Identification and validation of mitochondria-associated endoplasmic reticulum membrane-associated genes as diagnostic biomarkers for preeclampsia

doi:10.21203/rs.3.rs-4927724/v1

Download PDF

Article

Identification and validation of mitochondria-associated endoplasmic reticulum membrane-associated genes as diagnostic biomarkers for preeclampsia

https://doi.org/10.21203/rs.3.rs-4927724/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background

ER-mitochondria Ca²⁺ transfer abnormalities by MAMs and subsequent resulting in mitochondrial autophagy contribute to trophoblast apoptosis and may be involved in the pathogenesis of PE suggesting a crucial role of MAMs in PE development. However, detailed investigations into the specific mechanisms and roles of MAMs in PE remain limited.

Methods

This study began with a search for PE-related datasets and MAMs-related genes. Candidate genes identified and analyzed by differential expression analysis and WGCNA. ROC analysis was conducted to evaluate the capacity of biomarkers to differentiate between PE and controls. GSEA was employed to understand the biological functions and immunoinfiltration analysis was utilized for revealing role of the immmunological system of biomarkers in the advancement of PE. Biomarker-disease association predicting and constructing of molecular regulatory networks were implemented to explore the mechanisms by which biomarkers affect PE. Expression of hub genes was further verified by RT-qPCR.

Results

ABCD3, CAST and PAWR were considered as latent diagnostic biomarkers for PE, and the AUCs representing the ability to diagnose PE were 0.8-1.0.GSEA found spliceosome, proteasome and ubiquitin-mediated proteolysis were co-enriched by biomarkers. Immunoinfiltration analysis certified negative correlations between biomarkers and differentially infiltrated immune cells. Using the NetworkAnalyst database, 21, 9 and 20 TFs that might regulate the level of ABCD3, CAST and PAWR. RT-qPCR verified down-regulation of CAST and PAWR in the PE placenta, but ABCD3 validation results was the opposite.

Conclusion

CAST and PAWR function as latent MAMs-related biomarkers diagnosing and affecting PE. These findings provided insights to enhance the diagnosis and treatment of PE.

Health sciences/Diseases

Health sciences/Health care

Health sciences/Medical research

Health sciences/Pathogenesis

Preeclampsia

Mitochondria-associated endoplasmic reticulum membranes

Biomarkers

Machine learning

Immunoinfiltration analysis

Preeclampsia(PE) is defined as a multifactorial pregnancy disorder that affects approximately 2–8% of pregnancies. The main clinical manifestations are high blood pressure and proteinuria, even organ failure such as hemolysis, pulmonary edema, liver and kidney dysfunction, heart failure, blurred vision, fetal growth restriction and ultimately eclampsia in severe cases, probably occurring after 20 weeks of pregnancy¹. Previous studies suggest the main dogma in PE thought to be inadequate placental perfusion because of poor trophoblastic invasion causing release of inflammatory factors and vascular endothelial dysfunction leading to maternal syndrome ². It has lasting effects with increased risk for atherosclerotic cardiovascular disease, and changes are most pronounced in early- onset PE (EOPE) or PE with severe features³. However, some points are still missing, and there is no reliable method for early prediction and treatment of EOPE⁴. The identification of biomarkers represents a significant step forward in the improvement of maternal and fetal prognosis in cases of PE.

Mitochondria have multiple functions such as supplying energy, regulating the redox status and producing proteins encoded by an independent genome⁵. Mitochondrial fission/fusion is observed to be impaired in PE. This phenomenon has been linked to an increase in mitochondrial oxidative stress, which subsequently affects the placental development resulting in PE⁶. Endoplasmic reticulum (ER) stress is crucial for placental development, and is involved in angiogenesis, cell proliferation, autophagy and apoptosis, and ER stress induced unfolded protein response (UPR) activation as well as a lack of secreted factors, driven by glycosylphosphatidylinositol (GPI) pathway disruption, may contribute to the systemic inflammation, endothelial dysfunction, and oxidative stress characteristic of PE^7,8. Although mitochondria and the ER have been showed to play a key role in the development of PE, they are indispensable organelles that do not work independently, instead, there's something called the endoplasmic reticulum membrane(MAMs) comprised of mitofusin1/2 (Mfn1/2), glucose-regulated protein 75 (Grp75), dynamic-related protein 1(Drp1), and so on, which plays an important role in bridging them⁹. It affects mitochondrial quality control system, oxidative stress, and cell apoptosis by regulating downstream signaling pathways. At the same time, it plays a key role in the pathogenesis of many diseases, including cardiovascular disease and diabetes ¹⁰.ER-mitochondria Ca²⁺ transfer abnormalities by MAMs and subsequent mitochondrial function contribute to trophoblast apoptosis and may be related to the pathogenesis of PE¹¹. However, further mechanism research on MAMs in PE is still limited and needs to be studied further.

This study identified biomarkers related to MAMs in PE by differential expression analysis and weight gene co-expression network analysis (WGCNA) and two machine learning algorithms and gene expression analysis. Gene set enrichment analysis (GSEA) was employed to understand the biological functions of biomarkers. Immunoinfiltration analysis was utilized for revealing the role of immunity in the development of PE. Moreover, subcellular localization, biomarker-disease association predictions and construction of biomarker-related molecular regulatory networks were implemented to explore the latent mechanisms by which biomarkers affect PE.

