Necroptosis in recurrent implantation failure: A bioinformatics analysis of key genes and therapeutic targets.

This study, through an integrated bioinformatics analysis, has elucidated a possible connection between necroptosis and RIF, with the identification of key genes and potential therapeutic targets. Although the data in this study are dependent on the GEO database and require further validation through in vitro and in vivo experimental studies, they provide reliable preliminary findings in exploring the molecular mechanisms of RIF. These findings represent a major step towards understanding the intricate molecular landscape of RIF, ultimately opening avenues for novel treatment strategies.

Recurrent implantation failure (RIF), a major challenge in assisted reproductive technology (ART), instills considerable emotional distress in couples aspiring to conceive. RIF is generally defined as the failure to achieve a clinical pregnancy after the transfer of at least 4 good-quality embryos in a minimum of 3 fresh or frozen cycles in a woman under the age of 40. [ 1 , 2 ] Despite technological advancements and increased understanding, the etiology of RIF remains complex and elusive, encompassing factors such as endometrial receptivity, embryonic competency, and maternal systemic factors. [ 1 ] Recently, necroptosis, a distinct form of programmed cell death separate from classical apoptosis, has garnered substantial attention in disease pathogenesis. Unlike apoptosis, which is a caspase-dependent and often immunologically silent process, necroptosis is caspase-independent and typically associated with inflammation due to the release of cellular contents. [ 3 , 4 ] This process is regulated by specific genes such as receptor-interacting protein kinase 1 ( RIPK1 ), RIPK3 , and mixed lineage kinase domain-like pseudokinase ( MLKL ). [ 4 ] Emerging research has highlighted potential roles for necroptosis across diseases such as neurodegenerative diseases, cancer, and more recently, conditions related to female reproduction including ovarian aging, follicular atresia, and endometriosis. [ 5 , 6 ] The potential involvement of necroptosis in the pathogenesis of RIF has not been thoroughly explored. However, some studies suggest a correlation between RIF and necroptosis. For example, sirtuin 1 ( SIRT1 ), a gene that plays a crucial role in necroptosis, has been reported to be significantly upregulated in RIF. [ 7 – 9 ] Additionally, levels of cell-derived microparticles (cMPs), released during cell activation or apoptosis, are found to increase significantly in RIF patients. [ 10 ] Moreover, trophoblast necroptosis plays a complex role in pregnancy disorders such as preeclampsia [ 11 , 12 ] and recurrent abortion. [ 13 ] Thus, a potential link exists between RIF and necroptosis, highlighting necroptosis as a promising target for understanding and treating RIF. Given the pro-inflammatory nature of necroptosis, deregulated necroptotic pathways are hypothesized to disrupt the immune environment necessary for successful embryo implantation. [ 14 ] A balanced immune response, teetering between tolerance and activation, is vital for establishing and maintaining pregnancy. Disruptions to this balance, perhaps through necroptosis-induced inflammation, could impair endometrial receptivity, contributing to RIF. In this study, we aimed to identify and analyze differentially expressed necroptosis-related genes (DENRGs) in RIF through an integrated bioinformatics approach. We further intended to construct protein–protein interaction (PPI) networks, perform immune infiltration analysis, and develop a diagnostic model for RIF based on these genes. The findings of our study predict potential therapeutic drugs for RIF, providing new insights into the underlying molecular mechanisms.

Conceptualization: Xiuye Xing. Data curation: Xiaoxiao Ni, Jiaojiao Wang. Formal analysis: Xiaoxiao Ni, Jiaojiao Wang, Junmei Shi. Funding acquisition: Xiuye Xing. Investigation: Xiuye Xing. Software: Xiaoxiao Ni, Jiaojiao Wang. Supervision: Xiuye Xing. Visualization: Jiaojiao Wang, Junmei Shi. Writing – original draft: Xiuye Xing. Writing – review & editing: Xiuye Xing, Junmei Shi.

