Patients with SIRS and sepsis were recruited from Department of Emergency and Critical Care at the Shanghai General Hospital. The inclusion criteria for patients with SIRS were the meeting of at least two of the four following standards: (1) body temperature $>$38 °C or 36 °C, (2) heart rate $>$90 breaths/min, (3) breathing $>$20 breaths/min or PCO${}_2$ 32.33 mmHg, and (4) white blood cell count $>$12 $\times$ 10${}^9$/L or 4 $\times$ 10${}^9$/L ($>$12,000/µL or 4000/µL, or immature granulocyte count $>$10%). The inclusion criteria for patients with sepsis included diagnosis according to SEPSIS 3.0. Blood samples were collected from the recruited patients. All treatments were applied in accordance with the recent guidelines of the Surviving Sepsis Campaign. Patients with autoimmune diseases, malignancies, immune deficiencies, metabolic and chronic diseases, endocrine diseases, and pregnancy were excluded. Healthy controls were recruited from a physical examination center at the Shanghai General Hospital. The healthy controls had normal clinical test results without a disease history. Clinical characteristics and demographics of the patients are shown in the result.

Clinical blood samples were collected into heparin-coated tubes and centrifuged at 2200 rpm for 10 min at room temperature. The supernatant was collected, then immediately stored in separate tubes at –80 °C until GC-MS analysis. Next, 20 µL of 2-chloro-l-phenylalanine (catalog number C105993, Aladdin, Shanghai, China) was dissolved in methanol as an internal standard, and 80 µL of serum sample was added. Next, 240 µL of methanol (catalog number 1.06035, Supelco, Bellefonte, PA, USA): acetonitrile (catalog number 1.00029, Supelco, Bellefonte, PA, USA) (2:1 v:v) was added to each sample. The mixtures were extracted by ultrasonication for 8 min in an ice-water bath. Next, the samples were centrifuged at 12,000 rpm for 10 min at 4 °C. An aliquot of the supernatant (150 µL) was transferred to vacuum at 24 °C. Then, 80 µL of methoxylamine (catalog number 89803, Supelco, Bellefonte, PA, USA) was added, followed by vortexing and incubation for 90 min at 37 °C. After that, 50 µL of BSTFA (with 1% TMCS) (catalog number B-023, Supelco, Bellefonte, PA, USA) and 20 µL n-hexane (catalog number 1.00795, Supelco, Bellefonte, PA, USA) were added and mixed, then vortexed and derivatized for 60 min at 70 °C. The samples were placed at 24 °C for 30 min prior to GC-MS analysis.

The samples were acted on by an Agilent 7890B-5977B system (Agilent, Palo Alto, CA, USA). A 30 m $\times$ 0.25 mm $\times$ 0.25 µm HP-5MS fused-silica capillary column (Agilent J & W Scientific, Folsom, CA, USA) was used to separate the derivatives. Helium ($>$99.999%) was used as the carrier gas at a rate of 1 mL/min, and the injector temperature was maintained at 260 °C. The temperature of the mass spectrometry (MS) quadrupole was set to 150 °C, the ion source was 230 °C, and the collision energy was 70 eV. Mass spectrometric data were obtained in full-scan mode (m/z 50–500). The samples were individually run, and quality control was performed. Partial least squares discriminant analysis (PLS-DA) and orthogonal partial least squares discriminant analysis (OPLS-DA) were used to visualize metabolic differences between the two groups using the SIMCA version 14.1 software (Umetrics, Umeå, Sweden). Metabolites were screened using the National Institute of Standards and Technology database (NIST). Enrichment of metabolic pathway was based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) database.

The parametric Student’s t-test or chi-square test were used to compare the clinical data between the two groups. Statistical significance of the metabolites was determined using one-way Analysis of Variance (ANOVA). Receiver operating characteristic (ROC) curve analysis was used to evaluate the diagnostic values of the significant differential metabolites. A two-tailed Student’s t-test was used to verify whether the metabolite differences between groups were significant. Differential metabolites were selected with variable importance of projection (VIP) values $>$1.0 and p-values 0.05. Statistical tests were performed using GraphPad Prism (v8.2.1, GraphPad Software, San Diego, CA, USA). Data were assessed for normality and analyzed using parametric or nonparametric tests as appropriate. Statistical significance was set at p 0.05.

