Background
Follicle-stimulating hormone (FSH), one of the gonadotrophins synthesized by the pituitary gland, plays a pivotal role in reproduction. This gonadotrophin binds its cognate receptor, the follicle-stimulating hormone receptor (FSHR), in the granulosa cells of the ovarian follicles and the Sertoli cells lining the seminiferous tubules of the testes, to regulate an array of biological effects associated with reproductive competence. In the ovary, FSH stimulates follicle growth and maturation, as well as the synthesis of estrogens, whereas in the testes it supports spermatogenesis [
1,
2].
Of the nearly 2000 single nucleotide polymorphisms (SNPs) of the
FSHR, five are located in exon 10 [
3]. Four of these SNPs are non-synonymous and lead to amino acid substitution, resulting in the T307A, R524S, A665T, and N680S FSHR protein variants [
4,
5]. The most common and best studied SNPs of this receptor are c.919A > G (rs6165) and c.2039A > G (rs6166), which are inherited in strong linkage disequilibrium [at least in Caucasians and Asians and less in Africans [
6]] and whose most common FSHR variants, T307/N680 and A307/S680, are almost equally distributed among Europeans [
3,
6‐
9].
A number of studies indicate that FSHR function is influenced by the p.T307A and p.N680S polymorphisms. In particular, the p.N680S SNP has received special attention because of its association with variations in the sensitivity of the FSHR to its cognate agonist and the ovarian response to FSH stimulation as disclosed by in vitro [
10,
11] and in vivo studies [reviewed in [
4,
7,
8,
12]]. More specifically, young women with the GG genotype tend to present lower ovarian sensitivity to endogenous FSH, which apparently leads to higher pituitary FSH secretion and longer duration of the menstrual cycle compared to women with the AA genotype [
13]. This altered FSHR sensitivity to agonist, frequently makes necessary personalization of the controlled ovarian stimulation (COS) protocol, usually by administering higher FSH doses to overcome the decreased ovarian response provoked by the N → S substitution at position 680 of the FSHR [
8,
14‐
17]. Moreover, this particular SNP has been proposed as a predictive biomarker for determining the optimal FSH dose to be used in COS protocols [
8,
9,
13,
14,
18]. Apparently, the negative effect of the S680 variant on the FSHR response to agonist decreases with age and fertility status [
19‐
21]. The N680S SNP also has been linked with other abnormalities [
22‐
26], including lower testicular volume in selected North European populations bearing the S680S genotype variant, particularly when it coexists with the
FSHB -211G > A SNP [
27].
The less studied −29G > A polymorphism, located in the core promoter region of the
FSHR (rs1394205), has been associated with reduced transcriptional activity of the receptor gene in women with the -29AA genotype, as well as to primary or secondary amenorrhea and poor response to exogenous FSH in selected populations [
28‐
31]. The major A allele frequency of this SNP ranges from 50 to 70% in East Asia and Europe [
6].
Data on these
FSHR SNPs in Hispano-American population are rather scarce, and the only available data comes from the HapMap and the 1000 Genomes Project database obtained in a small cohort of Mexican-American subjects with Mexican origen residing in Los Angeles, CA, USA [
32]. In this particular population and according to the HapMap and the 1000 Genomes Project databases, the allele and genotype frequencies of the c.2039 GG SNP variant ranges from 33 to 34.0% and from 6.0 to 7.8%, respectively, whereas for the -29G > A SNP, these data bases indicate frequencies of 26% to 33% for the AA genotype and 49% to 55% for the A allele (
http://grch37.ensembl.org), respectively.
The primary objective of the present study was to analyze the frequency distribution of these common
FSHR SNPs in Mexican subjects of Hispanic origin, based on data obtained in larger populations than those previously included in reported databases [
32]. For this purpose, we analyzed samples and data from three distinctly different groups of subjects in order to obtain the most accurate prevalence values and also to examine the influence of ancestry on the frequency estimates observed in Mexican mestizos. As secondary objectives, we examined the potential associations between the c.2039A > G and -29G
> A SNPs genotypes with various outcomes of the COS protocol applied to women belonging to one of the study groups, as well as with some reproductive parameters extracted from the large database used as a reference.
Subjects and methods
Three different groups of Mexican subjects with Hispanic ancestry were included in the study: a cohort of normal and infertile Mexican mestizo women attending a private assisted reproduction clinic in Mexico City (IVF group); a group of 100 normal Mayan mestizo women with low Mayan-Spaniard miscenegation; and a population belonging to a large database of Mexican mestizo subjects in whom data on allelic and genotype frequencies of these SNPs were available.
