An Italian association study and meta-analysis with previous GWAS confirm WNT4, CDKN2BAS and FN1 as the first identified susceptibility loci for endometriosis
- Luca Pagliardini1,
- Davide Gentilini2,
- Paola Vigano’1,
- Paola Panina-Bordignon3,
- Mauro Busacca4,
- Massimo Candiani5,
- Anna Maria Di Blasio2
- 1Obstetrics and Gynecology Unit, San Raffaele Scientific Institute, Milano, Italy
- 2Molecular Biology Laboratory, Istituto Auxologico Italiano, Milano, Italy
- 3Reproductive Sciences Laboratory, Division of Genetics and Cell Biology, San Raffaele Scientific Institute, Milano, Italy
- 4Department of Obstetrics and Gynecology, Ospedale Macedonio Melloni, Milano, Italy
- 5Obstetrics and Gynecology Unit, Vita-Salute University, San Raffaele Scientific Institute, Milan, Italy
- Correspondence to Dr Paola Vigano’, U.O. Obstetrics and Gynecology, Scientific Institute San Raffaele, Via Olgettina, 60, Milano 20132, Italy;
- Received 22 August 2012
- Revised 3 October 2012
- Accepted 8 October 2012
- Published Online First 9 November 2012
Background Although endometriosis may benefit from primary prevention measures, the epidemiological risk factors identified are equivocal. Two genome-wide association studies (GWAS) have been conducted for endometriosis in two different ethnic populations but results are still to be replicated consistently and across various ethnicities. To confirm the association of GWAS-derived susceptibility loci, we conducted a replication Italian case-control study and a meta-analysis.
Methods An independent set of 305 laparoscopically-proven endometriosis patients and 2710 controls were recruited. Four SNPs—CDKN2BAS rs1333049, rs7521902 close to WNT4, rs12700667 in an inter-genic region on 7p15.2 and fibronectin 1 rs1250248—were selected for this association study.
Results Rs1333049 risk allele G frequency resulted significantly higher in endometriosis patients compared with controls (OR 1.32, 95% CI 1.11 to 1.57), confirming the role of this locus also in the Caucasian population. The meta-analysis showed that rs7521902 was associated with endometriosis at a genome-wide significance (pmeta=2.23×10−9) while for rs1250248, a genome-wide significant pmeta value of 3.89×10−9 was detected only in association with severe forms. An epistatic interaction between rs7521902 and rs1250248 (OR 1.56, p=1.19×10−2) was found especially in presence of ovarian disease (OR=2.15, p=3.12×10−4).
Conclusions We confirm WNT4, CDKN2BAS and FN1 as the first identified common loci for endometriosis.
Endometriosis is an oestrogen-dependent, chronic inflammatory disease that affects 2%–10% of reproductive age women, and associates with a significantly decreased quality of life and a high social economic burden.1 The definitive diagnosis of the disease, which is associated with an increased risk of ovarian cancer, can be obtained only by surgical examination of the pelvis. Primary prevention might be considered a valid health intervention in endometriosis based on the observed 40% risk reduction in oral contraceptive users.2 However, in the absence of specific risk factors, such a policy would imply systematic prescription of oral contraceptives to all adolescent women until they want to conceive, leading to major organisational problems and healthcare resources expenditure.2 ,3
In this context, the characteristic of familiar aggregation makes endometriosis worthy of genetic investigation. However, no candidate genes for the disease have been consistently replicated in different populations in previous association studies.4 Two genome-wide association studies (GWAS), the first in the Japanese population and the second from an Anglo-Australian consortium, have been recently conducted.5 ,6 The meta-analysis of the two studies produced a genome-wide significant p value for Single Nucleotide Polymorphism (SNP) rs7521902 on 1p36 close to WNT4, a critical regulator of proper postnatal uterine development, and of ovarian antral follicle development. However, other loci identified separately by the two GWAS remain intriguing and may also signal potential risks for endometriosis. The marker SNP rs10965235 that showed the strongest association (p=5.57×10−12) in the Japanese GWAS was located in intron 16 of CDKN2BAS that produces an antisense long non-coding RNA. In the Caucasian study, the strongest signal was observed for rs12700667 in an inter-genic region on chromosome 7p15.2 (p=1.4×10−9) and a second strong association was found for rs1250248 within the fibronectin 1 (FN1) gene (p=3.2×10−8 for more severe forms), although this association was not confirmed in their replication series.
Given the need to identify genetic predisposition factors for endometriosis with the aim of favouring primary prevention measures, and based on previous GWAS results, four SNPs were genotyped in an independent sample of laparoscopically-proven endometriosis patients, in women without laparoscopic evidence of endometriosis and in female general controls for replication study. Meta-analysis was applied to combine current replication data with GWAS results. Genotype–phenotype correlation between the SNPs and endometriosis was also analysed.
