersection of the genes implicated in the GO terms. Additionally, the pathways in which the EP-DEGs participated were predicted and mapped using the KEGG database in conjunction with DAVID.

The Search Tool for Retrieval of 31 Interacting EP-DEG (STRING; http://STRING-db.org;GeneMANIA; http://genemania.org/) online database was used to create protein network interaction maps. These two-genome PPI networks were predicted using the multi-protein online tool in the STRING database, and they were displayed using Cytoscape 3.6.1. The most critical modules were then filtered using the Complex Detection (MCODE) plugin, a Cytoscape App, which recognizes clusters in the produced network based on topology and detects strongly connected regions. Then, the MCC and Radiality algorithm in Cytohubba, an APP in Cytoscape, was utilized to screen out the hub genes (23). To map the protein interaction network and further investigate which biological processes the gene cluster is engaged in, the genes in the gene cluster with the highest score were screened and imported into STRING.

As previously reported (24), sepsis was induced using cecal ligation and puncture (CLP) in male Sprague-Dawley rats weighing 280 ± 30 g under sterile conditions with the use of anesthesia. After midline laparotomy, the cecum was exposed, ligated below the ileocecal valve, and punctured twice with a sterile 18-gauge needle. The abdominal cavity was closed with sutures following CLP, and normal saline (5 ml/250 g body weight) was injected subcutaneously. The control group underwent a sham operation with laparotomy and suturing only. The experimental protocol was approved by the Animal Use and Ethics Committee of Xiangya Third Hospital, Central South University (Approval NO. XMSB-2023-0018).

The induction of sepsis in rats leads to varying degrees of cognitive impairment and subsequent behavioral changes. It is critical to observe the behavior of rats to validate biomarkers and evaluate the successful establishment of the SAE model (25, 26). The modeled rats were randomly divided into two groups, one for behavioral testing (8 in the sham group and 11 in the CLP group) and the other for sample collection testing (14 in the sham group and 12 in the CLP group). Behavioral tests, including novel object recognition (NOR) and the Barnes maze test, were performed to determine the cognitive functions of rats after modeling. Referring to previous reference (27), rats were habituated to a NOR test arena without objects for one day, and then placed in a testing arena containing familiar and novel objects. These objects were randomly used as either familiar or novel, without repetition. Exploration behavior was recorded during each session. By calculating the novel object recognition index, the experiment assesses the memory capacity of mice in recognizing new objects. This index is obtained by dividing the time spent exploring the new object by the total time spent exploring both the new and familiar objects. The assessment of spatial learning ability in rats is performed using the Barnes maze, according to earlier research (28). Rats are positioned on a circular platform that contains 20 equally spaced holes, with one of them leading to a dark chamber referred to as the target box. To locate the target box, rats are stimulated with powerful light (200 W) while on the platform. The rats’ performance is evaluated in the reference memory phase on the first day following the training sessions, which involves recording the number of errors they make during each attempt using a video tracking system.

After 6 days of CLP modeling, cerebrospinal fluid and blood samples were collected from the rats under deep anesthesia using a stereotactic instrument, with the rats placed in a 45° prone position with their heads facing upward. The skin and muscle were incised after neck dissection, and a needle was inserted through the dura mater to collect 100-200 μl of CSF from the medullary canal (approximately 0.2 mm in depth). Subsequently, the chest was opened, and 4 ml of blood was collected from the left ventricle. The collected CSF was centrifuged at 2000g for 20 minutes at 4°C, and the collected blood was centrifuged at 3000g for 10 minutes. The obtained supernatants were collected and stored at -80°C for further analysis. ELISA was performed to validate the expression levels of EP-DEG in the control and septic blood samples. The levels of MMP8 and S100A8 were determined using commercial ELISA kits (MMP8: Cat. #ER1166, Wuhan Fine Biotech Co., Ltd., China; S100A8: Cat. #ER0592, Wuhan Fine Biotech Co., Ltd., China) according to the manufacturer’s instructions.

