Background

Male pigs used for meat production are castrated at an early age to avoid boar taint, which is an unpleasant odour and flavour of the meat from some boars. Due to animal welfare concerns, castration will be prohibited in Norway, and possibly EU countries and others. Alternative methods are therefore needed to prevent tainted meat. Because of available testicular steroids, entire male pigs also have better feed conversion and carcass traits compared to barrows and this makes them more advantageous for the pig industry [1]. Reduction of boar taint levels without castration is therefore of interest for pig breeders worldwide. Identification of genetic factors controlling boar taint may be implemented in breeding programmes to select animals that produce low levels of taint. However, selection for low boar taint generally coincides with selection for low androgen production [2]. Before starting selection it is therefore important to understand the complex genetic system controlling boar taint and to take into account possible correlated effects on other traits in the breeding goal.

Boar taint is mainly caused by elevated levels of the compounds androstenone [3] and/or skatole [4] in adipose tissue. Androstenone (5α-androst-16-en-3-one) is a 16-androstene steroid metabolised from cholesterol through the C21 steroids pregnenolone and progesterone in boar testis [5]. It is further reduced to its alcohols α-androstenol and β-androstenol [6], which also, in a lesser degree, contribute to tainted meat [7]. Skatole (3-methylindole) is a metabolite of the amino acid tryptophan and is produced by intestinal bacteria in the gut [8, 9]. Skatole is only a problem in intact male pigs and not in gilts or barrows, and its levels increase at sexual maturity [10, 11]. Indole is another metabolite from tryptophan. It also contributes to boar taint levels, although to less extent than androstenone and skatole [12]. Both androstenone and skatole are degraded in the liver and a relationship between their metabolism has been found [13].

Pregnenolone and progesterone are not only precursors of androstenone, but also of testosterone and estrogens [14]. The influence of other sex steroids on levels of androstenone and skatole has been studied with ambiguous results. Most studies have found levels of testosterone in plasma not to be correlated to levels of androstenone in adipose tissue [13, 15, 16] or levels of skatole in adipose tissue [13, 15, 17, 18]. In some studies, however, levels of androstenone in adipose tissue [18, 19] have been found to be correlated (0.26 – 0.64) to levels of testosterone. Results for estrogens are more consistent, showing positive correlations to levels of both androstenone (0.42 – 0.93) [13, 15, 16, 18–20] and skatole (0.29 – 0.53) [13, 15, 17, 20] in adipose tissue. No correlation between levels of skatole and estrone sulphate was, however, found in one study [18]. The correlations between levels of androstenone in plasma and adipose tissue diverge from high (0.46–0.94) [18, 19, 21–23] to not significant [24, 25]. Levels of skatole and indole in adipose tissue are shown to be highly correlated (0.46–0.75) [26, 10]. Studies on correlations between levels of androstenone and skatole in adipose tissue show inconsistent results, from medium correlations around 0.3 [18, 27, 28] to higher correlations between 0.45 and 0.68 [15, 23]. Levels of androstenone in plasma has been found correlated (0.44) [13] and not correlated [18, 23] to levels of skatole, while levels of skatole in plasma has been shown correlated (0.76) [23] and not correlated [18] to levels of androstenone in adipose tissue. Diverging results might be explained by breed effects, which significantly affect levels of boar taint, e.g. [29]. Moreover, levels of androstenone, testosterone and estrogens in plasma can also be affected by diurnal variations during the day, e.g. [30, 25, 31].

Detection of single nucleotide polymorphisms (SNPs) associated with boar taint compounds may be applied in practical breeding to reduce boar taint in intact boars. However, before implementation, it is of importance to examine associations of these SNPs with phenotypes related to reproduction because of possible unfavourable correlations. Breed specific variations has been observed for both androstenone [29] and skatole [10] and it is therefore of interest to include different breeds in such an examination. The objective of this study was to genotype candidate gene SNPs in Duroc and Norwegian Landrace boars and to study associations with traits related to boar taint and reproduction. The phenotypic traits included were levels of androstenone, skatole and indole in adipose tissue, levels androstenone, testosterone, 17β-estradiol and estrone sulphate in plasma, and the length of bulbo urethralis gland. Estrogens and the length of bulbo urethralis gland are also indicators of sexual maturity, e.g. [32, 17, 33]. Moreover, haplotype analyses were carried out in genes with several SNPs.

