Diagnostic value of cuproptosis-related genes in sepsis: A novel model based on bioinformatics screening

doi:10.21203/rs.3.rs-3307190/v1

Download PDF

Article

Diagnostic value of cuproptosis-related genes in sepsis: A novel model based on bioinformatics screening

https://doi.org/10.21203/rs.3.rs-3307190/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background:

Sepsis is a life-threatening functional disorder of the organs resulting from a dysregulated host immune response to infection and is a leading cause of death and critical illness worldwide. Genetic diagnosis combined with big data analysis of existing biomarkers has great potential in the diagnosis and prognosis prediction of sepsis, and there is an urgent need to construct prognostic models that will improve the effectiveness of treatment decisions.

Methods:

We used data from the Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo/) database to perform a comprehensive analysis of differential gene expression profiles associated with cuproptosis in sepsis. Combining the sepsis datasets (GSE131761 and GSE54514) as test sets, a total of 208 sepsis samples and 69 normal samples were used for the analysis of cuproptosis-related differentially expressed genes (CRDEGs), weighted gene co-expression network analysis (WGCNA), Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG). The gene sets from the Molecular Signatures Database (MSigDB) and were used to perform GSVA (Gene Set Variation Analysis) and GSEA (Gene-set Enrichment Analysis). The prognostic performance of the hub genes in the CRDEGs diagnostic model was examined in the validation set (GSE25504 and GSE26378), and receiver operating characteristic curves (ROC) were plotted. We constructed a Cox regression model and drew a nomogram based on the final screened CRDEGs. The prognostic Calibration and decision curve analysis were used to evaluate the model. Finally, we constructed a protein-protein interaction network (PPI network) and performed ceRNA network analysis and immune cell infiltration abundance correlation analysis.

Results:

We obtained two sepsis disease subtype groups based on clustering analysis of differentially expressed cuproptosis hub genes (LIPT1, PDHB, MTF1, GLS, SLC31A1). GO and KEGG analyses indicated that sepsis-related cuproptosis alterations were primarily enriched in cellular copper ion homeostasis, regulation of respiratory gaseous exchange by neurological system process, suckling behavior, protein-cofactor linkage. WGCNA yielded six cuproptosis-related gene co-expression modules and 202 CRDEGs between subgroups of sepsis. A total of 32 CRDEGs were extracted additionally based on LASSO analysis calculations, of which 23 CRDEGs were included in the optimized diagnostic gene labels used for constructing Cox regression models and plotting nomogram. Finally, in the immune infiltration analysis, there was a statistically significant relationship between the abundance of immune infiltration of 16 immune cells and the expression of CRDEGs.

Conclusions:

The diagnostic model we constructed on CRDEGs has promising predictive power, paving the way for further exploration of the mechanisms related to cuproptosis in sepsis and providing new ideas for discovering potential biomarkers and diagnostic patterns for sepsis.

Biological sciences/Computational biology and bioinformatics

Biological sciences/Computational biology and bioinformatics/Predictive medicine

sepsis

cuproptosis

bioinformatics

diagnostic model

Sepsis, a dysregulated host response to microbial infection, typically triggers cytokine storm and leads to multiple life-threatening organ dysfunction¹. According to the World Health Organization (WHO), sepsis affects 47 to 50 million people annually, causes 11 million deaths, and 30 million people develop sepsis complications yearly, with higher and underestimated prevalence in low-income and middle-income countries^2–4. Septic shock is a subtype of sepsis characterized by underlying circulatory, cellular and metabolic abnormalities that carry a higher mortality risk than sepsis alone⁵. A large-scale systematic review of English language articles published from 2009 to 2019 showed that mortality among patients with septic shock remained high and did not decline significantly after 2011⁶. Sepsis poses a tremendous challenge to the global public health care system attributable to its mortality rate, complicated clinical treatment and poor prognosis. Sustained excessive inflammation and immune suppression are the pivotal mechanisms underlying sepsis-induced organ destruction and death⁷.

In sepsis patients, the plasma copper level usually increases to promote the liver synthesizing plasma ceruloplasmin and other anti-inflammatory cytokines and activating the body's immune response⁸. One study revealed that infection affects micronutrient homeostasis in newborns with lower serum zinc and higher serum copper and ceruloplasmin⁹. The other study also discovered that sepsis might increase plasma Cu concentration and relative Zn deficiency¹⁰. A recent single-center prospective observational study also showed that whole blood copper levels are significantly higher in patients with sepsis and are positively associated with the risk of death¹¹. However, specific studies about copper metabolism in sepsis are still relatively scarce. The uptake, distribution, sequestration, and excretion at both the cellular and systemic levels of metal micronutrients copper (Cu) are vital for life and active in a diverse array of biological processes, including mitochondrial respiration, antioxidant defense, and bio-compound synthesis¹². The majority is bound copper, with a small amount being free copper, which plays a vital role in maintaining normal hematopoiesis, promoting the formation of connective tissue, maintaining the function of the central system, and in anti-oxidation^13,14. However, latest research has shown that excess copper could be cytotoxic. Tsvetkov et al found a new style of cell death that occurs via the direct binding of copper to lipoylated components of the tricarboxylic acid (TCA) cycle, primarily dependent upon intracellular copper accumulation, finally involving the disruption of specific mitochondrial metabolic enzymes¹⁵. Sepsis-related mitochondrial damage might lead to increased generation of reactive oxygen species and decreased energy substrate availability to support homeostatic and cytoprotective responses.

Sepsis could present in subtle ways, and some symptoms that appear to be sepsis turn out to be non-infectious diseases. The diagnostic accuracy of available instruments varies widely, with the majority having poor predictive values. In this study, we used the sepsis dataset in GEO and combined with known cuproptosis-related genes to screen CRDEGs by bioinformatics, then construct ceRNA networks based on WGCNA, GO and KEGG analysis, and analyse the correlation between immune infiltrating cells and function to establish a Cox risk model in order to provide theoretical support for the identification of cuproptosis in sepsis and its mechanistic regulation study. Copper metabolic homeostasis and its mediation of programmed cell death confers potential targets for intervention in sepsis. Understanding the complex mechanisms of cellular damage in sepsis could contribute to uncovering new diagnostic modalities, optimizing preventive and therapeutic interventions, and better-protecting patients' cellular function and stemming organ failure.

The specific technical route of this study is shown in Fig. 1

Dataset Acquisition and Processing

We downloaded septic-related datasets GSE131761¹⁶, GSE54514¹⁷, GSE26378¹⁸ and GSE25504¹⁷ from the GEO¹⁹(Gene Expression Omnibus) database through the GEOquery²⁰ package. GSE131761 and GSE54514 were the test sets, and GSE26378 and GSE25504 were the verification sets. All datasets are from Homo sapiens. In this research, dataset GSE131761 contained 129 samples, including 81 sepsis samples and 33 normal samples; dataset GSE54514 contains 163 samples, of which 127 are from sepsis patients and 36 are from normal patients; dataset GSE26378 contains 103 samples, including 82 sepsis samples and 21 normal samples; and dataset GSE25504 contains 63 samples, all from sepsis patients. To evaluate the role of cuproptosis in sepsis, we obtained a set of 12 cuproptosis-related genes (CRGs)¹⁵ from previously published literature (FDX1, LIAS, LIPT1, DLD, DLAT, PDHA1, PDHB, MTF1, GLS, SSKN2A, SLC31A1 and ATP7B). In addition, we found 1 CRG (SLC25A3) by searching with the keyword "cuproptosis" in GeneCard²¹. The whole set of 13 cuproptosis-related genes is shown in Table S1.

Sepsis-related Differentially Expressed Genes

To identify the potential mechanism and related biological characteristics and pathways of differential genes in sepsis, we merged the SS datasets GSE131761 and GSE54514. We used R's (https://www.r-project.org/) “sva” package to remove the batch effect and the “limma” package”²² to standardize the merged dataset to get a sepsis dataset (merged dataset). Principal component analysis (PCA) ²³ was used to compare the dataset's expression matrices before and after removing the batch effect. PCA is a method of reducing data dimensions. It extracts high-dimensional data's feature vector (component), converts it to low-dimensional data, and displays its characteristics.

