1 Introduction

Bos taurus is a mainly uniparous specie in which twinning occurs at a low frequency. According to Vinet et al. (2012), twinning rate (TR: number of twin birth/100 births) ranged between 0.6 % (in North American Holstein–Friesian cattle breed) and 5 % (in French Maine-Anjou cattle breed). TR is the result of ovulation rate (OR), conception rate, and embryo survival. Gregory et al. (1997) estimated an heritability of 0.35 for OR and a genetic correlation of 0.75 between OR and TR. Ovulation rate is affected by age and parity of the dam, birth season, and feeding (Komisarek and Dorynek 2002). A complex regulatory endocrine network within the ovary and the hypothalamic–pituitary–ovarian axis controls the formation, growth, and selection of follicles due to ovulation (Eppig 2001; Scaramuzzi et al. 2011). Genetic improvement of TR is hampered by the low heritability level of the trait (Komisarek and Dorynek 2002), and few information is available in the literature on genes involved in TR (Silva del Rio et al. 2007).

In Homo sapiens, Mus musculus, and Ovis aries, three genes: BMP15 (bone morphogenetic protein 15), GDF9 (growth differentiation factor 9), and BMPR1B (bone morphogenetic protein receptor 1B) were investigated for their effects on reproduction. These genes, related oocyte-derived members of the transforming growth factor-β (TGF-β) superfamily, are essential factors for follicular growth and ovulation (Chang et al. 2002; Otsuka et al. 2011; Nilsson et al. 2013; Li et al. 2014).

In humans, a large number of mutations in GDF9 and BMPR1B genes have been identified in mothers of dizygotic twins (Montgomery et al. 2004; Palmer et al. 2006; Hoekstra et al. 2008; Painter et al. 2010; Luong et al. 2011), while Zhao et al. (2008) has demonstrated that variation in BMP15 gene is not associated with spontaneous human dizygotic twinning. However, the important role of this gene in reproduction has been ascertained by Di Pasquale et al. (2004, 2006) who identified mutations in BMP15 associated with premature ovarian failure.

In mouse, GDF9 and BMP15 genes play a crucial role in ovarian follicular development and they affect granulosa cell proliferation (Yan et al. 2001; Su et al. 2008, 2009).

In sheep, different mutations in BMP15 (named FecX gene) and in GDF9 (named FecG gene) genes were associated with increased ovulation rate in heterozygous animals and sterility in homozygous animals (Galloway et al. 2000; Hanrahan et al. 2004; Bodin et al. 2007; Monteagudo et al. 2009; Silva et al. 2011; Shabir and Ganai 2012; Demars et al. 2013; Våge et al. 2013; Mullen and Hanrahan 2014). In several sheep breeds known to be prolific, a single mutation in BMPR1B gene (named FecB gene) was associated with higher litter size (Mulsant et al. 2001; Wilson et al. 2001; Davis et al. 2002, 2006).

To date, genomic research on twinning in cattle consisted in studies on QTL associated with twinning rate and ovulation rate (Bierman et al. 2010; Vinet et al. 2012). Recently, Kim et al. (2009), by using 54 k bovine BeadChip, refined the position of QTL for TR in the North American Holstein population and found a significant association between SNP of IGF1 (insulin-like growth factor 1) gene and TR. Kirkpatrick and Morris (2012), by using 3 k bovine BeadChip, located on BTA10 a major gene for bovine OR.

Given the scarce knowledge about genetic component in bovine twinning and the association between GDF9, BMP15, and BMPR1B genes with reproductive traits in sheep, human, and murine species, this study aimed to investigate the same genes in one cattle breed to ascertain the association of their SNP, if any exists, on twinning rate. For this study, the autochthonous Italian breed called “Maremmana” was chosen, a breed typically kept at range, well adapted to harsh environments, with a strictly seasonal reproductive cycle.

2 Materials and methods

2.1 Animals

Data from 98 cows of Maremmana breed reared in Castelporziano Presidential Estate (Rome, Italy) with one or more recorded calvings from 1996 to 2008 were analyzed. During those 7 years, 74 cows had only single births, and 24 gave birth to twins at least once. These 24 cows had 142 calvings, 40 of which were twins.