Methods

Animals

A total of 1102 Duroc and 1726 Norwegian Landrace boars were included in this study. The boars were the sons of 81 Duroc and 90 Norwegian Landrace sires. The boars were raised at NORSVIN's boar test stations until 100 kg live weight, on average 156 and 143 days for Duroc and Norwegian Landrace, respectively, and slaughtered on average 15 days later. The days between 100 kg live weight and slaughter are due to the boar selection process, where some animals wait for their destiny as elite boars. Blood samples were taken three days before slaughter for plasma suspension and DNA extraction. Samples from subcutaneous fat were taken at the slaughter line. All the samples were stored at -20°C until chemical analyses or DNA extraction was performed. The length of bulbo urethralis gland was measured at the slaughter line.

Chemical analyses

Levels of androstenone, skatole and indole in adipose tissue and plasma were analysed at the hormone laboratory at the Norwegian School of Veterinary Sciences (NVH). Levels of androstenone were analysed by a modified time-resolved fluoroimmunoassay [34], using an antibody produced at NVH [35], whereas levels of skatole and indole in adipose tissue were analysed using high performance liquid chromatography [26]. Levels of testosterone, 17β-estradiol and estrone sulphate in plasma were analysed at the hormone laboratory at Aker University Hospital. Plasma levels of testosterone were measured by a radioimmunoassay (Orion Diagnostica, Espoo, Finland). The intra- and total assay coefficients of variation (CVs) were 7% and 9%, respectively. Plasma levels of estradiol were measured by a fluoroimmunoassay (Perkin Elmer, Turku, Finland). The intra- and total assay coefficients of variation (CVs) were 3% and 7%, respectively. Plasma levels of estrone sulphate were measured by a radioimmunoassay (Diagnostic System Laboratories, Inc., Webster, TX, USA). The intra- and total assay coefficients of variation (CVs) were 5% and 7%, respectively.

DNA extraction

DNA was isolated from leukocytes using the automated DNA extractor Bio Robot M48 from Qiagen (CA, USA) and the supplementary MagAttract DNA Blood Midi M48 protocol. Concentrations were measured on a 1420 Victor plate reader (Turku, Finland) using PicoGreen fluorescence (Molecular Probes, OR, USA) or on a NanoDrop ND-1000 Spectrophotometer (NanoDrop Technologies, DE, USA). Normalisation of DNA samples was done using the Biomek FX robot from Beckman (Beckman Coulter, CA, USA).

SNP discovery and validation

On the basis of their known or putative role in boar taint, candidate genes were chosen based on literature studies and on results from our previously published microarray results [36, 37]. SNP discovery was performed by PCR resequencing of genomic DNA and cDNA from Duroc and Norwegian Landrace boars. Primers were designed using Primer3 [38] or Oligo primer analysis software v.6 (Molecular Biology Insights, Inc., CO, USA). The programmes Phred, Phrap and PolyPhred (v.4.06) were used to identify putative SNPs from the PCR resequencing chromatograms [39, 40] and the Consed programme was used to visually confirm the putative SNPs [41]. Additionally, SNPs were provided from alignment of EST sequences produced in the Sino-Danish sequencing project [42]. The genotyping was done in a two-step approach. Before high throughput genotyping, all the SNPs (a total of 275 SNPs in 121 genes) were validated on 380 animals from each breed using the genotyping procedure described below (see additional file 1: Genotyped SNPs). The SNPs that were significant in at least one of the phenotypes using a non-stringent significance level (p < 0.1) were used for genotyping of all the boars. Of the selected SNPs, 135 were in successful assays (see additional file 1: Genotyped SNPs).

