Recurrent Miscarriage: A machine learning approach to uncover key genes and immune Infiltration

doi:10.21203/rs.3.rs-3240374/v1

Download PDF

Article

Recurrent Miscarriage: A machine learning approach to uncover key genes and immune Infiltration

https://doi.org/10.21203/rs.3.rs-3240374/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Objective: Recurrent miscarriage (RM), defined as the failure to maintain more than two clinical pregnancies beyond the 20th week of gestation, remains a complex pathological condition with unclear underlying mechanisms. This study aimed to elucidate potential biomarkers and explore the extent of immune infiltration in RM, to inform and facilitate effective clinical treatments.

Methods: Leveraging the GSE76862 and GSE26787 datasets from the Gene Expression Omnibus (GEO) database, we implemented Weighted Gene Co-expression Network Analysis (WGCNA) and Protein-Protein Interaction (PPI) networks to identify five key genes (F2, EGF, NGF, IL13, and FOXP3). These genes showed a robust correlation with RM. Receiver Operating Characteristic (ROC) curve analysis, coupled with validation from external datasets (GSE26787 and GSE22490), demonstrated the high diagnostic accuracy of these key genes for RM. RT-PCR was employed to validate the expression of these key genes in RM samples. We further evaluated immune cell infiltration in RM tissues using the CIBERSORT package and examined the relationship between the expression levels of the five key genes and immune cell infiltration. Furthermore, we interrogated correlations between key genes and immune factors from the TISIDB database to unravel the roles of these key genes in the immune mediation of RM. Finally, through Gene Set Variation Analysis (GSVA), Circos analysis, and GeneMANIA, we delved into the roles of the key genes, anticipated gene interactions, and gained insights into the molecular mechanisms driving RM.

Conclusion: Our findings underscore that the five identified key genes (F2, EGF, NGF, IL13, and FOXP3) have intricate links with RM and could play pivotal roles in deciphering the molecular mechanisms underlying RM.

Biological sciences/Computational biology and bioinformatics

Biological sciences/Genetics

Biological sciences/Immunology

recurrent miscarriage

key genes

WGCNA

immune infiltration

machine learning

Recurrent miscarriage (RM) is characterized by the occurrence of two or more consecutive pregnancy losses. This unfortunate complication affects 1–5% of women [1]. The causes of RM are multifactorial, encompassing autoimmunity, genetic or chromosomal abnormalities, anatomical abnormalities, endocrine or hormonal disruptions, and coagulation protein defects [2, 3]. Identifying the definitive etiology behind RM is challenging, and providing effective treatment remains complex. Different biomarkers, by reflecting disease-relevant biological activity and predicting disease progression, can provide valuable diagnostic insights. Consequently, identifying specific and sensitive biomarkers for RM can significantly aid early screening and diagnosis.

Immune cells such as natural killer cells (NK cells), regulatory T cells, dendritic cells, Th17 cells, and innate lymphoid cells (ILC) have been implicated in RM in recent studies [4–8]. Therefore, assessing immune cell infiltration and determining variations in the components of these infiltrating immune cells can help elucidate the underlying molecular processes of RM and identify new targets for immunotherapies.

Over recent decades, microarrays have rapidly developed as high-throughput platforms for gene expression analysis. Network topology, based on this method, has found extensive use in clinical and basic medical research [9]. The Gene Expression Omnibus (GEO) is a public, freely available database, offering vast amounts of information for bioinformatics exploration [10]. Weighted gene co-expression network analysis (WGCNA) involves the construction of a multi-gene co-expression model, allowing for the clustering of similarly expressed genes and the evaluation of associations between module characteristics and phenotypes [11]. WGCNA categorizes genes into modules or networks based on the similarity of gene expression profiles, facilitating the correlation of diverse information and datasets of RM.

In our study, we performed an integrated analysis of the GSE76862 and GSE26787 datasets, which were subsequently transformed into expression metrics via the surrogate variable analysis (SVA) algorithm [12]. The principal component analysis (PCA) of boundary scores across samples was represented in a two-dimensional PCA clustering graph, both before and after score normalization (Fig. 1A, B). The results suggested that the clustering of the two groups of samples was more prominent after normalization, indicating the reliability of sample sourcing. Using WGCNA, we analyzed RM and normal samples, constructed the co-expression network for RM, and identified the critical features of the gene module. This network facilitated the identification of key genes, and their diagnostic utility was assessed using the receiver operating characteristic (ROC) curve. The CIBERSORT algorithm was employed to explore immune cell infiltration in RM, and a correlation analysis was conducted between key genes and infiltrating immune cells and immune factors. Finally, the function of the key genes was annotated by GSVA, Circos analysis, and GeneMANIA analysis. Figure 2 illustrates the study workflow. Overall, we identified five potential biomarker genes and examined immune infiltration characteristics, offering fresh insights for future RM research.

Ethics Statement

The datasets used in this study were sourced from previously published literature, thereby ensuring that all necessary written informed consent was previously obtained.

Microarray Data

The keywords "RM" (title) and "Homo sapiens" (organism) were employed to search the GEO database (http://www.ncbi.nlm.nih.gov/geo/). Two datasets, GSE76862 and GSE26787, were downloaded. Subsequently, the surrogate variable analysis (SVA) algorithm, implemented in R, was used to transform the original GEO data into expression measures [12]. Two additional datasets, GSE26787 and GSE22490, were used for validation.