The single-sample gene-set enrichment analysis (ssGSEA) ²⁴ algorithm could quantify gene abundance. We calculated cuproptosis scores (CPs) using R's “GSVA” package. Then we calculated the group expression difference of the cuproptosis scores between the high and low CPs groups of the disease group in the merged dataset through the Mann-Whitney U test (Wilcoxon rank sum test), and P.value < 0.05 was considered to be statistically different.

We used the limma package to perform differential analysis on the disease groups in the merged dataset to obtain differentially expressed genes (DEGs). We set |logFC| > 0.5 and P.adj 0.05 as the standard to screen out genes as DEGs for further study. Up-regulated differentially expressed genes have logFC > 0.5 and P.adj 0.05, and down-regulated genes have logFC − 0.5 and P.adj 0.05. We analyzed DEGs with |logFC| > 0.5 and P.adj 0.05 to find cuproptosis-related differentially expressed genes (CRDEGs). The volcano plot was created for visualization using the R package ggplot2.

Differential Gene Function Enrichment (GO) and Pathway Enrichment (KEGG) Analysis

Gene Ontology (GO)²⁵ analysis is a common method for large-scale functional enrichment research, including biological process (BP), molecular function (MF) and cellular components (CC). The Kyoto Encyclopedia of Genes and Genomes (KEGG) ²⁶ is a widely used database that stores information about genomes, biological pathways, diseases and drugs. We used the R package clusterProfiler²⁷ to perform GO annotation analysis on CRGs. The item selection criteria were P.value < 0.05 and FDR value (q.value) < 0.05, which was considered statistically significant, and the P value correction method was Benjamini-Hochberg (BH).

Gene Set Enrichment Analysis

Gene Set Enrichment Analysis (GSEA) ²⁸ is a computational method to analyze whether a specific gene set is statistically different between two biological states, and is usually used to estimate the activity of pathways and the variety of biological processes in samples of expression datasets. In this research, we used the “clusterProfiler package” to perform enrichment analysis on GSE131761 and GSE54514. The parameters were as follows: the seed was 2,020, the number of calculations was 1,000, the number of genes in each gene set was 10 ~ 500, and the P value correction method was the BH method. We obtained the c2.cp.v7.2.symbols gene set from the Molecular Signatures Database (MSigDB)²⁹, and the screening criteria for significant enrichment were P < 0.05 and FDR value (q.value) < 0.05.

Gene Set Variation Analysis

To assess whether different pathways are enriched among different samples. We obtained the gene set “h.all.v7.4.symbols.gmt” from the MSigDB and performed Gene Set Variation Analysis (GSVA) ³⁰ on data set GSE54388 and GSE18520, then calculated the differences in functional enrichment between the healthy sample group (Normal group, NG) and sepsis patient samples (Sepsis group, SG).

Sepsis Disease Subtype Identification

Consensus Clustering³¹ is a resampling-based algorithm for identifying each sample and its subgroup number and verifying the rationality of the clustering. Consensus clustering is multiple iterations over subsamples of a dataset, providing indicators of cluster stability and parameter decisions by exploiting subsampling to induce sampling variability. We used R's “ConsensusClusterPlus package” to identify sepsis subtypes in the merged dataset SG based on screened CRDEGs. Set the number of clusters between 2 and 8, repeat 50 times to extract 80% of total samples, clusterAlg = "km", distance="Euclidean". The group expression differences of CRDEGs among samples of different disease subtypes in the merged dataset were calculated by the Mann-Whitney U test (Wilcoxon rank sum test), and P.value < 0.05 was considered statistically different.

Weighted Correlation Network Analysis

We used the R’s “WGCNA” package to perform Weighted gene correlation network analysis (WGCNA)³² analysis on the merged dataset. First, we calculated the correlation coefficient between any two genes in the DEGs screened out from the merged dataset |logFC| > 0.5 and P.Value < 0.05, and use the weighted value of the correlation coefficient to make the relationship between genes in the network obeys the scale-free network. Then, a hierarchical clustering tree was constructed according to the correlation coefficients among genes. Different branches of the clustering tree represent different gene modules, and different colors represent different modules, and then the module significance is calculated. In our study, WGCNA takes |logFC| > 0.5 and P.Value < 0.05 as the input of DEGs screened out, the minimum number of module genes is set to 25, softpower is set to the optimal soft threshold 6, and the module merge cut height is set to 0.2, the minimum distance is set to 0.2, which is used to measure the correlation between NG and SG and different modules and record the genes in each module at the same time. The genes in each module were regarded as module-characteristic genes. After we selected the modules of interest based on the correlation values, all genes within the modules were identified as DEGs highly correlated with sepsis.

Construction of the Diagnosis Model of Cuproptosis

To get the diagnostic model of CRDEGs in the merged dataset, we first used the “glmnet package” ³³ to perform Least absolute shrinkage and selection operator ( LASSO ) ³⁴ regression based on CRDEGs with the parameter family = “ binomial “ and ten-fold cross-validation method.

$$riskScore=\sum _{i}Coefficient\left({ℎub gene}_{i}\right)\ast mRNA Expression\left({ℎub gene}_{i}\right)$$

We used the “randomForest package”³⁵ to construct the model with the expression matrix of CRDEGs in the disease group of the merged dataset. The parameters are set.seed(234), ntree = 1000.

$$I\left(X=xi\right)=-{\text{log}}_{2}p\left({x}_{i}\right)$$

Then we took the intersection of the genes obtained by the two machine algorithms to obtain the intersection CRDEGs, extract the penalty coefficient (lambda) of each intersection CRDEGs in the LASSO regression diagnostic model, calculate the riskScore of the CRDEGs diagnostic model based on the intersection CRDEGs, and used the median riskScore value of all disease group samples as a critical value to divide the disease groups in the merged dataset into high- and low-risk groups.

Clinical Evaluation of CRDEGs

We performed univariate Cox regression analysis on the expression of CRDEGs in the sepsis dataset, and selected factors with p༜0.1 into the multivariate Cox regression analysis to construct a model to study the clinical diagnostic value of CRDEGs for sepsis. Then we used the R’s “rms package” to construct a nomogram³⁶ based on the results of multivariate Cox regression analysis. Finally, we used Decision curve analysis(DCA)³⁷ plots to evaluate the accuracy and resolution of the Cox regression model. We used the R package “ggDCA” to draw DCA diagrams to evaluate the effect of the Cox regression model.

CeRNA Network Construction

We used the ENCORI³⁸ database (https://starbase.sysu.edu.cn/) and the miRDB³⁹ database (http://mirdb.org/) to predict miRNAs interacting with hub genes (mRNA), and then took the intersection of the two databases’ results to draw the mRNA-miRNA interaction network. We used the ENCORI database to predict the lncRNAs that the above miRNAs might bind to construct miRNA-lncRNA interaction pairs. By integrating miRNA-mRNA and miRNA-lncRNA interaction pairs, we constructed the mRNA-miRNA-lncRNA ceRNA regulatory network of sepsis key genes.

Protein-protein Interaction Network Analysis

STRING database⁴⁰ (http://string-db.org/) was used to build a CRDEGs-related protein-protein interaction network (minimum required interaction score: low confidence (0.150)) from the CRDEGs in the CRDEGs diagnostic model and utilized Cytoscape⁴¹ (version 3.9.1) visualization.

Immunoinfiltration Analysis

We uploaded the gene expression matrix data to the CIBERSORTx ⁴²(https://cibersortx.stanford.edu/) online website, combined with the Homo sapiens characteristic gene matrix, screened out the data with immune cell enrichment scores༞0, and finally obtained and displayed immune cell infiltration specific results for the abundance matrix. The correlation difference of immune cells between high and low risk group samples of the sepsis dataset was calculated by the Wilcoxon test, and P.value < 0.05 was considered to be statistically different. The correlation between different immune cells was calculated by Spearman's test and visualized by the R package “ggplot2”. Then we combined the gene expression matrix of the sepsis dataset to calculate the correlation between immune cells and CRDEGs and drew a correlation heat map through the R package “pheatmap”.