IVF group
The first study group (IVF group) was conformed by a cohort of 224 Mexican mestizo women [80 normal oocyte donors aged 18 to 29 years (median, 24 years) and 144 infertile patients aged 22–43 years (median, 35 years)] who attended the Instituto Valenciano de Infertilidad-Mexico (IVI) and accepted to participate in the study. All participants in this group were unrelated and of self-reported Mexican mestizo ancestry (at least 3 generations), and both the treating physician and the volunteer were blind to the genotyping results until the end of the study. Women in the donor group were eligible whenever they met the criteria established by the IVI for oocyte donors, including normal karyotype, age between 18 and 30 years, and normal follicular reserve as assessed by intravaginal ultrasound. Inclusion criteria for the infertile group included:
a. Presence of both ovaries without morphological abnormalities, except when the diagnosis of polycystic ovary syndrome (PCOS) was established [according to the Rotterdam criteria [
33]];
b. Both ovaries adequately visible by intravaginal ultrasound;
c. Absence of any endocrinological disease or obesity, except hypothyroidism under treatment or PCOS; and
d. Any cause of infertility, including tubal factor, endometriosis, male factor, mixed (female/male) factor, and unknown cause.
For COS, women were prepared with an oral contraceptive (OC) (ethynilestradiol 30 μg plus drospirenone 3 mg; Bayer Schering Pharma, Berlin, Germany) for 10–21 days in the cycle preceding the COS cycle and then treated with gonadotrophin-releasing hormone antagonist (GnRHa) and menotropins (highly purified LH/FSH, 1:1; Merapur®, Ferring, Mexico) with or without recombinant human FSH (recFSH; Gonal F®, Merck-Serono, Mexico) as add-on treatment, or with GnRHa and recFSH. All women presented withdrawal bleeding after discontinuation of the OC. Ovarian stimulation was started 5 days after discontinuation of the OC with 150 to 225 IU menotropins, 75 IU menotropins plus 150 IU recFSH, or 150 IU recFSH after establishing ovarian and uterine quiescence by intravaginal ultrasound. Gonadotrophins were administered in a step-up fashion, adjusting the dose every 3 to 4 days depending on the ovarian response and the criteria of the treating physician, following stimulation protocols well established by the IVI. When the mean diameter of the leading follicle reached 14 mm as disclosed by intravaginal ultrasound, daily s.c. injections of 0.25 mg Cetrorelix (Cetrotide; Merck Serono S.A. Mexico) were added to gonadotrophin treatment until one or more follicles reached a mean diameter of 18 mm, time when 250 μg s.c. recombinant human chorionic gonadotrophin (hCG) (Ovidrel®, Merck Serono S.A., Mexico) was administered. Transvaginal ultrasound-guided oocyte retrieval was performed 36 h after hCG injection.
As secondary objective for the IVF study group, the response to COS was recorded and analyzed for differences among women with distinct N680S and -29G/A SNPs. To accomplish this, data containing total FSH and LH administered, serum estradiol [measured by a commercial chemiluminescence immunoassay (Beckman Coulter Life Sciences, Indianapolis, IN, USA)], number of oocytes recovered, and days of stimulus required to reach a mean follicle diameter of 18 mm, were collected from donors and patients who completed the stimulation cycle until oocyte retrieval. Women who did not complete the COS cycle for any reason (either voluntarily or because of risk of hyperstimulation, poor response in terms of number of growing follicles, low serum E2 levels, and/or asynchrony in follicular growth), as well as patients with the diagnosis of PCOS (a condition that may influence on the ovarian response to COS) were excluded from the secondary analysis. To explore for differences in response among women with different N680S FSHR variants and -29G/A SNP genotypes and to minimize bias in the analysis of the results, we first examined separately in the group of donors and infertile patients who completed the stimulation cycle for homogeneity in the distribution of gonadotrophin treatment, age, and diagnosis (in the infertile group) among the three genotypes of each SNP, and thereafter analyzed within each group the effect of the genotype on the secondary outcomes controlling for gonadotrophin treatment [grouped as follows: a. recFSH treatment; b. LH/FSH (menotropins) treatment; and c. Mixed (menotropins plus recFSH) gonadotrophin treatment] and the other parameters.
