Background

Deep infiltrating endometriosis (DIE) and adenomyosis (FA) are very common benign gynecological conditions in women of childbearing age: 10–15% of women undergoing laparoscopy for benign reasons are generally found to have endometriosis [1], and in patients with infertility the incidence is even higher at 30–50% [2]. DIE is thought to occur in at least 20% of women with pelvic endometriosis [3]. FA affects 19.5% of women of childbearing age [4] [5]. In hysterectomy specimens, the incidence of diagnosed FA is 10–35% [6].

Both conditions are defined by the presence of endometrial glands and stroma either in the myometrium or outside the uterus. Originally DIE was even called “adenomyosis externa” [7].

There is a major discussion, if FA and especially DIE are related diseases or not, as in women with both entities the phenotype of FA appears to be related to the severity of endometriosis, particularly as women with DIE had a significantly higher frequency of focal adenomyosis in the external myometrium than patients with ovarian endometriosis [8]. Clinically however, the two conditions differ markedly. FA is commonly associated with dysmenorrhea, infertility, repeated implantation failure and pregnancy loss [9,10,11]. Multiparity and previous uterine surgery are discussed as risk factors for this condition [12,13,14]. DIE, on the other hand, is known to cause extensive adhesions up to complete obliteration of the Douglas space, as well as constriction of affected organs such as the bowel and bladder, and thus can cause not only dysmenorrhea but also chronic abdominal pain and, in severe cases, bowel obstruction up to the point of an ileus.

Although histologically benign, both DIE and FA are characterized by their propensity for local tissue invasion and resistance to apoptosis [15]. Notably, DIE has been described as a “benign tumor” [16]. Recent work using next-generation sequencing (NGS) has demonstrated driver mutations in cancer associated genes such as PIK3CA, ARID1A, PPP2R1A and KRAS in both, ovarian endometrioma [17] and DIE [15].

The presence of PIK3CA- or KRAS-mutated clones in histologically normal uterine endometrium in endometriosis [17] but also in patients without endometriosis has also been demonstrated [18], so the theory of the cellular origin of endometriosis requires further investigation. In contrast, the discovery of identical mutations in the KRAS gene in coexisting adenomyotic and endometriotic lesions in several patients [11], supports the theory of a common pathogenesis of adenomyosis uteri and endometriosis and a common molecular mechanism in these diseases [11, 19].

In this study, we aimed to further characterize the molecular mechanisms involved in FA and especially DIE to find molecular similarities and differences in both diseases. To this end, we analyzed cancer-related signaling pathways at the gene expression level using a nanostring gene panel encompassing the major signaling pathways of carcinogenesis using epithelial cells of FA and DIE. Any relevant genomic changes at the DNA level should be reflected in their gene expression and provide insight into the common or different underlying pathomechanisms of the two diseases and provide clues to the pathomechanisms of tumor development in these diseases and treatment options.

Patients and methods

Study population

For this study, we collected formalin-fixed and paraffin-embedded (FFPE) tissue samples for the analysis of two independent cohorts of patients with DIE or FA. Patients underwent surgery at the University Hospital, Heidelberg or cooperating clinics and the samples were histologically examined and assessed at the Dept. of Pathology of the University Hospital, Heidelberg between 2003 and 2018. Clinical records and histology were reviewed. Exclusion criteria were histological indications of cancer or dysplasia or a lesion size too small to gain sufficient material for further analysis. The samples were provided by the Tissue Bank at the National Center for Tumor Diseases (Heidelberg, Germany) in concordance with the Ethics Committee of the University of Heidelberg (approval No. S-362/2017).

Staining and laser microdissection (LMD)

For the RNA extraction, FFPE tissue blocks from FA and DIE were selected after reviewing all original tissue slides and were recut for hematoxylin & eosin sections, to be used for reference and to determine the lesion size. RNA was extracted using 10–20 FFPE slides for each entity.

