Introduction

Endometriosis is a common non-malignant gynecologic disease that affects 5–10 % of women of reproductive age [1]. Pelvic pain, metastatic pattern, and infertility are the most common features of the disease. It is characterized by the presence of functioning endometrial-like tissue (epithelium and stroma) outside the uterus [1, 2]. Several models have been proposed to explain the pathogenesis of endometriosis. The most accepted model is the retrograde menstruation; however, other factors may play a role in the disease occurrence including abnormal immune response, hormonal factors, genetic predisposition, and exposure to environmental factors [1, 3, 4].

Although the increasing evidence supports a genetic component to this disease, the etiology and pathophysiology remains unclear. It is likely that endometriosis is a complex polygenic and multifactorial disease, caused by multiple genetic and environmental factors [5, 6]. Considering that endometriosis has failed to show classic Mendelian inheritance, different types of genetic approaches have been undertaken to study its molecular basis [7, 8].

Candidate genetic association studies comparing the frequency of single nucleotide polymorphisms (SNPs) in a patient vs. a control group have been widely conducted [9]. The choices of the candidate genes are based on biological mechanisms thought to contribute to the susceptibility, development, and progression of diseases, such as endometriosis [10]. Studies have demonstrated that structural variations on DNA, mainly copy number variations (CNVs), have been considered as important keys in the complex traits. The identification of disease-associated rare CNVs may help to explain some missing heritability that could not be explained by common SNPs [11].

Classical cytogenetic, molecular genetics, and molecular cytogenetic techniques applied to endometriosis have led to the identification of consistent somatic genetic alterations [7]. A linkage study that adopted a positional-cloning approach identified a significant susceptibility locus for endometriosis on chromosome 10q26 involving the genes EMX2 and PTEN [12]. Array-comparative genomic hybridization uncovered alterations at 20q13.33 that were associated with ovarian endometriosis [13]. Additionally, studies using high-density genotyping arrays have reported copy number variation loci in 6.9 % of affected women compared to 2.1 % in the general population [14]. Taken together, these findings strongly suggest the involvement of genomic aberrations in the development of the disease.

High-resolution genomic approaches, such as microarray-based genotyping, have been performed successfully in the mapping of hundreds of thousands of SNPs, and consequently, in the discovery of new genomic regions associated with multifactorial diseases [15]. In the present study, we used DNA pooling methodology and high-density SNP array as a cost-effective alternative to evaluate CNVs in endometriosis patients.

The aim of the current study was to search for genomic regions that might contribute to the development of endometriosis by identifying CNVs significantly associated to this disease compared to healthy controls.

Materials and methods

Subjects

Five hundred sixty-four infertile women with endometriosis (mean age 35.1 ± 3.9 years) from the Human Reproduction and Genetics Center of the Faculdade de Medicina do ABC (SP; in the southeast of Brazil), participated in this case-control study. The patients were diagnosed with endometriosis by laparoscopy/laparotomy and classified according to the American Society for Reproductive Medicine [16] with obligatory histological confirmation of the disease. In this group, minimal/mild (stage I and II) endometriosis was found in 229 patients (40.6 %) and moderate/severe (stage III and IV) endometriosis in 335 patients (59.4 %).

The investigation into the cause of infertility included a hormonal and biochemical profile, testing for sexually transmitted diseases, imaging examinations, investigation of genetic and/or immunological abnormalities, hysterosalpingography, hysteroscopy, laparoscopy, and semen analysis of the partner.

Six hundred fifty-two fertile women (mean age 39.2 ± 5.8 years), from the Family Planning Outpatient Clinic of the Faculdade de Medicina do ABC, who were among subjects evaluated for tubal ligation participated as control group. In all of them, absence of endometriosis was confirmed by inspection of the pelvic cavity during the laparoscopy.

Genomic DNA for each individual was extracted from peripheral blood according to as previously described [17].

DNA pool construction

Pooling consists of combining multiple DNA samples for the purpose of screening many subjects simultaneously [18]. The construction is based on the addition of equal amounts of DNA from each individual to either patient or control pools [19].