Mayan mestizo women
The second population group studied was conformed by 100 normal Mayan mestizo women, aged 16 to 37 years (median, 20 years), resulting from the admixture between Mayan and Spaniard population with at least one Mayan surname, and whose DNA was analyzed to determine the frequency of the N680S FSHR variant and the impact of the Spaniard ancestry on the presence of this particular
FSHR SNP in the Mexican mestizo population. Other
FSHR SNPs were not analyzed in this group due to insufficient DNA sample available. Data on this particular population has been previously reported [
34].
Mexican mestizo subjects from a large database
To compare the allelic and genotype frequency of the
FSHR SNPs found in the above described groups with those from an open Mexican population, a third group of data (SIGMA cohort) from a large database genotyped using the Illumina OMNI 2.5 array was analyzed. This reference sample was conformed by 8182 Mexican mestizo subjects participants in the Slim Initiative in Genomic Medicine from the Americas (SIGMA) Type 2 Diabetes Consortium [
35] [3515 (43%) male and 4667 (57%) female; 4366 (53%) non-diabetic and 3848 (47%) subjects with type 2 diabetes (T2D), all exhibiting Native American and European ancestry as determined by Principal Components Analysis [
36]]. Details on the selection criteria, quality control procedure, and estimation of Native American and European ancestry proportions have been reported elsewhere [
35]. In our secondary analysis of this reference database, information related with reproductive events such as age at menarche and menopause, and number of pregnancies in a subset of 520 women (aged 34 to 89 years, median 52 years) were extracted from this database (UIDS cohort; [
35]) and analyzed for potential associations with the
FSHR SNPs studied.
FSHR genotyping in samples from the IVF group and the Mayan women
Total DNA was extracted from peripheral blood lymphocytes employing the QIAamp DNA Blood Mini kit (Qiagen Inc., Valencia, CA, USA) and purified using the Wizard Genomic DNA Purification Kit (Promega, Madison, Wisconsin, USA) following the manufacturers’ instructions. Analysis of the
FSHR SNP at position 2039 (N680S) was carried out using a predesigned TaqMan allelic discrimination assay for the StepOne plus system (Applied Biosystems, Inc., Foster City, CA, USA). The results from the TaqMan assay were verified in all samples by PCR-restriction fragment length polymorphism (RFLP) as previously described [
37]. In this IVF group, SNPs at positions 919 (T307A) and -29 also were analyzed by PCR-RFLP following the methods and oligonucleotide primers reported by Sudo et al. [
37] and Achrekar et al. [
28], respectively. For the 3 polymorphisms, the specificity and validity of the TaqMan and RFLP procedures were confirmed in 10% of the PCR products obtained (randomly selected from all samples processed) by direct sequencing. The procedure employed for determining the SNP genotypes in the SIGMA cohort has been described in detail elsewhere [
35].
Statistical analysis
Data from the IVF group and Mayan mestizo women
Differences in allelic and genotype frequencies between women included in the IVF group (donors and infertile patients) and Mayan mestizo women were analyzed using the chi-squared test, with Yates’ correction for the case of the allelic frequency of the -29G > A SNP.
For the analysis of the secondary objectives in the IVF group and given that the study was not originally designed with the power to evaluate the above described associations among different COS outcomes and genotypes, we first determined whether gonadotrophin treatment and diagnosis (in the case of the group of patients) were homogeneously distributed among the different genotypes and then analyzed for the existence of significant differences in secondary outcomes. To test for homogeneity of gonadotrophin treatment, age, and diagnosis vs genotype, the chi-squared test was employed. Differences in secondary endpoints (dose of gonadotrophins administered, serum E2 levels, days of COS, and number of oocytes retrieved) among genotypes in the donors and infertile patients were then analyzed by a generalized linear mixed model (GLMM), considering genotype as the fixed factor and hormonal treatment, age and diagnosis as random factors [
38]. The GLMM test was chosen considering that the study was not originally designed to analyze for differences in secondary outcomes among genotypes and that this test allowed to control simultaneously for age, treatment, and diagnosis (in the case of infertile patients). A GLMM with gamma error was employed to seek for differences in LH doses administered and serum E2 levels, whereas a GLMM with Poison’s error was used to calculate for differences in the number of oocytes recovered and days of COS. The Tukey’s test was employed as post-hoc test for the effect of the genotype on oocyte number in donors. Although the number of secondary outcomes compared was relatively small, correction for multiple testing was anyway performed employing the Bonferroni’s correction procedure [
39].