Genotyping

SNPs were genotyped using matrix-assisted laser desorption/ionisation time-of-flight mass spectroscopy (MALDI-TOF MS) assays. Multiplex assays for use in the Sequenom MassARRAY system were designed using MassARRAY Assay Design software (Sequenom, San Diego, USA) at multiplexing levels between 7 and 35. Primers for the genotyping can be found in additional file 2: Primer sequences for genotyping. Genotyping was done by the IPLEX protocol using manufacturer's instructions (Sequenom, San Diego, USA) [43]. The MassARRAY Typer software was used for automated genotype calling.

Haplotype construction

Genes with more than one SNP were used for haplotype analyses if one of the SNPs was significant for one of the phenotypes. Haplotype construction and frequency estimation were done using two programmes in combination. CRIMAP v2.4 [44] used pedigree information while PHASE v2.1.1 [45, 46] used linkage disequilibrium information to determine haplotypes.

Statistical analyses

Statistical analyses were performed for the two breeds Duroc and Norwegian Landrace separately. Association studies were done using the likelihood ratio test [47] by average information restricted maximum likelihood (AI-REML) [48] combined with expectation maximisation (EM-REML) if an update goes outside parameter space [49]. The procedure is part of the package DMU, v.6 release 4.7 [47]. The fixed effects fitted were sire, herd-year-season, waiting in boar test station before slaughter or not, and pen. By having sire as a fixed effect we make sure that the sire family effect does not affect the SNP estimate. Covariates used were age at 25 kg (start of boar test), days from 25 kg to 100 kg (days in boar test), days from 100 kg to slaughter (days in waiting station) and number of live born in same litter. Number of live born was also included as a squared term as this was shown to have a significant effect in the model. Animal ID, sample date for adipose tissue or plasma, and SNP-genotype or haplotype-genotype were fitted as random effects. The model used was as follows

where trait is bulbo urethralis gland length (cm), ln(ppm levels of androstenone in adipose tissue), ln(ppm levels of skatole in adipose tissue), ln(ppm levels of indole in adipose tissue), ln(ppm levels of androstenone in plasma), ln(ppm levels of testosterone in plasma), ln(ppm levels of estrone sulphate in plasma) or ln(ppm levels of 17β-estradiol in plasma). The linear model used assumes normality. The distributions were, however, skewed for all the chemical compounds in the study and the data was log-transformed to reduce this problem.

In the haplotype analyses, SNP was replaced by haplotype. SNPs with a genotyping success rate of less than 90% were excluded from further analyses and missing data was considered as a separate class in the random effect of SNP or haplotype. For each trait the model was run without SNPs or haplotypes for log likelihood comparison. A log likelihood ratio (LRT) exceeding 5.4 units, corresponding p < 0.001, was considered significant (assuming that 2*LRT is approximately chi-squared distributed with one degree of freedom).

Results

A total of 135 SNPs from 57 candidate genes for boar taint were included in this study. The SNPs were distributed in exons, introns and untranslated regions of the genes and were genotyped in the two breeds Duroc (D) and Norwegian Landrace (NL). Out of these, 9 and 4 SNPs were monomorphic in D and NL, respectively (Table 1). SNPs with an estimated minor allele frequency (MAF) of less than 1% were excluded from further analyses. For the D breed, 19 SNPs were excluded due to low MAF while for NL this number was 14 (Table 1). The resulting SNPs with a genotyping success rate of more than 90% were used for association studies. A summary of these SNPs and their genes, alleles and frequencies are presented for D in Table 2 and for NL in Table 3. Descriptive statistics for the different phenotypes used in association analyses are presented for D in Table 4 and for NL in Table 5. Significant effects of SNPs on these phenotypes are shown in Table 6.