A total of 150 samples were analyzed according to DNA pooling methodology. To establish homogeneous pools, the samples selected from the endometriosis group included 100 patients who did not have ovulatory and endocrine disorders, Mullerian defects, or autoimmune diseases and whose partner presented any male factor associated with infertility. In relation to the control group, we selected 50 samples per pool without medical history of autoimmune diseases, pelvic pain complaints, and spontaneous miscarriage.

DNA concentrations in each sample were measured using the spectrophotometer NanoDrop 2000 (Thermo Scientific, CA, USA) for the quantification of double-stranded DNA. DNA samples were diluted to a final concentration of 50 ng/μL using 1x Tris–EDTA (TE). Pools were constructed by combining equal volumes (10 μL) of each DNA sample. Each pool was composed by 50 samples. A total of three pools were constructed for the group of endometriosis and subdivided according to the stages of the disease (pool 1: minimal/mild endometriosis, pool 2: moderate/severe endometriosis, and pool 3: control group). The scheme in Fig. 1 represents the distribution of individuals subjected to the SNP array experiments and the following validations.

Fig. 1
figure 1

Distribution of subjects analyzed in each step of the study. To establish homogeneous pools 50/229 patients with EDT I/II, 50/335 EDT III/IV and 50/652 controls were selected (EDT endometriosis, SNP single nucleotide polymorphism, MLPA multiplex ligation-dependent probe amplification, qPCR quantitative polymerase chain reaction)

Full size image

High-density SNP genotyping

Once equimolar amounts of each sample were combined, the genomic DNA pools were assayed using the Illumina protocol for individual genotyping. Briefly, 750 ng of pooled genomic DNA was labeled and hybridized to the Illumina HumanOmni 2.5 BeadChip (Illumina, San Diego, CA, USA), which interrogated ∼2.5 million SNPs. The pools were genotyped in duplicate, serving as technical replicates.

In addition, 48 samples that composed the pools were selected to be genotyped individually and used as experiment controls. Genotyping using the Illumina Human OmniExpress BeadChip, which interrogated >715,000 SNPs was performed in 24 samples of the control group and 24 samples of the patient group (only endometriosis stage IV).

Data analysis

BeadChip data was processed using GenomeStudio Software v.3.1 (Illumina Inc.) according to the manufacturer’s recommendations. Primary data analyses, including raw data normalization, clustering, and genotype calling were performed using algorithms in the genotyping module (GT). The software provided log R ratios (LRRs) and B allele frequencies (BAFs) for each probe on the SNP array.

The Software UPDG [20] was used to perform the data analysis of the pooled DNA. Raw data were manipulated for correcting the preferential allelic amplification, normalization, and analysis of variance for genetic association test.

Standard quality control

To allow cross-platform evaluation, we kept only common SNPs between Omni 2.5 and OmniExpress beadchips. Standard quality control steps were as follows: SNPs with minor allele frequency (MAFs) >0.05, missing rates <0.05, or P value >0.0001 for the Hardy–Weinberg equilibrium test were included. Principal component analysis (PCA) was applied considering all valid SNPs to calculate genetic distances and account for population stratification in subsequent association analyses. Data processing was performed using PLINK (version 1.07) [21].

CNV analysis and post-CNV calling

CNVs were called using PennCNV algorithm, which considers the total signal intensity and allelic intensity ratio at each SNP, the distances between SNPs, and the frequency of each SNP via a hidden Markov model (HMM) [22]. The software is capable of processing integer copy numbers, according to a six-state definition: state 1 = deletion of two copies (copy number 0), state 2 = deletion of one copy (copy number 1), state 3 = two-copy state (copy number 2), state 4 = two-copy state with LOH (copy number 2), state 5 = single-copy duplication (copy number 3), and state 6 = double-copy duplication (copy number 4). A copy number state of two per individual was considered normal (one copy per chromosome); CNVs with copy number >2 were defined as duplications, while those with copy number <2 were considered deletions.

The Software BEDTools [23] of genomic arithmetic integrated the CNV data from the individual and pooled sample analysis, merging the CNV regions that were common among them. The most relevant regions were saved for subsequent validation.