Analysis of data from the SIGMA cohort
Pairedwise proportions test was employed to compare genotype frequencies between the IVF group and SIGMA subjects. Logistic regression models adjusted for ethnicity were employed to explore potential associations between FSHR genotypes and some reproductive outcomes such as age at menarche and menopause, and number of pregnancies in the UIDS cohort (see above). When the latter outcome was analyzed, the age was added as covariate in the model.
Linkage disequilibrium in the
FSHR SNPs variants detected in the IVF group and SIGMA cohort was determined using the Haploview version 4.1 [
40], in which
D’ = D/Dmax (where D is the deviation of the observed from the expected) and
r2 is the correlation coefficient between pairs of loci. The maximum values of
D’ and
r2 are 1.000, which indicate complete linkage disequilibrium or pairwise correlation between the loci, respectively.
Since the Mayan population studied did not followed Hardy-Weinberg equilibrium for the c2039A > G SNP, and considering that women in the Mexican culture preserve both parents’ surnames and that even if married they inherit the parental surnames to their descendants, we further compared the surnames of the Mayan population studied and tested for population equilibrium following the method described by Lasker [
41].
Discussion
In the present study, we determined the frequency of three FSHR/
FSHR variants (p.T307A, p.N680S and -29G > A), in three groups of Mexican mestizo subjects. These populations, as well as those from other Latin American countries, are particularly unique in that their genetic structure contains an extensive, complex, and variable admixture between Africans, Native Americans, and Europeans (mainly Spaniards) that has significantly contributed to their corresponding phenotypic and genetic makeups [
42,
43]. In women from the IVF cohort, we found a higher GG genotype frequency of the
FSHR c.2039A > G SNP than those previously reported in placental samples from Mexican mestizo women (reported frequency, 5.9%) [
23] as well as in the 1000 Genomes Project Phase 3 database [
32](
http://grch37.ensembl.org/) for a small cohort of Hispanic subjects residing in Los Angeles, CA, USA, of presumptive Mexican ancestry (frequency, 7.8%), but still markedly lower than in Caucasians, in whom the frequency range from ~ 20% to ~ 36% [
6] (
http://grch37.ensembl.org/). Further, in a population of fertile egg donors from Mediterranean origin residing in Spain, the frequency of this genotype is among the highest reported in Western Europe (42%) [
15]. The frequency of the S680S FSHR variant observed in the present study (which was similar in normal oocyte donors and infertile patients), was also lower than that reported in Colombians (~ 14%)(
http://grch37.ensembl.org/), in whom the estimated African and European genetic admixture proportions are higher than in Mexicans (11% vs. 5% and 60% vs. 37%, respectively) [
42], thus emphasizing on the substantial impact of the admixture with Spaniards on the c.2039A > G
FSHR SNP in Latin America. The even lower frequency of the GG genotype in Mayan women with low genetic admixture also points towards the genetic influence of Spaniards on the expression of this particular
FSHR variant in Mexican mestizo women. If this assumption is correct and considering the similar frequency of the heterozygous (AG) genotype in the two populations studied (48% and 55% in the Mexican mestizo and Mayan women, respectively), then one might expect a progressive rise in GG genotype frequency as the admixure with non-Native American individuals increase in this particular Mayan population, which might confirm the Spaniard origin of this SNP in the Mexican mestizo population.
We additionally assessed the ovarian response to COS as well as the time and amount of gonadotrophins required to reach a mean follicle diameter of 18 mm in normal oocyte donors and infertile patients from the IVF group bearing different N680S FSHR variants. Despite the low number of oocyte donors with the GG genotype, we consistently detected an association of this genotype with a lower number of oocytes retrieved after gonadotrophin administration, thus confirming previous studies on the effect of the S680S phenotype on the ovarian sensitivity and response to exogenous FSH administration [
6,
7,
15‐
17,
44‐
46]. The lower number of oocytes retrieved in donors with the GG genotype was not apparent in the infertile women, finding that may be due to the age-dependent vanishing effect of the N680S polymorphism on the ovarian response to COS, as previously suggested [
6].
