| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
ANIMAL GENETICS |
Department of Animal Science and the Center for Integrated Animal Genomics, Iowa State University, Ames 50011
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
Key Words: Gametic Imprinting • Meat Quality • Quantitative Trait Loci • Swine
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
). Some examples of gametic imprinting effects in livestock have been described. De Vries et al. (1994) found evidence for genetic variance associated with gametic imprinting in three pig populations for growth rate and backfat thickness. Knott et al. (1998) were the first to search for imprinted QTL in a genome scan. They inferred imprinting when effects differed significantly from Mendelian expression. Jeon et al. (1999) and Nezer et al. (1999) found paternal expression for muscularity in the IGF2 region of chromosome 2 in pigs. De Koning et al. (2000, 2001a,b) modified the approach of Knott et al. (1998) and reported a large number of imprinted QTL for growth and meat quality traits in pigs. They inferred paternal (maternal) imprinting if only one parental allele (maternal or paternal) had a significant effect.
De Koning et al. (2002) demonstrated that successful detection and inference on the mode of QTL expression puts greater demands on statistical tests and their interpretation. Therefore, the purpose of this research was to further develop tests for parent-of-origin effects and to implement them to characterize QTL for growth and meat quality traits in an extended genome scan of a previously reported cross between two commercial breeds of pigs, the Berkshire and Yorkshire (Malek et al. 2001a,b).
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
. A total of 39 traits related to growth, carcass composition, and meat quality traits were collected on the F2 animals, as described in Malek et al. (2001a,b). Phenotypic relationships among traits are described in Huff-Lonergan et al. (2001).
A genome scan based on 125 microsatellite markers across the genome to detect QTL with Mendelian effects was conducted by Malek et al. (2001a,b). To increase the number of informative meioses in special regions of interest that appeared to harbor QTL in the initial scan or that had limited marker coverage in the original scan of Malek et al. (2001a), all animals were genotyped for 33 microsatellite markers (SSC01: S0316, SW781, SW373; SSC02: SW1686, S0565, S0404, SW2443, SWC9; SSC04: SW445, SW856; SSC05: IGF1, SW1954, SSC06: SW1302, SSC07: SW2040, SW632; SSC08: SW1843; SSC09: S0024, SW2401, SW174, SW1651; SSC10: SW1991; SSC11: S0391; SSC12: SWC23; SSC13: SWR428, SW2440; SSC14: SWC6, SW2515; SSC15: SW1262, SW120, SW1339; SSC17: SW2441, SW2431: SSC18: SW787, S0120; SSCXY: S0022). Additional markers were selected from previously published maps (Rohrer et al., 1996; http://www.marc.usda.gov/genome/genome.html, accessed February 28, 2004) based on location and informativeness.
QTL Analysis
Marker linkage maps were derived using Cri-map (Green et al., 1994). The flips and all options were used to get the best order of the markers and the fixed option was used to obtain map distances.
Statistical Models.
The line-cross approach of Haley et al. (1994) was used to calculate probabilities Pkl(j) of parental and breed origin of alleles for each individual j in the F2 generation at every 1-cM position along the genome based on multimarker data, where Pkl(j) is the probability that the jth F2 individual’s paternal and maternal alleles for the putative QTL at the position originated from breeds k and l (k, l = B [Berkshire] or Y [Yorkshire]). These probabilities were then used to derive the following coefficients for fitting additive, dominance, paternal, and maternal effects in a least squares regression interval mapping approach, following de Koning et al. (2000):
 |
Following de Koning et al. (2000), these coefficients were then used to fit the following models at each 1-cM position across the genome using least squares linear regression:
 |
where yj is the trait phenotype of animal j; µj represents fixed effects, which were as described by Malek et al. (2001a,b); apat, amat, and d are paternally inherited, maternally inherited, and dominance effects of the QTL, which are to be estimated; and ej is a residual.
 |
where a is the additive effect of the QTL and other variables are as described previously.
 |
Note that the Mendelian, paternal, and maternal models are special cases of the full model by setting apat = amat= a, amat = d = 0, and apat = d = 0.
Statistical Tests.