WGCNA Analysis

The weighted gene co-expression network analysis (WGCNA) is a biological application used to construct the co-expression network, identify co-expression gene modules, and analyze the association between gene networks and phenotypes. For further analysis, 5000 genes with the most significant variance were selected. The weighted adjacency matrix was transformed into the topological overlap measure (TOM) matrix to analyze network connectivity. Then, a hierarchical clustering method was used to construct the cluster tree structure of the TOM matrix. The soft threshold was set to 9.

Functional Enrichment Analysis and Protein–Protein Interaction (PPI) Networks

The differentially expressed genes (DEGs) from the modules closely associated with RM were subjected to enrichment analysis using the online tool, Metascape (http://metascape.org) [13]. In this study, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses were conducted [14]. A p-value of 0.01 and a minimum count of 3 were deemed significant. A PPI biological network was built using the STRING online database (v10.54), with a node association reliability score > 0.4 [15]. The Cytoscape software (v3.5.15) was used to visualize and analyze interactions and identify central genes in functional networks [16]. The five nodes with the highest correlation were selected as the key genes of the PPI network.

The Receiver Operating Characteristic (ROC) Curve Analysis

The area under curve (AUC) of the receiver operating characteristic (ROC) curve is widely regarded as the optimal indicator of diagnostic system performance [17]. We analyzed the gene expression data and corresponding clinical information using R software, and the efficacy of these methods was further verified in the validation datasets GSE26787 and GSE22490.

RT-PCR Analysis

Total RNA was extracted from tissues using TRIzol Reagent (Invitrogen, Carlsbad, CA, USA). A Quantscript Reverse Transcription Kit (TransGen Biotech, Beijing, China) was used with 1 µg of total RNA for DNA synthesis. Real-time polymerase chain reaction (PCR) was conducted on the 7500 Real-Time PCR System (Applied Biosystems, Waltham, MA, USA) using SYBR Green PCR Master Mix (TransGen Biotech, Beijing, China). Primer sequences are available in Supplementary material: Table S1.

Evaluation of Immune Cell Infiltration

CIBERSORT (https://cibersort.stanford.edu/) is a tool for deconvoluting immune cell subtype expression matrices based on the concept of linear support vector regression [18]. We imported gene expression matrix data into CIBERSORT and filtered samples with P < 0.05. The proportions of immune cells in each RM sample were depicted using the "barplot" function. We used the "corrplot" package to create a correlation heatmap displaying the correlation of 22 types of infiltrating immune cells, and the "ggplot2" package was used to create a violin plot visualizing differences in immune cells infiltration [19]. Each sample's deconvolution was also assigned a P-value.

Identification of Immune Factors and Key Genes

We used the TISIDB database (http://cis.hku.hk/TISIDB) to investigate the associations between key genes and immunomodulators and chemokines [20]. The resulting graph was created using the "Hmisc" package in R.

Correlation Analysis between Identified Genes and Infiltrating Immune Cells

We utilized Spearman's rank correlation analysis in R to investigate the relationship between the identified key genes and the amount of infiltrating immune cells. The associations were visualized using the “ggplot2” package, with a P-value < 0.05 indicating statistical significance.

Gene Set Variation Analysis (GSVA)

To understand more about the biological role of the key gene in the development of RM, we used the "GSVA" package in R to compare the GSVA between high-expression and low-expression groups [21, 22]. Each gene set associated with a pathway was trimmed to contain only one gene to reduce pathway overlap and redundancy. The hallmark gene set was used as a reference gene set, with a False Discovery Rate (FDR) < 0.05 acting as the cut-off standard.

Gene Interaction Analysis

A circular plot of the key genes was created using the Corrplot and Circlize packages [23]. GeneMANIA (http://www.genemania.org) is a user-friendly PPI network creation database used to visualize gene functional networks and evaluate gene function and interaction[16]. GeneMANIA was used to create a network of key genes, investigate their interactions, and predict the potential gene regulation network in RM.

Statistical Analysis

Statistical analysis was performed using R (https://www.r-project.org/) (version 3.6). All statistical tests were two-sided, and P < 0.05 was considered statistically significant.

Construction of WGCNA network and identification of key modules

In an attempt to explore the potential mechanisms that underpin the pathogenesis of RM, we performed a Weighted Gene Co-expression Network Analysis (WGCNA) on data from both RM and normal samples. The clustering trees of the RM and normal samples are displayed in Fig. 3A, demonstrating that all samples were included in the cluster under both conditions. To establish scale-free networks, we calculated the soft power threshold β using the function “SFT $power estimation”. This calculation informed our decision to use the power of = 9 as the soft threshold (refer to Figs. 3B, C for details).Interestingly, the module eigengenes (MEs) found within the green-yellow module displayed a significant correlation with the histological grading of RM (R2 = 0.74, P = 9e-04) as illustrated in Figs. 3D, E. This compelling evidence led us to focus on the green-yellow modules for subsequent investigation.Fig. 3F presents the connection between module connectivity and gene significance. Gene modules were identified using the Topological Overlap Measure (TOM) matrix, setting a foundation for the forthcoming steps in our research. This streamlined and logical methodology allows us to make solid inferences and contributes to the credibility of our findings.