A number of 694 DEGs and 6,099 key module genes were obtained

Totally 694 DEGs between PE and controls were acquired by differential expression analysis of the GSE190971 dataset, with 447 over-expressed and 247 down-regulated DEGs in the PE group (Fig. 1A-B). The AMS-RGS score significantly differed between PE and control groups, with PE patients exhibiting a higher MAMs-RGs score (p = 0.0235) (Fig. 1C). MAMs is involved in cellular processes such as Ca²⁺ signaling, inflammation, lipid metabolism, ER stress, autophagy, and apoptosis¹². So we inferred that higher MAMs-RGs scores might reflect multiple pathological changes such as cellular stress, oxidative stress, metabolic disorders, and inflammatory responses in patients with PE. Clustering analysis manifested that there were no obvious outliers in the GSE190971 dataset, as the sample clustering dendrograms presented in Fig. 1D, included all samples within the clusters. The best β of 18 was selected when the R² was 0.8 and mean connectivity neared zero (Fig. 1E). Then, five co-expressed modules were identified by merging genes into different modules (Fig. 1F). Of note, the MEyellow (cor = -0.450, p < 0.01), MEblue (cor = -0.80, p < 0.001) and MEturquoise (cor = -0.44, p < 0.01) modules were deemed as key modules as a result of their strong correlations with MAMs-RGs score (Fig. 1G). The genes in the three key modules were then combined, generating 6,099 key module genes.

Totally 43 candidate genes might be involved in transcriptional regulation, energy metabolism, and nutrient response

By intersecting 694 DEGs with 6,099 key module genes, 43 candidate genes were identified (Fig. 2A). Subsequent GO analysis revealed that these candidate genes were involved in 87 BPs, 14 CCs and 24 MFs. The top 5 items were shown in the bar chart. Obviously, BPs in which candidate genes were involved included regulation of GTPase activity, the cell growth and muscle tissue development. CC entries such as cell cortex, phosphatase complex and actin filament were also enriched by candidategenes. Additionally, candidate genes might also be involved in actin binding, transcription corepressor activity, GTPase activator activity and other MFs (Fig. 2B). Furthermore, in sum of 80 functional entries enriched by candidate genes were found by Metascape database. The pathway-pathway interaction network showed complex interactions between metabolic pathways, e.g. fatty acid metabolism, tryptophan metabolism and response to nutrient levels (Fig. 2C). In short, candidate genes might be involved in several important biological functions and pathways, including transcriptional regulation, regulation of cytoskeletal dynamics, energy metabolism, and nutrient response, which were essential for maintaining normal cellular function and physiological status. The PPI network containing 31 proteins encoded by candidate genes after removing the discrete protein was constructed. These proteins interacted with each other, forming 39 interaction pairs. Obviously, PDPR interacted more closely with other proteins, such as ACSS1, APMPA and FOXJ2. In addition, interaction pairs encompassed CAST-NEB and PAWR-TCHP (Fig. 2D).

ABCD3, CAST and PAWR were considered as latent diagnostic biomarkers for PE

Two machine learning (ML) algorithms were executed to screen for candidate biomarkers from candidate genes. These 43 candidate genes were downscaled by the LASSO regression model, and a total of 12 LASSO-featured genes were selected when the minimum lambda value was 0.009234441 (Fig. 3A). The SVM-RFE algorithm singled out 15 SVM-RFE-featured genes, at which time the model was most accurate (Fig. 3B). The same genes in the two models were selected, obtaining five candidate biomarkers (ABCD3, CAST, IFRD1, TXNRD1, and PAWR) (Fig. 3C). Gene expression analysis found that ABCD3, CAST and PAWR were down-regulated in PE, and their expression trends were consistent in GSE190971 and GSE24129 datasets (p < 0.01) (Fig. 3D). These three genes were considered as latent diagnostic biomarkers for PE due to their property of distinguishing PE and control samples. A receiver operating characteristic(ROC) analysis was further proceeded to appraise the diagnostic performance of biomarkers. The AUCs of ABCD3, CAST, and PAWR were 0.957, 0.973, and 0.884 in the GSE190971 dataset, singly. And they had an AUC of more than 0.9 in the GSE24129 datasets. This highlighted that all three biomarkers had the ability to precisely diagnose both PE and control samples (Fig. 3E-F).

Biomarkers might synergistically regulate cell metabolism

Gene set enrichment analysis (GSEA) was applied in this study to reveal the functions of biomarkers. In the first 10 pathways in which each biomarker was involved, biomarkers were co-enriched in spliceosome, proteasome and ubiquitin-mediated proteolysis. And biomarkers were positively correlated with these pathways (Fig. 4A-C). As demonstrated in the Sankey diagram, among the total pathways enriched by biomarkers, the top 5 identical pathways embraced endocytosis, ubiquitin mediated proteolysis, spliceosome, RNA degradation and proteasome (Fig. 4D). Taken together, the co-enrichment of biomarker pathways suggested that they might hold a synergistic part in the regulation of cellular metabolism, protein explication and gene expression regulation. A query of the GeneMANIA database revealed 20 genes maintaining similar biological functions to biomarkers. They were displayed in the GGI network. The majority of these genes exhibited a strong tendency for physical interactions with three biomarkers. Intriguingly, ABCD3, along with PEX19, ABCD1, ABCD2, and PEX3, might perform analogous functions, particularly in peroxisomal membrane transport and related cellular processes (Fig. 4E).