To identify QTL and to determine their mode of inheritance, the alternative models were tested against each other in a sequence of tests following the decision tree in Figure 1. Statistical tests—based on an F-statistic that compared the residual sums of squares of a reduced model to those of the larger model—were used at each point in the decision tree. Significance thresholds for each test were derived using the permutation tests described below. The rationale behind the sequence of tests conducted in the decision tree of Figure 1 is that Mendelian expression can be considered the a priori model for gene expression and that parent-of-origin effects should be inferred only if the effect of the maternal and paternal alleles is significantly different from each other within the QTL region. Thus, as a first step, the Mendelian model was tested against the null model (Figure 1). For chromosomal regions where this test was significant, the full model was then tested against the Mendelian model at each position within that region to identify evidence of parent-of-origin effects. If not significant, a Mendelian QTL was declared. If significant, tests of the full against the paternal and maternal expression models were conducted within the region to determine the nature of the parent-of-origin effects. If the full model explained significantly more variation than the paternal model and the full model did not explain more variation than the maternal model, then the paternal contribution was not significant (apat = 0) and a maternally expressed QTL was inferred. A paternally expressed QTL was inferred vice versa (amat = 0). Partial expression was inferred if both tests were significant, which, combined with tests conducted earlier in the tree, indicates that the parental alleles were expressed but to a different degree (apat 
amat but apat 0 and amat 0) (de Koning et al. 2002). In principle, both tests can also be not significant, although this is not likely if a significant QTL was detected in the region using either the Mendelian or the full model.
|
). If significant, the nature of parent-of-origin effects were determined as described previously. In all cases, final estimates of QTL effects and position were derived using the inferred mode of expression.
Significance Thresholds.
Empirical significance thresholds were based on 10,000 data permutations and derived at the chromosomewise and genomewise levels for the following five representative traits: carcass weight, last-rib backfat, loin muscle area, cholesterol content, and marbling. Average thresholds across these five traits were used for significance testing of all traits, following Malek et al. (2001a). Thresholds for tests against the model with no QTL (Mendelian/null, full/null, paternal/null, and maternal/null) were obtained by shuffling all QTL coefficients at each position (Padd(j), Pdom(j), Ppat(j), and Pmat(j)) against the phenotypic data and their corresponding fixed effects and covariates. For tests of the full model against the Mendelian model, permutated data were created by switching the paternal and maternal breed-origin coefficients within each individual j, Ppat(j) and Pmat(j), with 50% probability. This created data under the Mendelian model, in which the paternal and maternal effects are expected to be equal (apat = amat) but any Mendelian QTL effects, if present, were unaffected. For tests of the full against the paternal model, coefficients Pmat(j) and Pd(j) were shuffled across individuals but paternal coefficients, Ppat(j), were not changed. This created data under the paternal model (amat = 0 and d = 0). Similarly, coefficients Ppat(j) and Pd(j) were shuffled for the test of the full against the maternal model.
Simulation.
To validate the significance thresholds derived by the permutation tests, a pedigree and data structure similar to that of the Berkshire-Yorkshire population was simulated with a total of 512 F2 individuals. Marker data were simulated for all animals for a chromosome of 90 cM with seven evenly spaced markers. Each marker locus had four line-specific alleles with frequencies between 0.05 and 0.80, which were chosen to result in similar information content as observed in the data. A biallelic QTL was simulated between the fifth and sixth marker. Data sets with an additive Mendelian or a paternally expressed QTL were simulated, with effects (a or apat) equal to 0.25 phenotypic standard deviation. Animals of the founder generation were either fixed for alternate QTL alleles or segregating at frequencies of 0.7 and 0.3 in the two parental breeds. A total of 10,000 replicates was simulated and analyzed using the statistical models described previously to derive empirical chromosomewise significance thresholds for tests of the full model against the Mendelian and paternal models.
Three randomly chosen simulated data sets with a paternally expressed QTL were used to obtain significance thresholds for the test of the full against the Mendelian model by the permutation test described previously, using 20,000 shuffles of the data. Similarly, three random data sets from the Mendelian QTL simulation were permuted to obtain significance thresholds for the test of the full against the paternal model. Thresholds derived by permutation were compared with those derived by simulation.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
. Map length only changed significantly for SSC2 and was extended by 13 cM because two additional markers were added to the proximal end. Generally, marker order did not change and only in cases where markers were tightly linked. In what follows, the terms proximal, central, and distal will be used to describe chromosomal positions relative to the zero point of the linkage map, with the orientation of the linkage maps corresponding to those given by Rohrer et al. (1996).