We obtained the GSE111974 and GSE92324 datasets from the Gene Expression Omnibus (GEO) database ( https://www.ncbi.nlm.nih.gov/geo/ ). The GSE111974 dataset, which includes RNA-sequencing data from endometrial tissues of 24 RIF patients and 24 controls, served as the training set (Table S1, Supplemental Digital Content, http://links.lww.com/MD/N197 ). The GSE92324 dataset, comprising RNA-sequencing data from ten patients and 8 controls, was used as the validation set (Table S2, Supplemental Digital Content, http://links.lww.com/MD/N197 ). Additionally, 159 necroptosis-related genes (NRGs) were identified from existing literature for analysis (Table S3, Supplemental Digital Content, http://links.lww.com/MD/N197 ). [ 15 ] The limma package (version 3.42.2) was used to detect differentially expressed genes (DEGs) in the GSE111974 dataset, using weighted least squares and a linear model fit. [ 16 , 17 ] DEGs were selected based on thresholds of |log 2 FC| > 0.5 and P  < .05 and visualized using volcano plots and heatmaps. DENRGs were then identified by intersecting DEGs with NRGs using jvenn online tools ( http://jvenn.toulouse.inra.fr/app/index.html ). Heatmaps of DENRGs were created using the pheatmap package (version 0.7.7). [ 18 ] Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed using the Cluster Profiler package (version 3.14.3) to categorize DENRGs into molecular function (MF), biological process (BP), and cellular component (CC). [ 19 ] The STRING database (version 11.0) was used to construct a PPI network to explore interactions among DENRGs. [ 20 ] The network was visualized in Cytoscape (version 3.8.0), and hub genes were identified using maximal clique centrality (MCC) in CytoHubba. [ 21 ] Key genes were determined by overlapping hub genes, and a key submodule was identified using the Molecular Complex Detection (MCODE) plugin. A diagnostic model was developed using key genes identified from the PPI network in the STRING database. [ 13 ] Receiver operating characteristic (ROC) curves were used to calculate the area under the curve (AUC) for these genes in the GSE111974 dataset to assess model specificity and sensitivity. [ 22 ] External validation was conducted using the GSE92324 dataset. Transcription factors (TFs) were downloaded from the hTFtarget, RegNetwork, and TRRUST databases. Correlations between TFs and key genes were computed using Pearson correlation in the GSE111974 dataset, with multiple testing corrections applied ( q   0.7). The TF regulatory network was then visualized in Cytoscape. The single-sample gene set enrichment analysis (ssGSEA) algorithm was used to assess immune cell infiltration using 28 immune cell types in the GSE111974 dataset. [ 23 ] Boxplots were generated using the ggplot2 package (version 3.3.2) via Wilcox tests to compare immune cell differences between RIF and controls. Correlations between key genes and immune cells were evaluated using the Spearman method. Expression levels of key genes were compared between the GSE111974 training set and GSE92324 validation set using t tests. The Drug-Gene Interaction Database (DGIdb) was used to predict potential drugs for RIF, and interactions between these drugs and key genes were visualized in Cytoscape.

