Multitrait meta-analysis identified genomic regions associated with sexual precocity in tropical beef cattle

Abstract

Multitrait meta-analyses are a strategy to produce more accurate genome-wide association studies, especially for complex phenotypes. We carried out a meta-analysis study for traits related to sexual precocity in tropical beef cattle (Nellore and Brahman) aiming to identify important genomic regions affecting these traits. The traits included in the analyses were age at first calving (AFC), early pregnancy (EP), age at first corpus luteum (AGECL), first postpartum anoestrus interval (PPAI), and scrotal circumference (SC). The traits AFC, EP, and SC_N were measured in Nellore cattle, while AGECL, PPAI, and SC_B were measured in Brahman cattle. Meta-analysis resulted in 108 significant single-nucleotide polymorphisms (SNPs), at an empirical threshold P-value of 1.39 × 10⁻⁵ (false discovery rate [FDR] < 0.05). Within 0.5 Mb of the significant SNP, candidate genes were annotated and analyzed for functional enrichment. Most of the closest genes to the SNP with higher significance in each chromosome have been associated with important roles in reproductive function. They are TSC22D2, KLF7, ARHGAP29, 7SK, MAP3K5, TLE3, WDR5, TAF3, TMEM68, PPP1R15B, NR2F2, GALR1, SUFU, and KCNU1. We did not observe any significant SNP in BTA5, BTA12, BTA17, BTA18, BTA19, BTA20, BTA22, BTA23, BTA25, and BTA28. Although the majority of significant SNPs are in BTA14, it was identified significant associations in multiple chromosomes (19 out of 29 autosomes), which is consistent with the postulation that reproductive traits are complex polygenic phenotypes. Five proposed association regions harbor the majority of the significant SNP (76%) and were distributed over four chromosomes (P < 1.39 × 10⁻⁵, FDR < 0.05): BTA2 (5.55%) from 95 to 96 Mb, BTA4 (5.55%) from 94.1 to 94.8 Mb, BTA14 (59.26%) from 24 to 25 Mb and 29 to 30 Mb, and BTA21 (5.55%) from 6.7 Mb to 11.4 Mb. These regions harbored key genes related to reproductive function. Moreover, these genes were enriched for functional groups associated with immune response, maternal–fetal tolerance, pregnancy maintenance, embryo development, fertility, and response to stress. Further studies including other breeds and precocity traits could confirm the importance of these regions and identify new candidate regions for sexual precocity in beef cattle.

INTRODUCTION

Reproductive traits are important for economic success in beef cattle production (Formigoni et al., 2005; Brumatti et al., 2011). Despite its economic importance, genetic gains of reproductive traits are generally slow. These traits are considered polygenic and highly influenced by environmental factors (Cardoso et al., 2015). Genomic selection has allowed higher genetic gains for fertility traits in beef and dairy cattle (Zhang et al., 2014; Garcia-Ruiz et al., 2016). In parallel with genomic selection, genome-wide association studies (GWAS) have detected quantitative trait loci (QTL) affecting reproductive traits aiming to better understand their genetic mechanisms (Hawken et al., 2012; Irano et al., 2016; Melo et al., 2017). However, these traits are controlled by many QTL of small effect and GWAS can produce high false-positive rates. Validation for the identified QTL requires further investigation. To improve QTL detection, GWAS meta-analyses corroborate evidence from independent studies of related traits by identifying polymorphisms that produce variation in common among complex traits (Bolormaa et al., 2014; Ramayo-Caldas et al., 2016). The aim of this study was to carry out meta-analysis across two independent Bos indicus populations to improve QTL detection for reproductive traits. Specifically, we aimed to: 1) identify genomic regions associated with sexual precocity traits measured relatively early in a cow’s life and the related bull trait scrotal circumference (SC); 2) identify candidate genes in those genomic regions; and 3) investigate biological roles of the identified candidate genes. The traits measured in Nellore were age at first calving (AFC), early pregnancy (EP), and scrotal circumference (SC_N) and in Brahman were age at the first corpus luteum (AGECL), first postpartum anoestrus interval (PPAI), and scrotal circumference (SC_B).

