| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
ANIMAL GENETICS |
* Área de Producció Animal, Centre UdL-IRTA, 25198 Lleida, Spain;
and
Departament de Ciència Animal i dels Aliments, Universitat Autònoma de Barcelona, 08193 Bellaterra, Barcelona, Spain; and
and
Centre de Tecnologia de la Carn, IRTA, 17121 Monells, Girona, Spain
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
Key Words: Androstenone • Bayes Factor • Boar Taint • Quantitative Trail Loci • Skatole
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
), an unpleasant odor and flavor in the meat (Babol and Squires, 1995). The main causes of boar taint are the high concentrations of androstenone and skatole in fat (Lundström and Bonneau, 1996; Bonneau et al., 2000). Androstenone is a steroid pheromone synthesized in the testis (Gower, 1972; Bonneau, 1982), and skatole is produced in the colon from trypthophan, and metabolized in the liver or deposited in fat tissue (Agergaard and Laue, 1993). Genetic variation has been reported for fat androstenone and skatole concentrations, both between breeds (Willeke, 1993; Squires and Lou, 1995; Xue et al., 1996) and within a breed (Willeke, 1993; Sellier, 1998).
Segregation of major genes has been proposed for fat androstenone (Fouilloux et al., 1997) and skatole (Lundström et al., 1996). In addition, QTL have been detected for fat androstenone (Quintanilla et al, 2003; Lee et al., 2004) and skatole concentrations (Lee et al., 2004) in crosses between Meishan and Large White breeds. Furthermore, Lin et al. (2004) suggested that a mutation of the porcine SULT1A1 gene (thermostable phenol sulfotransferase) is responsible for higher skatole concentrations.
Quantitative trait loci detection in pigs generally has been carried out in F2 experiments (Bidanel and Rothschild, 2002). Nevertheless, several authors (Evans et al. 2003; Nagamine et al., 2003; de Koning et al., 2003) have detected QTL segregating in European outbred pig populations. Analyses of QTL in outbred populations have been carried out using simple sib-pair analysis (Knott and Haley, 1998) or variance component analysis (George et al., 2000). Recently, Varona et al. (2001) developed a procedure to detect QTL in outbred populations through a Bayes factor (BF).
The objective of this study was to map QTL affecting androstenone and skatole concentrations in a Landrace population using the procedure described by Varona et al. (2001).
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
Pigs were from a Landrace line belonging to the experimental research farm of Nova Genètica in Solsona (Lleida, Spain). A sample of 217 males, from a parental generation comprising five boars and 63 sows randomly selected from the whole parental population (25 boars and 550 sows), was used for this study. These animals were slaughtered at 175 d of age, and backfat samples were taken. Fat androstenone and skatole concentrations from the backfat samples were determined with the following protocols.
Determination of Fat Androstenone Level
Internal standards 5
-androstan-3-one (2 µg) and 5-androstan-3-ol (2 µg) were added to 1 ± 0.005 g of backfat (fresh basis). Extraction was achieved with 50 mL of dichloromethane, and 5 mL was taken and evaporated to dryness. The residue was dissolved in 2 mL of methanol to remove the fat by precipitation at room temperature. A clean-up step was performed with a solid-phase extraction column (octadecyl) prewashed with 10 mL of methanol. The methanol extract was applied at the top of the column and washed with 2 mL of methanol. The methanol fractions were collected and evaporated to dryness. The residue was dissolved in 20 µL of isooctane and injected into a gas chromatography–mass spectrometry system (Hewlett Packard 5890–5970, Agilent Technologies, Palo Alto, CA). The separation of compounds was performed in an HP-5MS column (30 m, 250 µm x0.25 µm, J&W Scientific, Folsom, CA). The temperature program began at 70°C, followed by an increase of 10°C/min to 190°C, and by an increase of 5°C/min to 270°C, where the temperature was held for 5 min. The temperature of the injector and detector was set at 270 and 280°C, respectively. The detection was performed in the selective ion monitoring mode, and the selected ions had mass-to-charge ratios of 274, 272, 258, 257, 243, 241, and 202 (Rius and García, 1998).
Determination of Fat Skatole Concentration
The internal standard 7-ethylindole (0.5 µg) was added to 1 ± 0.005 g of backfat (fresh basis) and dissolved in 10 mL of hexane:2-propanol (92:8) for 30 min. The extract was filtered (0.45 µm) and injected into an HPLC system (LKB, Stockholm, Sweden) consisting of a Rheodyne injector with a loop of 100 µL and a 2150 HPLC pump. The detector used was a HP-1046A fluorimeter (Agilent Technologies). An Hypersil aminopropylsilica column (APS2 5 µm, 250 mm x4.6 mm, Tecknokroma, Spain) was used to achieve the separation of compounds and the mobile phase was hexane-2:propanol (92:8) at 1.5 mL/min. The detection was carried out by fluorescence (
excitation 280 nm/emission 360 nm; Garcia and Rius, 1998).