Functional and pathway enrichment of gene module

To comprehend the biological roles and signal pathways involved in gene modules, the Metascape database was used for annotation and visualization. To evaluate specific genes, we performed Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) path analysis. Minimum overlap≥3 and P≤0.01 were considered statistically significant. Based on the green-yellow module, the percentage of related genes in the important pathway of the module was determined. In this study, the GO biological process and KEGG pathways were selected. Fig.4A displays the top five most abundant terms, including receptor regulator activity, cilium, anterior/posterior axis specification, neural crest cell migration, and ubiquitin ligase complex. The cytokine-cytokine receptor interaction and Ras signaling were the enriched in KEGG pathway. Additionally, we built a network of enriched related terms (Figure 4B).

Identification of key genes in functional modules

To further explore the gene constitution of the particular module, which was primarily related to the RAS pathway, a PPI system was established to discover key genes in the green-yellow module. key genes were found significantly related to the nodes in the module and considered functionally significant. We submitted the genes enriched in RAS-correlated pathways to the STRING database and retrieved 419 PPI nodes and 628 edges with a confidence threshold > 0.4 in the green-yellow modules. Thereafter, the top five nodes in the PPI network with the highest connectivity were used as central agents to be analyzed by Cytoscape. Resultantly, five key genes (F2, EGF, NGF, IL13, and FOXP3) were identified (Fig. 5).

Evaluation the diagnostic accuracy of key genes

In clinical diagnostics, the AUC value is used to evaluate and compare different biomarkers. We used a ROC curve method to examine the diagnostic accuracy of key genes and found that the AUC value of F2, EGF, NGF, IL13, and FOXP3 was 0.828, 0.906, 0.898, 0.922, 0.922, respectively (Fig.6). Hence, all key genes had high diagnostic accuracy for assessing the potential application of metabolites in the diagnosis of RM [24]. To further validate the accuracy of diagnostic biomarkers, we were further validated for differences in the validation set GSE26787 and GSE22490. All five key genes had significantly higher expression levels in RM than in controls (p < 0.05) in the set GSE26787 (Figure 7A). In the validation set GSE22490, IL13 and NGF were expressed at significantly higher levels in RM than in controls (p <0.05) (Figure 7B). To further validate the expression of key genes in RM, we conducted RT PCR detection on both RM and normal tissues and found that all five key genes were specifically overexpressed in RM( Figure 8).

Immune cell infiltration analysis results

We employed the CIBERSORT algorithm to assess immune infiltration in 22 immune cells in RM. The relative proportions of immune cells in each sample are displayed in Fig. 8A. According to the correlation heatmap of the 22 kinds of cells in RM, activated mast cells correlated positively with neutrophil cells (r=0.98) and regulatory T cells (r= 0.95); regulatory T cells correlated positively with neutrophil cells (r=0.92); and activated CD4 memory T cells correlated positively with naive B cells (r= 0.81). Resting mast cells showed a strikingly negative connection with neutrophils (r= -0.73) and NK cells (r= -0.67). (Fig. 9B). The violin plot revealed that M0 macrophage infiltration was considerably lower in RM samples (p 0.05). (Figure 9C).

Spearman correlation analysis between the five key genes and infiltrating immune cells

The Spearman correlation between of the correlations between 5 key gene expressions and various immune cells, we discovered distinct patterns. F2, EGF, IL13, and FOXP3 gene expressions were positively associated with NK cells but negatively associated with a suite of immune cells including immature dendritic cells, Th1 cells, Th2 cells, and co-stimulation T cells. Intriguingly, NGF expression exhibited an opposite trend, being positively correlated with an array of immune cells and negatively correlated with NK cells. These findings contribute to our understanding of the complex interplay between these key genes and immune cell infiltration, providing a potential pathway for further investigation in the pathogenesis of RM.

Identification the correlation of immune factors and the five key genes

Utilizing the TISIDB database, we examined the relationship between key genes and a variety of immune factors, including immunosuppressive and immunostimulatory factors, chemokines, and receptors. EGF, F2, FOXP3, and IL13 demonstrated significant positive correlations with certain chemokines (e.g., CXCL3, CXCL2, CXCL1, CCL28, CCL22, CCL19, and CCL17), immune inhibitors (e.g., VTCN1, IL10RB, and BTLA), and immunostimulators (e.g., TNFSF9, TNFSF13, MICB, and IL6). Conversely, they negatively correlated with chemokines such as CXCL14, immune inhibitors like CD96 and CD244, and several immunostimulators. Notably, NGF exhibited inverse correlations compared to the other key genes. This compelling evidence suggests these key genes play a crucial role in the immune-mediated mechanisms of RM.

GSVA Analysis for key genes

Gene set variation analysis (GSVA) was used to explore the signaling pathways implicated in the five key genes to determine the potential molecular mechanisms of RM. Figure 12 demonstrates that several hallmark gene-sets, including “CHOLESTEROL-HOMEOSTASIS,” “NOTCH SIGNALING,” “PEROXISOME,” “HEDGEHOG SIGNALING” “PANCREAS-BETA-CELLS” “P53 PATHWAY,” “PI3K AKT MTOR SIGNALING,” “WNT-BETA-CATENIN SIGNALING,” “FATTY-ACID-METABOLISM,” were observed in all high expression groups of F2, EGF, NGF, IL13 and FOXP3. Moreover, genes in low expression groups of F2, EGF, NGF, IL13 and FOXP3 were enriched in the “KRAS-SIGNALING” and “COMPLEMENT” “ESTROGEN-RESPONSE-LATE” “EPITHELIAL-MESENCHY MAL-TRANSITON” “INTERFERON-GAMMA-RESPONSE” “ALLOGRAFT-REJECTION” pathways. Remarkably, IL13 and FOXP3 showed similar molecular mechanisms, while NGF had the opposite signaling pathway with the other four key genes. Similarly, some immune-related pathways, such as the IL2 and interferon response pathways, were found to be enriched in the high expression groups of these key genes, indicating that they may play a significant role in the RM-associated immune response.