Significance Thresholds
Chromosomewise and genomewise thresholds derived by permutation of the actual data are in Table 1 for each of the seven tests conducted. For a given test, thresholds were fairly uniform across the five traits and across chromosomes (results not shown). Significance thresholds for the test of the Mendelian against the null model were similar to those obtained by Malek et al. (2001a). Thresholds were similar for tests with equal numerator degrees of freedom and were greater for tests with a lower number of degrees of freedom.
|
). Also, significance thresholds derived by permutation from three randomly chosen data sets were similar to those derived by simulation (Table 2). Thresholds by permutation are specific to the data set and, therefore, can vary to some degree from replicate to replicate. When applying thresholds derived by permutation to the 10,000 simulated replicate data sets under the null hypothesis, Type I error rates also showed good agreement with the desired Type I error rates of 1, 5, and 10% (Table 2). Segregation at the QTL within the parental breeds, which can result in spurious detection of imprinted QTL (de Koning et al., 2002), tended to increase Type I error rates for the simulated data, but not to a large degree.
|
, 14 new growth and composition QTL were detected, of which two were significant at the 5% genomewise level (Table 3). On the other hand, four QTL detected by Malek et al. (2001a) were no longer significant (for average backfat on SSC1, last-rib backfat on SSC12, carcass weight on SSC7, and for carcass length on SSC11), although most were close to significant.
|
). Nine of the original meat quality QTL (Malek et al. 2001b) were not significant anymore at the 5% chromosomewise level (for color and flavor on SSC2, drip loss and glycogen on 11, chewiness on 12, 24-h ham pH on 14, juiciness on 17, cholesterol on 18, and for off-flavor on X).
Compared with Malek et al. (2001a,b), new QTL were generally in regions in which markers were added. On SSC1, new QTL were detected for loin pH and color score (Table 3), in the same region where Malek et al. (2001b) detected a QTL for drip loss. After increasing the map length of SSC2 in the proximal area by 13 cM, new QTL were detected in that region for traits related to backfat, of which QTL for average and lumbar backfat were significant at the 5% genomewise level (Table 3; Figure 2A). Adding another marker to position 63 on SSC2 resulted in detection of a new QTL for early growth (Table 3), near a previously detected QTL for growth on test (Malek et al., 2001a). Addition of this marker also resulted in detection of new QTL for firmness (5% chromosomewise), proximal to a suggestive firmness QTL detected by Malek et al. (2001b), and substantial increases in significance of a QTL for off-flavor score (1% genomewise significance) and of a QTL for drip loss (5% genomewise; Table 3; Figure 2A), proximal to a previously detected QTL for water-holding capacity (Malek et al., 2001b). To increase informativeness, a marker was also added at the distal end of SSC2, where several traits showed QTL at the end of the marker map (Malek et al. 2001a,b), but no markers were available to extend the map. This resulted in small changes in significance of QTL and the loss of a suggestive QTL for color score. A new QTL for Hunter reflectance in the loin was also identified at the distal end of SSC2 (Table 3), in the same region as previously identified QTL for other reflectance related traits.
|
), in the same region where suggestive QTL for loin pH and reflectance were identified by Malek et al. (2001b). Significance of QTL for backfat traits in this region increased to the 1% genomewise level (Table 3). On SSC6, a QTL for chewiness score became significant in the same region as a previously detected QTL for pH in the ham (Table 3, Malek et al., 2001b). Addition of two markers in the central region of SSC7 resulted in loss of a suggestive QTL for carcass weight but the addition of a QTL for reflectance in the loin measured by Minolta, in the same region as a previously detected QTL for loin reflectance using Hunter (Malek et al., 2001b). On SSC8, addition of a central marker resulted in a QTL for Minolta reflectance in the ham reaching significance (Table 3), whereas a similarly positioned QTL for Hunter reflectance remained just below significance.