Another SNP that has been reported to influence the ovarian response to COS is the -29G > A polymorphism [
4,
6,
7,
28]. In some studies the AA variant has been associated with reduced transcriptional activity of the
FSHR and altered level of mRNA and receptor protein expression in vitro [
47,
48] as well as with poor ovarian response to FSH during COS [
4,
28,
47], although the latter has not been consistently found in other studies [
31,
49]. The prevalence of the AA genotype varies depending on the geographic region considered, being relatively low in Caucasians [despite a relatively high frequency of the A allele in some European countries [
31]], Africans, and Central-South Asians, and high in East Asians and Americans from both the USA and some Latin America countries [
6] (
http://grch37.ensembl.org/). In the present study, we found a relatively high prevalence of the AA genotype (~ 20% to 27%) in both donors and infertile patients of the IVF cohort, which was lower than that reported in the 1000 Genomes Project Phase 3 database for Mexican-American residents of the USA (~ 33%) (
http://grch37.ensembl.org/). This difference in AA genotype frequency may be due to the relatively low number of samples genotyped and/or the particular genetic structure of the population included in that particular Project database. The frequency of the AA genotype detected in our normal oocyte donors also contrasts with those found in normo-ovulatory and infertile women from India (1% and 14%, respectively) [
28], in whom the AA genotype was associated with poor ovarian response to COS as well as with primary and secondary amenorrhea [
29]. More vividly, in the population of donors analyzed in the present study, those with the AA genotype did not show any significant difference in response to gonadotrophin administration compared with women exhibiting the GG or GA variants. Coexistence of and interactions with other ethnically-related SNPs at the
FSHR or other genes involved in the ovarian response to gonadotrophins, may explain these apparent discrepancies among the various studies ([
28,
29,
31], and present study).
Data extracted from a large database of SNPs in Mexican individual mestizo confirmed the allelic and genotype frequencies of the
FSHR SNPs found in samples from the IVF group. Further, data on the number of pregnancies reported by carriers vs no carriers of the G allele at the c.2039A > G SNP, strongly suggests that the fertility potential of carriers of the G allele could be compromised. This might be due to the decreased sensitivity of the S680 FSHR variant to the gonadotrophic stimulus [
10,
50] and the failure of the slight to moderate elevations in FSH levels to compensate for the abnormal function of the FSHR S680 variant, particularly in young women [
9], as suggested by the longer menstrual cycle length exhibited by women homozygous for the G allele [
13,
51]. The finding that the trend to decreased pregnancies in women bearing the G allele persisted even after stratification by the presence of more or less Native American/European ancestry, suggests that the effect of the Ser680 FSHR variant on reproductive potential results from the effect of Ser680 on FSHR function, rather than its interaction with other ethnically-related SNPs at the
FSHR or other genes implicated in fertility. Overall, the results indicate that the frequency of the c.919A > G and c.2039A > G GG genotypes in Hispanic mestizo subjects are among the lowest reported [
6] and remarkably similar to those found in large cohorts of Chinese women [
16,
17], and that the presence of the Ser680 FSHR variant may impact on the reproductive potential of women when present in the homozygous state.
A major limitation of the present study is that the sample size in both IVF groups was not sufficiently powered to allow for detection of statistically significant differences in all secondary outcomes as it was not originally designed for this purpose. Nevertheless, we found that age, COS protocols and diagnosis were homogeneously distributed among all genotypes studied and that even after controlling for all these factors the significant difference on the number of oocytes retrieved from donors with the S680S FSHR persisted. Another drawback is that in the IVF groups, ethnicity index was not available and thus models were not adjusted for admixture. Nonetheless, using the UIDS cohort we found that ethnicity was not a confounder in the association between outcomes and genotypes, making valid these findings. Although this is the first large-scale analysis of the fertiliy potential in women with the
FSHR SNPs analyzed, the information on reproductive events (mainly fertility potential as defined by the number of reported pregnancies) extracted from the large database of hispanic women also should be taken with caution as the questionaire applied was designed to obtain information on several metabolic aspects related to T2D, rather than on reproductive events and parameters that may influence, directly or indirectly, on the reproductive potential of the population studied. Thus, the data on the effect of the G allele at position c.2039 of the
FSHR on fertility potential in the general population should be confirmed in other populations, particularly in those with a higher prevalence of this particular
FSHR SNP variant than that reported herein. In this vein, Zilaitiene and colleagues [
52] recently reported a significant association between the S680S FSHR variant and lower possibility of natural conception during the first 12 months of planned conception and other fertility parameters in a large population of young Caucasian women.