Genotyping
Ten regions on SSC 1, 2, 3, 4, 6, 7, 8, 9, 10, and 13 were chosen as candidate regions, on the basis of previously detected QTL for growth and fatness (Evans et al., 2003). For each region, two or three markers were chosen to increase the chances of capturing heterozygous sires. The chromosomal regions and the micro-satellite markers selected, their locations at the USDA v2 map and the polymorphism information content calculated in this population are presented in Table 1.
|
Statistical Analyses
The records for androstenone and skatole concentrations from 217 males were normalized using a logarithm transformation and subsequently tested for normality using a Wilk-Shapiro test (Shapiro and Wilk, 1965). After transformation, the statistical analysis for detection of QTL segregation was performed following the method of Varona et al. (2001).
We compared two models using a BF. The BF is the ratio between the marginal probabilities of data between two candidate models, and it is directly related with the posterior probability of the models.
Firstly, we consider a mixed inheritance model (QTL model):
 |
where ß= the systematic effects (contemporary group and RYR1 genotype). Pigs tested at the same time and in the same fattening building were considered as one contemporary group, and RYR1 is the genotype for the halothane gene—homozygous and heterozygous. Moreover, u = the polygenic effects, q = the effects associated with a genome segment, e = the residuals, and X, Z1, and Z2 = the corresponding incidence matrices. The assumed distribution for random effects is u ~N(0, A
2u), where A = the polygenic relationship matrix and u2 = the polygenic genetic variance; and q ~N(0, Q q2), where Q = the relationship matrix associated with the genome segment and q2 = the variance caused by the QTL effects. This Q matrix was calculated using the algorithm described by Pérez-Enciso et al. (2000). Finally, e ~N(0, I2e), where 2e is the residual variance.
This model can be reparameterized as y = Xß+ e*, where e* = Z1u + Z2q + e, and consequently e* ~N(0, V).
 |
where 2p = the phenotypic variance (2u + 2q + 2e), h2u = 2u/2p is the proportion of phenotypic variance due to polygenic variation, and h2q = 2q/2p is the proportion of variation caused by the QTL. The total heritability is defined as h2u + h2q = (2u + 2q)/2p.
Records and parameters are jointly distributed for this QTL model (p1) as follows:
 |
where priors for ß, 2p, h2u, and h2q are:
 |
 |
 |
The hyperparameters k1 and k2 are arbitrary constants to ensure the property of the prior distributions. Note that the parametric space for both heritabilities (h2q, h2u) is a triangle, and the flat density to assure a volume equal to one must have a height of two. Note also that, assuming prior independence, marginal priors of h2q and h2u are:
 |
 |
The second model (no-QTL model) that we have assumed is y = Xß+ Z1u + e, which can be reduced to y = Xß+ e*, where e* = Z1u + e, and consequently:
 |
 |
with the joint distribution of records and parameters for this no-QTL model (p2) being:
 |
where priors for ß and 2p are the same as in the QTL model, whereas prior distribution for hu2 is:
 |
with U denoting a uniform distribution. It must be noted that the no-QTL model is a particular type of QTL model when hq2 = 0.
According to Varona et al. (2001), the Bayes factor of the QTL model against the no-QTL model is:
 |
As previously reported by Varona et al. (2001), only the analysis with the complex model (QTL model) is required. A BF >1 indicates that the complex model (QTL model) is more probable. The Bayesian calculations with the complex model were performed using a Gibbs sampling (Gelfand and Smith, 1990), with Metropolis-Hastings steps (Hastings, 1970) for sampling both heritabilities (h2q, h2u). The analyses were performed separately for each trait and at each chromosomal region (see Table 1
). For each analysis, a total of 25,000 iterations were performed after discarding the first 5,000. All correlated samples were used to calculate the posterior distributions using the ergodic property of the chain (Gilks et al., 1996). Convergence was checked using the algorithm of Raftery and Lewis (1992).
|
Results and Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
The mean and variance for raw and log-transformed data, along with the Wilk-Shapiro statistics, are presented in Table 2. Figures 1 and 2 show the ranking plots associated with the Wilk-Shapiro statistic for fat androstenone and skatole concentrations, respectively. The cumulative plot of data after transformation is almost a straight line, indicating that the transformed data are close to a Gaussian distribution, and the Wilk-Shapiro statistic is above the 5% significance level for both traits. These results indicate that the logarithm transformation can be used to normalize the data.
|
|
|
. As mentioned above, most of the analyses in outbred populations have been half-sib analyses (Knott and Haley, 1998), which takes into consideration the segregation from sire families, and assumes that maternal alleles are randomly distributed. In contrast, the two-step variance component approach (George et al., 2000) also considers the paternal and maternal segregation, but it needs to invoke asymptotic assumptions. The BF approach makes use of both maternal and paternal information, and provides a clear framework to compare competing models in terms of probability, so it is reliable even with small samples (Varona et al., 2001). Genetic Determinism for Fat Androstenone and Skatole Concentrations
Tables 3 and 4 show results obtained for heritabilities and QTL mapping of fat androstenone and skatole concentrations, respectively. The posterior mean estimates for total heritability (h2q+ h2u in Tables 3 and 4) ranged from 0.59 to 0.73 for fat androstenone and from 0.74 to 0.89 for fat skatole. Despite the small variation between the estimates obtained in the 20 different analyses performed, one for each trait and chromosomal region, the results are consistent and suggest an important genetic determinism for these traits. These results are also consistent with previous heritability estimates for these traits (see reviews of Willeke, 1993; Sellier, 1998). Although the heritability estimates were obtained from a small sample size, and thus displayed a very large posterior standard deviation, these results allow us to conclude that concentrations of fat androstenone and skatole in the analyzed population are under strong genetic control.