Addition of four markers across SSC9 resulted in new QTL for early growth in proximal and distal regions that had previously detected QTL for 16-d weight and average daily gain on test (Table 3). A new QTL for loin muscle area was also detected in the proximal region, and a new QTL for off-flavor score was detected in the central region. Addition of a central marker on SSC10 resulted in borderline significance of new QTL for carcass weight and last-rib backfat and in an increase in significance to the 5% genomewise level of a QTL for Star probe force (Table 3). Addition of a distal marker on SSC12 resulted in a suggestive distal QTL for carcass weight but in loss of previous suggestive QTL for last-rib backfat and chewiness score (Table 3). Addition of two markers to SSC14 resulted in a suggestive QTL for Star probe force in the central region that had a previous QTL for cooking loss percent (Table 3). A distal QTL for ham pH was lost. On SSC15, addition of three markers to the second half of the chromosome resulted in new central QTL for backfat thickness, average drip loss, and off-flavor score (Table 3; Figure 2B). These QTL were in a region around and proximal to previously detected strong QTL related to pH, glycolytic potential, and tenderness, and near PRKAG3, which has been shown to have a significant effect on meat quality (Ciobanu et al., 2001). Finally, addition of markers to chromosomes 17, 18, and X, resulted in lost suggestive QTL for juiciness, cholesterol, and off-flavor score.
QTL with Parent-of-Origin Effects
The main emphasis of this study was to determine the mode of expression of QTL. Details on QTL for which parent-of-origin effects were detected are given in Table 4. A total of 33 QTL with parent-of-origin effects were detected at the 5% chromosomewise level, of which 11, 6, and 9 were significant at the 1% chromosomewise, and at the 5 and 1% genomewise levels based on their inferred mode of expression. All but 12 of these QTL were not detected by the Mendelian model.
|
). For average and last-rib backfat, the Mendelian model was significant at the 5% chromosomewise level (Table 4, Figure 3A). Following the decision tree in Figure 1, the full model was significant over the Mendelian model, which implied a QTL with parent-of-origin effects for this region. Paternal expression was inferred because the full model did not explain significantly more variation than the paternal model in the same chromosomal region, but the full model was significant over the maternal model (significant at the 1% genomewise level). Effects on 10th-rib backfat and loin muscle area were not detected by the Mendelian model and a parent-of-origin effect was inferred based on significance of the full over the null model.
|
, Figure 3C). Backfat traits showed significance at the 1% chromosomewise level for the Mendelian model. Parent-of-origin effects were inferred because the full model was significant over the Mendelian model. Corresponding tests of the full over the paternal model were not significant, but tests of the full over the maternal model were significant, at the 1% chromosomewise level. The final test, of the paternal over the null model, confirmed a paternally expressed QTL that was significant at the 1% genomewise level. The same region showed a paternally expressed QTL for loin muscle area, which was not detected by the Mendelian model (Table 4; Figure 3D). This region is known to contain the IGF2 gene near the second marker on our map (SWC9), which has been shown previously to be paternally expressed (Morison et al., 2001). Pig chromosome 2 also showed two maternally expressed QTL in the distal part of SSC2, for reflectance and pH in the loin (24-h loin Hunter and 48-h loin pH; Table 4).
Parent-of-origin effects were also be identified on SSC3 for off-flavor score (Table 4, Figure 3B). Two separate QTL are reported in Table 4, one maternally expressed and an adjacent QTL that is paternally expressed. Following the decision tree, this could also represent a single QTL that is partially expressed. The distinction between these two possibilities is not clear because significance of the full over the paternal vs. the maternal models switched right next to the QTL region, where the test of the full over the null model showed significance (Figure 3B).
Several QTL with maternal expression were detected in the same marker interval in the central region of SSC9 for traits related to meat quality, including drip loss, light reflectance in the loin, and off-flavor score (Table 4). Only the QTL for off-flavor score was detected under the Mendelian model. Maternally expressed QTL were identified on SSC10 for 10th-rib backfat and marbling in the proximal region and for cholesterol in the distal region (Table 4). A paternally expressed QTL was identified for loin muscle area in the central region (Table 4 and Figure 4A). Only the QTL for marbling was also detected using the Mendelian model.