MATERIALS AND METHODS

Nellore Phenotypes and Genotypes

Information of animals from Alliance Nellore dataset, born in eight different farms distributed in Northwest, Southwest, and Midwest of Brazil were used. Heifers were either artificially inseminated or naturally mated. Usually farmers apply two breeding seasons, in the early breeding season females are exposed to the first mating at around 16 mo of age and, those females that failed to conceive have a second chance in the later breeding season, at about 26 mo of age. Females that did not conceive in the first neither in the second chance were culled and were not considered in the analysis of AFC. More details about the dataset are described in Costa et al. (2015) and Irano et al. (2016).

Summary statistics of Nellore phenotypes—AFC, EP, and SC_N—are presented in Table 1. The phenotype AFC was defined as the number of days from birth to first calving; EP was defined as success (1) for heifers that had the first calf with less than 31 mo of age, i.e., heifers that got pregnant in the first breeding season, or failure (0) for those heifers that had the first calf after 31 mo of age. In both populations, SC was measured at about 18 mo of age, at the widest point of the scrotum with a standard metal tape, in centimeter.

Table 1.

Summary statistics of the traits AFC, EP, and SC_N in Nellore cattle and AGECL, PPAI, and SC_B in Brahman cattle

Trait	Number of observations1	Mean ± SE	Minimum	Maximum
AFC, d	1,796	1068.2 ± 118.5	778.9	1292.8
EP, %	1,849	28.3	—	—
${\text{SC}}_{\text{N}}^2$, cm	4,248	26.57 ± 2.56	17.47	36.17
AGECL, d	1,007	750.6 ± 142.14	394	1211
PPAI, d	629	180.11 ± 108.71	17	484
${\text{SC}}_{\text{B}}^3$, cm	1,203	26.51 ± 2.76	18.25	39.25

¹Number of observations considered in analyses: animals with available genotypes and phenotypes.

²SC_N measured at approximately 521 d of age.

³SC_B measured at approximately 540 d of age.

The contemporary groups (CGs) for AFC, EP, and SC_N were defined by concatenating the information of herd, year and season of birth, weaning, and yearling management groups. Groups with less than four animals for AFC and EP or less than three animals for SC_N were excluded. For AFC and SC_N, we considered only data in the interval of CG ± 3 SD from the mean of each group. For EP CG without variability, i.e., presenting only females with the same categorical response were excluded. For SC the linear effect of the age of the animal at the recording was included as a covariate in the model. It was considered in the analysis just animals presenting age at recording from 10 to 24 mo of age. It was excluded a total of 20 animals that presented ages out of this interval.

For Nellore traits, phenotypes were previously corrected for CG effects to avoid biased fixed effect estimates because the whole Nellore dataset contained both genotyped and non-genotyped animals. CG estimates used to pre-correct the phenotypes were obtained under a regular mixed animal model (Henderson, 1984). Then, these corrected phenotypic traits for genotyped animals were used in GWAS approach.

Animals were genotyped with high-density Illumina Bovine HD Assay (Illumina, San Diego, CA) and GeneSeek Genomic Profiler Indicus HD—GGP75Ki (Neogen Corporation, Lincoln, NE), which have 777,962 and 74,677 (73,941 in common with HD) SNP markers, respectively. Genotype imputation to Bovine HD was performed for animals genotyped with GGP75Ki using FImpute software (Sargolzaei et al., 2014), taking into account pedigree information. The imputation accuracy is expected to be higher than 0.98 (Carvalheiro et al., 2014). Quality control was performed, removing animals with call rate <0.90 and removing SNP with minor allele frequency (MAF) < 0.01, call rate < 0.95, GC score < 0.15, Hardy–Weinberg equilibrium test P-value < 10⁻⁵, SNP in non-autosomal regions, and unmapped SNP. After quality control 2,923 females with genotypes for 412,993 SNP and 5,078 males with genotypes for 477,317 SNP remained for the analyses. To construct the genomic relationship matrix (G), SNP with MAF >0.1 were considered.