Our analysis yielded 3289 DEGs between RIF patients and controls, with 1567 upregulated and 1722 downregulated genes (Fig. 1 A, B and Table S1, Supplemental Digital Content, http://links.lww.com/MD/N197) . Among these DEGs, we identified 20 DENRGs by intersecting DEGs with 159 previously reviewed NRGs (Fig. 1 C and Table S1, Supplemental Digital Content, http://links.lww.com/MD/N197) . The expression patterns of these DENRGs are depicted in a heatmap (Fig. 1 D). GO and KEGG enrichment analyses revealed significant involvement of these DENRGs in various biological processes and pathways, including necroptosis and inflammatory signaling (Fig. 1 E, F and Tables S2 and S3, Supplemental Digital Content, http://links.lww.com/MD/N197) . Identification and Functional Enrichment of 20 DENRGs. (A) Volcano plot displaying 3289 DEGs between RIF cases and controls from the GSE111974 dataset, with black dots on the left representing downregulated genes, black dots on the right indicating upregulated genes, and grey dots showing no significant difference. (B) Heatmap illustrating the expression patterns of DEGs, where each block denotes the gene expression in individual samples. (C) Venn diagram demonstrating the selection of 20 DENRGs through the intersection of DEGs and NRGs. (D) Heatmap depicting the expression profiles of the 20 DENRGs. (E) Bar graph showing the top 10 GO terms associated with DENRGs. (F) Bubble chart representing the top 10 KEGG pathways linked to DENRGs. DENRGs = differentially expressed necroptosis-related genes, DEGs = differentially expressed genes, NRGs = necroptosis-related genes, GO = gene ontology, KEGG = Kyoto Encyclopedia of genes and genomes. The PPI network, constructed from STRING database analysis, comprised 20 DENRGs, forming a network of 18 nodes and 34 edges (Fig. 2 A). MLKL showed the highest connectivity. The top ten hub genes identified through MCC were BIRC3, MLKL, FASLG, CASP1, XIAP, TLR3, RNF31, TNFRSF10A, CYBB , and BAX (Fig. 2 B). A key submodule comprising 6 genes ( MLKL, FASLG, XIAP, CASP1, BIRC3 , and TLR3 ) was delineated (Fig. 2 C, D). Determination of 6 key genes. (A) PPI network constructed from 20 DENRGs. (B) Identification of 10 hub genes using MCC analysis. (C) Network visualization of a critical submodule. (D) Six key genes identified through a Venn diagram. DENRGs = differentially expressed necroptosis-related genes, MCC = maximal clique centrality, PPI = protein–protein interaction. ROC curve analysis revealed that the 6 key genes had high diagnostic potential (AUC > 0.696) in the GSE111974 dataset (Fig. 3 A). A diagnostic model incorporating these genes exhibited an AUC of 0.911, demonstrating excellent predictive accuracy (Fig. 3 B). The efficacy of this model was further validated in the GSE92324 dataset, with AUC values exceeding 0.8 for 5 of the 6 genes (Fig. 3 C, D). Diagnostic model development and validation using six key genes. (A) ROC curves evaluating the diagnostic performance of 6 key genes in the GSE111974 dataset. (B) ROC curves assessing the diagnostic model constructed from these 6 key genes. (C) External validation of the diagnostic performance of the 6 key genes in the GSE92324 dataset. (D) External validation of the diagnostic model in the GSE92324 dataset. ROC = receiver operating characteristic. From the integration of 3 human TF databases, a regulatory network comprising 185 nodes (six key genes and 179 TFs) and 232 edges was established (Fig. 4 ). The regulatory interactions of TLR3 were particularly notable, exhibiting both positive and negative correlations with various TFs, thereby highlighting its complex regulatory role in RIF. Furthermore, NFIA was identified as an overlapping TF for BIRC3, CASP1, FASLG , and MLKL . TF regulatory network based on six key genes. The network illustrates six key genes and their interactions with 179 TFs. Positive correlations are indicated by red lines and negative correlations by purple lines, with line thickness representing the strength of correlations. TFs = transcription factors. Immune cell infiltration analysis revealed significant differences in the prevalence of 16 immune cell types between RIF patients and controls, with lower levels in the RIF group (Fig. 5 A). Correlation analysis between these immune cells and the 6 key genes identified 80 significant relationship pairs, indicating a complex interplay between these genes and the immune system (Fig. 5 B). Immune cell infiltration analysis. (A) Box plot comparing the proportions of 28 immune cell types between control and RIF samples in the GSE111974 dataset, with significance levels marked (* P  < .05, ** P  < .01, *** P  < .001). (B) Heatmap depicting correlation strengths between immune cell types and the 6 key genes, with red indicating positive correlations, blue indicating negative correlations, and color intensity denoting correlation strength (* P  < .05, ** P  < .01). RIF = recurrent implantation failure. Significant differences in the expression levels of key genes were observed between RIF patients and controls in the GSE111974 dataset. BIRC3, TLR3 , and XIAP expression were higher in the RIF group than in the control group (Fig. 6 A). The DGIdb identified 49 potential therapeutic drugs interacting with these key genes, including 23 drugs targeting CASP1 (Fig. 6 B, C). Among these drugs, AT-406, LCL161, and birinapant were associated with both XIAP and BIRC3 . Validation of key gene expression and potential therapeutic drug identification. (A) Box plot of expression levels of the six key genes in the GSE111974 dataset, with significance levels indicated (* P < 0.05, ** P < 0.01, **** P < 0.0001). (B) Expression validation of the six key genes in the GSE92324 dataset, with significance levels denoted (* P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, ns = not significant). (C) Drug-gene interaction network, showcasing six key genes (circles) and potential therapeutic drugs (squares).