Biomarkers were negatively related to activated B cells, activated CD4 T cells and activated DCs

The immunoinfiltrating landscape of cells in PE was characterized. In the GSE190971 dataset, although activated B cells and activated CD4 T cells had negative ssGSEA scores in the samples, they had relatively high scores in PE (p < 0.05). Activated dendritic cells (DCs) scored significantly higher than the scores of activated B cells and activated CD4 T cells. Moreover, the score of activated DCs in the PE group was also markedly higher compared to the control samples (p < 0.001). This implied that B-cell activation, CD4 + T-cell activation as wellas DCs activation were significantly increased in PE (Fig. 5A-B). Activated DCs were positively correlated with activated B cells using Spearman's correlation analysis. (cor = 0.420, p < 0.05) and activated CD4 T cells (cor = 0.400, p < 0.05) (Fig. 5C). There were clear negative correlations between biomarkers and differentially infiltrated immune cells. Activated DCs showed the most negative correlations with ABCD3 (cor = -0.505, p < 0.001), CAST (cor = -0.636, p < 0.001) and PAWR (cor =-0.472, p < 0.01) (Fig. 5D).

Abrine, temozolomide and acetaminophen were co-acting compounds of biomarkers

A large number of compounds were screened using the CTD to identify potential drugs acting with biomarkers for the treatment of PE. A number of 149, 129 and 144 drugs interacting with ABCD3, CAST and PAWR, respectively were predicted in CTD. The drug-biomarker network displayed that a total of 34 drugs comprising abrine, temozolomide and acetaminophen were co-acting compounds of biomarkers (Fig. 5E).

Characterization of protein localization, structure, and disease relevance of biomarkers

Subcellular localisation unveiled that the proteins corresponding to ABCD3 and PAWR were predominantly found in the nucleus, whereas the proteins of CAST were more active in the cytoplasm (Fig. 6A). The human protein structures corresponding to biomarkers were further predicted. The structure of the human protein corresponding to ABCD3 was relatively complex compared to that of CAST and PAWR, due to its diverse disc region folding (Fig. 6B). Besides, the relevance of biomarkers with other pregnancy-related diseases was explored. Prenatal exposure delayed effects exhibited the highest interaction scores with each biomarker among the top five diseases concurrently associated with three biomarkers, whereas fetal death showed the lowest interaction scores with each biomarker (Fig. 6C).

Biomarker expression was regulated by multiple miRNAs and TFs

A total of 11, 30, and eight miRNAs acting with ABCD3, CAST and PAWR were identified in the miRWalk website. When clipExpNum was greater than 20, 11 and 3 lncRNAs targeting CAST and PAWR were obtained, severally. The lncRNA-miRNA-mRNA showing the interaction pairs formed by biomarkers, miRNAs and lncRNAs was synthesized. For instance, regulation of CAST and PAWR by hsa-miR-455-5p and hsa-miR-625-3p, respectively, might be simultaneously affected by MALAT1 (Fig. 6D). Using the NetworkAnalyst database, 21, 9 and 20 TFs that might regulate the level of ABCD3, CAST and PAWR were predicted, respectively. As demonstrated in the mRNA-TF network, ZNF610 might regulate the expression level of ABCD3 and CAST, and expressed content of ABCD3 and PAWR might be affected by FOXA3, CEBPG, MXD4 and ZNF76 (Fig. 6E).

Clinical validation of biomarkers in PE placenta

The placentas of the patients were tested for the expression of distinctive genes using real-time reverse transcriptase-polymerase chain reaction (RT-qPCR)The relative mRNA expression of of the biomarkers were down-regulation in PE patients than in control patients for PAWR and CAST (P = 0.02; P = 0.04).However,ABCD3 was up-regulation (P = 0.007) (Fig. 7A-C).

PE remains one of the leading causes of morbidity and mortality for both mothers and newborns worldwide and the only effective therapy remains termination. Much effort has been put into the understanding the pathology to find strategies for prevention, early detection, treatment, and prognosis¹³. It has been previously reported impaired mitochondrial function and persistent ER stress in the syncytiotrophoblast drive the onset and persistence of PE¹⁴. However, as a bridge between them, MAMs involved in important cellular functions are rarely studied in PE.

Therefore, in this study gene expression analysis found that ABCD3, CAST, and PAWR were down-regulated in PE. These biomarkers associated with MAMs were considered as latent precisely diagnostic biomarkers for PE, and they all had higher AUC values of more than 0.85. GSEA found the three biomarkers were co-enriched in spliceosome, proteasome and ubiquitin-mediated proteolysis. The immunoinfiltrating analysis showed negative correlations between biomarkers and activated B lymphocytes, CD4 T cells and DCs. Subcellular localisation unveiled that the proteins corresponding to ABCD3 and PAWR were predominantly found in the nucleus, whereas the proteins of CAST were more active in the cytoplasm.