|
, Figure 4B), where maternally expressed QTL were detected for 16-d weight and average daily gain to weaning (not shown). Chromosome 18 also showed parent-of-origin effects for two traits but in different regions (Table 4); a paternally expressed QTL for glycogen in the center of the chromosome and a maternally expressed QTL for 10th-rib backfat in the proximal region. Single parent-of-origin effects were identified on chromosomes 4, 5, 6, 11, 12, 14, and 16. |
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
,b) for Mendelian QTL in one of the first crosses between two commercially used breeds of swine, Berkshire and Yorkshire, by genotyping another 33 markers in regions of interest; 2) to further develop tests for detection of QTL with parent-of-origin effects; and 3) to implement these methods to further characterize QTL for growth and meat quality traits in the Berkshire-Yorkshire cross. Results obtained for these objectives will be further discussed in the following.
Mendelian QTL Scan
Results of the new genome scan were in general consistent with the previously reported results on the same cross by Malek et al. (2001a,b), where a less complete marker map was used. The increase of markers in regions of interest helped to uncover several QTL that did not reach the suggestive level of significance (5% chromosomewise) under the previous scan and increased the level of significance of others. On the other hand, several QTL that reached significance in the previous scan dropped below the level of suggestive significance, indicating that they may have been false-positives. Nevertheless, the new scan resulted in a net increase in the number of significant QTL at the chromosome- and genomewise levels. Positions of most QTL did not change at all or only slightly.
The increase in the number of QTL detected was achieved by increasing the information content in some areas. The most notable example is in Figure 2A for the marker interval SW2445-SW766 on SSC2, where F-values for off-flavor score and drip loss increased compared with the initial results of Malek et al. (2001b) and reached genomewise significance after adding marker SW1686. A new QTL for firmness was also detected in the same region. Significant QTL for backfat traits were also detected in the proximal area of SSC2 after extending the map by genotyping two additional markers (Figure 2A). This region harbors the IGF2 gene, which has strong effects on muscle mass and fat deposition (Nezer et al., 1999; Jeon et al., 1999; de Koning et al., 2001a). New markers in the distal region of SSC5 allowed identification of a new QTL for water-holding capacity and an increase in significance to the 1% genomewise level of QTL related to backfat. Knott et al. (1998) also identified QTL for backfat traits on SSC5, although not in exactly the same region, but Milan et al. (2002) found QTL for leanness traits on SSC5 in the same chromosomal region as in the current study. New QTL were also discovered on several other chromosomes. Although not every newly detected QTL can be confirmed by literature because the definition of traits are different or the traits have not been analyzed in other studies, our findings are consistent with previous results from this cross, as reported by Malek et al. (2001a,b) regarding the segregation of important QTL for growth and meat quality traits between the Berkshire and Yorkshire breeds.
QTL with Parent-of-Origin Effects
Methods for Detection.
The standard strategy for QTL detection using least squares regression involves fitting the Mendelian model at each position along the chromosome, identifying the position with the highest test statistic, and determining its level of significance. With the advent of searches for QTL with parent-of-origin-effects (e.g., Knott et al., 1998; de Koning et al., 2000), a complication is added by the need to choose among several alternative models, each of which can be fitted along the chromosome and tested against alternative models. This opens a debate about the best sequence and type of tests to conduct, at which positions to conduct these tests, and of the significance thresholds levels to use.
Knott et al. (1998) proposed to detect QTL with parent-of-origin effects by testing a full model against the Mendelian model based on a comparisonwise test at the best position for the Mendelian model. Their full model was the same as the Mendelian model but with addition of an effect for the contrast between the two types of heterozygotes (12 vs. 21). De Koning et al. (2000) solved for the inability of the imprinting model of Knott et al. (1998) to identify the mode of expression (paternal or maternal) by reparameterizing this model by fitting separate effects of paternal and maternal alleles, equivalent to the full model fitted in this study. They compared their full model with models that contained only the paternal or the maternal effect. Imprinting was inferred if one of the parental contributions was significant from zero and the contribution from the other parent and dominance were not significant. This test does, however, not evaluate whether the two parental contributions are different from each other, which is needed to test for significant deviations from Mendelian expression. In addition, the test was conducted at a comparisonwise level at the best position for the imprinting model, which may be different from the best position for the Mendelian model. The same test was applied in de Koning et al. (2001a,b).