Brahman Phenotypes and Genotypes

Data from the Cooperative Research Centre for Beef Genetic Technologies (Beef CRC) were used in this study. The Beef CRC data for male and female Brahman cattle were previously described (Johnston et al., 2009; Burns et al., 2013; Corbet et al., 2013). The number of observations and descriptive statistics for each trait are presented in Table 1. The traits evaluated were AGECL, PPAI, and SC_B. When the heifers achieved the weight of 200 kg, an ovarian ultrasound was carried out at every 4 to 6 wk to detect the first corpus luteum (CL) and calculate AGECL. To get PPAI phenotypes, the number of days from calving to the first ovulation postpartum was observed. Ovarian ultrasound was used to observe CL presence after calving to indicate an ovulatory event.

For AGECL and PPAI, CGs were formed by concatenating information of cohort, year of birth, and management group. For SC_B, CG included the effects of cohort and year of birth. Age of young bull at recording was used as a covariate for SC_B.

Brahman animals were genotyped using the Illumina BovineSNP50 version 1 or 2, and posteriorly imputed for high-density panel using Beagle software (Browning and Browning, 2009). To allow imputation, representative animals of the Beef CRC population were genotyped using the HD chip from Illumina as described by Fortes et al. (2013). Quality control excluded animals with call rates < 98%, SNP with call rates < 85% and MAF < 0.02. After quality control remained 660,433 SNP for PPAI, 659,845 for AGECL and 465,644 for SC_B.

Genome-Wide Association Methods and Meta-analyses

A single-trait single-marker model was used to perform GWAS for all traits using the genomic mixed model proposed by Kang et al. (2010), which accounts for population substructure, fitted as follows for Nellore (1) and Brahman (2):

$$\mathit{y}=1\text{µ}+\mathit{sa}+\mathit{Zu}+\mathit{\epsilon },$$ (1)

$$\mathit{y}=\mathit{X\beta }+\mathit{sa}+\mathit{Zu}+\mathit{\epsilon ,}$$ (2)

where y is a vector containing the phenotypic information, corrected for the fixed effects in Nellore data (equation 1), $1\mathit{ }$ is a vector of ones, $\text{µ}$ is an overall mean, X is an incidence matrix relating fixed effects (CG) in β with the phenotypes in y, s is the vector containing the genotypes coded as 0, 1, or 2 according to the number of B allele copies, a is the vector with the SNP additive genetic substitution effects, Z is the incidence matrix of polygenic random effects of the animals in u and ε is the vector of residuals. u and ε followed normal distribution with u~N (0, G ${\sigma }_a^2$) and ε~N (0, I ${\sigma }_e^2$), respectively, where G is the genomic relationship matrix for all individuals and SNP (except the SNP considered in a), calculated as described in the first method proposed by VanRaden (2008), ${\sigma }_a^2$ is the additive genetic variance, I is an identity matrix, and ${\sigma }_e^2$ is the residual variance. The SNP effect estimates were computed using the SNP & Variation Suite v8.3.0 software (Golden Helix, Inc., Bozeman, MT) under the model previously described and the EMMAX method (Kang et al., 2010).

Multitrait meta-analyses were performed using the method described by Bolormaa et al. (2014). This method consists of a statistic test following a χ² distribution with n degrees of freedom, where n is the number of traits included in meta-analysis, calculated as:

$$\text{Multi-trait }{\chi }^2={t^{\prime }}_i{V}^{-1}{t}_i,$$ (3)

where t_i is a vector 6 × 1 of the ith SNP effect estimates for the six traits divided by their respective standard errors, t′_i is the transpose vector of t_i, and V⁻¹ is an inverse of the 6 × 6 correlation matrix of the correlations between the t-values of the six traits across the 387,971 SNP considered in the model. The t-values vector was computed as:

$$t_i=\frac{a_i}{\text{SE}(a_i)}$$ (4)