2.2 Blood collection and DNA isolation

Approximately 10 ml of blood was collected from the tail vein of the 98 cows using Vacutainer tubes with sodium heparin as anticoagulant. Samples were stored at −20 °C until the isolation of genomic DNA. Genomic DNA was extracted from frozen blood samples using Wizard DNA Extraction kit (Promega, Madison, WI, USA). DNA was quantified by DTX 800 Multimode detector (Beckman Coulter, CA, USA) using Quant-iT™ PicoGreen dsDNA kit (Invitrogen, UK), and DNA quality was assessed by the spectrophotometer 260/280 ratio.

2.3 PCR amplification

PCR amplification of different fragments of BMP15, GDF9, and BMPR1B genes (Table 1) was performed using 16 DNA samples: eight from cows with single birth and eight from cows with at least one twin birth. The structure of the three genes in bovine was taken in account to select the fragments (http://www.ncbi.nlm.nih.gov/gene/): (a) the I and II exon both for BMP15 and GDF9 and VI exon for BMPR1B were amplified, because in sheep causative mutation in these regions was found (Galloway et al. 2000; Mulsant et al. 2001; Hanrahan et al. 2004); (b) the 3′UTR for GDF9 was amplified, because in bovine a SNP C/T (rs17871989) was identified (http://www.ncbi.nlm.nih.gov/snp?term=rs17871989); (c) the IV, V, VII, and X exon for BMPR1B was amplified, because these regions correspond to parts of the receptor involved in signal transduction (http://www.ncbi.nlm.nih.gov/gene?term=BMPR1B%20bos%20taurus).

Table 1 Primer designed for PCR amplification of BMP15, GDF9, and BMPR1B genes
Full size table

For primer design, the following bovine gene sequences extracted by Ensembl database were used: GDF9 (ENSBTAG00000009478), BMP15 (ENSG00000130385), and BMPR1B (ENSBTAG00000002081).

PCR amplification was performed in 30 ul reactions containing 30 ng of genomic DNA, 1× PCR buffer, 0.2 mM each of the four dNTP, 0.8 pmol of each primer, and 0.08 U of GoTaq (Promega). A touch down PCR amplification was performed with an initial denaturation 5 min at 95 °C, followed by 14 cycles of 30 s at 94 °C, 30 s at 65 °C (−0.5 °C/cycle), 1 min 30 s at 72 °C; 25 cycles of 30 s at 94 °C, 30 s at 58 °C, 1 min 30 s at 72 °C; a final extension 5 min at 72 °C.

2.4 SNP identification

PCR products were purified using QIAquick PCR Purification kit (Qiagen, Italy) and quantified by DTX 800 Multimode Detector (Beckman Coulter) using Quant-iT™ PicoGreen dsDNA kit (Invitrogen). MWG-Biotech outsourcing service was used to sequence the purified PCR products. The sequences obtained for each fragment were aligned using BioLign software (http://en.bio-soft.net/dna/BioLign.html) to identify putative SNPs.

2.5 SNP genotyping

Genotyping was performed by KBioscience using the proprietary Kaspar methodology. The putative non-confirmed SNPs were discarded. The remaining SNPs were genotyped in all other animals.

2.6 In silico analysis

BLASTP software was used (http://www.ncbi.nlm.nih.gov/BLAST/) to search the homology of the predicted protein sequences with other species. MEGA5 software was used for protein alignment (Tamura et al. 2011, http://www.megasoftware.net/). To determine the potential effect of the amino acid changes on structure and function of the proteins investigated, we used SIFT software (http://blocks.fhcrc.org/sift/SIFT.html). This software uses the protein sequence similarity of different species and the hydrophobic characteristics of amino acids to calculate the probability of a phenotypic effect of a specific amino acid variant (Ng and Henikoff 2003). Wild and mutant sequences were analyzed with the Human Splicing Finder (HSF) (Desmet et al. 2009), which includes several matrices to analyze splice sites, to identify potential SNPs with impact on splicing of the three genes.

2.7 Statistical analysis

For each SNP, the genotype frequencies were calculated and the Hardy–Weinberg equilibrium was verified using the ALLELE procedure in SAS (SAS Institute Inc. 2007).