Eight patients with SIRS and eight with sepsis were included in this study. Clinical symptoms and Computed Tomography (CT) scan showed that two of SIRS patients have pneumonia, three of them have abdominal infection, other three patients have urethra infection. Two of sepsis patients have pulmonary infection, two patients have abdominal infection, four patients have urethra infection.

Eight healthy individuals were also included as controls. The general characteristics and laboratory test results of patients with sepsis are shown in Table 1. There were no significant differences in general characteristics between patients with SIRS and the healthy controls, nor between patients with sepsis and the healthy controls. There were significant differences in c-reactive protein (CRP), interleukin (IL-6), IL-1$\beta$ between patients with SIRS and sepsis.

To determine the metabonomic changes, the peripheral blood of patients was collected, with eight samples in each group and eight healthy individuals as controls. Serum was extracted and analyzed by GC-MS. The PLS-DA method is a supervised discriminant statistical method using partial least squares regression for predictive and descriptive modeling, as well as for discriminative variable selection, to explain and predict the differences between the two groups. In our PLS-DA model, owing to the different metabolites, patients with SIRS were significantly distinguished from healthy controls, and patients with sepsis were significantly separated from healthy controls (Fig. 1A). To further improve the analytical ability and effectiveness of the model and maximize the differences between different groups in the model, we modified the PLS-DA to filter out noise irrelevant to the classification information. As a result, OPLS-DA showed that SIRS split with healthy controls. Similarly, there was a split between sepsis and healthy controls between SIRS and sepsis (Fig. 1B).

Fig. 1.

The PLS-DA and OPLS-DA models distinguished patients between groups. (A) The green triangles represent total patients with SIRS, the blue squares represent patients with sepsis, and the red triangles represent heathy controls in the PLS-DA score plots. (B) The green circles represent total patients with SIRS, the blue squares represent patients with sepsis, and the red circles represent heathy controls in the OPLS-DA score plots (PLS-DA, partial least squares discriminant analysis; OPLS-DA, orthogonal projection latent to structure discriminant analysis).

In the OPLS-DA model, the VIP value can be used to evaluate the effects of the influence intensity and interpretation ability of the expression pattern of each metabolite on the classification and discrimination of each group, mine the differential metabolites with biological significance, and then use a t-test to verify whether the differential metabolite is significant between groups. The screening criteria were that the VIP value of the first principal component of the OPLS-DA model was $>$1 and the p-value of the t-test was 0.05. According to these criteria, there were 51 differential metabolites between SIRS patients and healthy controls, and 47 differential metabolites were upregulated. Additionally, there were 63 differential metabolites between patients with sepsis and healthy controls, and 47 differential metabolites were upregulated. Compared to SIRS, there were only four differential metabolites in sepsis patients, all of which were downregulated (Table 2, Fig. 2).

Fig. 2.

Differential metabolites were detected by GC-MS. (A) The histogram shows the quantity of differential metabolites. (B) The volcanic map visualizes changes in differential metabolites.

The load map can be used to identify the strength of influence of the metabolites in the comparison group. The abscissa of the S-plot is the characteristic value of the effect of the metabolite on the comparison group, and the ordinate is the correlation between the sample score and the metabolite. As the eigenvalues and correlations are both positive and negative, all points in the visualized graph are distributed in the first and third quadrants (similar to the “s” shape) and are called S-plots. The closer the metabolite is to the upper-right corner and lower-left corner, the more significant the difference. In the S-plot of the diagnostic model, we identified methyl 3-o-acetyl-d-galactopyranoside and N-acetylputrescine as candidate biomarkers for distinguishing between SIRS and healthy controls, as well as between sepsis and healthy controls (Fig. 3A). Correlation analysis using the Pearson correlation coefficient was used to evaluate the correlation between the different metabolites. The results showed a significant positive correlation between methyl 3-o-acetyl-d-galactopyranoside and N-acetylputrescine levels (Fig. 3B).

Fig. 3.

The S-plot was used to screen the potential biomarkers distinguishing between SIRS, sepsis, and healthy controls and the correlation between differential metabolites. (A) The S-plot was used to select the potential biomarkers distinguishing between SIRS, sepsis, and healthy controls. (B) Correlation analysis was used to measure the degree of correlation between differential metabolites. The red dot indicates positive correlation, and the blue dot indicates negative correlation.

