Introduction

Endometriosis is a common estrogen-dependent gynecological condition of unknown etiology characterized by the extrauterine growth of endometrial-like epithelium and stroma. Endometriosis affects 1–2% of all females and 10% of women during their reproductive lifespan, up to 50% of infertile women, and 87% of women with chronic pelvic pain [1–3]. On average a definitive diagnosis is typically achieved by laparoscopy 5.4–12 years after the onset of symptoms [4–6]. Laparoscopy remains the gold standard for diagnosis, and women with endometriosis can expect to undergo multiple diagnostic and operative laparoscopies over the course of their disease [7, 8]. Diagnostic delay, cost, surgical risk, and poor correlation between symptoms and extent of disease are the basis for arguments to shift from surgical to clinical diagnosis [9, 10]. Thus, there is an urgent need to identify novel clinical markers of endometriosis [9, 11], which remains a high-priority research objective [9, 11, 12].

Many biomarkers have been proposed in the literature; however, none have been found, either alone or combined in a panel of markers, to yield sensitivity and specificity equivalent to laparoscopy [13–15]. In recent years, exosomes and their cargo [16], tissue concentrations of circular RNA [17], and long noncoding RNA [18] as well as circulating concentrations of microRNAs (miRNA) [19–21] have been documented in women with endometriosis compared with control populations and thus are thought to have potential value as candidate diagnostic markers of endometriosis [22, 23]. Of these biomarkers, miRNA have received the greatest attention in the literature. Furthermore, miRNA are thought by some to show the greatest diagnostic promise. Briefly, miRNA are very short fragments of RNA ∼19–25 nucleotides long, which post-transcriptionally regulate protein expression through binding to and inhibiting the translation of mRNA transcripts into protein [24] or enhancing mRNA instability leading to its degradation. MiRNAs are synthesized in the cytoplasm from nucleic hairpin intermediates (pre-miRNA) [25], which are then processed to yield mature miRNA that resist RNase degradation [26]. MiRNA form an RNA-induced silencing complex with Argonaute, Dicer, TAR RNA binding protein, and protein activator of PKR to post-transcriptionally regulate genes by binding to the 3′ region of the mRNA transcript and inhibiting translation [27]. While increased miRNA expression is associated with mRNA downregulation, lower miRNA expression is expected to be associated with increased target gene mRNA expression and translation resulting in increased protein expression. MiRNAs are present in body fluids including blood, either contained in exosomes [16] or bound to protein complexes that make them more stable than circulating protein concentrations and thus better candidate markers of disease [28]. One miRNA can target several mRNAs and one mRNA can be targeted by several different miRNAs [29]. Greater than 2000 mature human miRNA sequences have been identified and are thought to regulate ∼50% of all protein-coding genes. Although widely studied in cancer [30–32], the potential of miRNAs as diagnostic markers of endometriosis remains to be established.

Circulating levels of several miRNAs differ in women with endometriosis compared with control groups [19–21, 33–38]. Despite initial enthusiasm for the diagnostic potential of circulating miRNA levels, important weaknesses have been noted [37, 39]. Specifically, of the numerous miRNAs reported to differ in women with endometriosis compared with controls, very few miRNAs [19, 20, 34, 35] have been replicated in more than one study. The lack of replication and poor sensitivity and specificity are reasons creating perception that miRNAs have questionable value as diagnostic markers of endometriosis [37, 39]. Although this opinion is understandable, we postulate that differences in study design and methodological issues are substantial, and thus, the true potential of circulating miRNA for the diagnosis of endometriosis remains to be defined. While recent reviews have summarized differential expression of miRNA in the circulation of women with endometriosis compared with controls [23, 40, 41], levels of miRNA differentially expressed in recent studies suggest greater replication of results than previously appreciated. Moreover, the reasons for the lack of replication of results across studies have not been critically evaluated previously. Therefore, the objective of this review was to update the literature and critically evaluate study design and methodological approaches used to identify miRNA as potential diagnostic markers of endometriosis. Herein, important issues contributing to uncertainty and future directions to enhance diagnostic reliability of circulating miRNA are described.

