Endometriosis is characterized by chronic painful symptoms and infertility (1
). There is strong evidence suggesting that endometriosis is a multifactorial disease resulting from interactions of multiple susceptibility genes and environmental factors (5
). Following recent reports of genetic associations to endometriosis, there has been a surge in the interest to uncover the role of genetic variants in this disease (13
). Significant associations to endometriosis have been documented in dozens of genes thus far (15
) but for most of these genes association was highly dependent on ethnicity (17
). None of these studies have been conducted in a Hispanic population.
We have previously identified genes that were upregulated in human and experimental endometriosis, including lysyl oxidases (LOXs) and complement factor 3 (C3)
). LOXs are a family of copper-dependent amine oxidases that are involved in collagen and elastin crosslinking at the extracellular matrix (20
), and play roles in inflammation, cell migration and invasion (21
expression is upregulated in endometriotic lesions (18
) and LOXL1
is downregulated during secretory phase in endometrium from healthy women (22
). Interestingly, LOXL4
gene is located near regions shown to be linked to endometriosis (13
). The important roles of LOXs in extracellular matrix stability make these proteins potential candidates to be studied in the context of infertility (25
is also dysregulated in women and animal models of endometriosis (19
). Infertile patients with endometriosis expressed high C3 levels in the peritoneal fluid (29
). In normal endometrium, high expression of C3
was observed during the secretory phases of the menstrual cycle (30
). Moreover, C3
expression is upregulated in endometrium of baboons treated with hCG compared to controls, and C3 levels in eutopic endometrium predict pregnancy success, which indicates that regulation in C3
expression is critical for implantation (31
). Since C3
is marker of chronic inflammation, a hallmark of endometriosis, we hypothesized that SNPs in C3
could play a role in the pathogenesis of endometriosis and infertility (34
Study subjects were recruited by the Endometriosis Research Program at PSMHS. Patients were premenopausal women with surgically-confirmed endometriosis. Controls were women without endometriosis or known infertility who had surgery for benign gynecologic conditions. Documentation of the revised American Fertility Society criteria severity (rAFS) was obtained during surgery, and demographic, gynecologic, obstetric and clinical data were collected (35
). Subjects (n=384) were categorized as follows: 1) Control group: subjects without endometriosis or infertility (n=147); 2) Endometriosis group: subjects with endometriosis, with or without infertility (n=214; 3) Endo-only: subjects with endometriosis without known infertility problems (n=72); 4) Endo-associated infertility: subjects with endometriosis-associated infertility (n=73). Women were considered “infertile” if they answered positively to “Have you experienced problems getting pregnant?”. Most infertile women were nulliparous (73.4%), although some have had children (n=26). These were surveyed for the specific cause of their perceived infertility. The majority of them (19/26) reported suffering from either miscarriages and/or a history of infertility (average: 4.9 yrs; range: 1-10yrs). One reported being infertile due to endometriosis and six did not report a specific reason or problem. All participants read and signed an informed consent form prior to enrollment. Protocols were approved by the PSMHS IRB Committee.
Isolation of genomic DNA from lymphocytes was conducted using standard procedures (QIAamp DNA Blood kit, Qiagen, Valencia, CA), genotyping was conducted on 12 SNPs in C3
, on a custom-made Golden Gate chip from Illumina, Inc. (San Diego, CA) as previously described (36
). Genotyping was conducted at the Southern California Genotyping Consortium, UCLA, CA.
All statistical analyses were performed using SYSTAT 13.0 (Cranes Software Int. Ltd.). Demographic, gynecologic and clinical parameters are expressed as mean ± standard errors for quantitative variables or as percentages for frequency data. Differences in clinical characteristics between groups were compared using chi-square (χ2) for frequencies or Student's t-test for quantitative variables. SNP genotypes of controls were assessed for departures from Hardy-Weinberg equilibrium (HWE) using χ2 test. Associations between SNP and risk of endometriosis and/or endometriosis-associated infertility were performed by χ2 test. Odds ratios (OR) and 95% confidence interval (CI) were calculated using a logistic regression model adjusted for confounding factors.
Gene-gene interaction was modeled using the multifactor dimensionality reduction (MDR) approach. MDR is a non-parametric genetic model-free approach for detecting epistatic interactions between SNPs that combines a cross-validation and permutation procedure to assess the significance of the detected interactions. Data were first divided into a training set (9/10 of the data) and a testing set (1/10) (10-fold cross-validation). Performance was estimated by cross-validation consistency and model's prediction accuracy. Subsequently, permutation analysis was performed in which case-control status was randomized 1000 times to assess significance of the modeled interaction.
Group characteristics, allelic frequencies and genotype distribution of significant SNPs associated to endometriosis or endometriosis-associated infertility are shown in . The prevalence of symptoms such as pain and infertility was significantly higher in patients than controls (37
). Thirty percent of patients with endometriosis reported having fertility problems, further demonstrating that infertility commonly coexists with endometriosis (2
). Several studies suggest a possible link between endometriosis and miscarriages, but there is insufficient evidence for this association (39
). In our population, 40% of patients with endometriosis reported having at least one miscarriage, which was significantly different from controls. These data add to the current state of knowledge in this controversial issue by providing support for a possible link between pregnancy loss and endometriosis in a Hispanic population.