Statistical Analysis

All data processing and analysis in this article are based on R software (Version 4.1.2). For the comparison of two groups of continuous variables, the statistical significance of normally distributed variables is estimated by the independent Student t-test, and the Mann-Whitney U test (ie Wilcoxon rank sum test) was used to analyze the differences among non-normally distributed variables. The Chi-square test or Fisher's exact test was used to compare and analyze statistical significance between two groups of categorical variables. If not specified, the results were calculated by Spearman correlation analysis between the correlation coefficients between different molecules, all statistical P values are two-sided, and P < 0.05 is considered statistically significant.

Technology Roadmap:

Analysis of genes related to cuproptosis

We first used the R package sva to remove the batch effect on the two sepsis datasets GSE131761 and GSE54514, and then used the limma package to standardize the merged datasets to obtain a sepsis dataset (merged dataset) (Fig. 2A-B), the merged dataset includes 69 sepsis patient samples (SG) and 208 normal control samples (NG). Then we performed Principal Component Analysis (PCA) on the expression matrix of the merged dataset before and after removing the batch effect according to the source of the sample to verify the effect of removing the batch effect (Fig. 2C-D), the results show that after removing batch processing, the batch effect of samples from different sources in the merged dataset is basically eliminated.

To obtain the difference of the cuproptosis-related genes in the sepsis data set (merged dataset) between the SG and NG, we analyzed the CRDEGs in the disease group and the normal group (Fig. 3A), the results showed that the expression levels of MTF1 and SLC31A1 in different groups of the merged dataset had statistically significant differences (P < 0.001), and LIPT1 and PDHB were significantly different in merged dataset. It has a highly statistically significant difference (P < 0.01) in the expression levels of different groups, and a certain statistically significant difference (P < 0.05) in the expression level of GLS in different groups of the merged dataset. To further obtain the correlation between cuproptosis-related genes, we drew the correlation heat map between genes (Fig. 3B), and we drew the correlation scatter diagram for the gene expression with an extreme statistical difference ( Fig. 3C-G), the results showed that there was a statistically significant difference between LIPT1 and GLS (P < 0.001), there was a statistically significant difference between LIPT1 and MTF1 (P < 0.001), there was a very statistically significant difference between LIPT1 and PDHB statistically significant difference (P < 0.001), very statistically significant difference between MTF1 and SLC31A1 (P < 0.001), very statistically significant difference between PDHB and SLC31A1 (P < 0.001).

Then, based on the expression of 5 CRGs (LIPT1, PDHB, MTF1, GLS, SLC31A1) in the samples of patients with sepsis in the merged dataset, we obtained each sepsis patients' cuproptosis scores (CPs) in the merged dataset through the ssGSEA algorithm. To analyze the difference in gene expression between the high and low CPs groups in the disease group patient samples in the merged dataset, we used the limma package to perform differential analysis on the merged dataset to obtain the DEGs between the high and low CPs groups in the sepsis dataset. The results of our differential analysis on this dataset were plotted for volcanoes (Fig. 3H). A total of 44 genes in the merged dataset meet the threshold of |logFC| > 0.5 and P.value < 0.05. At this threshold, the number of genes that are highly expressed in the high CPs group (low expression in the low CPs group, logFC is positive, and up-regulated genes) There are 3, and the number of low expression in high CPs group (high expression in low CPs group, logFC is negative) is 41.

Construction of associated disease subtypes of sepsis

To explore the expression differences of the module CRDEGs in the merged dataset in sepsis patient samples, we used the R package ConsensusClusterPlus to identify sepsis by consensus clustering method based on five CRDEGs (LIPT1, PDHB, MTF1, GLS, SLC31A1). Associated with different disease subtypes, two sepsis disease subtypes (cluster1 and cluster2) were finally identified (Fig. 4A). sepsis disease subtype 1 (cluster1) contained 115 samples, and sepsis disease subtype 2 (cluster2) contained 93 samples. Then we performed PCA on the expression matrix of the dataset of the two sepsis disease subtypes in sepsis, PCA Clustering results revealed significant differences between the two sepsis subtypes’ samples (Fig. 4B). Then we also showed the consistent clustering cumulative distribution function (CDF) graph of different cluster numbers in the consistent clustering results (Fig. 4D). It can be seen from the figure that when k = 2 is used as the cluster number of unsupervised clustering, The consistent clustering results of the merged dataset are the best. Then we used the R package pheatmap to draw a heat map to show the expression differences of the five CRDEGs in the two disease subtype samples of the merged dataset (Fig. 4C). The results showed that the 5 CRDEGs in the two diseases subtype samples of the merged dataset showed certain differences in expression levels.

Then, we used the Mann-Whitney U test (Wilcoxon rank sum test) to analyze the expression levels of 5 CRDEGs (LIPT1, PDHB, MTF1, GLS, SLC31A1) in the merged dataset of two sepsis disease subtypes (cluster1 and cluster2) and showed the results of expression difference analysis through the group comparison chart (Fig. 4E). It can be seen from Fig. 4E that the expression of 3 CRDEGs (LIPT1, MTF1, GLS) among the 5 CRDEGs between the two sepsis disease subtypes (cluster1 and cluster2) in the merged dataset has at least a statistically significant difference (P < 0.05)). Among them, MTF1 was a significantly up-regulated gene, and LIPT1 and GLS were significantly down-regulated genes.

Determination of co-expression modules in the merged dataset by WGCNA analysis

We performed WGCNA to screen the co-expression modules. In the process of WGCNA, we first clustered the SG and NG in the merged dataset through the clustering tree and marked the grouping information (without setting cut height), and then we set the screening standard to 0.9 to determine the best module Quantity, the DEGs between different groups in the merged dataset are gathered in 12 modules including MEblack, MEred, MEyellow, MEbrown, MEblue, MEturquoise, MEgrey, MEpurple, MEgreenyellow, MEpink, MEgreen, MEmagenta (Fig. 5A). Then we set the module merge cut height to 0.2, cut and merged the modules whose cut height is lower than 0.2 (Fig. 5C), and clustered the DEGs between different groups in the merged dataset again to visualize the relationship between genes and corresponding new modules. Finally, according to the expression patterns of the module genes and the grouping information of the different groups of the merged dataset, we obtained the correlation between the six modules (MEbrown, MEgreen, MEturquoise, MEblack, MEblue, MEgrey) and the different groups of the merged dataset (Fig. 5B).

First of all, we had statistically significant differences (P < 0.05, absolute value of correlation ≥ 0.3) among the 6 modules (excluding the useless module gray module: MEgrey) and different groups in the merged dataset. The differentially expressed genes (DEGs) contained in the most relevant MEbrown(|COR|=0.58, P = 6e-20) in cluster1 and the most relevant MEturquoise(|COR|=0.51, P = 3e-15) in cluster2 were analyzed subsequently. First, we intersected the cuproptosis-related differentially expressed genes (CRDEGs) in the merged dataset with the DEGs contained in the MEbrown module (Fig. 5D) and the MEturquoise module (Fig. 5E), respectively, and draw a Venn diagram to obtain Module CRDEGs, as shown in Fig. 4, we have obtained a total of 202 CRDEGs, the specific gene information is shown in Table S2.

GO and KEGG enrichment analysis of cuproptosis-related genes

To analyze the biological processes, molecular functions, cellular components, biological pathways and the relationship between the five CRGs (LIPT1, PDHB, MTF1, GLS, SLC31A1) and cuproptosis, we first performed GO and KEGG enrichment analysis. We showed the results of GO functional enrichment analysis (Fig. 6A, 6C, Table S6) and KEGG enrichment analysis (Fig. 6B, 6D, Table S7). In addition, we displayed the relationship between CRGs and GO and KEGG enrichment analysis results in a circular network diagram (Fig. 6E-F). The results showed that CRGs were mainly enriched in cellular copper ion homeostasis, regulation of respiratory gaseous exchange by neurological system process, suckling behavior, protein-cofactor linkage and other biological processes (BP). In mitochondrial matrix, oxidoreductase complex, late endosome, recycling endosome and other cellular components (cellular component, CC). oxidoreductase activity, acting on the aldehyde or oxo group of donors, transition metal ion transmembrane transporter activity, oxidoreductase activity, acting on the aldehyde or oxo group of donors, NAD or NADP as acceptor, histone acetyltransferase binding, Mlecular f) middle. CRGs are significantly enriched in Central carbon metabolism in cancer, Mineral absorption, Glycolysis/Gluconeogenesis, Pyruvate metabolism, Alanine, aspartate and glutamate metabolism, Citrate cycle (TCA cycle), Proximal tubule bicarbonate reclamation, Arginine biosynthesis pathway.

