Transcriptomic Analysis of the Aged Nulliparous Mouse Ovary Suggests a Stress State That Promotes Pro-Inflammatory Lipid Signaling and Epithelial Cell Enrichment

Abstract

Ovarian cancer (OC) incidence and mortality peaks at post-menopause while OC risk is either reduced by parity or increased by nulliparity during fertile life. The long-term effect of nulliparity on ovarian gene expression is largely unknown. In this study, we describe a bioinformatic/data-mining analysis of 112 coding genes upregulated in the aged nulliparous (NP) mouse ovary compared to the aged multiparous one as reference. Canonical gene ontology and pathway analyses indicated a pro-oxidant, xenobiotic-like state accompanied by increased metabolism of inflammatory lipid mediators. Up-regulation of typical epithelial cell markers in the aged NP ovary was consistent with synchronized overexpression of Cldn3, Ezr, Krt7, Krt8 and Krt18 during the pre-neoplastic phase of mOSE cell cultures in a former transcriptome study. In addition, 61/112 genes were upregulated in knockout mice for Fshr and for three other tumor suppressor genes (Pten, Cdh1 and Smad3) known to regulate follicular homeostasis in the mammalian ovary. We conclude that the aged NP ovary displays a multifaceted stress state resulting from oxidative imbalance and pro-inflammatory lipid signaling. The enriched epithelial cell content might be linked to follicle depletion and is consistent with abundant clefts and cysts observed in aged human and mouse ovaries. It also suggests a mesenchymal-to-epithelial transition in the mOSE of the aged NP ovary. Our analysis suggests that in the long term, nulliparity worsens a variety of deleterious effects of aging and senescence thereby increasing susceptibility to cancer initiation in the ovary.

Keywords: age; mouse model; nulliparity; ovarian cancer; risk.

Figure 1

Previous study, data processing and canonical analysis of DEGs in the aged NP ovary. (A) Two C57BL6 female mice cohorts were maintained in nulliparous (NP) and multiparous (MP) breeding regimens from 4 through 16 months old. Short overlapped white arrows depict gestation plus lactation periods in MP mice. (B) Total ovarian RNA from the two conditions was analyzed with Illumina™ beadchip expression microarrays resulting in 177 differentially expressed genes (DEGs) between the NP and MP conditions [13]. (C) Summary of a gene set enrichment analysis (GSEA) of the 112 DEGs of higher expression in NP ovaries, using the MSigDBv7.3 database (details in Methods and Table 1). The gene number associated to each functional theme is indicated in the respective bar, and their identities described in Table 1.

Figure 2

Overlap and gene network among xenobiotic metabolism, ion/small molecule transport and electrolyte homeostasis. (A) Venn diagram to determine the gene coincidence between 3 of the functional themes of Figure 1C. The gene lists of each theme are described in Table 1. (B) Known and predicted relationships among the 20 coincident genes shown as a gene network obtained with STRING v12.0 (0.35 confidence score; unconnected nodes removed). The meaning of color lines depicting gene interactions is shown at the bottom-right and was adapted from the software’s output.

Figure 3

Overlap and gene networks among oxidative metabolism and epithelium related themes. Venn diagrams showing the gene coincidence between themes related to oxidative metabolism (A) and to epithelium (C). Detailed gene groups are shown in Table 1. Gene networks among the 20 coincident genes of themes related to oxidative metabolism (B) and among the 11 coincident genes of themes related to epithelium (D). Parameters of STRING v12.0 analysis and meaning of color-coded line interactions as in Figure 2B.

Figure 4

GEO datamining of KO mouse models. The 112 DEGs were mined in the Enrichr database crowd tab, gene perturbations from GEO up output. (A) The top 10 mouse KO datasets ranked by adj-p values. KO gene symbol and GEO accession is indicated within each bar. The number at the right indicates the gene count from the 112 DEGs which are contained in each dataset. (B) Overlap of the top 20 genes in the 10 KO datasets shown in (A). (C) Gene coincidence between the top 4 KOs; (D) Selected heatmaps of 46/112 genes that were found in the transcriptome dataset of pre-malignant, cultured mOSE cells. Labels p2 to p28 indicate passage 2 through passage 28 log₂ fold change averages of 4 replicate two-channel arrays [53].

Figure 5

Selected gene expression in human ovarian carcinomas. A subset of 12 DEGs were queried in TNMplot as their respective human orthologs (see Methods). Gene expression density plots are shown for the indicated genes (left) in 133 normal and 374 human ovarian cystadenocarcinoma samples (“Tumor” label). The fold change (FC) corresponds to the direct ratio between mean expression values for each group; p values were obtained from a Mann–Whitney U test. Density plots are supported by boxplots of Figure S2.