where SE(a_i) is the SE of a_i (SNP effect vector). Just common SNP in both Nellore and Brahman panels were considered in the analysis. False discovery rate (FDR) (f) was calculated as:

$$f=\frac{m\alpha }{s},$$ (5)

where m corresponds to the number of tests (markers) considered in the meta-analysis, α is the significant threshold, and s is the number of tests, where p < α, where p is the correspondent observed p-value for each marker. To select α values producing an FDR lower than 0.05, the p-value rank position procedure (Benjamini and Hochberg, 1995) was used. In summary, equation 5 was changed to obtain α =fs/m, and s was defined as the largest p-value rank position i that satisfies p_i ≤ f_i/m with f = 0.05 (Pereira et al., 2016).

Gene Annotation and Enrichment Analysis

The genes within 0.5 Mb of significant SNP (P < 1.39 × 10⁻⁵) were annotated as candidate genes. These candidate genes were used to perform gene ontology (GO) and enrichment analyses, using DAVID bioinformatics resources (Huang da et al., 2009).

RESULTS AND DISCUSSION

Significant SNP

Heritabilities were previously reported for the traits and populations studied here. They were 0.08 for AFC, 0.30 for EP, 0.41 for SC_N, 0.57 for AGECL, 0.52 for PPAI, and 0.75 for SC_B (Forni and Albuquerque, 2005; Johnston et al., 2009, 2010; Corbet et al., 2013; Irano et al., 2016).

Meta-analysis resulted in 108 significant SNPs, at an empirical threshold P-value of 1.39 × 10⁻⁵ (Figure 1). This number of significant SNP was the maximum rank position, which P-value satisfied the condition P ≤ α with FDR = 0.05 (Supplementary Table S1). Most of the significant SNP (59.3%) were mapped to chromosome 14, producing a strong association peak between 24.3 and 26 Mb (Figure 2), in which 14 genes were harbored (XKR4, TRNAT-AUG, TMEM68, TGS1, LYN, RPS20, MOS, PLAG1, CHCHD7, SDR16C5, SDR16C6, PENK, LOC101907667, IMPAD1). Another region in BTA14, from 29.5 to 29.6 Mb presented a smaller peak, in which four SNPs were significant. All those SNPs (rs109465877, rs111023138, rs134875123, and rs42534153) were located in an intronic region of the gene NKAIN3, which was in a genomic window explaining 1% of the genetic variance of early pregnancy in Nellore heifers (Oliveira Junior et al., 2017).

Figure 1.

Manhattan plot of the meta-analysis for sexual precocity traits in Nellore and Brahman cattle. The y-axis represents the log inverse P-values for SNP associations and the x-axis represents the position in base pairs from chromosome 1 to 29. The blue line indicates genome-wide significance, a P-value cutoff of P < 1.39 × 10⁻⁵, which is equivalent of an FDR lower than 5%.

Figure 2.

Zoom in BTA14, region of the strongest peak. SNPs above the red line were significant (P < 1.39 × 10⁻⁵). In the bottom, genes mapped in this region (total = 87).

The peak observed in chromosome 14 is close to significant regions that were associated with Brahman and Nellore reproductive traits (Fortes et al., 2012; Fortes et al., 2013; Irano et al., 2016). The PLAG1 gene (mapped to 25,000,459 to 25,052,403 bp), a key gene associated with several traits of economic importance in different cattle breeds (Karim et al., 2011; Littlejohn et al., 2012; Nishimura et al., 2012; Fortes et al., 2013; Utsunomiya et al., 2013; Saatchi et al., 2014; Juma et al., 2016; Pereira et al., 2016) is located in this region. One of the most significant SNPs (P-value = 6.39 × 10⁻⁶), the rs109815800 was mapped in an intronic region of PLAG1.

Fortes et al. (2012) also found a region in chromosome 14 associated with AGECL and the age in which bull reach 26 cm of SC—phenotypes that represent the age at puberty for both females and males—for the same Brahman population used in this study. One of the most significant SNPs also was close to SDR16C6 and PLAG1 genes.