GSEA and GSVA of Sepsis Dataset

To determine the impact of gene expression levels in the normal sample group (NG) and sepsis patient sample group (SG) on sepsis in the data set, we analyzed the data set GSE131761 and the data set GSE54514 through GSEA (Fig. 7A and 7F, Table S8 and S9). The link between gene expression and the biological processes involved, the cellular components affected, and the molecular functions exerted. The results showed that the differentially expressed genes in the data set GSE131761 were significantly enriched in pathways such as Hedgehog pathway (Fig. 7B), MAPK pathway (Fig. 7C), TP53 pathway (Fig. 7D), and Glycolysis pathway (Fig. 7E). The differentially expressed genes in the data set GSE54514 were significantly enriched in Hedgehog pathway (Fig. 7G), Autophagy pathway (Fig. 7H), TGFBETA pathway (Fig. 7I), Wnt pathway (Fig. 7J) and other pathways.

To explore the variability of dataset GSE131761 and GSE54514 between the sepsis samples (group: sepsis) and the corresponding normal samples (group: Normal), We performed GSVA enrichment analysis on datasets GSE131761 (Fig. 8A, Table S10) and GSE54514 (Fig. 8B, Table S11) to calculate the functional enrichment differences. The results show that the data set GSE131761 and the data set GSE54514 show differences between sepsis samples (group: sepsis) and corresponding normal samples (group: Normal). The differential genes in the dataset GSE131761 were significantly enriched in Glycolysis pathway, PI3K AKT pathway, Wnt BETA pathway, IL6 JAK pathway and other pathways. The differential genes in the dataset GSE54514 are significantly enriched in Notch pathway, Oxidative pathway, PI3K AKT pathway, TGF BETA pathway and other pathways.

Construction of CRDEGs diagnostic model and its prognostic performance

To determine the diagnostic value of 202 CRDEGs in the merged dataset, we used Lasso regression analysis to construct a CRDEGs diagnostic model (Fig. 9A). Lasso regression is based on linear regression. By increasing the penalty item (lambda × absolute value of the slope), the overfitting of the model is reduced and the generalization ability of the model is improved. In addition, we also visualized the results of Lasso regression, and the Lasso variable locus diagram (Fig. 9C) was obtained. The results showed that the genes changed with the coefficient of the lambda coefficient (log post) of the Lasso penalty term, and the number of genes gradually increased as the lambda decreased. Then we visualized the expression of CRDEGs in different groups in the CRDEGs diagnostic model through a forest plot (Fig. 9B). It can be seen from Fig. 9C that we constructed 32 CRDEGs in the CRDEGs diagnostic model, see Table S3.

After the 32 CRDEGs in the CRDEGs diagnostic model were calculated based on the coefficients of the 32 genes calculated by LASSO analysis, the expression of each gene was multiplied by the corresponding coefficient and added to establish the prediction score of the CRDEGs diagnostic model. The final prediction score of each sample, and then through the Wilcoxon signed rank test (Wilcoxon signed rank test) to analyze the expression of the CRDEGs diagnostic model in the samples of different groups (sepsis/Normal) in the merged dataset and compare the results through the group comparison chart As shown in Fig. 10A, the results showed that the expression of the CRDEGs diagnostic model constructed by us was statistically significantly different among different groups in the merged dataset (P < 0.001). We divided the final prediction score of each sample in the disease group of the 32 CRDEGs in the merged dataset into high and low risk groups (High/Low group) to draw a heat map (Fig. 10B), and then we based on the 32 CRDEGs in the merged dataset. The final prediction score of each sample in the group was divided into high and low risk groups (High/Low group), and then we passed the Wilcoxon signed rank test (Wilcoxon signed rank test) on the 32 CRDEGs in the CRDEGs diagnostic model in the merged dataset disease group The expression difference between the high and low risk groups (High/Low group) was analyzed (Fig. 10C), and the results showed that 20 CRDEGs (AK1, BLVRB, CACNA1E, CD82, GBA, HBG1, HBQ1, IL1RN, LGALS3, LTB4R, MCOLN1, MTF1, PIK3R2, RPS3A, RUNDC3A, SGSH, TMEM145, VSTM1, VTI1B, XPO7) have extremely statistically significant differences (P < 0.001) in the expression levels of the high and low risk groups in the merged dataset disease group. Among them, CACNA1E, CD82, GBA, IL1RN, LTB4R, MTF1, SGSH and VSTM1 were significantly up-regulated genes, and AK1, BLVRB, HBG1, LGALS3, MCOLN1, PIK3R2, RPS3A, RUNDC3A, TMEM145, VTI1B, XPO7, and HBQ1 were significantly down-regulated genes. The expression levels of FLOT2 and FOXO4 in the high and low risk groups of the merged dataset disease group have highly statistically significant differences (P < 0.01), and the expression levels of MAF1 in the high and low risk groups of the merged dataset disease group have a certain statistically significant difference (P < 0.05 ). Among them, FLOT2 was a significantly up-regulated gene, and FOXO4 and MAF1 were significantly down-regulated genes.

Finally, the expression of 32 CRDEGs in different groups (sepsis/Normal) in GSE25504 was verified for expression difference analysis (Fig. 11A), and the results showed that 23 CRDEGs (BLVRB, CACNA1E, CD82, CYP1B1, FLOT2, FOXO4, GBA, GRN, HBQ1, HSPA1A, IL1RN, LTB4R, MAF1, MCOLN1, MTF1, RNF10, RPL35, RPS3A, RUNDC3A, SGSH, SLC1A5, TMEM158, VTI1B) have statistically significant differences in the expression levels of different groups in the merged dataset (P < 0.001), the expression levels of 3 CRDEGs (HBG1, XPO7, SPRYD3) in different groups of the merged dataset had highly statistically significant differences (P < 0.01). The expression levels of the three CRDEGs (LGALS3, TUBB2A, VSTM1) in different groups of the merged dataset had statistically significant differences (P < 0.05). It was also verified that the expression levels of CRDEGs between different groups (Sepsis/Normal) in GSE26378 were analyzed for expression differences (Fig. 11B). The results showed that 12 CRDEGs (CACNA1E, CD82, CYP1B1, FLOT2, GRN, IL1RN, LTB4R, MCOLN1, MTF1, RNF10, SGSH, VSTM1) in different groups of the merged dataset had statistically significant differences (P < 0.001). Among them, CACNA1E, CD82, CYP1B1, FLOT2, GRN, IL1RN, LTB4R, MTF1 were significantly up-regulated genes, and MCOLN1 and RNF10 were significantly down-regulated genes. The expression levels of the three CRDEGs (AK1, FOXO4, LGALS3) in different groups of the merged dataset had highly statistically significant differences (P < 0.01). Among them, AK1 was a significantly up-regulated gene, while FOXO4 and LGALS3 were significantly down-regulated genes.