Birth recording data were used to estimate the allelic substitution effect on number of births (Sherman et al. 2008). The number of copies of each allele of each SNP was regressed against each of the traits separately, using a mixed-effects model with the following equation:

$$ Y_{\text{i}} = \, \mu \, + \, b_{ 1} P_{\text{i}} + \, b_{ 2} {\text{SNP}}_{\text{i}} + \, e_{\text{i}} $$

where, Y i was the trait being modeled per cow: prolificacy (total number of birth/total number of parturition), fertility (total number of birth/total number of mating cows), twinning rate (total number of twin birth/total number of parturition); µ is the overall mean; b 1 is the regression coefficient of the number of births (P i); b 2 is the regression coefficient of the number of copies of each allele; e i is a residual random error associated with the individual observation. Models were fitted using the MIXED procedure of SAS software (SAS Institute Inc. 2007). Significance of the allele substitution effect of SNP was tested by Student “t” test in the MIXED procedure.

A pedigree analysis was done using Par3 software of PEDIG package (Boichard 2002) to calculate the relationship coefficients within and between single and twin birth groups. Differences were analyzed and tested by one-way ANOVA in the SAS procedure.

3 Results

3.1 PCR fragments

In total, 10 amplicon were amplified. For BMP15 gene, the following amplicons were obtained: (1) I exon, length 327 bp and (2) II exon, length 853 bp. For GDF9 gene, the following amplicons were obtained: (1) I exon, length 456 bp, (2) II exon, length 952 bp, and (3) 3′UTR, length 281 bp. For BMPR1B gene, the following amplicons were obtained: (1) IV exon, length 305 bp, (2) V exon, length 307 bp, (3) VI exon, length 183 bp, (4) VII exon, length 296 bp, and (5) X exon, length 304 bp.

The sequences of all fragments were submitted to GenBank: BMP15 accession number EU712722; GDF9 accession number GQ922451, and BMPR1B accession number EU712721.

3.2 SNP discovered

In total, 9 new SNPs were discovered and were submitted to dbSNP. The position (in bp) for each SNP in the gene, the GenBank accession numbers, and amino acid change, when occurred, are listed in Table 2.

Table 2 SNP identified in BMP15, GDF9, and BMPR1B genes
Full size table

The silico analysis highlighted that only two of the nine SNPs were in coding region and were non-synonymous mutations. The polymorphism g.6045 G>A encodes for an amino acid change from Val to Ile (V204I), while the polymorphism g.231 T>C encodes for an amino acid change from Leu to Ser (L66S).

The bovine protein encoded by the BMP15 gene showed a 98 % identity with ovine, 74 % with human, and 69 % with murine proteins. Protein encoded by the GDF9 gene showed a 95 % identity with ovine, 79 % with human, and 68 % with murine proteins. Protein encoded by the BMPR1B gene showed a 99 % identity with ovine, human, and murine proteins. The amino acid substitution V204I was located in the propeptide region, and from the alignment with the protein sequences from other species resulted that the amino acid isoleucine was present. The amino acid substitution L66S was located in the propeptide region, and the alignment with protein sequences from other species showed that the amino acid serine was never displayed.

Three of the nine SNPs (g.2630 C>T, g.23650 C>T, g.49215 C>T) were in noncoding region, and only two of these were in introns. The possible influence of the SNPs localized in the introns on the splicing event was analyzed, but none of the SNPs influenced the canonical donor, acceptor, and branch point sites.

3.3 SNP genotyping

Only 89 animals out of 98 were completely genotyped (the genotype of the remaining 9 animals was lost for technical inconvenients).

Out of the nine studied SNPs, two were monomorphic (GDF9 r903T>C and GDF9 r1272T>C) and seven polymorphic but in linkage disequilibrium. Allele substitution effects at polymorphic SNPs of BMP15, GDF9, and BMPR1B genes were estimated for fertility, prolificacy, and twinning rate (Table 3). Only SNP g.231 T>C in GDF9 gene was significantly associated with prolificacy parameter (P = 0.007) and with TR (P = 0.03).

Table 3 Estimates of the allele substitution effect of BMP15, GDF9, and BMP15 SNPs
Full size table

The estimated average of relationship within single birth group was 10.41 % and within twin birth group was 12.90 %, while between these groups it was 11.20 %. Between-group variance was 0.0100, and within-group variance was 0.0008, with significant F statistic (P ≥ 0.0006).

4 Discussion

Reproduction is a fundamental trait for the efficiency of animal production. Nevertheless, little knowledge is available about the genes involved in the expression of reproductive traits, especially in bovine. Twinning has the potential to increase efficiency of beef production, by obtaining more calves from fewer cows (Gregory et al. 1997).

