BMP-15: Bone Morphogenetic Protein; POR: Poor Ovarian Response; ORT: Ovarian Reserve Test; AFC: Antral Follicle Counts; AMH: Anti-Mullerian Hormone; IVF: In Vitro Fertilization; SNP: Single Nucleotide Polymorphism(s); OD: Odds Ratio

Poor ovarian response (POR) is a term denoting the reduced number of ultrasound-detectable growing follicles during gonadotropin stimulation (apart from specification of underlying functional condition, which represents a special issue). By European Society of Human Reproduction and Embryology consensus on POR definition (the Bologna criteria), at least two of the following three features must be present: (i) advanced maternal age (≥ 40 years) or any of the other specific risk factors, (ii) ≤ 3 oocytes retrieved using a conventional controlled ovarian stimulation protocol, and (iii) an abnormal ovarian reserve test (ORT; i.e., low antral follicle counts (AFC; less than 5–7) or low blood levels of anti-mullerian hormone (AMH; <0.5–1.1 ng/ml)). Thus, patients over 40 years of age with an abnormal ORT may be a priori classified as poor responders (or, more properly, the expected PORs); on the other hand, two episodes of poor response after maximal stimulation are considered sufficient to define a patient as a poor responder in the absence of advanced maternal age or abnormal ORT [1]. The prevalence of poor responders varies in the literature between 9 and 24%; POR is accounting for approx. 50% of total number of cancelled In vitro fertilization (IVF) cycles, and it is a major negative contribution to the efficacy of IVF programs [2].

The exact causes of POR are not very well understood, although its high prevalence probably correlates with the advanced maternal age. In younger women, POR is frequently conditioned by premature ovarian failure, a disease of complex epidemiology with up to 30% of the cases constituted by familial forms [3]. Associations between nucleotide sequence variation and the premature ovarian failure susceptibility have been studied for a number of genes [4]. Among these BMP15 occupies a special place because it was cloned as a marker for premature ovarian failure [5].

The gene encodes bone morphogenetic protein 15 (BMP-15), a growth factor of the TGF-β superfamily. Specific expression of BMP15 in oocytes begins on the early stages of follicle development. Forming complexes with another factor, GDF-9, the BMP-15 secretory proteins promote early follicle growth by enhancing granulosa cell proliferation, modulate hormone production by granulosa cells, and subsequently restrain dominant/pre-ovulatory follicle development by inhibiting the FSHR gene activity [6,7]. Expression of BMP15 in oocytes is critical for recruitment and growth of follicles, and alterations in its nucleotide sequence are associated with ovulation rate abnormalities [8,9]. Moreover, some of the missense mutations behave as dominant (this may be due to X-chromosome localization of BMP15) causing the defective production of bioactive dimers and, ultimately, the premature ovarian failure [10]. Although associations of BMP15 sequence variants with premature ovarian failure have been reported for various ethnic groups [11,12], relatively few studies focus on BMP15 in association with outcomes in IVF programs [13-15].

We describe two novel single nucleotide polymorphisms (SNP) of BMP15: a substitution found only in patients with the normal response to stimulation (therefore, it may contribute to the maintenance of ovarian reserve), and a deletion that is significantly more common in POR patients. Additionally, associations of POR with two known variants of BMP15, rs41308602 and rs3897937, are demonstrated. We discuss possible direct molecular effects of these variations.

A total of 150 patients (65 women with POR and 85 women with the normal ovarian response as a control group) participated in the retrospective case-control study. All of them were recruited while receiving IVF treatment in a hospital of the Research Center for Obstetrics, Gynecology and Perinatology between September 2012 and August 2014. Inclusion criteria were as follows: age under 40, normal karyotype, integrity of both ovaries, and no history of ovarian surgery or pelvic radiation therapy. Thus, we excluded the gross risk factors (age in the first place), so no one of the patients was subject to the 1st of the Bologna criteria. It is an important restriction because the gene we are dealing with is specifically expressed in growing oocytes and almost nowhere else, and this was done to bring to the forefront some pathologic processes in the ovary per se but not reflections of major organic disturbances. The written informed consent was obtained from all participants; the study was approved by the Institutional Review Board at the Research Center for Obstetrics, Gynecology and Perinatology.

Ovarian stimulation was commenced using recombinant folliclestimulating hormone, FSH (Gonal-F up to 300 IU/day, Merck Serono SA, Geneva, Switzerland) from day 3 of the cycle, and gonadotropinreleasing hormone antagonist (Cetrotid 0.25 mg, Merck Serono SA) was started when at least 1 follicle of ≥ 14 mm could be observed by ultrasound. [16]. The AMH blood levels were measured using AMH Gen II ELISA kit (A79765, Beckman Coulter, Brea, CA, USA). The genotyping was carried out using Prep-GS-Genetics kits for DNA extraction and original labeled oligonucleotides for the modified kissing probe assay (all consumables were provided by DNA-Technology JSC, Moscow, Russia). DNA amplification, detection of fluorescence, and digital analysis of melting curves were done with DT-96 Real-Time PCR Cycler (DNA-Technology JSC).

Sequencing of both exons of BMP15 with 3130 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA) was performed using BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems). Corresponding genomic fragments, 650 and 900 base pair long were amplified with specific primers. The sequences were analyzed using BioEdit Sequence Alignment Editor (Ibis Biosciences, Carlsbad, CA, USA) by alignment to BMP15 reference sequence NC_000023.11 (NCBI, Bethesda, MD, USA).