Defining the CNV Loci

CNV calls were grouped into loci having at least 1 kb in length. A confidence score of 10 probes minimum was used as threshold to classify reliable CNV calls [24]. To identify if the CNV loci were common or novel, we compared our results with those published in the Database of Genomic Variants (DGV) [25]. We used the gene annotation of the University of California Santa Cruz (UCSC) Genome Browser [26] to identify genes that were located within or partially overlapped with CNV loci. All annotated CNV genomic regions were based on the Human genome build 19 (hg19).

Validation of CNV calls

For validation of the selected CNVs, the loci were subdivided into subtelomeric and intrachromosomal, considering that the validation strategy for each of these findings was performed differently. The same 150 subjects (100 patients and 50 controls) used in the SNP array were analyzed in the MLPA experiments. For qPCR, all 564 endometriosis samples and 652 controls included in the study were analyzed.

Multiplex ligation-dependent probe amplification

Deletions and duplications in subtelomeric regions were analyzed using the multiplex ligation-dependent probe amplification (MLPA) technique with the kit SALSA P070 Human Telomere probe mix, according to the manufacturer’s protocol (MRC Holland, Amsterdam). The probe amplification was performed in the thermocycler MasterCycler Gradient (Eppendorf) and the fragment analysis through capillary electrophoresis using the ABI 3500 Sequencer (Applied, Life Technologies). MLPA results were analyzed through GeneMarker Software (Softgenetics®, LLC, State College, PA, USA).

Real time qPCR

Four intrachromosomal loci were selected and subjected to TaqMan CNV detection analysis. Real-time qPCR was performed to validate the candidate CNVs in the FCGBP, NR5A1, PTGES2, SLC25A25, and CXXC5 gene regions, and Taqman assays Hs02788922_cn, Hs00492415_cn, Hs01128655_cn, Hs02406760_cn and Hs06063192_cn were used for genotyping each locus, respectively. TaqMan Copy Number Reference Assays (RNaseP and TERT) were used as references (internal control). Reaction plates (384-well) with a mixture of TaqMan Fast Advanced Master Mix (Applied Biosystems, USA), target and reference TaqMan Assays (Applied Biosystems, USA) and 10 ng DNA/well were prepared using a Biomek FX (Beckman Coulter, Fullerton, CA, USA). Experiments were performed according to the manufacturer’s instructions and ViiA 7 Real-Time PCR System (Life Technologies, CA, USA) was used for the quantitative PCR reactions (qPCR). The reactions were done in triplicate and the results analyzed based on the delta-delta Ct method were imported to Copy Caller Software v.2.0 (Applied Biosystems Inc., USA) to determine copy numbers. A confidence level of 95 % and |z-score| value of <1.75 was applied to call the CNVs.

Statistical analysis

Fisher’s exact test was used to carry out the case-control association analysis of CNVs identified. The significance threshold was chosen as 0.05. For comparison of CNV sizes and mean numbers between patients and controls, two-tailed Mann–Whitney U test was used.

Results

A basic association check was performed between the SNPs and phenotype to investigate possible systematic biases in the association results, revealing a genomic inflation factor (GIF) of 1.41082. Outliers were subsequently dismissed, and the data was corrected by three components of ancestry that reduced the GIF value to 1.11251, which we considered acceptable.

The combined analysis of individual and pooled samples using BedTools Software was performed to integrate the CNV regions that were common among them, revealing 49 CNV loci present in patients with endometriosis and absent in the control group. Of these, 33 were kept and 16 were excluded from the study—6 due to size smaller than 1 kb and 10 for absence of the minimum number of probes required to be considered a reliable CNV.

The CNV sizes ranged from 4 to 61 kb with a median size of 19 kb. These regions were studied and screened according to length, presence of overlapping regions in DGV, chromosomal location, and overlapping of genes and its functions. The results were divided into subtelomeric and intrachromosomal CNVs.