Recently, de Koning et al. (2002) used simulation to compare three alternative tests to identify QTL with parent-of-origin effects. All tests were applied to cases in which a significant QTL was detected using either the paternal or the maternal expression model. Their first test was identical to our test of the full model against the Mendelian model and to the test of Knott et al. (1998), except it was carried out at the position of the best QTL using a comparisonwise threshold. In contrast, Knott et al. (1998) conducted the test at the best position for the imprinted QTL, whereas our test of the full against the Mendelian model was evaluated against a chromosomewise threshold. The second test evaluated by de Koning et al. (2002) was identical to the test used by de Koning et al. (2000, 2001a,b): the full model was evaluated against the significant imprinting model at the best position of the imprinted QTL. Their third test inferred imprinting only if both the first and second tests pointed toward imprinting. De Koning et al. (2002) showed that, although the first test is in general more liberal in declaring imprinted QTL, the first and second tests could lead to higher than desired rates of detection of spurious imprinting in some situations, in particular when the QTL was segregating within the parental breeds and the number of F1 sires was low. Applying both tests simultaneously was the least liberal in declaring imprinting.
The testing procedure for imprinting that was employed in the current study differs from the tests proposed by de Koning et al. (2002) in both the sequence of tests and the derivation of significance thresholds. Both will result in a more stringent test for declaring imprinting compared with the test used by de Koning et al. (2000, 2001a,b) and is the main reason for the lower proportion of QTL with parent-of-origin effects in this compared with their studies. The main justification for a somewhat conservative attitude toward declaring imprinting is that imprinting should be considered as the exception rather than the rule. Thus, substantial evidence is needed to reject Mendelian expression. This also provided the basis for the sequence of tests that was used to declare parent-of-origin effects in the current study (see Figure 1). Thus, the genome was first scanned using the Mendelian model, followed by tests of the full against the Mendelian model in regions where the Mendelian model was significant, and by tests of the full against the no QTL model in regions where the Mendelian model was not significant. Compared with the approach of de Koning et al. (2000, 2002), who conducted a complete scan of the genome using the Mendelian, paternal, and maternal expression models, this approach may identify slightly fewer QTL regions because the larger degrees of freedom of the full model reduces the power to detect QTL that are truly imprinted. Once imprinting has been declared based on the first set of tests, the remainder of the decision tree (Figure 1) involves determining the nature of the parent-of-origin effect, by testing the full model against both the paternal and maternal expression models.
The second difference between the tests for imprinting proposed here vs. those used previously is the way in which significance thresholds were determined. In contrast to the comparisonwise tests conducted at either the best position for the Mendelian QTL (Knott et al., 1998) or at the best position for the imprinted QTL (de Koning et al. 2000, 2001a,b, 2002), chromosomewise thresholds were used here. The rationale for using chromosomewise thresholds for comparison of alternative expression models is that the best position may differ between models and estimates of position have large confidence intervals, often covering a large part of the chromosome. Thus, models must be compared across the QTL region, which involves multiple correlated tests.
We developed specialized permutation tests to account for the multiple tests conducted when comparing two models. Simulation results (Table 2) demonstrated that these tests result in appropriate control of type I errors. One limitation of the derived thresholds is that the tests, as implemented in the decision tree, are not independent but thresholds were derived for each step in the decision tree, without taking results from previous steps into account. In principle, thresholds must be derived conditional on previous decisions but this overly complicates the permutation strategy. Despite this limitation, we can state that our strategy enables detection of both QTL showing Mendelian inheritance and QTL showing parent-of-origin effects and that it is able to distinguish QTL with parent-of-origin effects from those with Mendelian expression.
