Abstract

Aims

   To identify the common and specific molecular mechanisms of three
   well-defined subtypes of endometriosis (EMs): ovarian endometriosis
   (OE), peritoneal endometriosis (PE), and deep infiltrating
   endometriosis (DIE).

Methods

   Four microarray datasets: [32]GSE7305 and [33]GSE7307 for OE,
   [34]E-MTAB-694 for PE, and [35]GSE25628 for DIE were downloaded from
   public databases and conducted to compare ectopic lesions (EC) with
   eutopic endometrium (EU) from EMs patients. Differentially expressed
   genes (DEGs) identified by limma package were divided into two parts:
   common DEGs among three subtypes and specific DEGs in each subtype,
   both of which were subsequently performed with the Kyoto Encyclopedia
   of Genes (KEGG) pathway enrichment analysis. The protein-protein
   interaction (PPI) network was constructed by common DEGs and five hub
   genes were screened out from the PPI network. Besides, these five hub
   genes together with selected interested pathway-related genes were
   further validated in an independent OE RNA-sequencing dataset
   [36]GSE105764.

Results

   A total of 54 EC samples from three EMs subtypes (OE, PE, DIE) and 58
   EU samples were analyzed, from which we obtained 148 common DEGs among
   three subtypes, and 729 specific DEGs in OE, 777 specific DEGs in PE
   and 36 specific DEGs in DIE. The most enriched pathway of 148 shared
   DEGs was arachidonic acid (AA) metabolism, in which most genes were
   up-regulated in EC, indicating inflammation was the most common
   pathogenesis of three subtypes. Besides, five hub genes AURKB, RRM2,
   DTL, CCNB1, CCNB2 identified from the PPI network constructed by 148
   shared DEGs were all associated with cell cycle and mitosis, and
   down-regulated in EC, suggesting a slow and controlled proliferation in
   ectopic lesions. The KEGG pathway analysis of specific DEGs in each
   subtype revealed that abnormal ovarian steroidogenesis was a prominent
   feature in OE; OE and DIE seems to be at more risk of malignant
   development since both of their specific DEGs were enriched in the
   pathways in cancer, though enriched genes were different, while PE
   tended to be more associated with dysregulated peritoneal immune and
   inflammatory microenvironment.

Conclusion

   By integrated bioinformatic analysis, we explored common and specific
   molecular signatures among different subtypes of endometriosis:
   activated arachidonic acid (AA) metabolism-related inflammatory process
   and a slow and controlled proliferation in ectopic lesions were common
   features in OE, PE and DIE; OE and DIE seemed to be at more risk of
   malignant development while PE tended to be more associated with
   dysregulated peritoneal immune and inflammatory microenvironment, all
   of which could deepen our perception of endometriosis.

   Keywords: Bioinformatic analysis, Differentially expressed genes,
   Microarray, Endometriosis, Subtype

Introduction

   Endometriosis (EMs), characterized by the growth of endometrium-type
   tissue outside the uterine cavity, is a common and usually chronic
   (long-term) inflammatory disorder, affecting 5–10% of women in their
   reproductive years ([37]Zondervan et al., 2018). EMs is also considered
   as a phenotypically heterogeneous condition not only due to diverse
   symptoms, such as infertility, pelvic pain or dysmenorrhea but also
   different lesion locations, predominantly but not exclusively, in the
   pelvic compartment ([38]Vercellini et al., 2014). Since the classic
   retrograde menstruation hypothesis ([39]Sampson, 1927) that during
   menstrual uterine contractions, endometrial fragments via trans-tubal
   reflux flowed to implant onto the peritoneum and abdominal organs could
   not explain the fact that 76–90% of women experienced retrograde
   menstruation but not all of these women suffered from EMs
   ([40]Blumenkrantz et al., 1981), there must exist other mechanisms
   facilitating the development of EMs. On the other hand, as early as
   1997, Nisolle and Donnez provided morphological and histochemical
   evidence indicating that three main subtypes of endometriosis: ovarian
   endometriosis (OE), peritoneal endometriosis (PE), and deep
   infiltrating endometriosis (DIE), should be considered different
   entities, though they shared the histologic features of endometrial
   glands and stroma ([41]Nisolle & Donnez, 1997). Thus, investigating the
   common and specific mechanisms among different EMs subtypes may provide
   new insight into the pathogenesis of endometriosis. However, due to the
   limited information as well as samples available from single cohorts,
   few integrative analyses of EMs subtypes were conducted.

   Therefore, in this article, we analyzed the microarray datasets
   [42]GSE7305 and [43]GSE7307 of OE, [44]E-MTAB-694 of PE, and
   [45]GSE25628 of DIE to obtain common differentially expressed genes
   (DEGs) among three EMs subtypes along with specific DEGs in each
   subtype by comparing ectopic lesions (EC) with eutopic endometrium (EU)
   from EMs patients. Then, the Kyoto Encyclopedia of Genes (KEGG),
   protein-protein interaction (PPI) network and validation analysis were
   performed to analyze these common and specific DEGs ([46]Fig. 1).
   Overall, all results were combined to promote further understanding of
   different EMs subtypes and reveal a more thorough landscape of EMs.

Figure 1. The flowchart of integrative analysis of microarray datasets from
three different endometriosis subtypes.

   [47]Figure 1
   [48]Open in a new tab

   Abbreviations: OE, ovarian endometriosis; PE, peritoneal endometriosis;
   DIE, deep infiltering endometriosis; DEGs, differentially expressed
   genes; KEGG, KyotoEncyclopedia of Genes; PPI, protein-protein
   interaction.

Materials & Methods

Data resources

   The search for endometriosis-related microarray datasets was conducted
   in two public databases: Gene Expression Omnibus
   ([49]https://www.ncbi.nlm.nih.gov/geo/) and Array-Express
   ([50]https://www.ebi.ac.uk/ar-rayexpress/). The keywords:
   ‘endometriosis’, ‘endometrium’, ‘tissue’, ‘homo sapiens’ or ‘human’
   respectively with ‘ovarian’, ‘peritoneal’, ‘DIE’, ‘deep’ or
   ‘infiltrating’ were employed to mine the datasets for three EMs
   subtypes. Additionally, all selected datasets were based on Affymetrix
   platforms to reduce the ‘platform effect’ due to different probe
   designs among different companies. Finally, four datasets were
   included: [51]GSE7305 and [52]GSE7307 for OE, [53]E-MTAB-694 for PE,
   and [54]GSE25628 for DIE ([55]Table 1), all of which contained at least
   8 samples for both ectopic lesions (EC) and eutopic endometrium (EU)
   from the EMs patients. A total of 54 EC samples from these datasets of
   different EMs subtypes and 58 EU samples were included.

Table 1. Basic information of the microarray datasets of three different EMs
subtypes.

   Subtype Accession Platform No. of probes No. of samples (EU/EC, whether
   paired) Tissue stage References