Co-expression analysis and the molecular function-based relationship between key genes.

Thereafter, the Corrplot and Circlize packages were used to investigate the genetic interactions of key genes. Co-expression of five key genes was assessed. As shown in the circular diagram (Figure 13A), EGF, IL13, F2, FOXP3 showed closely positive relationships, but the NGF had negative associations with each other. Different processes required to be investigated further. Next, we predicted a potential interaction of all identified key genes and constructed a gene-gene functional interaction network by using the GeneMANIA algorithm (Fig. 13B). Results showed a condensed interaction and pathway relations between key genes and their 20 neighbors. NGF and EGF mainly participated in the neurotrophin signaling pathway and neurotrophin TRK receptor signaling pathways. F2 and FOXP3 mainly participated in the regulation of cell activation and negative regulation of cell activation signaling pathways. Besides, FOXP3 also took part in the regulation of B cell activation. Overall, the results provided new insights into the interaction of key genes and their co-operators mediating RM oncogenesis and development.

RM often gives immense physical and emotional pain to pregnant women. The emergence and evolution of RM is a complicated process. However, almost 50% of all cases remain unexplained. Better biomarkers for predictive and prognostic molecules for RM are urgently required. Recent studies indicate that more than half of RM cases are due to immune dysfunction [25, 26]. The immunological response of pregnant women is critical to the maintenance of pregnancy. Increasing number of studies have confirmed that immune infiltration is related to the RM. However, few research have employed CIBERSORT to compare the expressed gene biomarkers linked with immune infiltration in RM and normal tissues. As a result, the purpose of this study was to identify possible diagnostic biomarkers and investigate the function of immune cell infiltration in RM.

To assess the connections between genes and modules, as well as the relationships between modules and clinical features, we first used WGCNA to build a gene co-expression network. Five key genes with high connectivity with RM (F2, EGF, NGF, IL13, and FOXP3) were identified.

F2 gene, also named Coagulation factor II encodes prothrombin protein [27]. Mutations in F2 expression always lead to various forms of thrombosis and roxithromycin [28]. Previous studies revealed that F2 gene mutation was the most common genetic factor of hereditary thrombophilias acting early or late during pregnancy [29, 30]. EGF belongs to the epidermal growth factor superfamily, which plays an essential role in the growth and the proliferation of numerous cell types [31]. Studies implicated that EGF receptors had a crucial role in the regulation of endometrial function during early pregnancy [32]. NGF is a neurotrophic factor that regulates target neuron neurite development, maintenance, proliferation, and survival [33, 34]. Evidence suggests that a local threshold of NGF expression is required to ensure the tolerance of healthy pregnant women, when the threshold is exceeded, it may lead to fetal rejection due to the aggravation of inflammation [35]. Interleukin-13 (IL13) was an effective growth-promoting cytokine, supporting the proliferation of a variety of hematopoietic cells [36]. It is involved in cell growth, differentiation, and apoptosis. Previous research showed that regulating the expression of IL-13 in spleen cells could reduce pregnancy loss [37]. FOXP3 was a member of the forkhead/pterygoid helix transcriptional regulation family [38]. Defects in this gene led to immune deficiency, multiple endocrine diseases, intestinal diseases, and X-linked syndrome (IPEX). The study showed that FOXP3 might affect cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) and regulate the occurrence and development of RM [39].

Based on the ROC curve analysis, we validated the reliable values of five key genes in RM patients in our study. Following that, the infiltration of immune cells in RM tissues was assessed using the CIBERSORT software. The findings showed particulars of 22 kinds of immune cell infiltration in RM. We documented that M0 macrophages cells infiltration was significantly lower in RM samples. The lower M0 macrophage cells infiltration in RM patients suggests that M0 macrophages immune cells play an essential role in RM progression. Furthermore, we performed the correlation analysis between the expression and infiltrating immune cells of five key genes. The results showed a significant correlation of the expression level of the key genes with most immune cells. For instance, previous research demonstrated that innate ILCs (lymphoid cells) are crucial in the maintenance of the maternal-fetal interface [40]. The quantity and viability of peripheral blood NK cells have also been associated to RM development [41]. Moreover, a recent study suggested women with RM may have immunological abnormalities such as disordered NK cells or aberrant NK cell subpopulations [42]. Elshal et al. discovered that myeloid dendritic cells (DCs) and CD86 + DCs were considerably higher in RM women. [43]. Tsao FY et al. found a close association between a decreased proportion of M2 macrophage with RM [44]. Furthermore, Treg cells have been shown to be critical in self-tolerance, allograft tolerance, and the formation of immunological tolerance at the maternal-fetal interface [45, 46], and a reduction in Treg cell levels may result in unfavourable pregnancy outcomes [47]. However, the specific mechanism of these immune cells in RM remains poorly understood. Therefore, more comparative exploration is still needed.