Despite this evident peak in chromosome 14, the most significant SNP (P-value = 3.21 × 10⁻¹⁴) of the meta-analysis was located in chromosome 9 at 75.6 Mb. This SNP is in an intronic region of the MAP3K5 gene, which was identified as an estrogen receptor in human (Mirkin et al., 2005).

In Table 2 are presented the most significant SNP of each chromosome, its position and the closest gene. In BTA1, TSC22D2 emerged as a candidate gene in our meta-analysis study. This gene was high-expressed in human oocytes (Kakourou et al., 2013). In BTA2, KLF7 was the closest gene to the top SNP (rs134877457). It is associated with growth traits in Chinese cattle (Ma et al., 2011); presents an important role in mouse neurogenesis (Laub et al., 2001; Caiazzo et al., 2010) and olfactory neurogenesis (Kajimura et al., 2007); and plays a role in adipogenesis regulation in humans (Kawamura et al., 2006).

Table 2.

Number of significant SNP (P < 1.39 × 10⁻⁵) per Bos taurus autosome (BTA), the most significant SNP in each autosome (top SNP), its position, and its distance from the closest gene in base pairs

BTA	Number of SNP	Top SNP	Position, bp	Gene Symbol	Distance, bp
1	1	rs110366479	118,606,312	TSC22D2	5,852
2	6	rs134877457	95,917,958	KLF7	14,648
3	3	rs109478958	49,436,731	ARHGAP29	0
4	3	rs136050748	110,440,534	Uncharacterized Protein (ENSBTAG00000048097)	530,702
6	1	rs43492923	118,436,689	7SK – misc RNA 1	2,828
7	6	rs135631400	94,710,749	7SK – misc RNA 1	510,435
8	1	rs136590180	68,307,110	Uncharacterized Protein (ENSBTAG00000012266)	16,590
9	1	rs110257163	75,613,158	MAP3K5	0
10	3	rs110458186	16,769,219	TLE3	132,436
11	2	rs108980439	104,936,387	WDR5	0
13	2	rs136854801	16,097,639	TAF3	0
14	64	rs109748092	24,710,609	TMEM68	718
15	1	rs110493922	9,064,376	Uncharacterized Protein (ENSBTAG00000047425)	141,535
16	1	rs109871859	1,929,989	PPP1R15B	36,732
21	7	rs110478544	11,435,365	NR2F2	628,702
24	1	rs133759831	2,276,768	GALR1	129,893
26	2	rs135708259	23,405,679	SUFU	47,301
27	1	rs135961785	31,926,152	KCNU1	0
29	2	rs109184359	9,174,027	Uncharacterized Protein (ENSBTAG00000046374)	1,952

¹miscRNA, miscellaneous RNA.

In BTA3 the top SNP (rs109478958) suggested ARHGAP29 as a candidate gene. Its protein was upregulated in bull semen with good quality, compared with bad-quality semen (Singh, et al., 2018). The top SNPs for chromosomes 6 (rs43492923) and 7 (rs135631400) were close to 7SK, a small nuclear ribonucleoprotein-coding gene (snRNA). This snRNA plays a role in DNA transcription and was found close to a significant SNP for bull fertility in chromosome 10 (Suchocki and Szyda, 2015); has a critical role in the control of primordial germ cell proliferation of mouse embryos (Okamura et al., 2012); and was detected in seminal plasma, ejaculated sperm, and epididymal sperm in pigs with a significant biological role in spermatogenesis (Chen et al., 2017).

In BTA9, the gene MAP3K5 was the closest to the top SNP (rs110257163) and was previously associated with regulation of apoptosis in porcine ovarian granulosa cells; the expression of MAP3K5 was increased when the levels of FSH hormone increased in those cells (Sirotkin et al., 2008). Also, this gene was a strong candidate associated with growth and reproduction traits in transcriptome analysis of bovine pituitary gland (Pareek et al., 2016), and it presented some interactions with estrogen receptors in rat primary cortical neurons (Singer et al., 1999).