Table 1 SNPs excluded from further analysis due to low variation.
Full size table
Table 2 SNPs used for the final association analyses in Duroc.
Full size table
Table 3 SNPs used for the final association analyses in Norwegian Landrace.
Full size table
Table 4 Descriptive statistics for phenotypes in Duroc.
Full size table
Table 5 Descriptive statistics for phenotypes in Norwegian Landrace.
Full size table
Table 6 Significant results for SNP associations in Duroc and Norwegian Landrace.
Full size table

In Duroc, 4 SNPs were associated with levels of androstenone in adipose tissue and they were found to explain 5.1% (NGFIB_in4), 5.6% (CYP2D6_1276(ex7)), 12.3% (CYP2C49_1083(ex7)) and 16.3% (CTNND1_3'UTRa) of the total variation. In NL, none of the SNPs examined in this study was significantly associated with levels of androstenone in adipose tissue, but 4 SNPs were significantly associated with levels of androstenone in plasma. These SNPs explain 1.3% (BAP1_3'UTRb), 1.6% (HYAL2_583(ex1)), 2% (BAP1_3'UTRa) and 2.1% (SRD5A2_3'UTRd) of the total variation. Levels of skatole in adipose tissue were significantly associated with 5 SNPs within two different genes in D and 3 SNPs from one gene in NL. In D, SNPs within CYP21 explained 12.4% and 13.4% of the total variation while SNPs within CYP2E1 explained between 2.5% and 3.1% of the total variation. CYP2E1 SNPs in NL explained between 2.1% and 3% of the total skatole variation. Association results for levels of indole in adipose tissue show that 7 SNPs from two genes (CYP21 and CYP2E1) explain between 6.3% and 7.5% of the total variation in D and that 5 SNPs from CYP2E1 explain between 3.1% and 4.9% of the total variation in NL. Some SNPs were also associated with levels of other steroids but none of them were associated with the length of bulbo urethral gland (Table 6).

Haplotypes were constructed for genes with more than one SNP and an additional association analysis was performed on the phenotypes found to be significant in the single SNP analysis. This resulted in 5 significant genes for D (Table 7) and 5 for NL (Table 8). Only haplotypes with frequencies of more than 1% were included in the statistical analyses. Significant haplotype effects on the phenotypes investigated are shown in Table 9. The haplotype analyses of CYP2E1 were significant in both breeds for levels of skatole and indole and analyses suggest that as much as 12% of the total variation in indole and 6% of the total variation in skatole can be explained by this gene in D. In NL, 9.5% and 4.6% of the total variation in indole and skatole, respectively, was explained by CYP2E1 haplotypes. The effect of different CYP2E1 haplotypes is presented in Figure 1a for D and in Figure 1b for NL. Significant effects were also found for BAP1, HYAL2 and SRD5A2 and levels of androstenone in Norwegian Landrace and these haplotypes are presented in Figure 1c, d and 1e, respectively.

Figure 1
figure 1

Significant effects of haplotypes. Significant effects were found for CYP2E1 haplotypes in Duroc (a) and Norwegian Landrace (b), BAP1 haplotypes (c), HYAL2 haplotypes (d) and SRD5A2 haplotypes (e) in Norwegian Landrace.

Full size image
Table 7 Haplotypes and their frequencies in Duroc.
Full size table
Table 8 Haplotypes and their frequencies in Norwegian Landrace.
Full size table
Table 9 Significant results for haplotype associations.
Full size table

Discussion

Unfavourable correlations between boar taint compounds and phenotypes related to reproduction make selection against high levels of boar taint challenging. Analysing the effect of candidate genes on androstenone, skatole and sex steroids will give new information about which genes that affect different compounds. It may also reveal polymorphisms that can be used for breeding against high levels of boar taint without simultaneously affecting fertility and reproduction in the pigs. In this study we report a larger set of SNPs in selected genes involved in biosynthesis and metabolism of androstenone and skatole, in addition to regulatory factors likely involved in boar taint. The SNPs were genotyped in boars from Duroc (D) and Norwegian Landrace (NL) and tested for associations to 8 different phenotypes related to boar taint and reproduction.

