P-372 RNA sequencing studies evaluating oocyte quality in women with endometriosis: a meta-analysis

Abstract Study question Can a network-based integrative analysis (NBIA) meta-analysis approach be used to identify clinically relevant cumulus-oocyte complex biological pathway and gene targets in women with endometriosis? Summary answer Bi-level meta-analysis (BLMA) of multiple RNAseq studies shows enrichment of both inflammatory and mitochondrial pathway-related differential gene expression within oocyte, granulosa and cumulus cells. What is known already Endometriosis-related infertility is multifactorial, and studies suggest that impaired oocyte quality plays a significant role. The close relationship between granulosa, cumulus cells and developing oocytes makes them ideal candidates for studying non-invasive biomarkers for oocyte quality. As high-throughput sequencing becomes more accessible, the volume of RNAseq molecular data in public databases such as Gene Expression Omnibus (GEO) has increased. Combined analysis of multiple datasets theoretically increases study power due to larger number of samples. Comparison of differentially expressed genes (DEGs) and biological pathways across multiple independent studies therefore requires a ‘meta-analysis’ framework that can effectively integrate data from different sources. Study design, size, duration We identified relevant RNAseq datasets through NCBI public functional genomics data repository Gene Expression Omnibus (GEO). Datasets were split into “test” and “control” groups referring to models of oocyte incompetence and competence. Accession numbers GSEXXX or PRJNXXX represent unique identifiers for RNAseq data files contained within a given NCBI GEO series or BioProject record, respectively. Five human RNAseq datasets (PRJNA514416, GSE168214, PRJNA216966, PRJNA727838 and GSE81579) were identified and DEG lists generated for each individual dataset. Participants/materials, setting, methods Normalisation and differential expression between RNAseq datasets was carried out in R using the Bioconductor DESeq2 package. DEGs were identified as those whose P-value was adjusted using a false discovery rate less than 0.5, adjusted using Benjamini-Hochberg procedure. Log2 fold-change ratio (logFC) of each DEG was used to identify significantly up and downregulated genes. The NBIA pipeline (Nguyen et al, 2020) examines over-representation of significantly impacted genes in a pathway as per KEGG pathway analysis. Main results and the role of chance Upregulated DEGs were identified in mitochondrial membrane ATP synthase pathways, (ATP5G1, ATP5G2, ATP5J2 and ATP5H). KEGG Pathway terms “Parkinson’s”, “Huntington’s”, and “Alzheimer disease” were also enriched, due to many genes partaking in oxidative phosphorylation in the context of these pathologies. Functional analysis also revealed significantly down-regulated genes related to inflammatory pathways (CXCL1, 2, 3, 5,6 and 8), which have a role in innate immunity. Although surprising, in our analysis, downregulated functions based on KEGG functional pathway analysis, observed that innate immune response, cellular defense response, chemokine-mediated signaling were significantly downregulated. BLMA over-representation analysis (ORA) and impact analysis (IA) analysis showed that the Amoebiasis pathway was enriched expressed across all datasets. Genes involved in this pathway include interleukin (IL), CXCL1, CXCL2, CXCL3, SERPINB, TGFB1 and TGFB2, and its biological process is related to natural killer (NK) cell mediated cytotoxicity. Overall, the enrichment of immune and inflammation-related pathways revealed by BLMA meta-analysis underscores the relevance of inflammation to follicular function. Indeed, ovulation is inextricably linked with inflammation. Other significantly enriched pathways highlighted by this meta-analysis are related to infection (path:hsa05134 Legionellosis; path:hsa05140 Leishmaniasis) immune response (path:hsa04145 Phagosome; path:hsa04610 Complement and coagulation cascades) and inflammation (path:hsa05323 Rheumatoid arthritis). Limitations, reasons for caution Methods used availed of publicly available GEO datasets and bioinformatics software (Bioconductor, R and DAVID). Although useful tools, these approaches are open to inter-user variation. Cell types compared across studies were oocytes, granulosa and cumulus cells. Controlled ovarian stimulation may also be open to influence based on stimulation protocol used. Wider implications of the findings Enrichment of inflammatory and mitochondrial pathways in our GO analysis is compelling. Given the role of mitochondria in oocyte developmental competence, this reinforces the role of impaired oocyte competence in women with endometriosis. Our study identifies several genes that require further elucidation as potential predictors of oocyte quality in endometriosis. Trial registration number N/A