Patients 18 years of age or older who met the clinical criteria for severe sepsis (1) within the previous 24 hours at any time during their stay in the ICU were enrolled in the study after being screened for eligibility criteria. Non-infected critically ill patients within 24 hours of admission were enrolled as the control group. All participants provided written informed consent, and our study was approved by the Ethical Review Committee of the third Xiangya Hospital, Changsha, China (Approval number: 22139) and registered at the Chinese Clinical Trial Registry (ChiCTR2300069425). The following were the exclusion criteria: primary central nervous system disease, such as cerebral hemorrhage; pre-existing psychiatric or neurological disorders; use of immunosuppressive medications, such as antibiotics and hormones before admission; recent history of blood transfusion; clear evidence of encephalitis/meningitis; altered consciousness and mental status due to other causes of acute encephalopathy; renal replacement therapy, such as hemofiltration and hemodialysis; cancer. Clinical data were collected on all recruited cases, including vital signs, blood gas analysis, blood routine, liver and kidney function, infection site, mechanical ventilation duration, and more. The patients’ SOFA, APACHE-II, and GCS scores were evaluated. 28-day mortality, ICU length of stay, and hospitalization time were tracked. All individuals underwent a neuropsychiatric evaluation for brain damage. A thorough history from the patient or family was acquired to look for early clinical indicators of brain dysfunction, such as disorientation, agitation, or a loss of consciousness. All SAE diagnoses come from the records of ICU patients’ medical histories by ICU physicians. In the ICU, SAE refers to the clinical manifestations of concomitant extra-infection and neurological dysfunction in sepsis patients, which can present as impairment of consciousness ranging from delirium to coma. Doctors need to perform neurological system examination and consciousness state assessment for sepsis patients, and exclude other acute cerebrovascular diseases, metabolic encephalopathy, intoxication, alcohol or drug withdrawal, pulmonary brain disease, non-convulsive epileptic seizures, and failure of peripheral organs such as heart, liver and kidney that could cause similar symptoms, to make an accurate diagnosis of SAE. Non-SAE patients are those who have sepsis but have not developed encephalopathy.

Peripheral venous blood was drawn three milliliters within 24 hours after admission. Following the manufacturer’s recommendations, an ELISA was carried out to verify the expression levels of EP-DEG in the control and sepsis blood samples. The levels of IL6, MMP8, CSF3, and S100A8 were determined by the following commercial ELISA kits, according to manufacturer’s recommendations: MMP8 (Cat.#CSB-E04680h, Cusabio Biotech Co. Ltd, Wuhan, China), S100A8 (Cat. #CSB-E11833h, Cusabio Biotech Co. Ltd, Wuhan, China), IL-6 (Cat. #CSB-E04638h, Cusabio Biotech Co. Ltd, Wuhan, China), and CSF3 (Cat. #EK0360, Boster Biotech Co. Ltd, USA). The flow chart for this study is shown in Figure 1 .

The pivot gene’s diagnostic value was estimated using ROC analysis, which was carried out in R using the pROC package (29). The hub gene is regarded to have strong specificity and sensitivity in differentiating between non-SAE and SAE when the AUC value is more than 0.7.

The Biobase package, limma package, and ClusterPro “ler package of R were used to evaluate mRNA expression from GEO database. SPSS 26.0 and GraphPad V9.1.1 software were utilized to do statistical analysis on clinical data. The mean differences between the two groups were compared using a two-tailed independent samples t-test. Receiver operating characteristic (ROC) curves were used to obtain the performance characteristics of diagnostic SAEs. The area values under the curve were used in the ROC curve study. Pearson correlation analysis was used to assess the relationship between the levels of different EP-DEGs and clinical indicators as well as prognostic factors, including mechanical ventilation duration, 28-day mortality, ICU length of stay, and hospitalization time. All p values were two-tailed, and statistical significance was set at p< 0.05.