MLPA was used to confirm the CNVs located at subtelomeric regions that represented 48 % (16/33) of the CNVs of interest (Table 1). Of these, six showed agreement between the MLPA and SNP Array analysis. The regions 1p36.33, 16p13.3, 19p13.3, and 20p13 represented gains while 17q25.3 and 20q13.33 showed losses. The non-validated regions were considered to be false positives due to the limited specificity of the DNA-pooling methodology that tends to overestimate the effect size and were not pursued any further [27].

Table 1 Characteristics of the loci located in the subtelomeric regions
Full size table

Among the 33 CNV loci, 17 (52 %) were intrachromosomal. Two of them were not involved with any gene, while 15 were located within intergenic and gene coding regions. All investigated regions were previously reported on DGV (Table 2).

Table 2 Detailed CNV findings and characteristics of intrachromosomal regions
Full size table

Of note, a 4678-bp duplication (chr19 40,360,845–40,365,523) involving the FCGBP gene (Fig. 2) was observed in 8 patients and was absent in controls, thus conferring significantly increased endometriosis risk (p = 0.007).

Fig. 2
figure 2

Graphical representation of the duplication located within the FCGBP gene visualized at the UCSC genome browser view (version hg19)

Full size image

We detected a 38,721-bp deletion, involving the SLC25A25 and PTGES2 genes (chr9 130,850,879–130,889,600), impacting 4 patients and in 6 controls. Thus, we could not validate this locus (p = 0.92). The results of NR5A1 and CXXC5 were also not validated by qPCR.

Discussion

Large-scale genotyping using SNP microarrays has been performed successfully in the discovery of new genomic regions that are associated with multifactorial diseases [15, 28]. The methodology assesses variation across the genome using computational models to compare the genotypes of people with and without disease to identify variants that are associated with disease.

Even though the cost of SNP arrays was significantly reduced in the last years, it remains prohibitive for numerous research centers to genotype a large number of samples [18, 19]. One way to reduce the cost is to use the DNA pooling strategy, an alternative and attractive method that can unveil loci that are highly informative. However, DNA pooling has limitations and the most common is the difficulty to access genotyping frequencies. Instead, this methodology is used to estimate allele frequencies, a process referred as “allelotyping” [29].

DNA pooling has been originally used as an economic and alternative tool for genotyping and GWAS; therefore, it can be applied to discover common CNVs in certain studies [30, 31]. However, the copy number estimation represents a challenge in terms of false-positive rates, so it has been suggested that CNVs identified from SNP genotyping data must always be validated with an alternative method to avoid erroneous calls [31, 32]. A past study with schizophrenia in Brazilian samples also used pooling strategy to evaluate the CNVs and the possible role in the disease. As in our study, they were able to identify few CNVs with relevant association with the phenotype, but they also revealed the limitations of pooling in admixture population and the risk of false-positive calls [33].

In the present study, we have carried out a SNP array analysis through a pooled-sample strategy in a Brazilian cohort with endometriosis. The high-resolution SNP array used was able to produce enough data to support CNV detection, thus enabling integrated analysis of SNPs and CNVs in the same patients.

Our findings revealed 33 CNV loci in common between-individual analysis (1 array per individual) and pooled samples, which were present in patients with endometriosis and absent in the control group. Of these, 16 were located in the subtelomeric regions with the 1p36.33, 16p13.3, 17q25.3, 19p13.3, 20p13, and 20q13.33 loci confirmed through MLPA validation.

Our results are in agreement with previous studies that demonstrated an association between 1p36.33 and endometriosis [34–36]. This locus contains markers located in or near the WNT4 gene, which plays an important role on the development of the female reproductive tract and steroidogenesis [37, 38]. The literature emphasizes the role of the WNT4 as a regulator of cell proliferation and differentiation, in which the signaling pathway involves proteins that directly participate in both gene transcription and cell adhesion [39]. Based on these biological functions, the duplication involving the 1p36.33 locus is likely to be involved on endometriosis development.

Genomic imbalance at the 17q25 and 19p13 loci was also found to be associated with endometriosis by Veiga-Castelli et al. [40] in a comparative genomic hybridization (CGH) study conducted in Brazilian women. In another study, using the same samples, the authors observed a differential expression of the genes MXRA7 and UBA52, located respectively in 17q25 and 19p13, suggesting that these alterations could lead to the development, establishment, and maintenance of the ovarian endometriomas [41].