Then we plotted the Receiver operating characteristic curve (ROC) curve of 23 CRDEGs (see Table S4) with statistically significant differences in the expression levels in the high and low risk groups of the merged dataset disease group. The results showed that: 23 differentially expressed genes related to cuproptosis were compared one by one Draw the ROC curve and display the results (Fig. 12). From the ROC curve in Fig. 12, we can see: AK1 (AUC = 0.657, Fig. 12A), BLVRB (AUC = 0.727, Fig. 12B), CACNA1E (AUC = 0.686, Fig. 12C), CD82 (AUC = 0.664, Fig. 12D), FLOT2 (AUC = 0.614, Fig. 12E), FOXO4 (AUC = 0.620, Fig. 12F), GBA (AUC = 0.704, Fig. 12G), HBG1 (AUC = 0.637, Fig. 12H), HBQ1 (AUC = 0.635, Fig. 12I), IL1RN (AUC = 0.646, Fig. 12J), LGALS3 (AUC = 0.668, Fig. 12K), LTB4R (AUC = 0.731, Fig. 12L), MAF1 (AUC = 0.600, Fig. 12M), MCOLN1 (AUC = 0.645, Fig. 12N), MTF1 (AUC = 0.713, Fig. 12O), PIK3R2 (AUC = 0.656, Fig. 12P), RPS3A (AUC = 0.696, Fig. 12Q), RUNDC3A (AUC = 0.652, Fig. 12R), SGSH (AUC = 0.676, Fig. 12S), TMEM145 (AUC = 0.643, Fig. 12T), VSTM1 (AUC = 0.651, Fig. 12U), VTI1B (AUC = 0.657, Fig. 12V), XPO7 (AUC = 0.652, Fig. 12W) were significantly associated with high and low risks of sepsis sex.

We then performed single/multivariate Cox regression analysis on the expression levels of 23 CRDEGs and constructed a Cox regression model, and then we conducted nomogram analysis to judge the prognostic ability of the model and drew a nomogram (Fig. 13A). In addition, we performed prognostic Calibration analysis on the nomogram (nomogram) of the univariate and multivariate COX regression models and drew a calibration curve (Fig. 13B), and finally we used decision curve analysis (decision curve analysis, DCA) The effect of the constructed Cox regression model on clinical utility was evaluated and the results were displayed (Fig. 13C). The x-axis in the DCA chart represents the probability threshold or Threshold Probability, and the y-axis represents the net benefit. The result could be judged by observing that the model line can be stably higher than the All positive line and the All negative line's x-value range. The larger the x-value range, the better the model effect.

Construction of protein-protein interaction (PPI) network and ceRNA network analysis

We used the STRING database to analyze the protein-protein interaction of 23 CRDEGs (all parameters in the STRING database keep the default value and do not change. It was equivalent to the minimum required interaction score: medium confidence (0.400)), and constructed the 23 CRDEGs Protein-protein interaction (PPI) network using Cytoscape software to visualize the interaction relationship (Fig. 14A).

To further obtain the correlation between the 23 CRDEGs, we drew a heat map of the correlation between genes (Fig. 14C-F), the results showed that there was a very statistically significant difference between the gene BLVRB and AK1 (P < 0.001), a very statistically significant difference between FOXO4 and AK1 (P < 0.001), a very statistically significant difference between GBA and AK1 There was a statistically significant difference between MTF1 and FOXO4 (P < 0.001), and there was a statistically significant difference between MTF1 and FOXO4 (P < 0.001).

Then we used the ENCORI database and the miRDB database to predict miRNAs interacting with hub genes (mRNA), and then screened the mRNA-miRNA interaction pairs in the miRDB database according to the Target Score > 90 standard, and then took the pairs found in the two databases. The mRNA-miRNA action pair is intersected to obtain the mRNA-miRNA action pair, and the miRNA in the obtained mRNA-miRNA action pair is used to predict the LncRNA that interacts with the miRNA in the ENCORI database, and the LncRNA corresponding to the miRNA is screened based on the clear sequence of miRseq and targetSeq. Results The ceRNA interaction network consisted of 7 CRDEGs (FLOT2, CACNA1E, MTF1, VTI1B, IL1RN, TMEM145, XPO7), 17 miRNA molecules, and 30 LncRNA interactions, and the specific mRNA-miRNA-lncRNA interaction relationship was visualized display (Fig. 14G), the relationship is shown in Table S5.

Immune Infiltration Analysis of Sepsis Dataset (CIBERSORTx)

We sorted the expression profile data of the merged dataset and uploaded it to the CIBERSORTx online website, and used the CIBERSORTx algorithm to calculate the expression profile data of the 22 kinds of immune cells and the sepsis patient sample group (group: SS) in the merged dataset Correlation between high and low risk groups, according to the results of immune infiltration analysis, we plotted the infiltration of immune cells in the merged dataset in the form of a histogram (Fig. 15A).

To investigate the connection between the expression of seven CRDEGs (XPO7, VTI1B, TMEM145, MTF1, ILIRN, FLOT2, CACNA1E) and the abundance of immune cell infiltration in the dataset, we combined the expression profile data and the abundance of immune cell infiltration and calculated the correlation coefficient. The heat map displays the connection between CRDEG expression levels and the quantity of 22 immune cell infiltrates (Fig. 15B). The results show that there are 16 kinds of immune cells in the merged dataset (Memory B cells, Naïve B cells, Dendritic cells activated, Macrophages M0, Macrophages M2, Mast cells resting, Monocytes, NK cells activated, NK cells resting, Plasma cells, T cells CD4 memory activated, T cells CD4 memory resting, T cells CD4 naive, T cells CD8, T cells gamma delta, T cells regulatory (Tregs)) There is a statistical relationship between the abundance of immune infiltration and the expression of differentially expressed genes related to cuproptosis learning meaning.

The definition of sepsis is a systemic inflammatory response to the dysfunction of the host, causing life-threatening organ dysfunction. Although research on the subject has been intensifying in recent years and treatment options are becoming increasingly refined, patients with sepsis still have difficulty identifying early and given targeted treatment. As the stage of septic shock advances, the fatality rate could reach 50%⁴³, and patients present with a persistent and irremediable hypotensive state, severe tissue hypoperfusion, and multiple organ failure. The metabolism of divalent metal ions exhibits high complexity, with numerous interactions with central physiological mechanisms in organisms. Dissecting the physiopathological mechanisms of their metabolism and further targeting interventions is a promising new treatment option for sepsis.

As one of such metal ions, copper exerts an enormous influence on the maintenance of many physiological functions in the organism by acting as a structure-active center for relevant enzymes⁴⁴ or as an electron transport chain carrier⁴⁵. Copper, as a transition metal, causes cell lesions through Fenton chemistry which facilitates the generation of products such as ROS, nitrogenous substances, and free radicals that could damage proteins and lipids, even DNA^46–48. As a result, the body has evolved a sophisticated mechanism for the dynamic balance of copper ion homeostasis to maintain them in a relatively beneficial state for its development.

Such a situation, however, presupposes the normal functioning of homeostatic mechanisms, which are clearly not present in septic patients^8,49. At normal pH, plasma proteins such as ceruloplasmin and albumin combine with free copper ions in the plasma to maintain a specific concentration⁵⁰. Sepsis patients have varying degrees of hypotension, tissue hypoperfusion and hypoxia, resulting in increased anaerobic cellular respiration and elevated blood lactic acid. The binding capacity of copper to these transport proteins is decremented in an acidic environment. In an ischemia-reperfusion injury model, serum copper concentrations in mice increased nearly 50-fold within minutes of the reperfusion molding operation⁵¹. It is consequently critical to comprehend the function that copper disturbance plays in the subsequent progression of sepsis.

We are certainly not the only ones to have noticed this, as many scientific teams have observed the imbalance in copper homeostasis in sepsis and have conducted studies on the correlation between aberrant serum copper levels and organ failure and immune dysfunction^11,52,53. Equally, the recent discovery by Tsvetkov P et al. of a novel pattern of cell death based on copper accumulation - "cuproptosis" - is becoming the focus of many scientists. To investigate the role of this mechanism in the course of sepsis, we used bioinformatics to explore whether cuproptosis is involved in the pathogenesis of sepsis and to identify subtypes by analysis of CRGs' specific expression, and then to establish a diagnostic model for cuproptosis in sepsis. To enhance the accuracy of our research, we combined two sepsis datasets (GSE131761, GSE54514) for differential expression of CRGs between the disease and control groups, ultimately obtaining five key genes (MTF1, SLC31A1, LIPT1, PDHB, GLS). The expression of MTF1 and SLC31A1 was significantly higher in the disease group than in the control group, and LIPT1, PDHB, and GLS were lower than in the control group.