A between-sample differential gene expression analysis was carried out to assess the variations in gene expression between the sepsis group and the critical ill control group. In the GSE135838 dataset, the median, upper and lower quartiles, maximum and minimum values of the 24 sample genes were virtually identical ( Figure 2A ). In the sepsis group, correlation analysis revealed a more significant intra-group association ( Figure 2B ). According to the sample information and data matrix analysis of GSE135838, 165 differentially expressed genes were extracted from the brain tissues. TM4SF19-TCTEX1D2, CPZ, ADAM6, ZC2HC1B, and ART1 were down-regulated genes, while SNORD115-27, RNU11, S100A8, SNORA45, LOC649330, S100A9, and IL6 were the top seven up-regulated genes with the least p-values ( Figure 2C ). The heat map of DEGs revealed that, in contrast to the control group, DEGs were consistently up or downregulated in the sepsis group ( Figure 2D ).

We used extracellular protein genes annotated in existing public libraries to filter EP-DEGs from DEGs. Extracellular protein genes that were identified in the HPA and Uniprot databases were intersected with DEGs, and 31 EP-DEGs were screened; 19 were found to be up-regulated and 12 to be down-regulated ( Figure 3A ). Among the top 14 up-regulated extracellular protein-encoding genes in the sepsis and control groups that had the highest logFC values were S100A8, IL6, MMP8, S100A9, CSF3, S100A12, FCGR3B, CD177, FCN1, ATP6V0A4, CXCL3, SLURP1, CXCL11, and LY96 ( Figure 3B ). To further analyze these genes, their mRNA and protein levels in healthy individuals were summarized. A three-dimensional histogram was used to show how the top 14 up-regulated genes’ mRNA expression varied in different parts of the normal brain ( Figure 3C ). MMP8, CXCL3, SLURP1, CXCL11, and FCN1 had relatively low amounts of mRNA in normal brain tissue according to HPA database. Still these levels were markedly increased in the brain tissues of sepsis patients. Except for CD177 and ATP6V0A4, which were not reported, all 12 extracellular proteins among the top 14 up-regulated genes were expressed at various levels in the plasma of healthy populations. FCN1 expression was the highest among all, reaching 1.7 mg/L. The scatter plot shows how these 12 plasma proteins are distributed in terms of concentration across the literature ( Figure 3D ). Among the extracellular molecules CSF3, IL6, MMP8, and S100A8, no significant correlations (p<0.05) as shown in Figure S1 .

To comprehend the function of the screened 31 EP-DEGs discovered in the brain tissue dataset, functional and pathway enrichment analyses were performed using DAVID. As seen in Figure 4A , these genes are primarily enriched in biological processes such as zinc ion binding, calcium-dependent protein binding, and cytokine activity. Inflammatory response, innate immune response, and neutrophil chemotaxis are primarily enriched in these genes. All the genes are enriched in extracellular space. All genes are enriched in the KEGG pathway of the IL-17 signaling pathway, cytokine-cytokine receptor interaction, and other signaling pathway. The top 5 processes with the minimum p-value of BPs, CCs, and MFs are displayed using the cnetplot function in the ClusterProfiler package. GDF15, LY96, CSF3, IL6, CD177, CXCL3, S100A8, S100A9, S100A12, and FCN1 are enriched in at least two BPs, CCs, and MFs ( Figure 4B ).

The significantly enriched hallmark terms revealed by GSEA included defense response to other organisms, regulation of response to stress, and others ( Figure S2 ).

We built a PPI network using the STRING database to investigate how the 31 EP-DEGs interact with one another. Utilizing the cytoscape application, the PPI network was seen ( Figure 5A ). The PPI network consists of 86 edges and 26 nodes using calculations of molecular pathways and processes linked to the 31 core DEGs. We developed working modules using the MCODE plug-in for Cytoscape. The findings showed a module of 16 genes ( Figure 5B ). Using ten topological techniques from the CytoHubba plug-in in Cytoscape, the top 10 hub genes were screened. The four genes MMP8, CSF3, IL-6, and S100A8 were included in each of the two methods ( Figure 5C ). Results of GeneMANIA also revealed that the functions of 31 EP-DEGs were primarily related to myeloid leukocyte, leukocyte migration, neutrophil migration, and response to molecules of bacterial origin ( Figure 5D ).

