Identification of iron metabolism-related genes as diagnostic signatures in sepsis by blood transcriptomic analysis

Abstract

Iron metabolism is considered to play the principal role in sepsis, but the key iron metabolism-related genetic signatures are unclear. In this study, we analyzed and identified the genetic signatures related to the iron-metabolism in sepsis by using a bioinformatics analysis of four transcriptomic datasets from the GEO database. A total of 21 differentially expressed iron metabolism-related signatures were identified including 9 transporters, 8 enzymes, and 4 regulatory factors. Among them, lipocalin 2 was found to have the highest diagnostic value as its expression showed significant differences in all the comparisons including sepsis vs healthy controls, sepsis vs non-sepsis diseases, and mild forms vs severe forms of sepsis. Besides, the cytochrome P450 gene CYP1B1 also showed diagnostic values for sepsis from the non-sepsis diseases. The CYP4V2, LTF, and GCLM showed diagnostic values for distinguishing the severe forms from mild forms of sepsis. Our analysis identified 21 sepsis-associated iron metabolism-related genetic signatures, which may represent diagnostic and therapeutic biomarkers of sepsis, and will improve our understanding of the molecular mechanism underlying the occurrence of sepsis.

Keywords: diagnostic signatures; iron metabolism; lipocalin 2; sepsis; transcriptomic analysis.

Figure 1

Identification of differentially expressed iron metabolism-related genes in sepsis by using the four transcriptomic datasets. A total of 5,294 DEGs among severe sepsis (Severe_S), mild sepsis (Mild_S), and healthy controls (H) were identified in GSE69063 dataset. A total of 4,040 DEGs among sepsis (S), sepsis shock (Shock_S), and healthy controls (H) were identified in GSE154198 dataset. A total of 2,166 DEGs between sepsis (S) and noninfectious controls (noninfect), or between sepsis (S) and healthy controls (H) were identified in GSE134347 dataset. A total of 7,567 DEGs between sepsis (S) and healthy controls (H) were identified in GSE185263 dataset. By extracting the iron metabolism-related genes, 21 genes were identified and differentially expressed simultaneously in all the four databases.

Figure 2

Heatmap of the expression levels of the sepsis related iron metabolism-related genes in the four transcriptomic datasets.

Figure 3

Enriched GO terms associated with the sepsis related iron metabolism-related genes. (a) GO results of biological process (BP), cellular component (CC), and molecular function (MF). (b) gene-concept network. Solid red circle represented up-regulated genes, while solid blue circle represented down-regulated genes.

Figure 4

PPI network of the sepsis-related iron metabolism-related genes. (a) The 21 nodes and 27 edges in a PPI network were built among the 21 common DEGs including 9 transporters, 8 enzymes, and 4 regulatory factors. (b) The ten hub genes were further identified by using the cytoHubba. (c) Gene–gene interaction networks of the ten hub genes in GeneMANIA. Another 20 nodes representing genes and 265 links associated with the hub genes in co-expression and physical interactions were identified. (d) Enriched pathways of the ten hub genes and their interacted genes.

Figure 5

Expression levels of the sepsis related iron metabolism-related genes in patients with sepsis and healthy controls. *p < 0.05.

Figure 6

ROC curves of top eight iron metabolism-related genes including SLC22A4, TSPO, NFE2, FLVCR2, LCN2, BCL2, FTO, and CYP4V2 in the diagnosis of sepsis compared with healthy controls. The AUC was >0.9 in GSE69063 (a), GSE154918 (b), GSE134347 (c) datasets, and AUC > 0.75 in GSE69063 (d) dataset (Table S2).

Figure 7

Expression levels of the sepsis-related iron metabolism-related genes in sepsis vs noninfectious disease (a), and in sepsis vs nonsepsis (uncomplicated) infectious disease (c). ROC curves of LCN2 and CYP1B1 which simultaneously differentially expressed in the two datasets of GSE134347 (b) and GSE154198 (d). The AUC of LCN2 and CYP1B1 were 0.744 (0.671–0.816) and 0.666 (0.586–0.746) in GSE134347 dataset, and 0.671 (0.489–0.853) and 0.662 (0.506–0.819) in GSE154198 dataset (Table S2). *p < 0.05.

Figure 8

Expression levels of the sepsis-related iron metabolism-related genes in severe sepsis vs mild sepsis (a), and in sepsis shock vs sepsis (c). ROC curves of GCLM, LCN2, LTF, and CYP4V2 which simultaneously differentially expressed in GSE69063 dataset (b) and GSE154198 dataset (d). The AUC of GCLM, LCN2, LTF, and CYP4V2 were 0.788 (0.667–0.909), 0.88 (0.796–0.965), 0.815 (0.698–0.932), and 0.717 (0.572–0.863) in GSE69063 dataset, and 0.65 (0.471–0.829), 0.745 (0.588–0.901), 0.697 (0.526–0.869), and 0.721 (0.556–0.886) in GSE154198 dataset (Table S3). *p < 0.05.

Figure 9

Expression levels of the sepsis-related iron metabolism-related genes between the different outcomes (a). ROC curves of MBOAT2, GCLM, ATP6V1C1, CYP1B1, LCN2, CLIC2, LTF, RHAG, and GYPA, which were up-regulated in the deceased patients than the survived cases in the GSE263298 dataset (b). The AUC values were MBOAT2: 0.656 (0.569–0.744), GCLM: 0.65 (0.555–0.745), ATP6V1C1: 0.664 (0.575–0.752), CYP1B1: 0.635 (0.543–0.727), LCN2: 0.631 (0.55–0.712), CLIC2: 0.637 (0.548–0.726), LTF: 0.632 (0.548–0.717), RHAG: 0.68 (0.602–0.758), GYPA: 0.626 (0.543–0.71) (Table S3). *p < 0.05.