Materials and methods

An electronic search of the literature published between the years 2000 and 2020 was carried out (August 2020) using PubMed and Web of Science. Keywords used included “endometriosis,” “biomarker,” “diagnosis,” “microRNA,” and “miRNA.” Inclusion criteria included articles describing primary research, publication in a peer-reviewed journal, and availability of full-text publication in English. Exclusion criteria employed included review articles, commentaries or letters to the editor, and meeting summaries. Titles of all articles identified in the electronic database searches were copied to an Excel spreadsheet and duplicate titles identified. Duplicate articles were removed and abstracts for each paper were independently reviewed by authors (A.L. and W.G.F.) for inclusion and/or exclusion. Disparate decisions were resolved through discussion until consensus was achieved. The full text of all included papers was printed for review. Reference lists from included articles and review papers were mined for additional titles potentially missed by our electronic search strategy (snowballing). Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were followed throughout the process [42].

Results

A total of 462 article titles were identified through PubMed and Web of Science search. Of the retrieved article titles, 241 were duplicate titles and removed. Review articles (n = 113), meeting summaries (n = 25), and patent summaries (n = 22) were also excluded resulting in a list of 61 articles that were examined for relevance, from which 18 studies met inclusion criteria (Figure 1).

Study design

Although all studies reviewed employed a case–control study design, the characteristics of the study population with endometriosis and control groups differed widely. In all studies included in our review (Table 1), the presence of endometriosis was confirmed by laparoscopy; however, American Fertility Society (AFS) stage was not reported in two studies [43, 44], whereas all disease stages were included in the majority (12/18, 66.7%) of studies [20, 34–39, 45–49]. Three studies [19, 21, 33] included only women with moderate–severe endometriosis (AFS Stages III–IV) or only women with minimal–mild disease (AFS Stages I–II) were included in one study [50]. The reporting of comorbidities was inconsistently described in the reviewed studies. Similarly, the characteristics of the reference populations used for comparison differed markedly across studies. Women symptomatic for endometriosis were included exclusively as the reference population in the majority (11/18, 61.1%) of the studies, whereas women with other benign conditions (e.g., prolapsed uterus, ovarian cyst, or urinary incontinence) were employed as a control population in several other studies. Infertility caused by tubal factors lacking surgical evidence of endometriosis served as the control group in one study [48]. In two studies [34, 35], circulating miRNA levels in women with endometriosis were compared separately with results from two distinct control groups separately (asymptomatic and symptomatic). One study employed three different control groups (healthy women, symptomatic, and endometriosis-associated ovarian cancer) for separate comparisons with the case group [43]. Therefore, it is challenging to compare results across studies with disparate inclusion and exclusion criteria.

Figure 1.

Search strategy employed for study identification and selection for exclusion vs. exclusion in accordance with PRISMA guidelines.

Differential miRNA expression

In the 18 studies included in our review, circulating levels of 63 distinct differentially expressed miRNAs were found in women with endometriosis compared with a control group (Table 1). Expression of miRNAs were both upregulated in plasma (n = 12) and serum (n = 22) as well as downregulated in plasma (n = 18) and serum (n = 25) in these studies (Table 2). While many of the reported differentially expressed miRNAs were unique to individual studies, we found that 17 of 63 differentially expressed miRNAs (27.0%) were significantly dysregulated in more than one study; however, the direction of dysregulated expression was common for just 14 (22.2%) miRNAs (Table 3). Specifically, circulating levels of miR-17, miR-20a, miR-139-3p, miR-141, miR-320a, miR-3613, and let-7b were downregulated in more than one study. In contrast miR-18a-5p, miR-122, miR-125, miR-199a, miR-451a, miR-150, and miR342 were significantly upregulated in women with endometriosis compared with controls in more than one study. Furthermore, miR-20a and miR-17 were downregulated in three studies [21, 49, 50] regardless of whether serum or plasma was used.