Metal regulatory transcription factor 1(MTF1) is ubiquitously expressed in different human tissues/cells. After activation, its primary location shifts from the cytoplasm to the nucleus, where it identifies and binds the metal response element (MRE) within heavy metal response gene promoter regions⁵⁴. Subsequently, it initiates the transcription of metallothioneins (MTs), which play a crucial role in metal homeostasis/detoxification, and binds DNA to regulate RNA transcription in response to alterations in cytosolic Zn and Cu ion levels⁵⁵. The metal-binding sites of MTF1 are relatively conserved across species. Drosophila MTF1 is enriched with a cysteine cluster structural domain that allows intracellular copper to bind and activate, and ones lacking MTF1 are more sensitive to copper toxicity than usual⁵⁶. A similar copper-binding center has also been identified in mammalian MTF1⁵⁷. A study found that mice lacking MTF1 died of liver failure during the embryonic period⁵⁸. These are presumably also possible responses to copper in human MTF1. In humans, MTF1 abnormalities can lead to pathological accumulation of copper and cause copper toxicity in all vital organs, mainly in the liver and kidney or CNS⁵⁹. All of the studies mentioned above highlight the pivotal contribution of MTF1 in the heavy metal stress response and normal physiology. Our study did not obtain serum copper ion concentrations in sepsis patients. Still, from the increased expression of MTF1 in the disease group, we could infer that the intracellular copper ion concentrations in these sepsis patients were already relatively high in sequence to force the organism to increase the expression of regulatory genes related to heavy metal metabolism.

Copper transporter 1 (CTR1, also known as SLC31A1), an essential copper influx transporter expressed on the cell membrane, is a transporter protein with a high affinity for monovalent copper⁶⁰. Increased expression of SLC31A1 produces significant copper accumulation in cells. There is evidence that the initial rate of Cu²⁺ uptake by HEK293 cells highly expressing human SLC31A1 is much faster than normal cells⁶¹. The high expression of SLC31A1 in diseased in this study further validates our suspicion that a state of high intracellular copper ions exists in sepsis patients and that the high SLC31A1 expression is also the culprit for the elevated concentration of intracellular copper, providing the copper accumulation basis for cuproptosis. Intriguingly, high extracellular copper ion concentrations did not induce high SLC31A1; on the contrary, it was reported that high extracellular Cu²⁺ reduced SLC31A1 mRNA expression by 30% (25 mM Cu²⁺) and 60% (100 mM Cu²⁺), respectively⁶². In sepsis patients with high expression of SLC31A1, we speculated on two possibilities: For one, the relative serum copper concentrations in sepsis patients might not be as high as reflected in the relevant data we have collected, or the copper ion concentrations around the cells of the associated tissues where major cuproptosis occurs are not as high. If our hypothesis is real, it would be inappropriate to correlate serum copper concentrations with the degree of cell cuproptosis in sepsis patients in a more subjective way in clinical studies. Redeveloping targeted and accurate diagnostic models for cuproptosis in sepsis patients is indispensable. Another possibility is that high serum copper concentrations in sepsis patients are accompanied by inhibiting SLC31A1 expression by specific mechanisms that counteract high extracellular copper ion concentrations. In a state of high extracellular copper ion concentration superimposed on the high efficiency of cellular uptake of copper ions, the rate and amount of intracellular copper ion accumulation in sepsis patients might be far greater than in other diseases. Therefore, we hypothesized that the probability of cuproptosis and the corresponding degree of damage to cells and tissues might be much higher than we predicted. Cuproptosis might be the most essential pathogenic mechanism of sepsis, not just one of several.

To verify whether varied serum copper concentrations exist in sepsis patients or whether unknown factors contribute to differences in the extent of cuproptosis, two subtypes (cluster1 and cluster2) emerged from the analysis of the hub CRGs in the dataset in sepsis patients using a consistent clustering approach. MTF1 was considerably increased in cluster 1 compared to cluster 2, while LIPT1 and GLS were dramatically downregulated. The function of the lipoyltransferase 1 (LIPT1) enzyme, which is encoded by the gene for LIPT1, is to transfer lipoyl moieties from the glycine cleavage H protein (GCSH) to the E2 subunits of the 2-oxoacid dehydrogenases, that are necessary for mitochondrial metabolism and the generation of lipoic acid^63,64. Evidence suggests that a LIPT1 deficit worsens the tricarboxylic acid cycle in mitochondria by preventing pyruvate oxidation, which might lead to defective lipid synthesis since pyruvate dehydrogenase complex(PDHC) is necessary for lipogenesis by supplying acetyl-CoA^65,66. Other severe mitochondrial dysfunction syndromes⁶⁷, such as fatal lactic acidosis⁶⁸, might still ensue from clinical patients' lack of LIPT1 expression. Down-regulation of GLS results in the same disruption of cellular metabolism as LIPT1 because Glutaminase (GLS) catalyzes the conversion of glutamine to glutamate, which is also a raw material for alpha-ketoglutarate implicated in tricarboxylic acid cycle events⁶⁹. We proposed that LIPT1 and GLS expression levels are vital mechanisms of the severity of cuproptosis in pathological tissues, as the essence of cuproptosis is cell death resulting from the disrupted tricarboxylic acid cycle. The difference in the degree of cuproptosis in the septic patients was the key distinction between the two sepsis subgroups in this investigation. Based on the proportion of the two subtypes and assuming that the findings of this study are consistent with the clinical profile of sepsis patients, we conclude that more than half of sepsis patients might be clinically affected by high levels of cuproptosis.

During the analysis of the sepsis dataset, the downregulation of the expression of several hub genes regulating pyruvate metabolism and the mitochondrial respiratory chain involved in cuproptosis led us to realize that cuproptosis might be more highly tied to septic shock. Alternatively, cuproptosis might be a main causative factor for developing severe lactic acidosis in septic shock patients. Furthermore, since medications cannot directly target the intracellular concentration of copper ions, lactic acidosis induced by cuproptosis is more chronic and challenging to correct than lactic acidosis produced by other reasons. Our earlier work on the GO and KEGG analyses of the five hub genes we found corroborated our hypothesis. As a corollary, we propose the novel notion of ‘sepsis-related cuproptosis’. We believed that early discovery and diagnosis are as valuable clinically as any of the recognized auxiliary diagnoses of sepsis. Therefore, we established a diagnostic model for sepsis-related cuproptosis conditions using a pooled dataset of 32 CRDEGs. The model was verified in datasets GSE25504 and GSE26378, and 23 CRDEGs that also statistically differed in the validation set were further tested for cuproptosis. The ROC curves revealed a substantial relationship between these CRGs and sepsis, enabling us to build a Cox regression model to identify patients at risk of sepsis, allowing for an accurate diagnosis of sepsis-related cuproptosis at the genetic level.

Our GSEA and GSVA of the sepsis dataset revealed that DEGs were primarily enriched in immune-related pathways like inflammation and apoptosis. (GSEA: Hedgehog pathway, MAPK pathway,TP53 pathway, Autophagy pathway, Wnt pathway; GSVA: PI3K_AKT pathway,Wnt_BETA pathway,IL6_JAK pathway,Notch pathway,Oxidative pathway,TGF_BETA pathway) Given the impact of immune cell infiltration on the body's immune response in sepsis, where the systemic inflammatory response is the central disease mechanism, we investigated the expression levels of seven CRGs (XPO7, VTI1B, TMEM145, MTF1, ILIRN, FLOT2, CACNA1E) in the sepsis dataset in relation to immune cell correlation of infiltration abundance, along with 16 immune cells potentially regulated by cuproptosis. Numerous studies have proven that the degree of CRG expression correlates strongly with the condition of immune cell infiltration in the microenvironment of sick tissues in different forms of cancer^70–73. Similarly, we discovered that CRGs have a role in altered immune infiltration of the organism in non-cancerous illnesses. Cuproptosis might not only trigger sepsis by disrupting mitochondrial respiration and causing problems in energy metabolism, but it also is a factor in the immunological dysfunction and systemic inflammatory response in septic patients.