It is important to note that evidence of parent-of-origin effects identified using this strategy does not necessarily imply imprinting. Parent-of-origin effects in an F2 design such as analyzed here, can also be caused by QTL that segregate within the parental breeds, which can cause a different frequency of QTL alleles among F1 dams vs. F1 sires, in particular if the number of F1 parents is small, as demonstrated by de Koning et al. (2002). Our simulation results (Table 2), however, demonstrate that for our design and testing strategy, segregation of QTL does not result in a large increase in false-positive rates for tests for imprinting. These simulations, as well as those of de Koning et al. (2002), however, were based on the assumption that markers and QTL are in equilibrium in the parental breeds. This is important when considering that evidence for parent-of-origin effects in an F2 design comes from markers that segregate within the parental lines; markers that are fixed for alternate alleles in the parental breeds, as is the case for inbred lines, do not allow distinction of parental origin of alleles in heterozygous F2 progeny because the progeny and both F1 parents have the same heterozyous genotype. Tracing parental origin of marker alleles in heterozygous F2 progeny, therefore, requires the F1 sire and dam to have inherited different marker alleles from at least one of the parental breeds. This can result in false detection of parent-of-origin effects for Mendelian QTL if the marker and QTL are in disequilibrium in the parental breed (M. Georges, University of Liege, Belgium, personal communication).
In this study, partial imprinting was inferred if both the maternal and paternal allele had significant effects but they were significantly different from each other, as suggested by de Koning et al. (2002). Although alternative hypothetical models could explain such findings, partial imprinting has been described by Jirtle (1999) for the M6P/IGF2R gene in humans.
Detected QTL.
Several regions with parent-of-origin effects were identified in the Berkshire x Yorkshire cross, using the proposed approach (Table 4). Although the number of QTL with parent-of-origin effects was much lower compared with results described by de Koning et al. (2000, 2001a,b), Knott et al. (1998) identified only four QTL with parent-of-origin effects in two different regions across the entire genome, whereas Quintanilla et al. (2002) detected three QTL showing parent-of-origin effects on SSC9 by using similar methods as Knott et al. (1998). Milan et al. (2002) also used the approach of Knott et al. (1998) and identified five regions showing parent-of-origin effects, on chromosomes 6, 7, 9, and 17. Differences in numbers of QTL with parent-of-origin effects are primarily determined by the approach for detection and the procedure of significance testing, as well as by the number and nature of traits and populations analyzed. Some QTL with parent-of-origin effects detected in the current study confirm results of de Koning et al. (2000, 2001a,b), Rattink et al. (2000), Nezer et al. (1999), and Jeon et al. (1999), as will be discussed below.
One of the major regions shown to harbor QTL with parent-of-origin effects in our study was on SSC1 for backfat traits (Table 4, Figure 3A). So far, no other study has detected parent-of-origin effects for backfat traits within this region. This region might be of special interest because this porcine genomic region may be conserved in HSA15 (Goureau et al., 1996), in which the Prader-Willi syndrome is harbored, which is known to be affected by imprinting in humans (Buiting et al., 2001).
The IGF2 region on SSC2 had several QTL with strong paternal expression effects on muscle mass and fat deposition (Table 4, Figure 2). These results confirm earlier studies by Nezer et al. (1999), Jeon et al. (1999), and de Koning et al. (2000). Recently, Georges et al. (2003) mapped the apparent causative polymorphism for this QTL to IGF2. Hirooka et al. (2001) identified a paternally expressed QTL for teat number in the same genomic region.
Our results also identified maternal expression of a QTL for lipid percent on SSC6, which partially confirms a maternally expressed QTL for drip loss in the same region by de Koning et al. (2001a). De Koning et al. (2001a) also identified a maternally expressed QTL for intramuscular fat content about 50 cM proximal to our lipid percent QTL on SSC6. Although not confirmed by other studies, we detected a Mendelian QTL for chewiness score at 54 cM on SCC6, which is correlated to sensory evaluations of the porcine longissimus muscle (Huff-Lonergan et al., 2001).
Several QTL with maternal expression for meat quality traits were mapped to the same region of SSC9 (Table 4). These results agree with de Koning et al. (2001a), who detected maternal expression for shear force and pH in a similar region. Milan et al. (2002) found QTL with parent-of-origin effects for belly weight and the percentage of ham and loin in the carcass in the same region, but did not provide information on the direction of expression (paternal vs. maternal). Quintanilla et al. (2002) revealed imprinted QTL affecting growth traits in the distal region of SSC9. We identified Mendelian QTL for the same type of traits in the same chromosomal region (Table 3).