A. Characteristics of patients with endometriosis (Endo) versus women without endometriosis (Controls). B. Allelic frequencies and genotype distribution of significant SNPs associated to endometriosis or endometriosis-associated infertility.
Genotypes of 12 SNPs were tested for associations using allelic, recessive, dominant and additive models. Genotype distributions were in HWE in the control group for all SNPs tested. The GG genotype at rs737657 in LOXL4 was more frequent in patients compared to controls, indicating an increased risk for endometriosis in individuals homozygous for the G allele (OR=1.68; 95%CI: 1.049-2.700; p=0.031). This risk increased when patients with endometriosis-associated infertility were compared to controls (OR=2.34; 95%CI:1.273-4.286; p=0.006). Also, significant differences were observed in the frequency of the AA genotype at rs17524355 in LOXL4 when patients with endometriosis-associated infertility were compared to controls (OR=5.23; 95%CI:1.434-19.072; p=0.012). In conclusion, these two polymorphisms in the LOXL4 gene (GG genotype at rs737657 and AA genotype at rs17524355) are associated with higher risk for infertility in the context of endometriosis. SNP rs737657 was also shown in this study to be associated with a diagnosis of endometriosis, suggesting that some of the risk factors or molecular mechanisms underlying endometriosis and infertility may be linked. Because our population was recruited for an endometriosis case-control study, and we did not focus on recruiting infertility cases, our study design would not allow distinguishing between infertility alone and infertility in the context of endometriosis. Further studies would be needed to assess whether the observed associations remain significant in patients suffering from infertility only.
Interestingly, the LOXL4
gene is mapped to 10q24.2, close to two regions of interest in endometriosis: 10q23.3 which had significant linkage to endometriosis in families from Puerto Rico, and 10q26 identified by the Endogene study of 1,176 families as associated with endometriosis (13
). Noteworthy, in a recent report, linkage analysis on families grouped by fertility status not only validated the previously observed linkage peak at 10q26 but also detected association of endometriosis to a SNP in that genetic region (24
). Our study, thus, confirms that there is a region of interest in chromosome 10 that is associated with endometriosis-associated infertility and, moreover, independently identified a genetic polymorphism in this region associated to increased risk for this disease.
These studies also showed that the frequency of the GG genotype at rs2241392 in C3 was significantly different between patients with endometriosis and controls (OR=0.369; 95%CI:0.144-0.942) indicating a decreased risk for endometriosis associated with this genotype in this population. Significance for this particular SNP was lost when all endometriosis patients (including those with infertility) or when patients with endometriosis-associated infertility were compared to controls; therefore, we speculate that association is driven by the endometriosis phenotype and not by infertility. It is known that C3 levels are dysregulated in patients with the condition and aberrant expression of this inflammatory molecule could explain some of the symptoms of endometriosis.
In order to determine if more than one SNP can interact to increase risk, we compared prevalence of endometriosis in the study population grouped by the number of risk genotypes at LOXL4 and/or C3 loci. The prevalence of endometriosis was of 44.7% in women carrying neither of the risk genotype at the C3 and LOXL4 loci. In a group of individuals bearing a risk genotype at either locus, C3 or LOXL4, the prevalence of endometriosis increased to 52.7%, (OR=1.361; 95%CI:0.674-2.748; p=0.39) in comparison to women not carrying any risk genotype. Interestingly, in the group of women having the risk genotype at both SNPs, the prevalence of endometriosis reached 66.1% and the risk for endometriosis nearly doubled (OR=2.247; 95%CI:1.106-4.785; p=0.0362). These results suggest that, in combination, SNPs in different candidate genes may improve the probability of correctly predicting genetic susceptibility to endometriosis. More studies are necessary to identify a possible functional effect of these SNPs in gene expression or protein levels/activity in particular during the window of implantation; alternatively, these SNPs may be in linkage disequilibrium with other SNP(s) that would prove to be causative of this disease.
Some of the p values we report here did not reach statistical significance after applying Bonferroni correction. Also, the SNPs analyzed did not cover all of the potential LD blocks in the genes. We also performed a multifactor dimensionality reduction (MDR) analysis to search for possible SNP-SNP interactions between rs737657 and rs2241392. The testing accuracy was 0.5134 with a cross-validation consistency of 10/10. However, when submitted to a permutation analysis, this interaction did not reach statistical significance (p=0.5). For those reasons, it would be important to validate these findings in a larger sample size and in other populations. In addition, future studies should incorporate a tag-SNP strategy to provide a better coverage of the whole genes.
Our results suggest that polymorphisms in LOXL4
can be associated with risk for endometriosis and for infertility in the context of endometriosis. To our knowledge, this is the first report of a genetic association between SNPs in these genes and risk of endometriosis or endometriosis-associated infertility, and between any SNP and these two diagnoses in a Hispanic population. Since the prevalence of SNPs may vary among ethnic groups, it is important to ascertain effects due to ethnic differences and validate these findings in other populations (41