LIMITATIONS

Firstly, this study lacked clinically definitive information on each sepsis sample. Due to the variability and complexity of sepsis, the sampling nodes and treatment regimens used in these patients are not homogeneous, leading to potential biases in the grouping of samples based on differences in gene expression. Second, it is evident from the discussion that the lack of multi-omics analysis and simply from a transcriptomic standpoint leads to some flaws in confirming our important hypothesis conjectures. Only gene expression of copper ion transport-related proteins, mitochondrial respiratory chain, and immune infiltration results could be utilized to speculate. Despite our somewhat favorable effects, we anticipate that we will be able to investigate further the mechanisms associated with sepsis and cuproptosis in the long term in a collaborative multi-omics study.

This study identifies a tight association between cuproptosis and sepsis. It suggests that five hub genes, MTF1, SLC31A1, LIPT1, PDHB, and GLS, are the basis for determining whether cuproptosis is present in sepsis patients, and elucidates two subtypes of sepsis that we might not have previously been aware of based on the differential expression of these five genes. We have also constructed a diagnostic model for sepsis based on CRGs, which might provide a new research tool for clinical trials related to the cuproptosis theory in sepsis.

DATA AVAILABILITY

The datasets analysed during the current study are available in Gene Expression Omnibus. The datasets GSE54514 and GSE131761 can be accessed from the following URLs: [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE54514] and [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE131761].

ETHICS STATEMENT

None of the researchers conducted any studies with humans or animals for this article.

CONFLICTS OF INTEREST

All authors declare no conflict of interest.

AUTHOR CONTRIBUTIONS

Xuan Jia and Yuan Tian co-designed the study, performed the bioinformatics analysis, and drafted the manuscript and participated in collating the data and proofreading and translating the manuscript, contributing equally to this study. Dian Zhang participated in the collection and collation of the data and provided the algorithm optimization. Siqia Chen and Dongmei Meng supplied the research with a comprehensive evaluation and methodological direction. This study was accomplished by all of the authors, and the final version was approved with one accord.

ACKNOWLEDGMENTS

Thanks to the helixlife team for providing the bioinformatics course series on the platform.