A Chinese study by Yang et al. [13] revealed an association between 20q13.33 duplication and ovarian endometriosis. Their array-CGH results were validated through qPCR, confirming association of the genes GATA5 and SLCO4A1 in the gain region. In the present study, we also reported an evident contribution of this locus, but instead of duplication, our subjects presented a 3-kb deletion (chr20:60,961,271–60,964,301) involving both the exonic and intronic parts of RPS21, a gene responsible for encoding ribosomal proteins. In Drosophila, the complete absence of the RPS21 gene leads to excessive cell proliferation in specific tissues [42].

The 16p13.3 and 20p13 loci have never been associated with endometriosis prior to our study. A report that evaluated a 16p13.3 duplication in 12 patients reported normal to moderate mental retardation, mild arthrogryposis-like anomalies of the musculoskeletal system, mild facial dysmorphism, and occasional anomalies of the heart. In 2 patients, the duplication was inherited from an apparently normal parent, with de novo occurrence found in 10 of 12 patients, indicating that this duplication is associated in most patients with a reduced reproductive fitness [43]. In relation to 20p13, a significant linkage was previously reported at this locus and associated with intellectual and developmental disabilities [44].

Between the 17 CNV loci located within intergenic and gene coding regions, 3 represented gains, whereas 14 were losses. From these 17 loci, 4 were selected and individually checked by qPCR. The selection parameters were based on the respective gene functions of the gene candidates present in each locus and their possible correlation with the development and progression of endometriosis.

The CN gain on chromosome 19q13.2 spans 4 kb and contained ten SNP probes. It was observed in eight of the endometriosis patients, while none of the controls displayed a gain in this region (p = 0.007), suggesting that this locus is a promising risk-conferring candidate. According to the UCSC Genome Browser and Database of Genomic Variants [25, 26], this duplicated region is located within the FCGBP gene. Characterized by the production of a large (encoded by a 17-kb mRNA) mucin-like protein that binds the Fc portion of IgG molecules, this gene has been reported as differentially expressed in gallbladder [45], prostate [46], thyroid [47], lung [48], colon [49], and ovarian cancer [50]. Present in serum, the protein can be found in higher levels in patients with autoimmune diseases [51]. It has been suggested based on its IgG Fc binding property and tissue distribution that FCGBP might play a role in cell protection and anti-inflammation in tissues. Our data suggest for the first time that the duplication on the FCGBP gene could predispose to the progression of endometriosis. Thus, functional studies will be required to elucidate the exact contribution of this variant to the disease risk.

Population stratification was a concern, particularly in the current study, because the Brazilian population is one of the most heterogeneous and admixed populations in the world, formed mainly by the admixture between European, African, and Native American populations [52, 53]. After removal of SNPs through the sequential QC processes and correction by three components of ancestry, we observed genomic inflation factor of 1.11251. Thus, the inflation of false-positive rates on genetic association is within an acceptable level (genomic inflation factor <1.1) [54] for a case–control association study, indicating that the current QC processes are successful.

It has been suggested that CNVs identified from SNP genotyping data must be validated to avoid false-positive results [32]. Therefore, the identified CNVs were further validated by MLPA and qPCR in an extended sample set. The discovery of how these CNVs act together can be an important step in discovering the susceptibility, establishment, and progression of endometriosis.

Despite extensive research, the varied clinical presentations among patients, such as staging, pain, and infertility remain unclear. Genetic modifying factors are thought to underlie this variability. This study represents an effort to enlarge our knowledge about endometriosis risk genes through a genome-wide copy number variation analysis in the Brazilian population. While requiring independent validation, our novel findings contribute to the understanding of the complex pathways leading to endometriosis.

Conclusion

In summary, we identified, CNVs previously associated with endometriosis and we have uncovered novel chromosome regions that may contribute to the pathogenesis of this disease. Further large-scale discovery and replication studies will help determine the biological impact of these CNVs and eventually provide clues for the underlying mechanisms of endometriosis.