RIF remains a major challenge in the field of ART. [ 1 ] A deep understanding of the underlying molecular mechanisms is paramount for developing effective diagnostic and therapeutic strategies. In this study, we used a comprehensive bioinformatics approach to elucidate the potential role of NRGs in RIF, thereby contributing to the limited body of literature on the subject. Necroptosis, a form of regulated necrotic cell death, has gained increasing attention for its roles in several disease processes, such as neuroinflammation. [ 24 , 25 ] In this context, our research identified 6 key NRGs ( MLKL, FASLG, XIAP, CASP1, BIRC3 , and TLR3 ) significantly associated with RIF, suggesting a possible involvement of necroptosis in the pathogenesis of this condition. MLKL, a pivotal executor of necroptosis, has been extensively studied in the context of inflammatory diseases. [ 26 ] When activated, MLKL compromises cellular membrane integrity, leading to necroptotic cell death. [ 27 ] Therefore, the increased expression of MLKL, as observed in our study, may lead to elevated necroptosis in the endometrium, promoting an inflammatory environment unfavorable for embryo implantation. The Fas ligand (FASLG) is a transmembrane protein that initiates apoptosis upon binding to its receptor FAS. However, under certain circumstances, FASLG can trigger necroptosis. [ 28 ] Overexpression of FASLG in our RIF samples could imply an altered balance between apoptosis and necroptosis within the endometrium, potentially disrupting the microenvironment required for successful implantation. [ 29 ] X-linked inhibitor of apoptosis (XIAP) and baculoviral IAP repeat-containing protein 3 (BIRC3) belong to the inhibitor of apoptosis (IAP) family, which is known to regulate both apoptosis and necroptosis. [ 30 ] Disruptions in the expression of XIAP and BIRC3 could tip the balance between these processes, impacting the endometrial viability and potentially contributing to RIF. [ 31 , 32 ] Caspase-1 (CASP1) primarily mediates the activation of pro-inflammatory cytokines but is also implicated in necroptosis induction. [ 33 ] Elevated expression of CASP1 could instigate a pro-inflammatory state in the endometrium, making it less receptive to implantation. [ 34 ] Lastly, Toll-like receptor 3 (TLR3) is an innate immune receptor that can trigger necroptosis and is involved in immune responses. [ 35 ] Dysregulation of TLR3, as suggested by our data, may lead to an intensified inflammatory response, thereby exacerbating RIF. [ 36 ] Our study provided valuable insights into the correlation between necroptosis and the transcription factors PER2, RORC, FOXO1, UHRF1, BRCA2, and GLI1, demonstrating the complexity of necroptosis regulation. For example, PER2, a circadian clock gene, has been shown to regulate apoptosis in certain contexts, [ 37 ] suggesting a potential role in necroptosis. Similarly, RORC and FOXO1 have been implicated in various immune processes, [ 38 ] hinting at their possible influence on immune interactions during implantation. NFIA, as a TF, may interact with four key genes, namely BIRC3, CASP1, FASLG , and MLKL . Moreover, the altered immune landscape, as demonstrated by the significant decrease in levels of 16 types of immune cells in the RIF group, raises crucial questions about the role of immunity in RIF. A complex balance exists between the involvement of the immune system in providing a receptive environment for implantation and avoiding an overly aggressive response that could harm the embryo. [ 14 ] This observed alteration in immune cell infiltration could suggest an imbalance in this delicate system, further complicating the pathogenesis of RIF. Additionally, the identification of potential therapeutic drugs, particularly those modulating the activity of CASP1, was a notable outcome of the study. CASP1 , a gene involved in inflammation and programmed cell death, could be a promising therapeutic target for treating RIF. [ 39 ] Given the major role of inflammation in successful implantation and the ability of CASP1 to modulate inflammatory responses, drugs targeting CASP1 could potentially provide a novel treatment approach for RIF. [ 40 ] AT-406, LCL161, and birinapant were identified as overlapping drugs for XIAP and BIRC3 . The IAP inhibitors, AT-406 and LCL161, are small-molecule SMAC mimetics used as cancer therapeutics. [ 41 , 42 ] Similarly, birinapant is an antagonist of the IAPs, which has been reported as a potential drug in some malignancies (e.g. triple-negative breast, colon, and gastric cancers). [ 43 – 45 ] Whether these drugs can be used as a treatment for RIF remains to be further explored. While the results of this study represent a major advancement in understanding the molecular mechanisms underlying RIF, they should be interpreted within the limitations of the research. This study relied heavily on data extracted from the GEO database, which, although reliable, may contain potential bias inherent in any secondary data source. Therefore, the findings need validation through in vitro and in vivo experimental studies to confirm the involvement of the identified NRGs in RIF. Additionally, while the bioinformatics approach used in this study provides robust initial findings, further extensive mechanistic studies are necessary to understand the role of these genes in RIF. Future research should focus on the clinical application of these findings, testing potential therapeutic interventions in randomized controlled trials.

We thanks the patients who participated in GEO database.