PAWR is a tumor suppressor, also called PRKC, apoptosis Wilm’s tumor 1 (WT1) regulator, which selectively induces cell apoptosis in cancer and highly expressed in several tissues such as prostate, lung, and endometrium¹⁵. PAWR was significantly enriched and involved necrotic pathways that regulate immune cell infiltration with underlying mechanisms of PE pathophysiology¹⁶. Yuan Zhang et al showed the expression of circTRRAP was upregulated by hypoxia treatment to promote the PAWR level via inhibiting miR-370-3p, which result in the apoptotic, inflammatory, and oxidative injury in cardiomyocytes¹⁷. ATP-binding subfamily D member 3 (ABCD3) is a member of the ABC superfamily that is capable of inducing peroxisomal autophagy, affecting the fetus by affecting peroxisome proliferators ¹⁸. ABCD3 is regarded as the primary transporter within the placental barrier, and altered transporter expression at the barrier could contribute to fetal exposure to chemicals and/or accumulation of xenobiotics within trophoblasts and cause placental and fetal diseases such as PE¹⁹. But the ABCD3 validation results was the opposite in this study, and revalidation or switching to other methods was needed. Since ABCD3 and PAWR proteins are mainly found in the nucleus, they may be involved in regulating gene transcription, DNA repair, or other biological processes related to intracellular activity. CAST, an endogenous calpain-specific inhibitor, is more active in the cytoplasm, involved in many important biological functions including cell metabolism, protein synthesis and signal transduction. It is reported to maintain hepatocyte viability, CAST alters mitochondrial morphology towards an elongated shape viability after I/R²⁰. That reducing mitochondrial damage suppressed the proteasomal degradation of calpastatin, and alleviate cellular injury had been observed in neurodegenerative diseases²¹. CAST overexpressing in myocytes prevented from Ca²⁺ overload-induced myocyte cell death in arrhythmogenic cardiomyopathy²². It has been shown that the PGE2-mediated movement of EVTs is dependent on a complex of Ca2+-dependent proteases known as

calpains and calpastatin, as its inhibitor, blocked EVT migration in PE²³. In the pathogenesis of PE, trophoblast over-apoptosis plays an important role.These three biomarkers are related to cell apoptosis, accumulation of xenobiotics in the trophoblast cells, and migration of EVT. The preceding research findings suggest that these three biomarkers are associated with cell apoptosis, the buildup of xenobiotics within the trophoblast layer, and the migration process of EVT. This indicates that ABCD3, CAST, and PAWR may have important role for the diagnosis and prevention of PE.

The three biomarkers were co-enriched in many pathways,the top 5 ones embraced endocytosis, ubiquitin mediated proteolysis, spliceosome, RNA degradation, and proteasome. Through a single-cell RNA-seqencing (RNA-seq) study, Zhang T et al found that the proteasomes, spliceosomes, ribosomes, and mitochondria were aberrantly activated in the cytotrophoblasts of PE²⁴. The proteasomal pathway may be dysregulated in PE, as evidenced by a significant increase in circulating levels and lytic activity of the proteasomal pathway, which was partially explained by an increase in placental immunoproteasomal activity²⁵. For the representative of the classical pathway of ubiquitin-mediated proteolysis, PINK1-mediated mitophagy as ubiquitin-mediated proteolysis was abolished in the PE placenta, while the levels of pyroptosis were induced²⁶. How these three markers through which specific pathways in pathogenesis of PE need further clinical verification.

In addition, placental dysfunction caused by immune overactivation caused by maternal immune tolerance to the fetus is also one of the causes of PE. Immune cells such as regulatory T cells, macrophages, natural killer cells, and neutrophils are known to have major causative roles in the pathology of PE, but the contributions of additional immune cells such as B cells, inflammatory cytokines and anti-angiotensin II type 1 receptor autoantibodies are also now recognized²⁷. Based on immunoinfiltration analysis, there were clear negative correlations between biomarkers and differentially infiltrated immune cells. The three differentially infiltrated immune cells were separately activated B lymphocyte cells, CD4 + T lymphocyte cells and dendritic cells, which were significantly activated in PE. Evangeline Deer et al suggests that CD4 + T cells, independently of NK cells, induce mitochondrial dysfunction that contribute to hypertension in response to placental ischemia during pregnancy²⁸. This suggests that CD4 + T cells, not NK cells, may play a key role in the development of PE.

For biomarker-disease association analysis, the interaction between prenatal exposure delayed effects and biomarkers scored highest, which means these prenatal exposures may alter health outcomes by influencing the way biomarkers express themselves. Study shows that demonstrate higher rates of neuropsychiatric disorders such as increased autism spectrum diseases, attention deficit hyperactivity diseases in children of preeclamptic pregnancies and high rates of cognitive impairments²⁹. This interaction between them may indicate that certain populations are more sensitive or susceptible to prenatal exposure, which is closely related to factors such as an individual's genetic background, developmental stage, and degree of exposure.

In the study, the combination of machine learning and WGCNA has identified 3 biomarkers namely ABCD3, CAST and PAWR in PE associated with MAMs. These markers were also well validated by RT-PCR in PE placenta except ABCD3. The biological verification of biomarkers requires further in vitro and in vivo studies to clarify their exact roles in the etiology of PE. In addition, the clinical applicability and utility of the identified biomarkers need to be strictly tested in a larger patient cohort.

Data gathering

PE-related high throughput sequencing data for the GSE190971 (platform: GPL11154) and GSE24129 (platform: GPL6244) datasets were obtained from the Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/) database. The GSE190971 dataset encompassed placental tissue from 23 patients with preeclampsia and 18 nomal pregnancy patients. The GSE24129 dataset encompassed placental tissue samples from 8 patients with PE and 8 healthy controls. Totally 59 mitochondria-associated endoplasmic reticulum membranes-related genes (MAMs-RGs) were garnered from a published literature³⁰.

Differentially expressed gene (DEG) identification

The DEGs in the PE and controls of the GSE190971 dataset were pinpointed through differential expression analysis with the R package 'DESeq2' (Ver. 1.38.0) ³¹(|Log₂fold-change (FC)| > 0.5, adj. p < 0.05). A volcano plot demonstrating the DEGs was created through R package 'ggplot2' (Ver. 3.4.4)³². Top 15 DEGs which upregulated and downregulated expression patterns was shown by a heatmap, mapped by the R package 'ComplexHeatmap' (Ver. 2.14.0)³³.