In addition, we performed GVSA to explore the significant metabolic pathways enrichment between low-expression and high-expression of key genes in RM. The results revealed a substantial difference between the high and low expression groups. An earlier study found that the metabolism of cholesterol and progesterone differed between primary trophoblast subtypes and may be disrupted in recurrent abortion [48]. Cuman et al. reported that the Notch signalling pathway is involved in the control of embryonic and placental development[49], and that abnormalities in the notch receptor-ligand system impacted placentation. The current study found that decreased hedgehog signal transmission in the uterine stroma leads in delayed implantation and embryonic loss [50]. P53 signaling pathway polymorphisms have also been reportedly associated with recurrent pregnancy loss, the high level of p53 expression in placental villi in RM patients. This may result in cell apoptosis, leading to recurrent spontaneous abortion [51]. Moreover, the PI3K/Akt pathway, an intracellular signalling mechanism, is important in the cell cycle, proliferation, apoptosis, and activation of the PI3K/Akt/mTOR pathway in RM patients [52]. Previous research has suggested that abnormal expression of the wnt/-catenin signalling pathway may be linked to RM pathogenesis, and that downregulation of the wnt/-catenin signalling pathway may be associated to the pathogenesis of recurrent spontaneous abortion[53]. Epithelial-mesenchymal transition (EMT) is a critical step in embryonic development. EMT can induce cell migration and invasion; inhibit cell apoptosis and senescence; promote immunosuppression, and induce stress resistance and stem cell characteristics. Interferon-γ was also found to be significantly associated with embryotoxicity [54]. In addition, inflammation response was shown to be a necessary process for early pregnancy, followed by the maintenance of anti-inflammatory response during pregnancy and the recovery of the inflammatory response during delivery or termination of pregnancy [55]. Modzelewski et al. demonstrated that rejection of fetal “semi-allograft” by maternal allograft rejection pathway could lead to pregnancy loss [56]; Allograft rejection pathway was found to be significantly screened out in RM patients. Generally, the enrichment analysis of the pivotal gene had a significant enrichment effect in the RM correlation process. Our study showed that IL13 and FOXP3 expression have the same metabolic pathway, while NGF has the opposite signaling pathway with other key genes in the low and high expression groups, these results supported the results of co-expression analysis. Additionally, GeneMANIA analysis revealed a condensed interaction and pathway relations between key genes and their 20 neighbors. Promisingly, we discovered that F2 and FOXP3 were predominantly involved in the control of immune-related signalling pathways.

The current study has few limitations. First, the number of patients was too small, so it is suggested to expand the sample set to verify the expression level of these key genes in future research. To solved this issue, we added external datasets to validate the efficacy of key genes. Second, the molecular mechanism of this study was based on network topology, and further research on molecular experiments of these key genes in RM in vivo and in vitro is required.

Overall, this study aimed to screen RM-related key genes from the GEO database by WGCNA and detected the expression of five functional genes (F2, EGF, NGF, IL13, FOXP3). The CIBERSORT algorithm was performed to understand the infiltration of immune cells in RM. Furthermore, GSVA was performed to reveal significant metabolic pathway enrichment between different expressions of different key genes in RM. Our results also showed the participation of five key genes primarily in RM regulation. Next, we predicted a potential interaction of all identified key genes and constructed a gene-gene functional interaction network by using the GeneMANIA.

Summing up, Our findings gave unique insights into the key genes that mediate RM. The special biological functions of these key genes in the pathogenesis of RM need further in-depth study.

Funding

This work was supported by the Municipal Health Commission of Nantong (Grant No: QA2021047).

Competing Interests

The authors declare that they have no conflict of interest.

Author Contributions

Author contributed to the study conception and design. Material preparation, data collection and analysis were performed by Mengsi Lin. The first draft of the manuscript was written by Mengsi Lin commented on previous versions of the manuscript. Author read and approved the final manuscript.

Ethics approval

Not applicable.

Consent to participate

Not applicable.

Consent to publish

Not applicable.