|
|
No evidence of a QTL for fat androstenone concentrations segregating in this Landrace population was found (see Table 3) in any of 10 chromosomal regions analyzed. The BF between the QTL and the no-QTL model in the 10 chromosomal regions had values between 0.089 and 0.661, indicating that the no-QTL model was more probable than the QTL model in all cases. According to these results, the significant QTL for fat androstenone concentrations previously found by Quintanilla et al. (2003) and Lee et al. (2004) are not segregating in our population, although they mapped to similar genome regions tested in this experiment (SSC2, 3, 4, 6, 7, and 9). Results obtained in the present study suggest that the genome regions that accounted for the differences in fat androstenone concentrations between Meishan and Large White breeds are fixed in our Landrace population, allelic differences are smaller in our population, or they were not detected with the available sample size in an outbred population.
With respect to the fat skatole concentrations, there was evidence of a QTL segregating in SSC6 (see Table 4), associated with a BF of 5.160 and a posterior probability of the presence of a QTL of 0.873. Following Kass and Raftery (1995), a BF of 5.160 indicates positive evidence of the QTL model. The posterior distribution obtained for the proportion of variation (hq2) caused by this SSC6 region is given in Figure 3. The posterior mean of this distribution was 0.26, whereas the posterior mode took a smaller value (0.19). The highest probability density region at 95% for the hq2 ranged from 0.01 to 0.53, indicating that the absence of effect of the QTL (hq2 = 0) was located outside of the more probable regions. Moreover, the probability of the No-QTL model is 0.127, which is related to the false discovery rate (FDR) under the a priori assumption of equal probability of both models.
|
), and it is located in SSC14 (not analyzed in our population). This is the first time that a QTL for fat skatole concentrations has been described in SSC6. This QTL maps to the region defined by markers S0059 and S0121, which is located at the 6q chromosomal region of the pig genome (http://www.toulouse.inra.fr/lgc/pig/cyto/cyto.htm). No positional candidate genes associated with metabolism of skatole have been previously reported in this chromosomal region in pigs. However, in the human genome, the sequence that encodes the enzyme cytochrome P450, subfamily IIA, polypeptide 6 (CYP2A6) maps to the region 19q13.2 that is orthologous to the region 6q21 in the pig genome. The oxidation reactions carried out by cytochrome P-450 in the liver affect skatole concentrations in fat (Babol et al., 1998; Diaz et al., 1999), and CYP2A6 in particular has been shown to play an important role (Diaz and Squires, 2000). Linkage disequilibrium between the microsatellite markers used in the analysis and different alleles of the gene that encodes the CYP2A6 enzyme may be responsible for the detected QTL. In addition, Cytochrome P450, subfamily IIIA, polypeptide 7 (CYP3A7), and Cytochrome P450, subfamily 2B, polypeptide 7 (CYP2B7), genes also map to the human orthologous region and could be implicated in the fat skatole concentrations.
The QTL model was also more probable than the no QTL model for the fat skatole concentrations in the regions of SSC1 and SSC10 analyzed. Nonetheless, following Kass and Raftery (1995), a BF lower than 3 is not worthy of more than a mention.
It is also worth mentioning that SULT1A1, proposed by Lin et al. (2004) as a candidate gene, is located in the 16p12.1 region of the human genome. This region is orthologous with the 3p11-12 region of the pig genome located inside the analyzed region in SSC3, but we were not able to detect a QTL in this region in our population.
For both androstenone and skatole concentrations, the sums of posterior estimates of the 10 analyzed regions are 1.11 and 1.44 because the analysis was done separately for each genome segment and trait. Because the lower limit of the parametric space is 0 for heritability, and these posterior mean estimates were obtained from a reduced data set, the estimates are biased upward. However, if we consider exclusively the genome segments where the BF is >1.0, only skatole concentrations have genetic determinism associated with the analyzed genome segments, and the sum of the posterior mean estimates of QTL heritability was 0.60. A joint analysis with these three genome segments fitted simultaneously should provide a more accurate estimate, but could not be performed with the limited available data.
The detection here of a QTL for skatole in SSC6 confirms the possibility of QTL detection in outbred populations, as previously suggested by Evans et al. (2003), Nagamine et al. (2003), and de Koning et al. (2003).
|
Implications |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
|
Footnotes |
|---|
2 Correspondence—phone: 34-973-702637; fax: 34-973-238301; e-mail: luis.varona{at}irta.es.
Received for publication April 20, 2004. Accepted for publication October 27, 2004.
|
Literature Cited |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
-androstenone content in pigs. Pig News Info. 13:31–33.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