Although dysregulated in more than one study, the direction of dysregulated expression for several other miRNAs varied across the studies. For example, in one study [20], circulating miR-145 expression levels were lower in women with endometriosis compared with controls, whereas expression was upregulated by another group of investigators [36]. Similarly, miR-22-3p [21, 51] and miR-143-3p [19, 49] were differentially expressed in independent studies, although the direction of dysregulated expression compared with controls was opposite, and thus, results were not considered to be replicated. MiR-199a was upregulated in two studies [20, 45] using serum-derived miRNAs but downregulated when collected from plasma [49]. We also identified one study that employed miR-199a as a housekeeping gene to normalize the expression of other miRNAs [46]. We note that the reference gene employed in the reviewed studies differed across studies (Table 1). Moreover, details concerning stable expression in cases and controls and validation steps used to select reference genes could not be discerned from the methods described. Exactly half (9/18, 50%) of the studies reviewed used U6 as a reference gene [19, 20, 33, 35, 38, 39, 46, 49], whereas miR-16 was the reference gene used in two other studies [21, 52]. Multiple different individual reference genes were used in the studies reviewed [36, 43, 46, 50], whereas several other groups used the average of two or more reference genes to normalize miRNA expression [35, 39, 49].

Table 1.

Summary of differentially expressed hsa-miRNA after RT-PCR validation in women with endometriosis compared with controls, reference miRNA employed, and study characteristics. The sensitivity and specificity of individual miRNA and panels of miRNA in discriminating between women with endometriosis and control groups

Continued

Table 2.

Magnitude and direction of hsa-miRNA dysregulated in women with endometriosis compared with controls

Table 3.

Experimentally confirmed expression in the eutopic and ectopic endometrium of the replicated differential hsa-miRNA expression in the circulation of women with endometriosis compared with controls and reported function relevant to the pathophysiology of endometriosis

Discussion

Results of the present review revealed that, despite differences in study design and methodologies employed, 63 miRNAs were differentially expressed in women with endometriosis compared with controls. From this population of miRNAs, circulating levels of 17 miRNAs were differentially expressed in women with endometriosis compared with controls; however, the direction of the differential expression was the same for just 14 miRNAs. The lack of reproducibility of results across studies remains a concern and the absence of independent validation in different cohorts of women with endometriosis supports the prevailing view that circulating miRNAs are currently of limited value for the diagnosis of endometriosis. Yet, important study design and analytical differences render direct comparisons of studies problematic and prevent firm conclusions about the clinical value of circulating miRNAs.

While the results of our review showed that 20% of the circulating levels for miRNA differentially expressed in women with endometriosis compared with controls were replicated in more than one study, the lack of broader replication of results across studies continues to be a source of concern. Critical appraisal of the available studies revealed several important study design and methodological issues that we postulate account for the poor reproducibility of results across studies. Inconsistent reporting of comorbidities and differences in characteristics of the case and control populations employed in the studies prevent direct comparison of results across studies. Enrolling self-reported healthy individuals into a control group neglects asymptomatic cases of endometriosis and thus confounds the data. Prevalence of endometriosis ranges between 2% and 43% of asymptomatic women undergoing surgery for tubal ligation, 5–50% among infertile women, and to 5–21% among women hospitalized for pelvic pain [69]. In addressing this issue, we suggest that a reference population should include women for whom symptoms suggest endometriosis that cannot be explained by other causes.

Failure to account or adequately control for the effect of the menstrual cycle stage on the circulating levels of miRNA potentially differentially expressed in women with endometriosis compared with controls is another important variable that we postulate accounts for the poor reproducibility of results in the literature to date. Although six studies in the present review concluded that circulating miRNA levels are unaffected by the menstrual cycle stage [19, 20, 34, 35, 37, 46], two studies found the opposite [33, 39]. We suggest that sample sizes employed in the studies reviewed in the current report are too small to allow for stratification of study results across the menstrual cycle stage for case and controls to convincingly resolve this issue. Further support for our proposal is derived from evidence that DROSHA and DICER, two enzymes central to the production of miRNA, and different miRNA are differentially expressed in the endometrium across the menstrual cycle [70, 71], suggesting that miRNA expression is under gonadal steroid control. Moreover, progesterone treatment was critical in regulating miRNA expression in endometrial cells in culture [53]. Consequently, the influence of circulating gonadal steroids on miRNA expression throughout the menstrual cycle cannot be underestimated and is likely a contributing factor to the poor reproducibility of circulating levels of differentially expressed miRNAs reported across studies to date.