Rai R, Regan L (2006) Recurrent miscarriage. Lancet 368(9535):601-611. https://doi.org/10.1016/S0140-6736(06)69204-0
Homer HA (2019) Modern management of recurrent miscarriage. Aust N Z J Obstet Gynaecol 59(1):36-44. https://doi.org/10.1111/ajo.12920
Elkarhat Z, Kindil Z, Zarouf L, Razoki L, Aboulfaraj J, Elbakay C, Nassereddine S, Nasser B, Barakat A, Rouba H (2019) Chromosomal abnormalities in couples with recurrent spontaneous miscarriage: a 21-year retrospective study, a report of a novel insertion, and a literature review. J Assist Reprod Genet 36(3):499-507. https://doi.org/10.1007/s10815-018-1373-4
Zhu L, Aly M, Kuon RJ, Toth B, Wang H, Karakizlis H, Weimer R, Morath C, Ibrahim E, Ekpoom N et al (2019) Patients with idiopathic recurrent miscarriage have abnormally high TGFss+ blood NK, NKT and T cells in the presence of abnormally low TGFss plasma levels. BMC Immunol 20(1):10. https://doi.org/10.1186/s12865-019-0290-3
Liu S, Wei H, Li Y, Huang C, Lian R, Xu J, Chen L, Zeng Y (2018) Downregulation of ILT4(+) dendritic cells in recurrent miscarriage and recurrent implantation failure. Am J Reprod Immunol 80(4):e12998. https://doi.org/10.1111/aji.12998
Liu C, Wang XZ, Sun XB (2013) Assessment of sperm antigen specific T regulatory cells in women with recurrent miscarriage. Early Hum Dev 89(2):95-100. https://doi.org/10.1016/j.earlhumdev.2012.08.003
Wang WJ, Hao CF, Lin QD (2011) Dysregulation of macrophage activation by decidual regulatory T cells in unexplained recurrent miscarriage patients. J Reprod Immunol 92(1-2):97-102. https://doi.org/10.1016/j.jri.2011.08.004
Vomstein K, Feil K, Strobel L, Aulitzky A, Hofer-Tollinger S, Kuon RJ, Toth B (2021) Immunological Risk Factors in Recurrent Pregnancy Loss: Guidelines Versus Current State of the Art. J Clin Med 10(4). https://doi.org/10.3390/jcm10040869
Grassmann F (2019) Conduct and Quality Control of Differential Gene Expression Analysis Using High-Throughput Transcriptome Sequencing (RNASeq). Methods Mol Biol 1834:29-43. https://doi.org/10.1007/978-1-4939-8669-9_2
Barrett T, Troup DB, Wilhite SE, Ledoux P, Rudnev D, Evangelista C, Kim IF, Soboleva A, Tomashevsky M, Edgar R (2007) NCBI GEO: mining tens of millions of expression profiles--database and tools update. Nucleic Acids Res 35(Database issue):D760-765. https://doi.org/10.1093/nar/gkl887
Langfelder P, Horvath S (2008) WGCNA: an R package for weighted correlation network analysis. BMC bioinformatics 9:559. https://doi.org/10.1186/1471-2105-9-559
Chakraborty S, Datta S, Datta S (2012) Surrogate variable analysis using partial least squares (SVA-PLS) in gene expression studies. Bioinformatics 28(6):799-806. https://doi.org/10.1093/bioinformatics/bts022
Zhou Y, Zhou B, Pache L, Chang M, Khodabakhshi AH, Tanaseichuk O, Benner C, Chanda SK (2019) Metascape provides a biologist-oriented resource for the analysis of systems-level datasets. Nat Commun 10(1):1523. https://doi.org/10.1038/s41467-019-09234-6
Zhang T, Jiang M, Chen L, Niu B, Cai Y (2013) Prediction of gene phenotypes based on GO and KEGG pathway enrichment scores. Biomed Res Int 2013:870795. https://doi.org/10.1155/2013/870795
Szklarczyk D, Morris JH, Cook H, Kuhn M, Wyder S, Simonovic M, Santos A, Doncheva NT, Roth A, Bork P et al (2017) The STRING database in 2017: quality-controlled protein-protein association networks, made broadly accessible. Nucleic Acids Res 45(D1):D362-D368. https://doi.org/10.1093/nar/gkw937
Montojo J, Zuberi K, Rodriguez H, Bader GD, Morris Q (2014) GeneMANIA: Fast gene network construction and function prediction for Cytoscape. F1000Res 3:153. https://doi.org/10.12688/f1000research.4572.1
Huguet J, Castineiras MJ, Fuentes-Arderiu X (1993) Diagnostic accuracy evaluation using ROC curve analysis. Scand J Clin Lab Invest 53(7):693-699. https://doi.org/10.3109/00365519309092573
Wang L, Yang Z, Cao Y (2020) Regulatory T cell and activated natural killer cell infiltration in hepatocellular carcinoma: immune cell profiling using the CIBERSORT. Ann Transl Med 8(22):1483. https://doi.org/10.21037/atm-20-5830
Maag JLV (2018) gganatogram: An R package for modular visualisation of anatograms and tissues based on ggplot2. F1000Res 7:1576. https://doi.org/10.12688/f1000research.16409.2
Ru B, Wong CN, Tong Y, Zhong JY, Zhong SSW, Wu WC, Chu KC, Wong CY, Lau CY, Chen I et al (2019) TISIDB: an integrated repository portal for tumor-immune system interactions. Bioinformatics 35(20):4200-4202. https://doi.org/10.1093/bioinformatics/btz210
Ferreira MR, Santos GA, Biagi CA, Silva Junior WA, Zambuzzi WF (2020) GSVA score reveals molecular signatures from transcriptomes for biomaterials comparison. J Biomed Mater Res A. https://doi.org/10.1002/jbm.a.37090