In BTA10, TLE3 (closest to the top SNP rs110458186) was expressed in mouse mesoderm embryonic cells (Pfeffer et al., 2017). The expression of this gene was associated with FSH hormone increased levels in bovine granulosa cells (Nivet et al., 2018) and with estrogen receptors in human breast cancer cell line (Jangal et al., 2014). This gene was associated with mouse embryo survival. TLE3 mutants presented placenta defects (Gasperowicz et al., 2013) and this gene acted as a co-regulator of adipogenesis in mouse (Villanueva et al., 2011). This gene also affects the embryonic stem cell differentiation (Laing et al., 2015).

In BTA11, the candidate gene WDR5 presents an important role in porcine early embryo development (Ding et al., 2017) and was essential for vertebrate development (Wysocka et al., 2005; Gori et al., 2006). In BTA13, the most significant SNP (rs136854801) was located in an intronic region of the TAF3 gene. This gene harbors an SNP marker associated with the major histocompatibility complex (MHC) in an Italian human population (Pistis et al., 2013). The importance of the MHC for reproduction was previously discussed in Ziegler et al. (2010). Besides that, this gene was associated with endoderm differentiation in mouse embryonic stem cells (Liu et al., 2011), was significantly upregulated in human oocyte, and was enriched in the estrogen receptor signaling pathway (Kocabas et al., 2006).

In BTA14, TMEM68 was identified as a candidate gene. This gene was expressed in human epididymis (Belleannée et al., 2012), was upregulated in heifer blastocyst (Carter, et al., 2010) and was associated with feed intake and growth traits in cattle (Lindholm-Perry et al., 2012). In BTA16, PPP1R15B was the gene closest to the top SNP (rs109871859). It was expressed in bovine cumulus oocyte cells (Abd El Naby et al., 2013) and in bovine granulosa cells under FSH stimulation (Nivet et al., 2018). PPP1R15 gene knockout mouse embryos survived, but had several problems as grown retardation (Harding et al., 2009). In BTA21, the emergent candidate, NR2F2, was previously associated with Leydig cell steroidogenesis in mice testis cells (Mendoza-Villarroel et al., 2014). Bauersachs et al. (2006) concluded that this gene was associated with embryo implantation in cows. In BTA24 the gene closest to the top SNP (rs133759831) was GALR1. This gene was differentially expressed in ovine hypothalamus in pre-pubertal age compared with fetal and adult ages, suggesting a regulation role in the beginning of reproduction (Whitelaw et al., 2009). Also, it was located near a significant SNP associated with oligozoospermia and azoospermia in humans (Aston and Carrell, 2009), was associated with the regulation of ovarian steroids in rats (Mitchell et al., 2004), and interacted with testosterone in male rat brain (Bouret et al., 2000). In BTA26, SUFU was mapped close to the top SNP (rs135708259). This gene was a candidate gene associated with spermatogenesis in mouse, and it was found in spermatocytes (Szczepny et al., 2005). In BTA27, the KCNU1 was the closest gene to the top SNP (rs135961785). This gene presents an important role in semen quality (Santi et al., 2010), which could be observed in a knockout mouse experiment (SLO3, null mutation). Also Zeng et al. (2011) reported the importance of SLO3 for male fertility, by evaluating its role in the normal morphology and sperm motility in mouse. This gene was enriched in the pathway “Vascular smooth muscle contraction,” which was activated by progesterone treatment in neonatal mouse uterus, inhibiting its development (Filant et al., 2012).

On chromosomes 4, 8, 15, and 29, the top SNPs were close to uncharacterized proteins, and their identification is presented in Table 2. No significant SNPs were observed in BTA5, BTA12, BTA17, BTA18, BTA19, BTA20, BTA22, BTA23, BTA25, and BTA28. Evidence from the current meta-analysis seems to corroborate with previous knowledge of gene function for the discussed genes. Although the majority of the significant SNPs were located in chromosome 14, significant SNPs were also mapped in multiple chromosomes, which is consistent with the idea that reproductive traits are complex polygenic phenotypes.