For mounting on Zeiss 1.0 PEN slides (Carl Zeiss, Oberkochen, Germany) and for better adhesion of the tissue to the membrane, slides were irradiated with UV light/254nm for 30 min before, FFPE tissue blocks were cut at 8 μm thickness and incubated overnight at 37 °C. They were dewaxed in xylene (100%), rehydrated through decreasing concentrations of ethanol (100, 95, 75%), stained in 1% cresyl violet acetate (Sigma-Aldrich, Taufkirchen, Germany) and again dehydrated in increasing ethanol concentrations (75, 95, 100%). After that, tissue sections were dried and stored at 4 °C. Using a ZEISS PALM LMD laser capture microdissection unit, regions of interest (epithelium of the adenomyosis or epithelium of deep infiltrating endometriosis) were microdissected. The isolated tissue fragments corresponded to an area of approximately 20.000.000 μm² for each sample. They were collected in AdhesiveCap 500 opaque tubes (Carl Zeiss) and stored at − 20 °C until further processing (Fig. 1).

Fig. 1
figure 1

Light microscopy of deep infiltrating endometriosis of the rectum (A) and of adenomyosis uteri (B) stained with cresylviolet (10x). In each case the second image in the row shows the marking of the tissue for laser dissection [2]. The third image shows the tissue after laser dissection [3]

Full size image

RNA isolation

Extraction of the total RNA from microdissected tissue samples was performed using the AllPrep DNA/RNA FFPE Kit (Qiagen, Venlo, the Netherlands) according to the manufacturer’s protocol. They were quantified with the Nanodrop ND-1000 spectrophotometer (NanoDrop Technologies, Rockland, DE, USA).

Gene expression analysis

We analyzed the expression of 770 genes (Codeset: Human PanCancer Pathways) and hybridisation counts were measured using the nCounter technology (both by Nanostring™ Technology, Seattle, WA). A minimum of approximately 50 ng of total RNA was used. Hybridization time per cartridge was 16 h before measurement. According to the manufacturer’s protocol, the examined genes were attached to specific tag sequences and hybridized for 16 h at 65 °C to a capture/reporter probe pair equipped with a fluorescent barcode. These gene-specific barcodes were then detected by the nCounter Digital Analyzer providing count of genes. No cases were excluded.

The 770 genes codeset included 730 genes from 13 canonical pathways (e.g., cell cycle, chromatin modeling, apoptosis, MAPK, and PI3K) and 40 housekeeping genes [20]. The raw data were pre-analyzed for consistency using the manufacturer’s software (nSolver version 4.0). The geNorm pairwise variation statistics was used for stepwise selection of normalization genes from the housekeeping genes [21]. Six genes with minimal pairwise variation statistics were finally selected for normalization (TLK2, VPS33B, TMUB2, C10orf76, SLC4A1AP, ERCC3).

Statistics

Differential expression analysis was carried out using a linear data model in limma [22, 23], and nominal p-values were corrected for multiple comparisons using Benjamini and Hochberg’s method [24]. All genes with an adjusted false discovery rate (FDR) of p < 0.05 and fold change of < 0.66 or > 1.5 were considered differentially expressed. Differentially expressed genes (DEGs) were subjected to functional annotation and clusterization using DAVID Bioinformatics Resources (version 6.8, https://david.ncifcrf.gov/ [25, 26]) after conversion of gene symbols to Entrez IDs and uploading to DAVID using the “RDAVIDWebService” BioConductor library [27]. Basal cytokeratin co-regulated genes were identified using DAVID analysis in the “Biological Process” category and the KEGG pathway enrichment function with a significance threshold of 0.05. The p-values of selected GO terms were corrected using Benjamini-Hochberg correction and described as adjusted p-values [24]. Otherwise, differences between samples were tested using Wilcoxon signed-rank test, and correlation was tested using Spearman’s rank correlation test, and p-value of 0.05 was considered significant. All statistical calculations were done using R version 4.0 [28]. For visualization, the R packages ComplexHeatmap and Ggplot2 were utilized.