One of the key enzymes in the metabolism of skatole is cytochrome P450 family member CYP2E1 [15, 50]. A total of 7 metabolites have been identified for skatole [51] and 5 of them were produced at decreased levels when a CYP2E1 inhibitor was present [52]. Moreover, levels of CYP2E1 mRNA [53], protein [15, 53], and enzyme activity [54] have been found negatively associated with levels of skatole in adipose tissue of boars. It has also been shown that CYP2E1 is induced by skatole [55] and inhibited by androstenone and 17β-estradiol [56]. In our study we found no association of CYP2E1 and androstenone. This is in accordance with the results of Zamaratskaia et al. [57]. We did, however, find significant associations between SNPs and haplotypes within CYP2E1 and levels of skatole and indole in both breeds. These results are in agreement with associations previously found between CYP2E1_1423(ex9) and levels of skatole [58], although in another study there was no significant associations detected in Large White × Meishan crossbreds [59]. Both SNP and haplotype analyses showed that the extent of variation in skatole and indole explained by CYP2E1 is higher in D than NL. Breed specific differences for CYP2E1 is in agreement with Doran et al. [53]. In both breeds we identified a haplotype that was associated with reduced levels of both skatole and indole, and this haplotype contains the favourable alleles from the SNP analysis.

Cytochrome P450 family member CYP21 is involved in the steroid biosynthesis pathway [60]. It has previously been suggested as a positional candidate gene for androstenone [61]. We identified 7 SNPs in the introns of this gene. None of these SNPs were found to be associated with levels of any of the steroids included in this study. On the contrary, 3 SNPs were significantly associated with levels of skatole and indole in D, but not in NL. Two haplotypes containing all the favourable alleles have an advantageous effect on levels of indole and skatole in D. This effect was, however, not significant (LRT of 3.4 (p < 0.01) for indole and 1.6 (p < 0.1) for skatole). The role of CYP21 regarding to levels of skatole and indole needs to be clarified.

Cytochrome P450 porcine isoform CYP2C49 is a member of the subfamily CYP2C, which metabolise drugs and steroids [62]. We previously identified this isoform as the most significant gene in a gene expression study including pigs with extreme high and low levels of androstenone [37] and in the current study we identified 4 SNPs in the CYP2C49 gene. A significant association was shown between CYP2C49_1083(ex7) and levels of androstenone in adipose tissue in D, and this SNP explain 12.3% of the total variation. This result suggests that the level of gene expression is regulated within the gene. The SNP was not significantly associated with any of the other phenotypes in the study. A microsatellite linked to CYP2C18, which is the human ortholog of CYP2C49, has previously been tested for association with levels of skatole in adipose tissue, in accordance with our results, no significant results were obtained [63].

Cytochrome P450 member CYP2D6 is involved in drug metabolism and has a broad substrate specificity [64]. Some studies have also suggested that CYP2D6 is regulated by steroid hormones [65, 66]. In this study we found significant association between the SNP CYP2D6_1276(ex7) and level of androstenone in adipose tissue of D boars. SNPs in CYP2D6 showed no significant associations to levels of skatole or indole. This result is supported by the study of Diaz and Squires [52], showing that inhibition of CYP2D6 do not to affect production of skatole metabolites. Moreover, we found no associations with levels of testosterone or estrogens.

Catenin delta, CTNND1, is part of the catenin family that functions in intracellular signalling and transcriptional regulation [67] and phosporylation of CTNND1 leads to regulation of several transcription factors [67]. A SNP in CTNND1 was significantly associated with levels of androstenone in adipose tissue of D boars. No associations were found for any of the other phenotypes analysed, which makes CTNND1 interesting for selection against high levels of androstenone in D. Such regulators, however, will likely affect several other biological processes and this need to be considered before implementation in the selection scheme.

Orphan nuclear receptor family member NGFI-B, also known as Nurr77, is involved in transcriptional regulation of several steroidogenic genes, including steroidogenic acute regulatory protein (StAR), 3β-hydroxysteroid dehydrogenase (3β-HSD), cytochrome P450 c17 (CYP17), and CYP21 [68, 69]. A SNP in NGFI-B was significantly associated with levels of androstenone in adipose tissue of D boars. It was not associated with any of the other phenotypes examined and is therefore also interesting as a possible genetic marker for selection against androstenone.