Subjects and methods
A total of 590 women were enrolled at the endoscopic surgical services of the Obstetrics and Gynecology of the San Raffaele Scientific Institute between December 2009 and December 2010 and endometriosis was laparoscopically and histologically documented in 305 (52%). Endometriosis was staged according to the revised American Society for Reproductive Medicine classification.7 Deep endometriosis was defined as previously reported.8 The remaining 285 women who underwent laparoscopy and in whom no endometriosis was found served as laparoscopic controls. Clinical characteristics of patients with endometriosis and diagnosis of laparoscopic controls are summarised in online supplementary table 1. The mean age±SD of patients with endometriosis and controls was 32.6±5.5 and 34.6±6.4 years, respectively. A second group of controls composed of healthy female blood donors was obtained from the genotype bank of Istituto Auxologico Italiano (general population controls, n=2425). All individuals studied were of Italian Caucasian origin. Approval for this study was granted by the local Human Institutional Investigation Committee.
We genotyped all the SNPs from GWAS-derived loci associated with endometriosis showing a genome-wide significant p value (p<10−7) in at least one stage of the previous GWAS analysis. For SNPs located in blocks of strong linkage disequilibrium (LD) (r2>0.8), we chose only the one with the highest OR. For the SNP rs10965235, which is monomorphic in the Caucasian population, we genotyped the SNP (rs1333049) located in the same block of LD presenting the highest minor allele frequency in the Italian population (based on HapMap frequency in TSI population). Ultimately, four SNPs—rs7521902, rs1333049, rs12700667 and rs1250248—were selected for replication study. Details of the investigated loci are summarised in online supplementary table 2.
For surgical samples, SNPs were genotyped using Taq-Man predesigned SNP genotyping assay from Applied Biosystem Inc. (ABI, Foster City, USA). Genotyping reactions were performed on an ABI 7900HT genetic analyser using 11 µl of genomic DNA (1.8 ng/µl) following the manufacturer's instructions. To validate the genotyping assay, 5% randomly selected samples were sequenced.
Controls from the general population were genotyped using Illumina 660W-Quad BeadChip (Illumina, Inc., San Diego, USA). Quality control of the data was performed according to the protocol described by Anderson et al.9
LD was assessed using the Haploview software (V.4.2, Broad Institute, Cambridge, USA). Goodness-of-fit χ2 test was applied to test for Hardy–Weinberg equilibrium in controls. PLINK (V.1.07, http://pngu.mgh.harvard.edu/~purcell/plink/) was used to perform the standard case–control association analysis and to analyse the presence of epistasis between SNPs with the binary test (SNP×SNP) implemented in the program. Meta-analysis was performed using METAL (http://www.sph.umich.edu/csg/abecasis/metal/). ORs with 95% CIs were used to measure the strength of the association. The Bonferroni correction was applied considering the number of statistical tests performed. Statistical power analysis was performed using G*power 3 (http://www.psycho.uni-duesseldorf.de/abteilungen/aap/gpower3/). Sample size was determined assuming an effect OR of 1.5 and types I and II error of 0.05 and 0.2, respectively.
Hardy–Weinberg equilibrium was performed separately in laparoscopic and general population controls by PLINK and resulted higher than 0.05 suggesting no deviation in control groups for the four SNPs.
Allele frequencies and OR (95% CI) for associations of rs1333049, rs7521902 and rs1250248 with endometriosis and the results of the meta-analysis with previous GWAS are presented in table 1.
We found genetic association between rs1333049 genotyped in the CDKN2BAS locus and endometriosis with similar OR for general and laparoscopically proven controls, although in the latter case association was only nominal (significant threshold after Bonferroni's correction for multiple testing p=1.25×10−2). Meta-analysis with previous GWAS could not be performed due to the monoallelic nature of the SNP identified in the Japanese study and the lack of evidence for any association in the same locus in the Caucasian study.
For rs7521902, close to WNT4, a higher frequency of risk allele A was found in endometriosis patients compared with general control group (OR 1.36, 95% CI 1.09 to 1.69; p=5.59×10−3). Meta-analysis with previous GWAS and the current replication data reached 2.23×10−9 (OR 1.20, 95% CI 1.13 to 1.27) implying a genome-wide significance.
We did not find statistical difference for rs12700667 in an inter-genic region on chromosome 7p15.2 between endometriosis women and the general population (OR 1.00, 95% CI 0.82 to 1.21) and for this reason we did not proceed with the meta-analysis with the reported Caucasian OR.
A borderline significant association was found between endometriosis women and general controls for the risk allele A frequency of rs1250248 in the FN1 locus (OR 1.20, 95% CI 0.99 to 1.45; p=5.64×10−2) and the meta-analysis reached only a pmeta value of 3.88×10−5.
However, when the allele frequencies were used to evaluate association with the different phenotypes of endometriosis (table 2), significant genetic association for rs1250248 in the FN1 locus was found with the manifestation of the disease in the ovarian site (OR 1.34, 95% CI 1.07 to 1.66; p=9.00×10−3).
Accordingly, meta-analysis with the reported Caucasian OR for more severe forms resulted in a stronger signal of association compared with overall cases, with a genome-wide significant pmeta value of 3.89×10−9 and an OR of 1.29 (95% CI 1.19 to 1.40) (see online supplementary table 3).