Functional Annotation and Enrichment Analyses

A total of 389 genes were located within 0.5 Mb of the 108 significant SNP (P < 1.39 × 10⁻⁵). From these genes, functional annotation and enrichment analyses were performed and 333 genes were identified by DAVID software (Huang da et al., 2009), using bovine genome as a background gene list. Enrichment analyses identified 48 functional gene groups as significant in DAVID resources, using high astringency criterion. The top five functional gene groups, with higher enrichment score (ES), are presented in Table 3.

Table 3.

Top five most significant functional gene groups enriched in DAVID v.6.7 bioinformatics resources (https://david.ncifcrf.gov/home.jsp)

Functional groups	Top GO term, code ~ name	Ontology	ES	FDR	Genes
1	IPR001751~ S100/CaBP-9k-type, calcium binding		7.06	5.2E-6	S100A13, S100A1, S100A14, S100A16, S100A2, S100A3, S100A4, S100A7
2	UP_SEQ_FEATURE~ domain: EF-hand 1		2.79	1.2E-1	S100A13, S100A1, S100A14, S100A16, S100A2, S100A4, CAPS, NCS1, S100A7
3	GO:0016477~ Cell migration	Biological process	2.59	2.2E0	FCER1G, cdk5, DBH, MP14, NOS3, NR2F2, pex7, LOC515718
4	GO:0019864~ IgG binding	Molecular function	1.99	1.7E0	FCER1G, FCGR3, FCGR2B
5	GO:0051047~ Positive regulation of secretion	Biological process	1.65	1.6E1	FCER1G, CHRNB2, cdk5, serp1

GO terms, ES, and FDR are reported for each functional group; genes in each group are also reported.

Groups 1 (S100/CaBP-9k-type, calcium binding) and 2 (domain: EF-hand 1) were related to calcium binding pathway. S100 proteins are a family protein expressed just in vertebrates. They participate in several biological events, as regulation of proliferation, differentiation, apoptosis, and inflammation (Donato et al., 2013). Calbindin-D9k (CaBP-9k-type) is a vitamin-D-dependent Ca²⁺ binding protein, member of the S100 family. This protein is mainly found in intestinal tissue, but was also found in other tissues as uterus and placenta (Mathieu et al., 1989; Krisinger et al., 1995; Emam et al., 2016). All the genes clustered in group 1 were members of S100A family. This gene family was expressed in ovary, prostate, and testis of human and rats (Wicki et al., 1996) and in thyroid gland, placenta, and prostate (Cannon et al., 2011). Their roles are generally associated with immunological function, promoting the sterility of reproductive tissues against pathogens (Germeyer et al., 2009; Teijeiro and Marini, 2015); however, they were also related to embryo implantation process (Gray et al., 2006; Smits et al., 2018), to response to temperature stress (Landriscina et al., 2001), cell proliferation, and apoptosis (Jin et al., 2011).

The majority of genes clustered in group 2 were also in group 1. This functional group, EF-hand 1, presents a group of proteins that share a similar structure, which is commonly found in calcium binding proteins. Besides of some genes in common with group 1, other two genes were enriched in this group, CAPS, NCS1. CAPS gene was expressed in human epididymis and endometrium during follicular phase (Thimon et al., 2007; Blockeel et al., 2011), while NCS1 was expressed in bovine blastocysts cultured in vitro under stimulation of thyroid hormones, presenting an important role in mammalian neuron-memory development (Ashkar et al., 2016).