Short-chain dehydrogenase/reductase family members SRD5A1 and SRD5A2 are enzymes that catabolise a number of steroids into their 5α-reduced metabolites [70]. SRD5A2 also catalyses the final step of androstenone formation [71]. No relationship was found between genetic variation of SRD5A2 and androstenone in a Large White × Meishan crossbreed [61]. In our study, however, 6 SNPs were detected in the 3' UTR region of porcine SRD5A2 and the SNP SRD5A2_3'UTRd was significantly associated with levels of androstenone and estrone sulphate in plasma of NL boars. The haplotype analyses confirmed these associations. The haplotype associated with reduced levels of androstenone, however, was also associated with reduced levels of estrone sulphate, which makes it less desirable for selection purposes. Aldo-keto reductase AKR1C isoforms work together with SRD5As in liver catabolism of steroids [72] and we previously found the isoform AKR1C4 differentially expressed regarding to levels of androstenone in pig testes [36]. SNPs in isoform AKR1C3 were associated with levels of both estrone sulphate and estradiol in D boars, but no association with levels of androstenone suggest different functions for the two isoforms AKR1C3 and AKR1C4.

The breast/ovarian cancer susceptibility associated protein-1 (BAP1) has been implied to alter substrate function through post-translational modifications [73]. It has the same structure and functions as the ubiquitin carboxy-terminal hydrolase (UCH) family, which are involved in ubiquitin-mediated regulatory pathways [73]. Two SNPs in BAP1 were significantly associated with levels of androstenone in plasma of NL boars. Moreover, a BAP1 haplotye was associated with decreased levels of androstenone. One of the SNPs in BAP1 was, however, also associated with levels of testosterone in plasma in the same breed. The favourable allele with respect to androstenone was also associated with reduced levels of testosterone, which suggest that this SNP is not appropriate for selection purposes.

Hyaluronoglucosaminidases are enzymes responsible for hyaluronan metabolism. It has been shown that hyaluronan metabolic products can induce expression of heat-shock proteins [74], which are known to activate for example steroid hormone receptors [75]. Metabolism of hyaluronan has also been shown to change during testicular development and is associated with testicular descent [76]. We have identified SNPs in HYAL1, HYAL2 and HYAL3. Among these a SNP in HAYL2 was associated with levels of androstenone in plasma of NL boars, and a haplotype with negative effect on the levels of androstenone was also identified. This HYAL2 SNP and the haplotype were also associated with levels of testosterone in NL. The favourable haplotype for levels of androstenone is also associated with reduced levels of testosterone, implying that it is not so suitable for breeding purposes.

Except from the associations found for SNPs in CYP2E1 with levels of skatole and indole in adipose tissue, our results show breed differences with respect to significantly associated SNPs. The breed differences are in agreement with results from our previous expression profiling for D and NL breeds [36, 37, 77]. Different associations suggest that D and NL have different linkage disequilibrium (LD) with the causative mutation. It is therefore likely that there is some distance between the SNP and the causative mutation. Moreover, fixation or nearly fixation was observed for many candidate gene SNPs and different SNPs were monomorphic in D and NL. It emphasises the importance of testing polymorphisms in the population in question before using them for breeding purposes. When including all the significant SNPs for a trait in the same model there were some problems of convergence, which might indicate that the SNP effects are correlated and partly explain the same variation. Although a rather strict significance threshold corresponding to p < 0.001 was used in this study, the large number of tests performed suggests that there might still be false positive results and the findings should be confirmed in another study. Moreover, SNPs that are significant for boar taint compounds and not for phenotypes related to reproduction might still have a small effect on reproduction even though this study was not able to detect it.

Conclusion

This study reports significant associations between SNPs within CYP21 and CYP2E1 and reduced levels of both skatole and indole. Moreover, a number of SNPs within CYP2C49, CYP2D6, NGFIB and CTNND1 were found significantly associated with levels of androstenone in adipose tissue. These SNPs did not reveal obvious associations with levels of testosterone or estrogens, which might indicate that they can be implemented in practical breeding to reduce levels of androstenone and skatole without causing simultaneous negative consequences on phenotypes related to reproduction.