Identification and experimental validation of mitochondria-related genes biomarkers associated with immune infiltration for sepsis

Abstract

Background: Sepsis remains a complex condition with incomplete understanding of its pathogenesis. Further research is needed to identify prognostic factors, risk stratification tools, and effective diagnostic and therapeutic targets.

Methods: Three GEO datasets (GSE54514, GSE65682, and GSE95233) were used to explore the potential role of mitochondria-related genes (MiRGs) in sepsis. WGCNA and two machine learning algorithms (RF and LASSO) were used to identify the feature of MiRGs. Consensus clustering was subsequently carried out to determine the molecular subtypes for sepsis. CIBERSORT algorithm was conducted to assess the immune cell infiltration of samples. A nomogram was also established to evaluate the diagnostic ability of feature biomarkers via "rms" package.

Results: Three different expressed MiRGs (DE-MiRGs) were identified as sepsis biomarkers. A significant difference in the immune microenvironment landscape was observed between healthy controls and sepsis patients. Among the DE-MiRGs, NDUFB3 was selected to be a potential therapeutic target and its significant elevated expression level was confirmed in sepsis using in vitro experiments and confocal microscopy, indicating its significant contribution to the mitochondrial quality imbalance in the LPS-simulated sepsis model.

Conclusion: By digging the role of these pivotal genes in immune cell infiltration, we gained a better understanding of the molecular immune mechanism in sepsis and identified potential intervention and treatment strategies.

Keywords: immune cell infiltration; machine learning algorithm; mito-chondrial quality imbalance; mitochondria; sepsis.

Figure 1

Diagram of the Study flow.

Figure 2

Identification of DEGs and GSEA functional enrichment analysis. (A) Volcano plot of DEGs in HC and SP groups. The threshold of screening DEGs is set at |fold change| ≥ 2 and p (p.adjust) < 0.05. Turquoise dots represent down-regulated genes and red dots represents up-regulated genes. (B) Analysis of top 25 up- and down-regulated genes in HC and SP group. (C, D) GSEA analysis of DEGs in HC and SP group.

Figure 3

WGCNA analysis to select characteristics gene module for SP. (A) Scale free topology model fit (R^2 = 0.85) and mean connectivity. (B) Clustering of module genes. (C) Cluster dendrogram for selecting gene modules. (D) Association between the gene modules. (E) Correlation analysis of transcriptome in different modules. (F) Heatmap analysis of 18 modules and clinical features (HC, SP). (G, H) Module membership vs. gene significance for SP and HC in black module.

Figure 4

DE-MiRGs screening and function enrichment analysis. (A) Identification of pivotal DE-MiRGs in black module. (B) GO enrichment analysis of DE-MiRGs. (C) KEGG pathway analysis of DE-MiRGs.

Figure 5

Feature biomarkers selection via machine language algorithms. (A, B) LASSO analysis to screen key DE-MiRGs. (C) RandomForest (RF) analysis of key DE-MiRGs, the filter condition for screening feature variates is set at: importance > 3. (D) Venn analysis of LASSO and RF. The overlapping genes are considered as feature biomarkers. (E) Correlation heatmap of BCKDHB, LETMD1, and NDUFB3. Green color represents negative correlation, red color represents positive correlation.

Figure 6

Validation of the expression of feature biomarkers and effectiveness evaluation. (A) The expression of BCKDHB, LETMD1, NDUFB3 in training cohort (GSE65682). (B) Validation of BCKDHB, LETMD1, NDUFB3 in validation cohort (GSE95233, GSE54514). (C, D) Nomogram construction and ROC curve of three gene signatures in GSE65682. (E, F) Nomogram construction and ROC curve of three gene signatures in GSE95233 and GSE54514.

Figure 7

Analysis of immune microenvironment landscape in HC and SP groups. (A) Immune cells assessment between HC and SP groups. (B) Analysis of correlation in 22 type immune cells. (C) Violin diagram of 22 type immune cells in HC and SP groups. (D) PCA plot showed a different distribution pattern in HC and SP. (E–G) Correlation analysis of three diagnostic biomarkers (BCKDHB, LETMD1, and NDUFB3) and immune microenvironment.

Figure 8

Subgroup analysis of SP samples based on three feature biomarkers. (A–C) Consensus clustering analysis. (D–F) The expression of BCKDHB and LETMD1, and NDUFB3 in both cluster subgroups. (G) Immune microenvironment analysis of subgroups. *p<0.05, **p<0.01, ***p<0.001.

Figure 9

NDUFB3 inhibition attenuated mitochondrial quality imbalance after sepsis. (A) qPCR was conducted to examine the expression of NDUFB3, LETMD1 and BCKDHB in 30 SP and 15 HC samples. ****P< 0.0001 as compared with the SP group. (B) TEM images showed mitochondria cristae damage in heart tissues (bar = 0.5 μm) (n=3). (C) ROS staining immunofluorescence reflected the oxidative stress in H9C2 cells (bar = 20 μm) (n=3). (D) The fluorescence intensity of ROS. (E) JC-1 aggregate/monomer reflected the mitochondrial membrane potential in H9C2 cells (bar = 20 μm) (n=3). (F) The concentration of ATP (n=3). (G) Representative images of mitochondrial morphology in H9C2 cells (bar = 10 μm) (n=3). (H) Ratio (long/short) of mitochondria (long (> 8 µm) and short ≤ 8 µm) was quantified by ImageJ. ^**** P< 0.01 as compared with the normal group, ^# P< 0.05 as compared with the LPS group, ^### P< 0.001 as compared with the LPS group, ^#### P< 0.0001 as compared with the LPS group. all data are presented as the mean ± SD.