Group 3 (Cell migration) is a fundamental process in the development and maintenance of multicellular organisms. Several reproductive events depend on this process, as embryo formation (Jovanović et al., 2010), immunological responses (Sánchez‐Madrid and del Pozo, 1999), spermatogenesis (Smith et al., 2012), and oogenesis (Rørth et al., 2000). Genes clustered in this group were associated with daughter pregnancy rate, cow conception rate (Ortega et al., 2016), development of nervous system (Kapur et al., 1991; Kanaani et al., 2005; Jessberger et al., 2009), embryo lethality (Nourizadeh-Lillabadi et al., 2010), postnatal mortality and growth retardation (Pallares and Gonzalez–Bulnes, 2010), immunity (Bogdan, 2015), embryo differentiation (Rosa and Brivanlou, 2011), and perinatal mortality (Brites et al., 2003).

The group 4 clustered genes associated with the immunoglobulin G (IgG). This is the most common isotype in human blood and extracellular fluid (Janeway et al., 2007), and the most abundant antibody class present in healthy individuals. This antibody was found in the mucus of woman reproductive tract in Fahrbach et al. (2013), which could suggest a role in the fortification of mucus barriers in the female reproductive tract. Besides that, mouse experiments were performed with a specific monoclonal IgM and demonstrated that females immunized with this antibody presented gestational failure (Sthoeger et al., 1993). A similar effect was observed in another mouse experiment, by Ornoy et al. (2003), where the administration of a purified IgG in pregnant mouse reduced the yolk sac and the embryo growth. This pathway enriched the genes FCER1G, FCGR3A, and FCGR2B, which were associated with daughter pregnancy rate, cow conception rate (Ortega et al., 2016), embryo–maternal recognition in cattle (Mamo et al., 2012), and infertility in female mouse (Wetendorf et al., 2017).

The fifth top group was “Positive regulation of secretion.” As group 3, this group is related to a large variety of biological mechanisms, mainly mediated by regulatory hormones. Genes clustered in this group were associated with positive regulation of immune response (Macen et al., 1993; Kiba, 2016), neurogenesis (Jessberger et al., 2009), and response to stress (Faria et al., 2012).

David results presented gene clusters that were mostly associated with immune events. Earlier studies with Bos indicus cattle have found genes playing roles in immune responses, adaptability, and reproduction (Bahbahani et al., 2018). Fayemi (2005) observed that the presence of antibodies (IgG) in bulls’ sperm was related with fertility. Our results are in agreement with previous studies with Bos indicus cattle, pointing the importance of the immune competence to guarantee reproductive success in Bos indicus cattle.

Besides most of significant SNPs were located in chromosome 14, our meta-analysis identified genomic regions harboring significant SNP associations in the majority of the autosome chromosomes (19 out of 29), as expected for polygenic traits. The genes mapped within these associated regions are plausible candidate genes because of their position and due to previous evidence suggesting their functional association with reproduction, as discussed above. The candidate genes identified could be grouped in functional categories that seem to affect immune system and consequently, mammalian reproduction.

In summary, five regions distributed over four chromosomes presented the majority of the most significant SNP (76%): BTA2 (5.55%) from 95 to 96 Mb, BTA4 (5.55%) from 94.1 to 94.8 Mb, BTA14 (59.26%) from 24 to 25 Mb and 29 to 30 Mb, and BTA21 (5.55%) from 6.7 Mb to 11.4 Mb. Two independent populations were studied here, animals from different breeds, born and bred in different geographic regions. The genes identified as candidates in the meta-analysis are probably affecting physiological mechanisms that control sexual precocity in both breeds, because we found key genes expressed in reproductive tissues and playing important roles in immune response (including in reproductive tract), maternal–fetal tolerance, pregnancy maintenance, embryo development, fertility, and response to stress, or were located in previously reported QTL regions for reproductive traits in bovine. Also, we found some candidate genes associated with important hormones for the regulation of puberty in mammalian, as the FSH, estrogens, and testosterone. The importance of these hormones for sexual precocity is expected to be conserved among mammalians.

In short, the genes presented here are both positional and functional candidates for sexual precocity in Bos indicus cattle. Future works could fine-map and validate the identified genomic regions and elucidate which genes are in fact harboring the mutations that could explain the proposed QTL.

SUPPLEMENTARY DATA

Supplementary data are available at Journal of Animal Science online.