The data were analyzed using Statistica 10 software (StatSoft, Tulsa, OK, USA); p-values <0.05 obtained by two-sided tests were considered significant. Categorical variables were converted into percent values and assessed by χ² test, while associations were measured by odds ratio within 95% confidence interval.

No significant differences in morbidity or obstetrical history were observed between the groups; however, the patients of the POR group turned to be older than the controls (34.7 ± 3.9 and 31.9 ± 3.9, respectively, p<0.05; Table 1) and to have significantly longer histories of infertility, as well as higher FSH levels and lower AMH levels. Ultrasound examinations revealed lower ovarian reserve (decreased ovarian volume and AFC ≤ 5–7) for most women of this group as compared to their control counterparts (Table 1).

Table 1: Clinical features of the patients with poor and normal ovarian responses.

Allele frequencies for seven BMP15 SNP are given in Table 2. Minor allele frequencies ranged from 0.007 to 0.526. Statistical analysis revealed an association of POR with the exonic SNP rs41308602 (c.308 A>G; the POR group was significantly enriched with the rare G allele: three cases of heterozygosity versus none in the control group, p<0.05; Table 3), and probably also with the intronic SNP rs3897937 (c.328+905 A>G; an increased proportion of the G allele was observed in the POR group: 0.62 versus 0.46 in the control group, p ≈ 0.05; Table 3). Combined frequency of c.308G and c.328+905G was 2.6 ± 0.1 times higher in POR group as compared to the control group, whereas frequencies of other previously known SNP did not differ significantly between the groups.

Table 2: SNP description and allele frequencies in combined sample.

Table 3: Comparison of allele frequencies between POR and control groups.

Two novel SNP of BMP15 were revealed by exon sequencing and submitted for registration in ClinVar database. These were c.607 C>T (it was assigned rs796052131) and c.-8 delC (assigned rs796052132). The c.607 C>T substitution was found only in the control group, thus it may associate with functional preservation of ovaries. In contrast, the c.-8 delC, a novel deletion in the 5’ non-coding region of BMP15 mRNA, was more common in the POR group (Table 3).

Haplotype analysis was done for rs41308602, rs3897937, rs104894763, rs138281369, rs104894767, and the novel rs796052131 (all mapping to BMP15 exons 1 or 2, or their vicinity). Odds of having any minor variant of this haplotype were 2.7 times higher for the POR patients as compared to the control group (Table 4). We also checked haplotype frequencies for combination of three common SNP mapping to exon 1 or its vicinity (rs41308602, rs104894763, and rs3897937), given the reported association of rs3810682/rs3897937 haplotypes with increased follicle numbers (≥ 12) during ovarian stimulation [17]. Similarly, the odds turned to be higher for the POR patients than for the control group (Table 4).

Table 4: Combined frequencies of minor variant-comprising haplotypes in POR and control groups.

The results indicate association of POR with two previously known BMP15 variants: the exonic SNP rs41308602 (c.308A>G) and the intronic SNP rs38979307 (c.328+905A>G). The c.308A>G substitution causes asparagine-to-serine switch in 103 protein position. According to literature, the G allele can be found both in amenorrhea patients and in healthy women, though in amenorrhea patients c.308 G is more frequently combined with G (c.-9C>G) or T (c.852C>T) than in the healthy controls [18].

The c.328+905A>G relationship to ovarian function is controversial. Hanevik et al. [19] demonstrate higher frequencies of the G allele in POR samples, Mendoza et al. [20] find that women with the G allele in combination with several genetic markers of ovarian ageing have shorter fertile period, whereas Moron et al. [17] report increased frequency of TGG haplotype of c.673C>T, c.328+905A>G, and ?.- 9?>G for patients with enhanced responses to ovarian stimulation, including patients with ovarian hyperstimulation syndrome. Gonzalez et al. [21] suggest that AG compared to ?? genotype may produce a protective effect, preventing anovulation and infertility.

Three other previously known SNP did not associate with POR in this study, despite some expectations based on published evidence. For example, a large cohort study by Di Pasquale et al. [22] demonstrates associations of premature ovarian failure phenotypes with two BMP15 variants comprised by missense substitutions c.202C>T → p.Arg68Trp and c.538G>A → p.Ala180Thr, although Ledig et al. [23] qualify the c.538G>A as a polymorphism rather than a mutation, represented both in premature ovarian failure patients and control groups. However, the increased proportion of c.538G/A heterozygotes among premature ovarian failure patients is further confirmed in several studies [10,18]. In our setting, these heterozygotes were found in both groups, revealing no association of the G allele with POR; however, as the obvious limitation of this study is the smaller sample size, the associations will need further refinement.

The control group-specific novel c.607T variant gives no amino acid substitution (CTA, corresponding to leucine, is changed to synonymous TTA). Overall, the usage of these codons in human genome is equal (each of them accounts for 7% of the total number of leucines). However, specific tRNA for CTA is less abundant in the ovary than tRNA for TTA [24], and the c.607 C>T substitution may locally increase BMP15 mRNA translation rates thus influencing the protein production. The novel single nucleotide deletion c.-8 delC, more common in the POR group than in the controls, is located in the 5’ untranslated region of BMP15 mRNA in position 42. It does not interfere with the reading frame, but may affect initial steps of the translation.

We report two novel polymorphic sites in BMP15 gene sequence, plausibly related to POR phenotypes. Examination of larger patient cohorts is required to further elucidate their connection with the phenomenon of POR in In vitro fertilization programs.

The study was supported by Ministry of Healthcare of the Russian Federation (grant #2012-02 reg. #01201256341). We acknowledge Prof Denis Rebrikov for participation in the writing and helpful discussions.