More severe stages of the disease were associated with a higher frequency of risk allele G of rs1333049 in the CDKN2BAS locus with an OR of 1.41 (95% CI 1.16 to 1.72; p=6.58×10−4). Conversely, the reported risk allele A for rs7521902 close to WNT4 was found more associated with peritoneal superficial implants with an OR of 1.51 (95% CI 1.14 to 1.99; p=3.56×10−3), supporting a role of the locus in the early steps of disease establishment.
We also investigated the potential presence of epistasis among the three SNPs that were associated with endometriosis in our study and found an epistatic interaction between the SNPs rs7521902 and rs1250248 (OR 1.56, p=1.19×10−2) that remained statistically significant after correction for multiple testing (significant threshold p=1.67×10−2). Analysis for presence of epistasis in the different phenotypes of endometriosis showed a stronger interaction between the same two SNPs in the presence of ovarian disease, with an OR for the interaction of 2.15 and a p value of 3.12×10−4 (significant threshold p=5.56×10−3).
Two recently published GWAS have identified susceptibility loci to endometriosis.
This association study represents:
the first evidence of association of the CDKN2BAS locus with endometriosis also in the Caucasian population in addition to the previously described Asian population5
the first replication study confirming FN1 rs1250248 as a susceptibility locus for endometriosis
a further confirmation that rs7521902, close to WNT4, is significantly associated with endometriosis.
The CDKN2BAS rs10965235 variant located on chromosome 9p21 has been previously identified in the Japanese GWAS. Although we could not replicate these data since this SNP is monomorphic in individuals of European descent, replication with rs1333049 located in the same LD block showed a statistically significant association with an increased OR for more severe stages of the disease. The 9p21 interval contains three candidate tumour suppressor genes: p15/cyclin-dependent kinase inhibitor (CDKN)2B, p16/CDKN2A and p14/ARF.10 The CDKN2BAS gene contains 19 exons and is transcribed in a 3834-bp long non-coding RNA in the antisense orientation of the p15/CDKN2B-p16/CDKN2A-p14/ARF gene cluster which is thought to be potentially regulated by this long non-coding RNA.11 ,12 Interestingly, homozygous deletion or methylation of p15/CDKN2B-p16/CDKN2A is a major hallmark of iron-induced carcinogenesis13 and the role of iron has been strongly emphasised in the context of endometriosis as peritoneal fluid, peritoneal tissues, macrophages of patients affected and endometriotic lesions appear to be loaded with iron.14
Present replication and meta-analysis of previously published studies also confirm rs7521902 in WNT4 region as a susceptibility locus for endometriosis since a genome-wide association significance was found. Although it is known that WNT4, a transcriptional factor from the WNT gene family, is involved in the genital tract development through WNT/β-catenin pathway,15–17 the functional significance of this SNP in endometriosis remains to be elucidated. The failure to find an increased association of rs7521902 with the more severe forms of disease does not support the theory that deep infiltrating disease implies the involvement of mullerian rests by a process of metaplasia.18
For FN1 rs1250248, although a statistical significant association could not be found with endometriosis cases in general probably due to the small sample size, there is evidence for contribution in risk for the ovarian disease. As a matter of fact, a significant association was found with the manifestation of the disease in the ovarian site and meta-analysis with the reported Caucasian OR for more severe forms including ovarian disease resulted in a genome-wide significance. Support to these findings derives from the consideration that this is the first replication study in which the diagnosis of cases was not self-reported and in which staging was very well documented. This indeed represents one of the strengths of the present association study. We deem important to emphasise these data since the ovarian disease is thought to have a specific pathogenesis different from the other manifestations.18
The inability to detect significant associations for rs12700667, even in more severe cases, is probably related to a limitation of the present study which is represented by the small sample size of cases; our study has indeed only 38% power to detect the previously reported effect of this SNP.6
Increasing evidence indicate that gene–gene interaction (epistasis) could be of major relevance in susceptibility to complex disease.19 ,20 For the first time, we have demonstrated an epistatic interaction between two SNPs associated with endometriosis, rs7521902 and rs1250248, where the presence of both risk alleles gives an increased susceptibility to develop endometriosis compared with the effects of the single SNP. Specifically, this interaction seems to play a pivotal role in the development of endometriosis at the ovarian site, although no common pathway for these two genes is known and more studies are needed to explain these results.
In conclusion, we support WNT4, CDKN2BAS and FN1 as common loci involved in the susceptibility to endometriosis. Since there are no known risk factors for the disease useful in the adolescent period and given the possibility of setting up primary prevention measures, the clinical implications of these findings may have a dramatic effect on the economic and social burden of this disease.
Contributors All authors contributed to conception and design of the study; PV, MB, MC and AMDB were involved in the enrolment of patients and in obtaining blood samples; LP, PPB, DG and AMDB were involved in performing genotyping and statistical analysis; all authors contributed to drafting or revising the content and approved the final version.
Funding This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
Patient Consent Obtained.
Data Sharing Statement All data included in this manuscript are available upon request.