31 sepsis patients, including 10 SAE patients and 21 non-SAE patients, were enrolled to confirm the main extracellular proteins found. Serum samples from the patients were taken to measure the levels of MMP8, CSF3, IL-6, and S100A8. Additionally, 31 sepsis patients’ clinicopathological characteristics, such as the source of the sepsis, the medical history, their comorbidities, and other variables, were compared between patients with and without SAE ( Table 1 ). The SAE group exhibited notably higher scores for SOFA compared to the non-SAE group (p<0.05). Between the SAE and non-SAE group, there were no appreciable differences in the source of sepsis as well as the medical history. Septic shock (p=0.049), acute renal injury (p=0.0067), and acute lung injury (p=0.0129) were all more prevalent in SAE patients. The 28-day mortality rate was higher in the SAE group, with a p-value of 0.0501 that was considered marginally significant. The concentrations of the proteins S100A8, MMP8, CSF3, and IL-6, were measured in peripheral blood samples from 31 sepsis patients to validate the critical extracellular proteins identified in the study. The SAE patients had higher levels of IL-6, MMP8, and S100A8 when compared to the non-SAE (n=10) group (all p<0.05) ( Figure 6A ). The predictive usefulness of MMP8 and S100A8 in the diagnosis and prognosis of SAE was investigated using ROC curves. Both S100A8 (AUC = 0.962, 95% CI = 0.9001-1.000) and MMP8(AUC = 0.791, 95% CI = 0.625-0.956) demonstrated high diagnostic accuracy for the diagnosis of SAE, with S100A8 having the highest AUC ( Figure 6B ). S100A8 had a cut-off value of 82.29 pg/ml and corresponding sensitivity and specificity of 80% and 100%, respectively. S100A8, MMP8, CSF3 and IL-6 protein levels in the peripheral blood of sepsis patients were revealed to be substantially associated with GCS scores following a correlation investigation of the four extracellular molecules with clinical indications (all p<0.05) ( Figure 6C ). S100A8 levels were significantly correlated with 28-day mortality (r=0.634, p<0.01) and IL-6 levels were positively correlated with mechanical ventilation time (r=0.360, p=0.046), according to a correlation analysis of the four extracellular molecules and clinical prognostic markers ( Figure 6D ).

Following a successful modeling procedure, a total of 45 rats, comprising the SHAM group (n=22) and the CLP group (n=23), were utilized for behavioral experiments and for obtaining blood and CSF samples ( Figure 7A ). The CLP model rats (n=11) exhibited a significantly lower novel object recognition index compared to the SHAM group rats (n=8) seven to eight days after modeling (p <0.05 vs. Sham) ( Figure 7B ). In the Barnes maze experiment, CLP model rats (n=11) made significantly more errors before locating the target hole on the 9^th and 10^th day after surgery than SHAM group rats (n=8) (p <0.01, p <0.05 vs. Sham) ( Figure 7C ). Modeling induced brain injury and septic encephalopathy in the rats. The concentrations of S100A8 and MMP8 proteins were measured in the rat blood and CSF samples obtained six days after modeling to confirm the key extracellular proteins identified in the study. In comparison to the SHAM group (n=14), CLP rats (n=12) had significantly increased serum MMP8 levels (p <0.05), with CSF levels also increasing (p = 0.0506) ( Figure 7D ). Similarly, compared to the SHAM group (n=14), the serum S100A8 levels in CLP rats (n=12) were higher (p <0.01), with significant increases in CSF levels (p <0.0001) ( Figure 7E ).