Results

Study population

The study cohort included 19 female patients with DIE. DIE lesions were defined as histologically confirmed endometriosis with infiltrative growth with a depth of more than 5 mm into the wall of pelvic organs, e.g. in the bowel or bladder (Tables 1 and 2). The control group included seven patients with FA and infiltration of less than half of myometrium (superficial) or more than one half of myometrium (deep adenomyosis) (Tables 1 and 2).

Table 1 Cohort description including age (years), symptoms/presentation in clinic, site of manifestation, and severity according to rASRM/Enzian classification for each of the 26 patients enrolled in the study
Full size table
Table 2 Overview of clinical and pathological patient data
Full size table

The mean age of the patients suffering from FA was 48.4 years, as compared to 33.4 years for patients suffering from DIE. Patients with adenomyosis had undergone surgery due to bleeding disorders, while DIE was resected for different reasons mainly including acute and chronic pain or incipient intestinal obstruction. The GI-tract (rectum, recto-sigmoid, colon and ileum, 15 cases) was mostly affected by DIE, followed by the vagina (7 cases) and the bladder (1 case), including overlapping sites. FA included cases with superficial and deep infiltration.

Dysregulated genes in DIE vs. adenomyosis uteri

A mathematical model was constructed for the analysis of differential gene expression in FA and DIE. When using a threshold for fold changes (FC) < 0.66 or > 1.5, a total of 162 genes were identified that were up- or downregulated (adj. p < 0.05). This analysis included significantly more genes with upregulation in DIE (116 genes), as compared to 46 genes with relative downregulation in DIE, compared to adenomyosis (FC < 0.66). When using a stricter threshold of significance (p < 0.001 was used), 15 genes were upregulated in DIE, and only one gene (FDZ2) was upregulated in adenomyosis uteri (Table 3; Fig. 2). With regards to the functional properties, no specific pathway could be assigned, and therefore a separate gene ontology analysis was performed (see below).

Table 3 Genes with significant relative upregulation (n = 15) or downregulation (n = 1) in DIE vs. adenomyosis (> 1.5 fold, Benjamini–Hochberg adjusted P ≤ 0.001)
Full size table
Fig. 2
figure 2

Volcano plot showing genes with most significant dysregulation in DIE vs. adenomyosis (adj. p-adjusted < 0.001 and log2 FC > 1.5). 15 genes were highly significantly upregulated in DIE, and one gene (FZD2) was downregulated (log2 FC < 0.66).

Full size image

In order to relate dysregulated genes to clinical characteristics of adenomyosis and DIE, an unsupervised heatmap was constructed. Here, clustering revealed a clear separation of DIE and adenomyosis cases with only one DIE case clustering within adenomyosis (Fig. 3). But generally, samples from DIE had generally higher pathway activity scores than samples from adenomyosis (FA). In this analysis, one larger gene group with upregulation in DIE could be separated from a smaller gene group with upregulation in adenomyosis (Fig. 3). However, in this clustering no correlation of gene expression with clinical characteristics (organ, BMI, depth of adenomyosis) was evident.

Fig. 3
figure 3

Heatmap of supervised hierarchical clustering of differential genes (adj. p < 0.05, log2 fold change > 1.5 or < 0.66 and adjusted p-value < 0.05, n = 162) for the FA versus DIE group. This includes 116 upregulated genes in DIE and 46 genes with upregulation in adenomyosis. Gene expression with cases of adenomyosis (FA) is clearly distinct from cases with deep infiltration endometriosis (DIE).

Full size image

Pathways with activation in adenomyosis uteri and DIE

In order to analyze the functional properties of dysregulated genes in both diseases we performed gene ontology analysis using the KEGG pathway analysis. This analysis revealed upregulating of several signaling pathways in DIE, and interestingly, the PIK3CA pathway was most significantly upregulated. Other gene ontology groups included pathways involved in virus infection, focal adhesion, endocrine resistance and malignancy (Fig. 4a).