Hanzelmann S, Castelo R, Guinney J (2013) GSVA: gene set variation analysis for microarray and RNA-seq data. BMC bioinformatics 14:7. https://doi.org/10.1186/1471-2105-14-7
Gu Z, Gu L, Eils R, Schlesner M, Brors B (2014) circlize Implements and enhances circular visualization in R. Bioinformatics 30(19):2811-2812. https://doi.org/10.1093/bioinformatics/btu393
Akobeng AK (2007) Understanding diagnostic tests 3: Receiver operating characteristic curves. Acta Paediatr 96(5):644-647. https://doi.org/10.1111/j.1651-2227.2006.00178.x
Li D, Zheng L, Zhao D, Xu Y, Wang Y (2021) The Role of Immune Cells in Recurrent Spontaneous Abortion. Reprod Sci. https://doi.org/10.1007/s43032-021-00599-y
Lin QD, Qiu LH (2010) Pathogenesis, diagnosis, and treatment of recurrent spontaneous abortion with immune type. Front Med China 4(3):275-279. https://doi.org/10.1007/s11684-010-0101-y
Demirci FY, Dressen AS, Kammerer CM, Barmada MM, Kao AH, Ramsey-Goldman R, Manzi S, Kamboh MI (2011) Functional polymorphisms of the coagulation factor II gene (F2) and susceptibility to systemic lupus erythematosus. J Rheumatol 38(4):652-657. https://doi.org/10.3899/jrheum.100728
Tang Y, Zhang L, Xie W, Jin J, Luo Y, Deng M, Liu Z, Pan HW, Zhang Y, Zheng Z et al (2020) A Novel Heterozygous Variant in F2 Gene in a Chinese Patient With Coronary Thrombosis and Acute Myocardial Infarction Leads to Antithrombin Resistance. Front Genet 11:184. https://doi.org/10.3389/fgene.2020.00184
Mulder R, Lisman T, Meijers JCM, Huntington JA, Mulder AB, Meijer K (2020) Linkage analysis combined with whole-exome sequencing identifies a novel prothrombin (F2) gene mutation in a Dutch Caucasian family with unexplained thrombosis. Haematologica 105(7):e370-e372. https://doi.org/10.3324/haematol.2019.232504
Xu Z, Zhang Y, Liu W, Liu Y, Su Y, Xing Q, He X, Wei Z, Cao Y, Xiang H (2018) Polymorphisms of F2, PROC, PROZ, and F13A1 Genes are Associated With Recurrent Spontaneous Abortion in Chinese Han Women. Clin Appl Thromb Hemost 24(6):894-900. https://doi.org/10.1177/1076029617750487
Eden ER, Huang F, Sorkin A, Futter CE (2012) The role of EGF receptor ubiquitination in regulating its intracellular traffic. Traffic 13(2):329-337. https://doi.org/10.1111/j.1600-0854.2011.01305.x
Large MJ, Wetendorf M, Lanz RB, Hartig SM, Creighton CJ, Mancini MA, Kovanci E, Lee KF, Threadgill DW, Lydon JP et al (2014) The epidermal growth factor receptor critically regulates endometrial function during early pregnancy. PLoS Genet 10(6):e1004451. https://doi.org/10.1371/journal.pgen.1004451
Ivanov AD (2014) [The role of NGF and BDNF in mature brain activity regulation]. Zh Vyssh Nerv Deiat Im I P Pavlova 64(2):137-146
Chaves RN, Alves AM, Lima LF, Matos HM, Rodrigues AP, Figueiredo JR (2013) Role of nerve growth factor (NGF) and its receptors in folliculogenesis. Zygote 21(2):187-197. https://doi.org/10.1017/S0967199412000111
Frank P, Barrientos G, Tirado-Gonzalez I, Cohen M, Moschansky P, Peters EM, Klapp BF, Rose M, Tometten M, Blois SM (2014) Balanced levels of nerve growth factor are required for normal pregnancy progression. Reproduction 148(2):179-189. https://doi.org/10.1530/REP-14-0112
Jin R, Hu S, Liu X, Guan R, Lu L, Lin R (2019) Intranasal instillation of miR410 targeting IL4/IL13 attenuates airway inflammation in OVAinduced asthmatic mice. Mol Med Rep 19(2):895-900. https://doi.org/10.3892/mmr.2018.9703
Williams TJ, Jones CA, Miles EA, Warner JO, Warner JA (2000) Fetal and neonatal IL-13 production during pregnancy and at birth and subsequent development of atopic symptoms. J Allergy Clin Immunol 105(5):951-959. https://doi.org/10.1067/mai.2000.106211
Jia H, Qi H, Gong Z, Yang S, Ren J, Liu Y, Li MY, Chen GG (2019) The expression of FOXP3 and its role in human cancers. Biochim Biophys Acta Rev Cancer 1871(1):170-178. https://doi.org/10.1016/j.bbcan.2018.12.004
Fan Q, Zhang J, Cui Y, Wang C, Xie Y, Wang Q, Wu L (2018) The synergic effects of CTLA-4/Foxp3-related genotypes and chromosomal aberrations on the risk of recurrent spontaneous abortion among a Chinese Han population. J Hum Genet 63(5):579-587. https://doi.org/10.1038/s10038-018-0414-2
Chang RQ, Zhou WJ, Li DJ, Li MQ (2020) Innate Lymphoid Cells at the Maternal-Fetal Interface in Human Pregnancy. Int J Biol Sci 16(6):957-969. https://doi.org/10.7150/ijbs.38264
Hosseini S, Zarnani AH, Asgarian-Omran H, Vahedian-Dargahi Z, Eshraghian MR, Akbarzadeh-Pasha Z, Arefi S, Jeddi-Tehrani M, Shokri F (2014) Comparative analysis of NK cell subsets in menstrual and peripheral blood of patients with unexplained recurrent spontaneous abortion and fertile subjects. J Reprod Immunol 103:9-17. https://doi.org/10.1016/j.jri.2014.03.002