In mammals, the tendency to spontaneously conceive and maintain embryos is a complex trait affected by environmental and genetic factors. The candidate gene approach was very successful to find causal mutations involved in reproductive traits, in sheep and humans. Up to now, GDF9, BMP15, and BMPR1B genes, which play a key role in regulating fertility in mammals, were mostly investigated for reproduction in this class of vertebrates.

Mulsant et al. (2001) identified a mutation in BMPR1B associated with increased ovulation rate in Booroola Merino ewes. Davis et al. (2006) confirmed this mutation in other prolific sheep breeds. Different authors (Galloway et al. 2000; Hanrahan et al. 2004; Bodin et al. 2007; Martinez-Royo et al. 2008; Monteagudo et al. 2009; Demars et al. 2013) discovered six different mutations in BMP15 gene in eight sheep breeds (Inverdale, Hanna, Belclare, Galway, Lacaune, Rasa Aragonesa, Grivette, and Olkuska). Three different mutations in GDF9 gene were identified in Belclare, Cambridge, Norwegian White, Thoka, and Embrapa sheep breeds (Hanrahan et al. 2004; Nicol et al. 2009; Silva et al. 2011; Våge et al. 2013).

In Homo sapiens, some mutations in BMP15 gene have been identified in women with premature ovarian failure (POF) (Di Pasquale et al. 2004, 2006; Dixit et al. 2006; Laissue et al. 2006). In addition, rare deletions and missense mutations in the BMP15 gene have also been identified in mother of dizygotic twins (Montgomery et al. 2004; Palmer et al. 2006). Furthermore, different mutations in GDF9 gene significantly increased ovulation rate in women (Palmer et al. 2006) or caused the POF phenotype (Dixit et al. 2010; Christin-Maitre and Tachdjian 2010; Persani et al. 2011).

Homozygous GDF9 knock-out female mice were found infertile, whereas heterozygous GDF9 knock-out female mice were found fertile (Dong et al. 1996). Yan et al. (2001) demonstrated that homozygous BMP15 knock-out female mice were found subfertile, with reduced litter size compared with heterozygous and wild-type females.

To date, few studies are available in the literature on these candidate genes in bovine. A research on BMP15 gene (Zhang et al. 2009) detected a deletion in five Chinese cattle breeds. This deletion alters the reading frame and introduces a stop codon in BMP15 mRNA, producing a shorter protein supposed non-functional by the authors. Hosoe et al. (2011) compared the expression patterns of BMP15 and GDF9 genes in young and adult bovine ovaries, and they showed that mRNA expression was different in calves and cows. Two polymorphisms detected on GDF9 gene have been associated with superovulation and sperm quality traits in Chinese Holstein cows and bulls (Tang et al. 2013a, b).

Our study ascertained that the Maremmana cattle breed is polymorphic in GDF9, BMP15, and BMPR1B genes. To the best of our knowledge, mutations in BMP15 and GDF9 genes have already been discovered by Zhang et al. (2009) and Tang et al. (2013a); moreover, these ones are different from the newly identified in our investigation.

Pedigree relationship within and between single and twinning birth groups showed in this trial a small difference. Even if significant, this difference should not pose any risk of false associations, since no selective mating for this trait has ever been performed in the past.

Two SNPs (g.6045 G>A in BMP15 and g.231 T>C in GDF9) of the nine identified caused non-synonymous mutations. The SIFT analysis indicated that the amino acid change (V204I) in BMP15 protein did not affect its structure and function, while the amino acid change Leu to Ser (L66S) in GDF9 protein caused by g.231 T>C SNP influenced the protein structure and function (SIFT score <0.05). From a multiple alignment with other mammalian, GDF9 protein resulted that this mutation occurs in a very conserved region (the proregion of protein). Interestingly, Shimasaki et al. (2004) referred that one of the important features in the posttranslational processing of the TGF-β superfamily members is that the proregion is necessary for the structure of the mature GDF9 protein. An interesting hypothesis to be investigated is whether the mutation occurring in GDF9 proregion, as found in two Maremmana cows with twin birth, could influence GDF9 protein three-dimensional folding and affect its function.

In conclusion, Maremmana cattle breed is polymorphic in GDF9, BMP15, and BMPR1B genes involved in reproductive traits in other species. One of the three polymorphisms of the GDF9 gene showed an effect on twinning. Due to the size of our sample and the limited number of homozygous detected for the g.231 T>C SNP, these results require to be verified on a wider scale.