WGCNA analysis

Utilizing the single-sample gene set enrichment analysis (ssGSEA) algorithm in R package 'GSVA' (Ver. 1.46.0)³⁴, the scores of the MAMs-RGs in each sample of the GSE190971 dataset were computed. The discrepancy in scores between PE and controls was assessed by the Wilcoxon rank sum test (p < 0.05). Based on the GSE190971 dataset, WGCNA was employed by R package 'WGCNA' (Ver. 1.71)³⁵ to ascertain key module genes associated with MAMs with MAMs-RGs score as a phenotypic trait. Clustering analysis across all GSE190971 samples was conducted to exclude outlier samples. The best soft threshold power (β) was selected when the scale-free index (R²) was 0.8 and mean connectivity approached zero. A scale-free network was then built based on β value with a minimum of 100 genes per module (mergeCutHeight = 0.3). Modules with a correlation |r| ≥ 0.4 and p < 0.05 to MAMs-RGs scores were considered as key modules and the genes within the key modules were aggregated to form the set of key module genes.

Acquisition and functional analysis of candidate genes

Candidate genes linked to MAMs and PE were pinpointed by intersecting DEGs with key module genes via R package 'VennDiagram' (Ver. 1.7.3)³⁶. To decipher the functional roles of these candidate genes, Gene Ontology (GO) analysis was conducted utilizing the R packages 'clusterProfiler' (Ver. 4.4.4) and 'org.Hs.eg.db' (Ver. 4.4.4)³⁷, focusing on biologic processes (BPs), molecular functions (MFs) and cellular constituents (CCs) (p < 0.05). (p < 0.05). In order to clarify the biological functions involved in candidate genes, the online platform Metascape (http://metascape.org) was accessed to identify functional entries significantly enriched by candidate genes. And a pathway-pathway interaction network was formed. To investigate protein-protein interaction (PPI) among candidate genes, the physical and functional interactions between proteins encoded by candidate genes were predicted using the search tool for identification of interacting genes (STRING, https://string-db.org) website (confidence = 0.15). A PPI network was established and through Cytoscape software (Ver. 3.8.2)³⁸.

Identification of biomarkers in PE

To obtain the fewest dimensional PE strong correlation features, fewest absolute shrinkage and selection operator (LASSO) regression analysis has been implemented using R package 'glmnet' (Ver. 4.1-4) ³⁹ performing 10-fold cross-validation, and the feature with the smallest misclassification error of the model were considered as LASSO-featured genes. SVM-RFE (Support Vector Machine Recursive Feature Extraction) was executed by 'caret' (Ver. 6.0–93)⁴⁰. After iterative feature elimination, the SVM-RFE model retained genes with the lowest generalization error. The overlapping genes of LASSO-featured genes and SVM-RFE-featured genes were deemed as candidate biomarkers. Furthermore, The difference expression levels of the target biomakers were estimated in the GSE190971 and GSE24129 datasets by the Wilcoxon rank-sum test (p < 0.05). Candidate biomarkers with variant expression levels in PE and controls as well as consistent expression trends in 2 datasets were further selected as biomarkers (p < 0.05). Moreover, to appraise the ability of biomarkers to distinguish PE and controls. In the GSE190971 and GSE24129 datasets, ROC (Receiver Operating Characteristic) curves corresponding to biomarkers were mapped via the R package 'pROC' (Ver. 1.18.4) ⁴¹. Biomarkers possessing area under curves (AUCs) of greater than 0.70 in two datasets were considered to have the ability to accurately discriminate between PE and control samples.

GSEA and analysis of gene-gene interaction (GGI) networks

For the sake of understanding the biological functions of biomarkers, correlation coefficients between each biomarker and all the genes in the GSE190971 dataset were calculated through Spearman's correlation analysis with the R package 'psych' (Ver. 3.4.4)⁴². A gene list for GSEA was generated subsequently by sorting the correlation coefficients in descending order. GSEA was accomplished through the R package 'clusterProfiler' with the c2.cp.kegg.v7.5.1. signatures derived the Molecular Signatures Database (MSigDB, https://www.gsea-msigdb.org/) as the reference gene set [|Normalized Enrichment Score (NES)| > 1, adj. p < 0.05]. In order to analyse the flow relationship of biomarkers with the enriched pathways, the pathways enriched by biomarkers were intersected, and the top 5 identical pathways were chosen to show the relationship with the biomarkers. The flow relationships were demonstrated by Sankey diagram, plotted via R package 'plotly' (Ver. 4.10.1) ⁴³. Biomarkers were entered into the GeneMANIA (http://genemania.org/) database to understand the biological function and role of genes by predicting and inferring other genes associated with the biomarkers. A gene-gene interaction (GGI) network was synthesized to show the relationships among these genes.

Immunoinfiltration analysis and drug prediction

The characterization of immune infiltration can reveal the role of the immune system in the occurrence and development of PE. The ssGSEA score of 28 kinds of immune cells⁴⁴ in the all samples of GSE190971 dataset was calculated using ssGSEA algorithm in the R package 'GSVA'. Differentially infiltrated immune cells between PE and controls were calculated through the Wilcoxon rank-sum test (p < 0.05). Moreover, the correlations between differentially infiltrated immune cells as well as between biomarkers and differentially infiltrated immune cells were explored by Correlation analysis by Spearman (p < 0.05). To develop latent drugs for the treatment of PE, drugs acting with biomarkers were predicted by Comparative Toxicogenomics Database (CTD, https://ctdbase.org/) and a drug-biomarker network was synthesized with Cytoscape software.