Yang SL, Tan HX, Niu TT, Li DJ, Wang HY, Li MQ (2021) Kynurenine promotes the cytotoxicity of NK cells through aryl hydrocarbon receptor in early pregnancy. J Reprod Immunol 143:103270. https://doi.org/10.1016/j.jri.2020.103270
Elshal MF, Aldahlawi AM, Saadah OI, McCoy JP (2015) Reduced Dendritic Cells Expressing CD200R1 in Children with Inflammatory Bowel Disease: Correlation with Th17 and Regulatory T Cells. Int J Mol Sci 16(12):28998-29010. https://doi.org/10.3390/ijms161226143
Tao Y, Li YH, Piao HL, Zhou WJ, Zhang D, Fu Q, Wang SC, Li DJ, Du MR (2015) CD56(bright)CD25+ NK cells are preferentially recruited to the maternal/fetal interface in early human pregnancy. Cell Mol Immunol 12(1):77-86. https://doi.org/10.1038/cmi.2014.26
Li L, Tu J, Jiang Y, Zhou J, Schust DJ (2017) Regulatory T cells decrease invariant natural killer T cell-mediated pregnancy loss in mice. Mucosal Immunol 10(3):613-623. https://doi.org/10.1038/mi.2016.84
Afzali B, Mitchell P, Lechler RI, John S, Lombardi G (2010) Translational mini-review series on Th17 cells: induction of interleukin-17 production by regulatory T cells. Clin Exp Immunol 159(2):120-130. https://doi.org/10.1111/j.1365-2249.2009.04038.x
Care AS, Bourque SL, Morton JS, Hjartarson EP, Robertson SA, Davidge ST (2018) Reduction in Regulatory T Cells in Early Pregnancy Causes Uterine Artery Dysfunction in Mice. Hypertension 72(1):177-187. https://doi.org/10.1161/HYPERTENSIONAHA.118.10858
Vondra S, Kunihs V, Eberhart T, Eigner K, Bauer R, Haslinger P, Haider S, Windsperger K, Klambauer G, Schutz B et al (2019) Metabolism of cholesterol and progesterone is differentially regulated in primary trophoblastic subtypes and might be disturbed in recurrent miscarriages. J Lipid Res 60(11):1922-1934. https://doi.org/10.1194/jlr.P093427
Cuman C, Menkhorst E, Winship A, Van Sinderen M, Osianlis T, Rombauts LJ, Dimitriadis E (2014) Fetal-maternal communication: the role of Notch signalling in embryo implantation. Reproduction 147(3):R75-86. https://doi.org/10.1530/REP-13-0474
Harman RM, Cowan RG, Ren Y, Quirk SM (2011) Reduced signaling through the hedgehog pathway in the uterine stroma causes deferred implantation and embryonic loss. Reproduction 141(5):665-674. https://doi.org/10.1530/REP-10-0468
Fraga LR, Dutra CG, Boquett JA, Vianna FS, Goncalves RO, Paskulin DD, Costa OL, Ashton-Prolla P, Sanseverino MT, Schuler-Faccini L (2014) p53 signaling pathway polymorphisms associated to recurrent pregnancy loss. Mol Biol Rep 41(3):1871-1877. https://doi.org/10.1007/s11033-014-3036-6
Feng X, Jiang S, Leung W, Wang L, Gober HJ, Chen L, Zhang Y, Wang L (2020) BuShen HuoXue Decoction Promotes Decidual Stromal Cell Proliferation via the PI3K/AKT Pathway in Unexplained Recurrent Spontaneous Abortion. Evid Based Complement Alternat Med 2020:6868470. https://doi.org/10.1155/2020/6868470
Li N, Li S, Wang Y, Wang J, Wang K, Liu X, Li Y, Liu J (2017) Decreased expression of WNT2 in villi of unexplained recurrent spontaneous abortion patients may cause trophoblast cell dysfunction via downregulated Wnt/beta-catenin signaling pathway. Cell Biol Int 41(8):898-907. https://doi.org/10.1002/cbin.10807
Kim DH, Xing T, Yang Z, Dudek R, Lu Q, Chen YH (2017) Epithelial Mesenchymal Transition in Embryonic Development, Tissue Repair and Cancer: A Comprehensive Overview. J Clin Med 7(1). https://doi.org/10.3390/jcm7010001
Straubinger K, Paul S, Prazeres da Costa O, Ritter M, Buch T, Busch DH, Layland LE, Prazeres da Costa CU (2014) Maternal immune response to helminth infection during pregnancy determines offspring susceptibility to allergic airway inflammation. J Allergy Clin Immunol 134(6):1271-1279 e1210. https://doi.org/10.1016/j.jaci.2014.05.034
Modzelewski J, Kajdy A, Rabijewski M (2019) The role of lymphocytes in fetal development and recurrent pregnancy loss. Ginekol Pol 90(2):109-113. https://doi.org/10.5603/GP.2019.0019

No competing interests reported.

SupplementaryTable1.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

Recurrent Miscarriage: A machine learning approach to uncover key genes and immune Infiltration

Status:

Version 1

Abstract

Figures

Introduction

Materials and Methods

Ethics Statement

Microarray Data

WGCNA Analysis

Functional Enrichment Analysis and Protein–Protein Interaction (PPI) Networks

The Receiver Operating Characteristic (ROC) Curve Analysis

RT-PCR Analysis

Evaluation of Immune Cell Infiltration

Identification of Immune Factors and Key Genes

Correlation Analysis between Identified Genes and Infiltrating Immune Cells

Gene Set Variation Analysis (GSVA)

Gene Interaction Analysis

Statistical Analysis

Results

Discussion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1