To determine a simplified diagnostic biomarker to use in differentiating between SIRS and sepsis, ROC analysis was carried out, and the area under the curve (AUC) of each candidate metabolite was calculated. 2-deoxy-d-erythro-pentofuranose-5-phosphate was the best predictor of SIRS and sepsis (AUC = 0.9297, 95% confidence interval: 0.8069–1.0000), indicating its strong diagnostic performance. The AUC of terbutaline was 0.9063, and that of allantoic acid was 0.7969, indicating that terbutaline and allantoic acid were good discriminators between SIRS and sepsis patients (Fig. 4A). Correlation analysis showed a significant positive correlation between 2-deoxy-d-erythro-pentofuranose-5-phosphate and terbutaline (r = 0.66108) and between 2-deoxy-d-erythro-pentofuranose-5-phosphate and allantoic acid (r = 0.55947) (Fig. 4B).

Fig. 4.

The ROC was used to analyze the potential biomarkers between SIRS and sepsis patients and the correlation between differential metabolites. (A) The ROC was used to analyze the potential biomarkers for differentiating patients with SIRS and sepsis (ROC, receiver operating characteristic). (B) Correlation analysis was used to measure the degree of correlation between differential metabolites.

To determine the metabolic pathways enriched in SIRS and sepsis, differential metabolites were subjected to pathway analysis using the Kyoto Encyclopedia of Genes and Genomes database (KEGG). These results are shown in Fig. 5. Eighteen pathways were enriched in SIRS patients compared to healthy controls. Nine metabolic pathways were related to amino acid biosynthesis and metabolism, including protein digestion and absorption; aminoacyl-tRNA biosynthesis; beta-alanine metabolism; phenylalanine, tyrosine, and tryptophan biosynthesis; glycine, serine, and threonine metabolism; cysteine and methionine metabolism; valine, leucine, and isoleucine biosynthesis; tryptophan metabolism; and glutathione metabolism. There were seven atypical pathways which are related to the use of glucocorticoids and catecholamines, including cocaine addiction; amphetamine addiction; alcoholism; prolactin signaling pathway; dopaminergic synapses; Parkinson’s disease and melanogenesis. The other two important pathways were the central carbon metabolism in cancer and the biosynthesis of pantothenate and coenzyme A (CoA).

Fig. 5.

The bubble diagram was analyzed using KEGG metabolism pathway enrichment. The p value is the significance of enrichment of this metabolic pathway. The ordinate is the name of the metabolic pathway; the abscissa is the rich factor (rich factor = the number of significantly different metabolites/the total number of metabolites in this pathway). The larger the rich factor, the more metabolites enriched in the pathway. The color from green to red indicates that p-value decreases in turn.

There were 21 enriched pathways in sepsis patients compared with healthy controls. Eight metabolic pathways were related to amino acid biosynthesis and metabolism, including protein digestion and absorption; aminoacyl-tRNA biosynthesis; beta-alanine metabolism; phenylalanine, tyrosine, and tryptophan biosynthesis; glycine, serine, and threonine metabolism; cysteine and methionine metabolism; glutathione metabolism; and thiamine metabolism. Two pathways related to glucose metabolism (glycolysis/gluconeogenesis and the pentose phosphate pathway), one pathway related to pyrimidine metabolism, and one was related to the biosynthesis of unsaturated fatty acids. Three other important pathways were identified: central carbon metabolism in cancer; pantothenate and CoA biosynthesis, and sulfur metabolism. There were six atypical pathways, including cocaine addiction; amphetamine addiction; alcoholism; prolactin signaling pathway; dopaminergic synapses; and Parkinson’s disease.

Only two pathways were enriched in sepsis compared to SIRS. One was the dopaminergic synapse, and the other was tyrosine metabolism.

In order to further verify the clinical value of 2-deoxy-d-erythro-pentofuranose-5-phosphate, we analyzed the correlation between 2-deoxy-d-erythro-pentofuranose-5-phosphate and some clinical indicators. Analysis showed no significant correlation between 2-deoxy-d-erythro-pentofuranose-5-phosphate and clinical indicators, such as IL-6, IL-1$\beta$, tumor necrosis factor-$\alpha$ (TNF-$\alpha$), CRP, platelet, Ca${}^{2+}$, the absolute value of monocytes, and the absolute value of lymphocytes. There was a significant negative correlation between 2-deoxy-d-erythro-pentofuranose-5-phosphate and Lactate (r = –0.8756, p = 0.0044) (Table 3).