Subcellular localization, protein structure and biomarker-disease association predictions

Proteins must perform their biological functions in specific subcellular organelles, and for understanding the biological functions the proteins encoded by the biomarkers, subcellular localisation of the proteins corresponding to the biomarkers was performed. Specifically, the FASTA sequences of the biomarkers were first queried through the National Center for Biotechnology Information (NCBI, https://www.ncbi.nlm.nih.gov/) website, after which, the FASTA sequences of the biomarkers were submitted to the mRNALocater database (http://bio-bigdata.cn/mRNALocater/) in which the five subcellular localisation prediction scores for each biomarker were obtained and specific accurate localisation was based on the highest score. To further reveal the protein structures of the biomarkers, human protein identifiers are available in the UniProt database (https://www.uniprot.org/, and then these identifiers were entered into AlphaFold2 (https://alphafold.ebi.ac.uk/search) to predict the protein structures. The protein amino acid sequences were displayed. To analyse the role of biomarkers in other pregnancy-associated diseases, the relationships between other pregnancy-associated diseases and biomarkers were analysed through the Comparative Toxicogenomics Database (CTD, https://ctdbase.org/). The top 5 diseases were demonstrated on bar charts.

Construction of molecular regulatory networks related to biomarkers

To understand the interaction between biomarker expression and its upstream molecules and to reveal how upstream molecules affect biomarker expression, molecular regulatory networks related to biomarkers were developed. First, miRNAs targeting biomarkers were identified by miRWalk database (http://129.206.7.150/). lncRNAs targeting miRNAs were subsequently predicted by Starbase database (https://rnasysu.com/encori/). lncRNAs satisfying clipExpNum greater than 20, miRNAs and biomarkers were opted for construction of a lncRNA-miRNA-mRNA network. Simultaneously, transcription factors (TFs) regulating biomarker expression were predicted via NetAnalyst (https://www.networkanalyst.ca/) website. A mRNA-TF network was then developed. The lncRNA-miRNA-mRNA and mRNA-TF internetworking was displayed by Cytoscape hardware.

RNA extraction and quantitative real time PCR

Twenty-four placenta tissues included 12 control and 12 PE patients were collected in Hebei general Hospital. The study passed the hospital's ethics (2024-LW-139).RNA was extracted and purifed from placental tissue using RNAiso Easy(Takapa code: TCH020, Japan). PrimeScript™ RT reagent Kit with gDNA Eraser(Code NO. RR047ATaKaRa, Japan) was used to RNA-to-DNA reverse transcription and RT-PCR was conducted on an LightCycler® 480II(Roche, Switzerland). Appropriate primers sequences were listed in Table 1. The relative expression levels of the biomarker genes were calculated using GAPDH as an internal reference gene.

Table 1

Candidate gene primer sequence
Gene	Primer sequence
GAPDH	Forward	GCACCGTCAAGGCTGAGAAC
GAPDH	Reverse	TGGTGAAGACGCCAGTGGA
ABCD3	Forward	ACATCCTTGAGCGAGAAGGC
ABCD3	Reverse	GGTGATGCCAACCTTTCGAC
CAST	Forward	AGGAGATGGAGAGGATGAAGCAGAC
CAST	Reverse	AAAGTCCCAGGCAGCAATCAGAAG
PAWR	Forward	CTGCCGCAGAGTGCTTAGAT
PAWR	Reverse	CATCTTCTCGTTTCCGCTCT

Statistical analysis

Statistical analysis was performed using R software (Ver. 4.2.3). Differences between two groups were analyzed by the Wilcoxon rank-sum test or t test. P < 0.05 was considered a significant difference.

Data availability statement

All data generated or analysed during this study are included in this published article (and its Supplementary Information files).

Acknowledgements

The authors would like to thank the families for their participation, and the staff of the Central Laboratory,the Second Hospital of Hebei Medical and the staff at the Hebei General Hospital.

Authors' contributions

All authors contributed to the study conception and design. Data analysis was performed by J.J. Jiang, literature review by H.F. Kong and X.Y. Ma; The charts and diagram are completed by Y. Su and L.L. Zheng;The initial draft was written by C. Zhang; Subject and paper reviewed were by H. Xin. Authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Conflict of interest

The authors declare no competing interests.

Funding

There is no funding for this paper.

Ethics declarations

Approval was obtained from the ethics committee of Hebei General Hospital.The procedures used in this study adhere to the tenets of the Declaration of Helsinki. standards.

Informed consent Statement

Informed consent was obtained from the participants.