Fig. 4
figure 4

(a) KEGG enrichment analyses for differentially upregulated genes (p < 0.05, n = 116), 12 most significantly upregulated pathways are shown. This analysis revealed upregulating of signaling pathways in DIE, most significantly the PI3K pathway, but also pathways involved in virus infection, focal adhesion, endocrine resistance and malignancy. (b) Same KEGG analysis, for differentially downregulated genes (p < 0.05, n = 46), 4 most significant pathways (p < 0.0001) are shown. Here, RAS, PI3K-AKT, RAP1 and Calcium signaling pathways are significant

Full size image

The identification of virus infection pathways was an unexpected finding, but further analysis revealed that these virus-pathway associated genes had 12 genes in common, and 11 of these genes were also common to the PIK3CA pathway. Therefore, it is believed that the PIK3CA pathway upregulation is the root cause for showing virus-pathway associated pathways in this analysis, and that this analysis does not point to a virus related cause of DIE. Upregulated pathways in adenomyosis included RAS, PI3K-AKT, RAP1 and calcium signaling pathways (Fig. 4b).

Discussion

To our knowledge, this is the first study comparing isolated epithelium cells of deep infiltrating endometriosis with epithelium of adenomyosis uteri using the nanostring technology. This is of particular interest, as adenomyosis and endometriosis lesions are often surrounded by many stromal and inflammatory cells, which cause some blurring of studies at the molecular level [29]. Anglesio et al., Inoue et al. and Moore et al. have recently shown that the somatic mutation occurs in the epithelial component of DIE [15], as well as of FA [11] and in the histologically normal endometrium of healthy patients [18]. In the latter studies, analysis of the laser microdissected epithelium has been shown to yield promising results [11, 15, 17, 18, 30]. Therefore, this method was also used in the current work.

In our studies, we show for the first time that the epithelium of DIE and FA differ significantly at the RNA expression level. Interestingly, these differences in RNA expression between both entities are independent of the site of DIE manifestation and the body mass index.

Looking first at the KEGG analysis of genes whose expression is elevated in DIE compared to FA, it appears that the PI3K pathway is significantly activated. This is consistent with the results of whole-exome studies by Anglesio 2017 and Suda 2018, which detected somatic driver mutations at the DNA level in the PIK3CA gene in the epithelium of DIE lesions [15] and in the epithelium of ovarian endometriosis, as well as in the eutopic endometrium of healthy patients [17], among others. Since most PIK3CA mutations in cancers show gain of function and growth advantages [31], it was concluded that the presence of the same mutations in endometriotic epithelial cells has functional significance in the pathogenesis of the disease [17, 32]. We show that this may also be reflected in the activation of the pathway at the RNA level. Previous work has also indicated dysregulation of the PI3K pathway in endometriosis: Yin et al. 2012 demonstrated an increase in pAKT, albeit in stromal cells from endometriomas, compared to cells from the eutopic endometrium of healthy women, and Guo et al. 2015 also showed that phosphorylated mTor is increased in ectopic endometrial lesions compared to eutopic endometrium from endometriosis patients [33, 34]. Also, in a recent work by Madanes, the authors are able to demonstrate increased expression of PI3K, reduced expression of PTEN, and increased levels of pAkt in the ectopic and eutopic endometrium of patients with peritoneal endometriosis [35].

The PI3K-AKT-mTOR pathway is one of the most frequently dysregulated signaling pathways in carcinoma diseases [36]. Its significant activation in DIE compared to FA may explain the different behavior of the two entities under study.

We believe, that the fact, that the KEGG analysis revealed activation of virus associated gene groups (HPV, KSAHV and CMV) is due to the large intersection of genes activated in these gene groups with genes that are significantly activated in the PI3K pathway, rather than a virus-related cause of DIE, as discussed in detail above.