Several QTL with parent-of-origin effects on growth and meat quality traits were also identified on SSC10 (Table 4; Figure 4A). Within the distal region of our paternally expressed QTL for loin muscle area, de Koning et al. (2001b) located a paternally expressed QTL for early growth. Another QTL with parent-of-origin effects on early growth was identified by Knott et al. (1998) in the distal region of SCC10.
We also detected two QTL with parent of origin-effects affecting meat quality traits on SSC15 (Table 4). Previous studies (Malek et al., 2001b) detected QTL for meat quality on this chromosome using a Mendelian model, but, so far, no other study has detected imprinted QTL on this chromosome.
Two maternally expressed QTL that influence performance traits were detected on SSC17 (Table 4, Figure 4B). De Koning et al. (2001b) also reported a maternally expressed QTL for lifetime growth on SSC17, but the identified genomic region was about 75 cM proximal to our QTL. Our results might be confirmed by findings in humans because this porcine genomic region is conserved in HSA20 (Lee et al., 2001). Davies and Hughes (1993) have implicated GNAS1 as a maternally expressed gene in this conserved region in humans for the inheritance pattern of Albright’s hereditary osteodystrophy.
Some QTL with parent-of-origin effects detected by other studies could only be identified as Mendelian QTL within our study. This includes QTL with parent-of-origin effects on SSC4 identified by Knott et al. (1998). We detected Mendelian QTL for lumbar and last-rib backfat in a nearby region. De Koning et al. (2001b) also detected a maternally expressed QTL for growth on test in a similar region. De Koning et al. (2000) also detected maternal expression of a QTL for muscle depth on SSC7, which could not be confirmed in our study. We did, however, detect several Mendelian QTL, with significant effects on growth and composition, in the same chromosomal region.
Nature of Parent-of-Origin Effects.
For most of the QTL showing parent-of-origin effects, biological reasons for the inherited mode are difficult to derive. Most of the traits investigated here, in particular those related to meat quality, have not been considered in the mouse or human, which are the two most extensively studied species with regard to imprinting. Evolutionary reasons behind the presence of parent-of-origin effects are also unclear, although several theories exist. One of the best accepted theories is based on the conflict hypothesis, as proposed by Tycko and Morison (2002), which seeks to explain the evolutionary development of imprinting mechanisms by positing opposite maternal vs. paternal "drives" to control allocation of maternal resources to the conceptus. The father will propagate his genome most efficiently if his germline imprints genes in a pattern that promotes growth of his offspring, both in utero and in the postnatal period. The mother, by contrast, is postulated to propagate her genome more successfully by imprinting genes to prevent undue metabolic demands on her resources by any single conceptus or by any single pregnancy.
The conflict hypothesis model makes a strong prediction that the mode of expression of a gene should correlate with its function. Genes that are paternally expressed are predicted to promote growth of the offspring or in some other way increase demands on maternal resources, whereas maternally expressed genes should have the opposite effect and tend to conserve resources to divide them among more offspring and to maximize reproductive performance on the female. In our study, QTL with parent-of-origin effects for growth-related traits tended to be paternally expressed (11 out of 17), whereas those for meat quality QTL tended to be maternally expressed. Hermesch et al. (2000) found nonzero genetic correlations between reproduction traits and meat quality traits for Australian pigs, which may explain the excess of maternally vs. paternally expressed QTL for meat quality traits. Further research is, however, needed to evaluate these conjectures.
|
Implications |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
|
Footnotes |
|---|
2 Present address: Vereinigte Informationssysteme Tierhaltung, Verden, Germany.
3 Present address: Hankyong National University, Anseong, South Korea.
4 Present address: ETH, Zurich, Switzerland.
5 Correspondence: 225C Kildee Hall (phone: 515-294-7509; fax: 515-294-9150; e-mail: jdekkers{at}iastate.edu).
Received for publication November 26, 2003. Accepted for publication April 20, 2004.
|
Literature Cited |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
This article has been cited by other articles:
|
|
 |
|
  V. Russo, L. Fontanesi, E. Scotti, F. Beretti, R. Davoli, L. Nanni Costa, R. Virgili, and L. Buttazzoni Single nucleotide polymorphisms in several porcine cathepsin genes are associated with growth, carcass, and production traits in Italian Large White pigs J Anim Sci, December 1, 2008; 86(12): 3300 - 3314. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
|