Burwick, R. M. & Rodriguez, M. H. Angiogenic Biomarkers in Preeclampsia. Obstetrics and gynecology 143, 515-523, doi:10.1097/aog.0000000000005532 (2024).
Pankiewicz, K., Fijałkowska, A., Issat, T. & Maciejewski, T. M. Insight into the Key Points of Preeclampsia Pathophysiology: Uterine Artery Remodeling and the Role of MicroRNAs. International journal of molecular sciences 22, doi:10.3390/ijms22063132 (2021).
Poon, L. C. et al. Hypertensive disorders of pregnancy and long-term cardiovascular health: FIGO Best Practice Advice. International journal of gynaecology and obstetrics: the official organ of the International Federation of Gynaecology and Obstetrics 160 Suppl 1, 22-34, doi:10.1002/ijgo.14540 (2023).
Zhang, Q. et al. Adrenomedullin has a pivotal role in trophoblast differentiation: A promising nanotechnology-based therapeutic target for early-onset preeclampsia. Science advances 9, eadi4777, doi:10.1126/sciadv.adi4777 (2023).
Song, N., Mei, S., Wang, X., Hu, G. & Lu, M. Focusing on mitochondria in the brain: from biology to therapeutics. Translational neurodegeneration 13, 23, doi:10.1186/s40035-024-00409-w (2024).
Nunes, P. R., Oliveira, P. F., Rebelo, I., Sandrim, V. C. & Alves, M. G. Relevance of real-time analyzers to determine mitochondrial quality in endothelial cells and oxidative stress in preeclampsia. Vascular pharmacology 155, 107372, doi:10.1016/j.vph.2024.107372 (2024).
Lorenzon-Ojea, A. R., Yung, H. W., Burton, G. J. & Bevilacqua, E. The potential contribution of stromal cell-derived factor 2 (SDF2) in endoplasmic reticulum stress response in severe preeclampsia and labor-onset. Biochimica et biophysica acta. Molecular basis of disease 1866, 165386, doi:10.1016/j.bbadis.2019.01.012 (2020).
Álvarez-Sánchez, A. et al. The GPI-anchor biosynthesis pathway is critical for syncytiotrophoblast differentiation and placental development. Cellular and molecular life sciences : CMLS 81, 246, doi:10.1007/s00018-024-05284-2 (2024).
Li, M. et al. Mitochondria-associated endoplasmic reticulum membranes tethering protein VAPB-PTPIP51 protects against ischemic stroke through inhibiting the activation of autophagy. CNS neuroscience & therapeutics 30, e14707, doi:10.1111/cns.14707 (2024).
Gao, P., Yang, W. & Sun, L. Mitochondria-Associated Endoplasmic Reticulum Membranes (MAMs) and Their Prospective Roles in Kidney Disease. Oxidative medicine and cellular longevity 2020, 3120539, doi:10.1155/2020/3120539 (2020).
Wang, A., Li, Z., Zhang, D., Chen, C. & Zhang, H. Excessive ER-phagy mediated by FAM134B contributes to trophoblast cell mitochondrial dysfunction in preeclampsia. Acta biochimica et biophysica Sinica, doi:10.3724/abbs.2024065 (2024).
Liu, Y. et al. Broadening horizons: the contribution of mitochondria-associated endoplasmic reticulum membrane (MAM) dysfunction in diabetic kidney disease. International journal of biological sciences 19, 4427-4441, doi:10.7150/ijbs.86608 (2023).
Abbas, R. A. et al. Preeclampsia: a review of early predictors. 3, 197-202 (2021).
Marín, R. et al. Oxidative stress and mitochondrial dysfunction in early-onset and late-onset preeclampsia. Biochimica et biophysica acta. Molecular basis of disease 1866, 165961, doi:10.1016/j.bbadis.2020.165961 (2020).
Subburayan, K. et al. Tumor suppressor Par-4 activates autophagy-dependent ferroptosis. Communications biology 7, 732, doi:10.1038/s42003-024-06430-z (2024).
He, L., Zhan, F., Lu, L., Zhang, X. & Wu, J. Role of necroptosis and immune infiltration in preeclampsia: novel insights from bioinformatics analyses. BMC pregnancy and childbirth 23, 495, doi:10.1186/s12884-023-05821-0 (2023).
Zhang, Y. et al. CircTRRAP Knockdown Has Cardioprotective Function in Cardiomyocytes via the Signal Regulation of miR-370-3p/PAWR Axis. Cardiovascular therapeutics 2022, 7125602, doi:10.1155/2022/7125602 (2022).
Liu, Y. et al. INTS7-ABCD3 Interaction Stimulates the Proliferation and Osteoblastic Differentiation of Mouse Bone Marrow Mesenchymal Stem Cells by Suppressing Oxidative Stress. Frontiers in physiology 12, 758607, doi:10.3389/fphys.2021.758607 (2021).
Kozlosky, D., Barrett, E. & Aleksunes, L. M. Regulation of Placental Efflux Transporters during Pregnancy Complications. Drug metabolism and disposition: the biological fate of chemicals 50, 1364-1375, doi:10.1124/dmd.121.000449 (2022).
Flores-Toro, J. et al. Critical Roles of Calpastatin in Ischemia/Reperfusion Injury in Aged Livers. Cells 10, doi:10.3390/cells10081863 (2021).
Hu, D. et al. Small-molecule suppression of calpastatin degradation reduces neuropathology in models of Huntington's disease. Nature communications 12, 5305, doi:10.1038/s41467-021-25651-y (2021).
Chelko, S. P. et al. Exercise triggers CAPN1-mediated AIF truncation, inducing myocyte cell death in arrhythmogenic cardiomyopathy. Science translational medicine 13, doi:10.1126/scitranslmed.abf0891 (2021).
Nicola, C., Timoshenko, A. V., Dixon, S. J., Lala, P. K. & Chakraborty, C. EP1 receptor-mediated migration of the first trimester human extravillous trophoblast: the role of intracellular calcium and calpain. The Journal of clinical endocrinology and metabolism 90, 4736-4746, doi:10.1210/jc.2005-0413 (2005).