Singer, M. et al. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA 315, 801–810 (2016).
Reinhart, K. et al. Recognizing Sepsis as a Global Health Priority - A WHO Resolution. N Engl J Med 377, 414–417 (2017).
Rudd, K. E. et al. Global, regional, and national sepsis incidence and mortality, 1990-2017: analysis for the Global Burden of Disease Study. Lancet 395, 200–211 (2020).
Stephen, A. H., Montoya, R. L. & Aluisio, A. R. Sepsis and Septic Shock in Low- and Middle-Income Countries. Surgical Infections 21, 571–578 (2020).
Shankar-Hari, M. et al. Developing a New Definition and Assessing New Clinical Criteria for Septic Shock: For the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA 315, 775–787 (2016).
Bauer, M. et al. Mortality in sepsis and septic shock in Europe, North America and Australia between 2009 and 2019- results from a systematic review and meta-analysis. Crit Care 24, 239 (2020).
van der Poll, T., van de Veerdonk, F. L., Scicluna, B. P. & Netea, M. G. The immunopathology of sepsis and potential therapeutic targets. Nat Rev Immunol 17, 407–420 (2017).
Akkaş, İ., Ince, N. & Sungur, M. A. Serum trace element and heavy metal levels in patients with sepsis. Aging Male 23, 222–226 (2020).
Wisniewska, M. et al. Copper to Zinc Ratio as Disease Biomarker in Neonates with Early-Onset Congenital Infections. Nutrients 9, 343 (2017).
Saadat, S. H. & Goodarzi, R. The Relation of Copper and Zinc with Incidence of Neonatal Sepsis and Possible Prediction Biomarker Role. Galen Med J 9, e1933 (2020).
Meng, J.-B. et al. The Correlation Between Whole Blood Copper (Cu), Zinc (Zn) Levels and Cu/Zn Ratio and Sepsis-Induced Left Ventricular Systolic Dysfunction (SILVSD) in Patients with Septic Shock: A Single-Center Prospective Observational Study. Int J Gen Med 14, 7219–7234 (2021).
Cobine, P. A. & Brady, D. C. Cuproptosis: Cellular and molecular mechanisms underlying copper-induced cell death. Mol Cell 82, 1786–1787 (2022).
Collins, J. F. Copper nutrition and biochemistry and human (patho)physiology. Adv Food Nutr Res 96, 311–364 (2021).
Chen, J. et al. The molecular mechanisms of copper metabolism and its roles in human diseases. Pflugers Arch 472, 1415–1429 (2020).
Tsvetkov, P. et al. Copper induces cell death by targeting lipoylated TCA cycle proteins. Science 375, 1254–1261 (2022).
Lin, S., Luo, B. & Ma, J. Multiple datasets to explore the molecular mechanism of sepsis. BMC Genom Data 23, 66 (2022).
Li, Z. et al. Identification of Potential Early Diagnostic Biomarkers of Sepsis. J Inflamm Res 14, 621–631 (2021).
Long, G. & Yang, C. A six‑gene support vector machine classifier contributes to the diagnosis of pediatric septic shock. Mol Med Rep 21, 1561–1571 (2020).
Barrett, T. et al. NCBI GEO: archive for functional genomics data sets--update. Nucleic Acids Res 41, D991-995 (2013).
Wang, X., Hu, S., Ji, W., Tang, Y. & Zhang, S. Identification of genes associated with clinicopathological features of colorectal cancer. J Int Med Res 48, 300060520912139 (2020).
Fishilevich, S. et al. GeneHancer: genome-wide integration of enhancers and target genes in GeneCards. Database (Oxford) 2017, (2017).
Ritchie, M. E. et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 43, e47 (2015).
Worley, B. & Powers, R. PCA as a practical indicator of OPLS-DA model reliability. Curr Metabolomics 4, 97–103 (2016).
Xiao, B. et al. Identification and Verification of Immune-Related Gene Prognostic Signature Based on ssGSEA for Osteosarcoma. Front Oncol 10, 607622 (2020).
Alkhouzaam, A. & Qiblawey, H. Functional GO-based membranes for water treatment and desalination: Fabrication methods, performance and advantages. A review. Chemosphere 274, 129853 (2021).
Kanehisa, M. & Goto, S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res 28, 27–30 (2000).
Yu, G., Wang, L.-G., Han, Y. & He, Q.-Y. clusterProfiler: an R package for comparing biological themes among gene clusters. OMICS 16, 284–287 (2012).
Powers, R. K., Goodspeed, A., Pielke-Lombardo, H., Tan, A.-C. & Costello, J. C. GSEA-InContext: identifying novel and common patterns in expression experiments. Bioinformatics 34, i555–i564 (2018).
Liberzon, A. et al. The Molecular Signatures Database (MSigDB) hallmark gene set collection. Cell Syst 1, 417–425 (2015).
Hänzelmann, S., Castelo, R. & Guinney, J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics 14, 7 (2013).
Brière, G., Darbo, É., Thébault, P. & Uricaru, R. Consensus clustering applied to multi-omics disease subtyping. BMC Bioinformatics 22, 361 (2021).
Langfelder, P. & Horvath, S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics 9, 559 (2008).
Engebretsen, S. & Bohlin, J. Statistical predictions with glmnet. Clin Epigenetics 11, 123 (2019).
Liaw, A. & Wiener, M. Classification and Regression by random Forest. 2, 5 (2002).
Alderden, J. et al. Predicting Pressure Injury in Critical Care Patients: A Machine-Learning Model. Am J Crit Care 27, 461–468 (2018).
Park, S. Y. Nomogram: An analogue tool to deliver digital knowledge. J Thorac Cardiovasc Surg 155, 1793 (2018).
Tataranni, T. & Piccoli, C. Dichloroacetate (DCA) and Cancer: An Overview towards Clinical Applications. Oxid Med Cell Longev 2019, 8201079 (2019).
Wu, D. et al. Exostosin1 as a novel prognostic and predictive biomarker for squamous cell lung carcinoma: A study based on bioinformatics analysis. Cancer Med 10, 2787–2801 (2021).
Chen, Y. & Wang, X. miRDB: an online database for prediction of functional microRNA targets. Nucleic Acids Res 48, D127–D131 (2020).
Szklarczyk, D. et al. The STRING database in 2021: customizable protein-protein networks, and functional characterization of user-uploaded gene/measurement sets. Nucleic Acids Res 49, D605–D612 (2021).
Shannon, P. et al. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res 13, 2498–2504 (2003).
Steen, C. B., Liu, C. L., Alizadeh, A. A. & Newman, A. M. Profiling Cell Type Abundance and Expression in Bulk Tissues with CIBERSORTx. Methods Mol Biol 2117, 135–157 (2020).
Font, M. D., Thyagarajan, B. & Khanna, A. K. Sepsis and Septic Shock - Basics of diagnosis, pathophysiology and clinical decision making. Med Clin North Am 104, 573–585 (2020).
Tikhonova, T. V. et al. Trinuclear copper biocatalytic center forms an active site of thiocyanate dehydrogenase. Proc Natl Acad Sci U S A 117, 5280–5290 (2020).
Szabo, R., Bodolea, C. & Mocan, T. Iron, Copper, and Zinc Homeostasis: Physiology, Physiopathology, and Nanomediated Applications. Nanomaterials (Basel) 11, 2958 (2021).
Halliwell, B. Free radicals and antioxidants: updating a personal view. Nutr Rev 70, 257–265 (2012).
Pohanka, M. Copper and copper nanoparticles toxicity and their impact on basic functions in the body. Bratisl Lek Listy 120, 397–409 (2019).
Johnson, M. D. L., Kehl-Fie, T. E. & Rosch, J. W. Copper intoxication inhibits aerobic nucleotide synthesis in Streptococcus pneumoniae. Metallomics 7, 786–794 (2015).
Zemrani, B. & Bines, J. E. Recent insights into trace element deficiencies: causes, recognition and correction. Curr Opin Gastroenterol 36, 110–117 (2020).
Kirsipuu, T. et al. Copper(II)-binding equilibria in human blood. Sci Rep 10, 5686 (2020).
Chevion, M. et al. Copper and iron are mobilized following myocardial ischemia: possible predictive criteria for tissue injury. Proc Natl Acad Sci U S A 90, 1102–1106 (1993).
Maung, M. T. et al. The molecular and cellular basis of copper dysregulation and its relationship with human pathologies. FASEB J 35, e21810 (2021).
Bar-Or, D. et al. Copper stimulates the synthesis and release of interleukin-8 in human endothelial cells: a possible early role in systemic inflammatory responses. Shock 20, 154–158 (2003).
Han, H. et al. The Hippo pathway kinases LATS1 and LATS2 attenuate cellular responses to heavy metals through phosphorylating MTF1. Nat Cell Biol 24, 74–87 (2022).
Malandrino, P. et al. Increased Thyroid Cancer Incidence in Volcanic Areas: A Role of Increased Heavy Metals in the Environment? Int J Mol Sci 21, E3425 (2020).
Slobodian, M. R., Petahtegoose, J. D., Wallis, A. L., Levesque, D. C. & Merritt, T. J. S. The Effects of Essential and Non-Essential Metal Toxicity in the Drosophila melanogaster Insect Model: A Review. Toxics 9, 269 (2021).
Tavera-Montañez, C. et al. The classic metal-sensing transcription factor MTF1 promotes myogenesis in response to copper. FASEB J 33, 14556–14574 (2019).
Schmidt, K., Steiner, K., Petrov, B., Georgiev, O. & Schaffner, W. Short-lived mammals (shrew, mouse) have a less robust metal-responsive transcription factor than humans and bats. Biometals 29, 423–432 (2016).
Chen, H. I. et al. An MTF1 binding site disrupted by a homozygous variant in the promoter of ATP7B likely causes Wilson Disease. Eur J Hum Genet 26, 1810–1818 (2018).
Kuo, M. T., Fu, S., Savaraj, N. & Chen, H. H. W. Role of the human high-affinity copper transporter in copper homeostasis regulation and cisplatin sensitivity in cancer chemotherapy. Cancer Res 72, 4616–4621 (2012).
Clifford, R. J., Maryon, E. B. & Kaplan, J. H. Dynamic internalization and recycling of a metal ion transporter: Cu homeostasis and CTR1, the human Cu⁺ uptake system. J Cell Sci 129, 1711–1721 (2016).
Is, S. et al. Transcription factor Sp1 plays an important role in the regulation of copper homeostasis in mammalian cells. Molecular pharmacology 74, (2008).
Pietikäinen, L. P., Rahman, M. T., Hiltunen, J. K., Dieckmann, C. L. & Kastaniotis, A. J. Genetic dissection of the mitochondrial lipoylation pathway in yeast. BMC Biol 19, 14 (2021).
Cronan, J. E. Progress in the Enzymology of the Mitochondrial Diseases of Lipoic Acid Requiring Enzymes. Front Genet 11, 510 (2020).
Soreze, Y. et al. Mutations in human lipoyltransferase gene LIPT1 cause a Leigh disease with secondary deficiency for pyruvate and alpha-ketoglutarate dehydrogenase. Orphanet J Rare Dis 8, 192 (2013).
Meldau, S. et al. Pitfalls of relying on genetic testing only to diagnose inherited metabolic disorders in non-western populations - 5 cases of pyruvate dehydrogenase deficiency from South Africa. Mol Genet Metab Rep 24, 100629 (2020).
Ni, M. et al. Functional Assessment of Lipoyltransferase-1 Deficiency in Cells, Mice, and Humans. Cell Rep 27, 1376-1386.e6 (2019).
Taché, V. et al. Lipoyltransferase 1 Gene Defect Resulting in Fatal Lactic Acidosis in Two Siblings. Case Rep Obstet Gynecol 2016, 6520148 (2016).
Zhang, J. et al. Inhibition of GLS suppresses proliferation and promotes apoptosis in prostate cancer. Biosci Rep 39, BSR20181826 (2019).
Lv, H. et al. Comprehensive Analysis of Cuproptosis-Related Genes in Immune Infiltration and Prognosis in Melanoma. Front Pharmacol 13, 930041 (2022).
Li, L., Li, L. & Sun, Q. High expression of cuproptosis-related SLC31A1 gene in relation to unfavorable outcome and deregulated immune cell infiltration in breast cancer: an analysis based on public databases. BMC Bioinformatics 23, 350 (2022).
Liu, H. & Tang, T. Pan-cancer genetic analysis of cuproptosis and copper metabolism-related gene set. Front Oncol 12, 952290 (2022).
Wang, X. et al. The copper-associated protein STEAP2 correlated with glioma prognosis and immune infiltration. Front Cell Neurosci 16, 944682 (2022).

No competing interests reported.

Supplymentarymaterial.zip

Download PDF

Version 1

posted

You are reading this latest preprint version

Diagnostic value of cuproptosis-related genes in sepsis: A novel model based on bioinformatics screening

Status:

Version 1

Abstract

Background:

Methods:

Results:

Conclusions:

Figures

INTRODUCTION

MATERIALS AND METHODS

Dataset Acquisition and Processing

Sepsis-related Differentially Expressed Genes

Differential Gene Function Enrichment (GO) and Pathway Enrichment (KEGG) Analysis

Gene Set Enrichment Analysis

Gene Set Variation Analysis

Sepsis Disease Subtype Identification

Weighted Correlation Network Analysis

Construction of the Diagnosis Model of Cuproptosis

Clinical Evaluation of CRDEGs

CeRNA Network Construction

Protein-protein Interaction Network Analysis

Immunoinfiltration Analysis

Statistical Analysis

RESULTS

Technology Roadmap:

Analysis of genes related to cuproptosis

Construction of associated disease subtypes of sepsis

Determination of co-expression modules in the merged dataset by WGCNA analysis

GO and KEGG enrichment analysis of cuproptosis-related genes

GSEA and GSVA of Sepsis Dataset

Construction of CRDEGs diagnostic model and its prognostic performance

Construction of protein-protein interaction (PPI) network and ceRNA network analysis

Immune Infiltration Analysis of Sepsis Dataset (CIBERSORTx)

DISCUSSION

LIMITATIONS

CONCLUSION

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1