The focal adhesion pathway plays an essential role in cell motility, cell proliferation and cell differentiation. Its increased activation in DIE compared to FA could also explain the more progressive behavior of DIE. Interestingly, in a recent analysis of the proteome of the eutopic endometrium of endometriosis patients, Méar et al. also demonstrated an increased activation of the PI3K pathway and the focal adhesion pathway compared to healthy controls [37].

The KEGG analysis of genes that are downregulated in DIE and upregulated in adenomyosis compared to DIE shows that the RAS pathway in particular is upregulated in adenomyosis. This fits well with the findings of Inoue et al., who demonstrated a mutation in the KRAS gene with consecutive activation of KRAS in 37.1% of adenomyosis cases [11]. They detected a mutation in PI3KCA in only two of 70 patients, which may explain the increased expression of PI3K pathway-associated genes in DIE in our study compared with FA. In contrast, in patients with both adenomyosis and endometriosis lesions, Inoue et al. were able to detect the same KRAS mutation in both lesions [11]. By comparing both tissues in our study and looking at the relative gene expressions of both entities, the absolute activation of the RAS pathway is shown to be lower in endometriosis patients, which would explain why it is not detected in the KEGG analysis. In direct comparison of the two entities at the level of gene expression however, the PI3K pathway appears to be the dominant pathway in DIE and the RAS pathway in FA.

This could explain, for example, the different sensitivity of the two entities to certain therapeutic approaches. While DIE responds well to therapy with the progestin dienogest, progesterone resistance is often described in adenomyosis patients, which according to Inoue et al. is due to the KRAS mutations [11, 32].

If we look at the changes of single genes, it is striking that Frizzled class receptor 2 (FDZ2) is the only gene that is significantly upregulated in adenomyosis compared to DIE. FZD2 is discussed as an important trigger of TGF-ß induced epithelial–mesenchymal transition (EMT) [38] and cell migration [39]. Accordingly, the induction of EMT and ultimate fibrosis by TGF-β1 appears to play a critical role in the pathogenesis of adenomyosis [40]. In addition, EMT promoted by FZD2 also plays an important role in the metastasis of endometrial cancer [41], therefore suggesting that the invasive behavior of epithelial cells in FA has a cancerogenic aspect.

The two genes upregulated most in DIE compared to FA, Heat Shock Protein Family B (Small) Member 1 (HSPB1) and Calpain 2 (CAPN2), have not yet been described in endometriosis, despite the fact, that they play a role in proliferation and invasion in various solid tumors [42,43,44,45] and may serve this same function in DIE.

Catenin Beta 1 (CTNNB1), on the other hand, which is likewise upregulated in DIE epithelial cells compared to cells from adenomyosis, is discussed as a key factor in the regulation of proliferation and invasion of endometriosis [46, 47]. Since both papers demonstrate an upregulation of CTNNB1 in endometrial stromal cells in endometriosis lesions, we can raise the question of an additional important role for CTNNB1 action in the epithelium of endometriosis lesions.

Limitations of this study may include the following: [1] The group of patients suffering from deep infiltrating endometriosis is younger than the group of patients with adenomyosis who underwent hysterectomy. However, some of the younger patients were also on therapy with a GnRH analogue, which hormonally corresponds to a menopausal status and therefore attenuates any age-related differences. [2] By comparing adenomyosis and DIE without comparison to normal endometrium, we can only show relative expression differences, but not absolute differences compared to healthy tissue. Nevertheless, especially in view of the frequent co-occurrence of the two entities and the presumed common molecular origin, we consider our results at the gene expression level to be of further value, particularly with regard to possible different therapeutic approaches. [3] The sample size is relatively small with 7 patients with adenomyosis and 19 patients with DIE. Further studies with a larger number of patients to evaluate the possible influence of age, therapy concept or site of manifestation of DIE would be useful.

Conclusions

Deep infiltrating endometriosis and adenomyosis uteri differ significantly at the RNA expression level: for deep infiltrating endometriosis, the genes most expressed were those belonging to the PI3K pathway, and for adenomyosis, those belonging to the RAS pathway.