Zhang, T. et al. Dissecting human trophoblast cell transcriptional heterogeneity in preeclampsia using single-cell RNA sequencing. Molecular genetics & genomic medicine 9, e1730, doi:10.1002/mgg3.1730 (2021).
Berryman, K. et al. Proteasome Levels and Activity in Pregnancies Complicated by Severe Preeclampsia and Hemolysis, Elevated Liver Enzymes, and Thrombocytopenia (HELLP) Syndrome. Hypertension (Dallas, Tex. : 1979) 73, 1308-1318, doi:10.1161/hypertensionaha.118.12437 (2019).
Sun, Y. et al. PINK1-mediated mitophagy induction protects against preeclampsia by decreasing ROS and trophoblast pyroptosis. Placenta 143, 1-11, doi:10.1016/j.placenta.2023.09.010 (2023).
Deer, E. et al. The role of immune cells and mediators in preeclampsia. Nature reviews. Nephrology 19, 257-270, doi:10.1038/s41581-022-00670-0 (2023).
Deer, E. et al. CD4+ T cells cause renal and placental mitochondrial oxidative stress as mechanisms of hypertension in response to placental ischemia. American journal of physiology. Renal physiology 320, F47-f54, doi:10.1152/ajprenal.00398.2020 (2021).
Gumusoglu, S. B., Chilukuri, A. S. S., Santillan, D. A., Santillan, M. K. & Stevens, H. E. Neurodevelopmental Outcomes of Prenatal Preeclampsia Exposure. Trends in neurosciences 43, 253-268, doi:10.1016/j.tins.2020.02.003 (2020).
Luo, F. et al. Down-regulation of the mitochondrial fusion protein Opa1/Mfn2 promotes cardiomyocyte hypertrophy in Su5416/hypoxia-induced pulmonary hypertension rats. Archives of biochemistry and biophysics 747, 109743, doi:10.1016/j.abb.2023.109743 (2023).
Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome biology 15, 550, doi:10.1186/s13059-014-0550-8 (2014).
Liu, S., Xie, X., Lei, H., Zou, B. & Xie, L. Identification of Key circRNAs/lncRNAs/miRNAs/mRNAs and Pathways in Preeclampsia Using Bioinformatics Analysis. Medical science monitor : international medical journal of experimental and clinical research 25, 1679-1693, doi:10.12659/msm.912801 (2019).
Li, J., Wang, H., Cao, F. & Cheng, Y. A bioinformatics analysis for diagnostic roles of the E2F family in esophageal cancer. Journal of gastrointestinal oncology 13, 2115-2131, doi:10.21037/jgo-22-855 (2022).
Hänzelmann, S., Castelo, R. & Guinney, J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC bioinformatics 14, 7, doi:10.1186/1471-2105-14-7 (2013).
Langfelder, P. & Horvath, S. WGCNA: an R package for weighted correlation network analysis. BMC bioinformatics 9, 559, doi:10.1186/1471-2105-9-559 (2008).
Pelikan, A., Herzel, H., Kramer, A. & Ananthasubramaniam, B. Venn diagram analysis overestimates the extent of circadian rhythm reprogramming. The FEBS journal 289, 6605-6621, doi:10.1111/febs.16095 (2022).
Yu, G., Wang, L. G., Han, Y. & He, Q. Y. clusterProfiler: an R package for comparing biological themes among gene clusters. Omics : a journal of integrative biology 16, 284-287, doi:10.1089/omi.2011.0118 (2012).
Chin, C. H. et al. cytoHubba: identifying hub objects and sub-networks from complex interactome. BMC systems biology 8 Suppl 4, S11, doi:10.1186/1752-0509-8-s4-s11 (2014).
Friedman, J., Hastie, T. & Tibshirani, R. Regularization Paths for Generalized Linear Models via Coordinate Descent. Journal of statistical software 33, 1-22 (2010).
López-Díaz JÓ, M. et al. Dummy regression to predict dry fiber in Agave lechuguilla Torr. in two large-scale bioclimatic regions in Mexico. PloS one 17, e0274641, doi:10.1371/journal.pone.0274641 (2022).
Robin, X. et al. pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC bioinformatics 12, 77, doi:10.1186/1471-2105-12-77 (2011).
Yang, H. et al. The Prognostic Role of Blood Inflammatory Biomarkers and EGFR Mutation Status in Stage IIIA/N2 Non-Small Cell Lung Cancer Patients Treated With Trimodality Therapy. Frontiers in oncology 11, 707041, doi:10.3389/fonc.2021.707041 (2021).
Mangiola, S., Doyle, M. A. & Papenfuss, A. T. Interfacing Seurat with the R tidy universe. Bioinformatics (Oxford, England) 37, 4100-4107, doi:10.1093/bioinformatics/btab404 (2021).
Charoentong, P. et al. Pan-cancer Immunogenomic Analyses Reveal Genotype-Immunophenotype Relationships and Predictors of Response to Checkpoint Blockade. Cell reports 18, 248-262, doi:10.1016/j.celrep.2016.12.019 (2017).

No competing interests reported.

SupplementaryTables.xlsx

Download PDF

Version 1

posted

You are reading this latest preprint version

Identification and validation of mitochondria-associated endoplasmic reticulum membrane-associated genes as diagnostic biomarkers for preeclampsia

Status:

Version 1

Abstract

Background

Methods

Results

Conclusion

Figures

Introduction

Results

Discussion

Materials and methods

Statistical analysis

Declarations

Data availability statement

Acknowledgements

Authors' contributions

Conflict of interest

Funding

Ethics declarations

Informed consent Statement

References

Additional Declarations

Supplementary Files

Status:

Version 1