In the present review, different RNA extraction and quality control methods and miRNA stability potentially contribute to differences in circulating levels of miRNA expression detected in different studies. Isolation methods and the amount of RNA input are known to affect study outcomes [72]. Additional critical issues affecting the reliability of miRNA data include miRNA sequence database errors, suboptimal RNAextraction methods, detection assay variability, the prevalence of online resources for bioinformatic analyses, and nonstandardized statistical analyses for miRNA clinical testing [73]. Furthermore, miR-122, which was dysregulated in two studies discussed above [20, 45], is a signature of sample contamination if found outside of the hepatic tissue [74]. MiR-21 is another ubiquitously expressed miRNA that is differentially expressed in women with polycystic ovarian syndrome [75], and thus, we suggest that this cannot serve as a reliable diagnostic marker of endometriosis [74]. We suggest that incorporation of spike-in markers, inactivation of nucleases, as well as protein digestion and collection of RNA in small volumes to avoid freeze–thaw cycles would be helpful. Sample quality was generally not considered in the studies identified in our search. Sample hemolysis is another potential explanation for the poor reproducibility of miRNA results in the reviewed studies. Several hemolysis susceptible miRNA [76] including miR-17 [21] and miR-451a [19, 34] have previously been advanced as potential diagnostic markers of endometriosis. Hence, dysregulated levels of these miRNA could be a consequence of sample handling rather than the disease state [35, 77]. Only three studies [39, 46, 49] out of eight that used plasma reported a check for sample hemolysis and exclusion of contaminated samples. Therefore, we suggest the incorporation of a hemolysis control when plasma is used.

While it is generally recognized that expression data must be normalized by use of a suitable reference gene, an agreement is lacking concerning the most appropriate miRNA reference to use. While the majority of studies considered in the present review employed U6 to normalize miRNA expression of circulating miRNA levels, U6 has been repeatedly shown to have high variation and is thought to be unsuitable for use to assess differential expression of circulating miRNA levels [78]. MiR-16 is another reference gene whose use in endometriosis studies is suspect since levels are affected by inflammation [52], a central feature of endometriosis, and thus may not be a suitable reference gene. We note that the Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) provides a set of guidelines designed to promote consistency between laboratories and increase experimental transparency [79]. Consistent with MIQE guidelines we suggest that two or more reference genes should be employed to normalize miRNA expression. The average of two or more reference genes was employed in several of the studies reviewed [35, 39, 49] and may be more reliable [80], although these reference genes have yet to be appropriately validated for this purpose as previously described [81, 82]. Hence, it is not surprising that, with wide disparities in study design, sample processing, and miRNA expression analysis, replication of study results is poor and the diagnostic potential of miRNA alone or in panels remains limited. However, we suggest that not all is lost and that studies to date have provided results that are both encouraging and highlight methodological issues that when addressed should improve the reproducibility of results.

Conclusions

Circulating levels of differentially expressed miRNAs in women with endometriosis compared with controls are poorly replicated across studies. Direct comparison of results across studies is problematic owing to important study design and preanalysis and analytical issues. Therefore, the clinical value of circulating levels of differentially expressed miRNAs for the diagnosis of endometriosis cannot be determined from the literature cited in this review. We suggest that replication of circulating levels of some differentially expressed miRNAs in more than one study together with the promise shown by several miRNA panels, expression in the endometrium, and experimental evidence for a regulatory role in the pathophysiology of endometriosis are encouraging developments. Thus, further studies in well-designed studies that address the weaknesses noted with the present body of literature are warranted.

Author contributions

The following authors contributed to project concept and study design (A.L., S.K.A., M.L., and W.G.F.), literature search and data analysis (A.L., V.E.T., and W.G.F.), and all authors contributed to the preparation, review, editing, and approval of the manuscript for submission.

Conflict of interest: The authors have declared that no conflict of interest exists.