| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
ANIMAL GENETICS |
Department of Animal Science, Michigan State University, East Lansing 48824
 |
Abstract |
|---|
 Top Abstract INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
Key Words: growth • pig • quantitative trait locus • random regression
|
INTRODUCTION |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
The Duroc and Pietrain breeds are utilized worldwide as sire breeds, and these breeds differ in growth phenotypes. In general, Duroc pigs and their offspring have been found to grow faster, but also have more backfat than other breeds (Kennedy et al., 1996
; Blanchard et al., 1999; Edwards et al., 2006). Quiniou and Noblet (1995) used Pietrain boars in their study of equations to predict composition because of their propensity toward leanness, and Edwards et al. (2006) reported slower but leaner growth in Pietrain-sired pigs vs. Durocsired pigs. Whereas each of these breeds has been utilized individually in resource populations with some native and some commercial breeds (Duroc: Grindflek et al., 2001; Sato et al., 2003; Stearns et al., 2005) (Pietrain: Nezer et al., 2002), few studies have reported a resource population utilizing both Duroc and Pietrain breeds. The objective of this study was to conduct a full genome scan by using microsatellite markers to search for QTL affecting growth traits in an F2 Duroc x Pietrain resource population.
|
MATERIALS AND METHODS |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
Population Development
A 3-generation resource population was developed at Michigan State University to study traits of growth, body composition, and meat quality. Semen from 4 F0Duroc sires from a closed, unselected control population (Kuhlers et al., 2003) and 16 F0 Pietrain dams from a closed herd propagated the F1 generation. All grandparents were confirmed to be homozygous normal for the RYR1 gene by a DNA test (Fujii et al., 1991). All animals were produced through AI at the Michigan State University Swine Teaching and Research Farm. From F1 progeny, 51 females and 6 males (sons of 3 F0 sires) were retained to produce the 1,259 F2 pigs born alive in 142 litters across 11 farrowing groups. Females were bred to boars that were not full- or half-siblings and were retained across multiple parities (with different sires for each litter) to produce F2 progeny.
Animal Management
All pigs were managed similarly in farrowing and nursery stages with dams placed into farrowing stalls 1 wk before farrowing. Baby pigs were processed (individually identified by ear tag, given 0.5 mL of penicillin and 1 mL of iron dextran subcutaneously, and tails clipped) at approximately 1 d of age. At 7 d of age, males not kept for breeding purposes were castrated. Pigs were weaned at 16 to 25 (mean of 19.8) d of age and sorted into nursery pens by sex and weight. All diets fed were Michigan State University standard swine farm diets that met or exceeded NRC (1998) requirements for all nutrients at each production stage. At 10 wk of age, F2 pigs were placed into 1 of 2 finishing facilities at the Michigan State University Swine Teaching and Research Farm. Farrowing groups 1, 3, 5, 7, 9, and 11 (n = 521 pigs) were placed into a modified, open-front building with two-thirds solid, one-third slatted floors and wet-dry feeders. Four larger pens (2.03 x 6.91 m) with 2-space feeders were targeted to contain 16 pigs per group. Four smaller pens (1.42 x 6.91 m) with 1-space feeders were targeted to contain 12 pigs per group. Farrowing groups 2, 4, 6, 8, and 10 (n = 465 pigs) were placed into a test station facility with solid floors bedded with straw or wood shavings with 1-space dry feeders and cup drinkers. All 25 pens used (1.42 x 4.93 m) were targeted to contain 4 pigs per group. Pigs in either facility had ad libitum access to feed and water.
Phenotype and Genotype Collection
Live animal traits collected on the F2 animals included BW at birth, weaning, and 6, 10, 13, 16, 19, and 22 wk of age. Additionally, B-mode ultrasound (Pie Medical 200SLC, Classic Medical Supply Inc., Tequesta, FL) estimates of 10th-rib backfat (BF10), last-rib backfat (LRF), and LM area (LMA) were recorded at 10, 13, 16, 19, and 22 wk of age. The ADG from 10 to 22 wk of age and the number of days to reach 105 kg were calculated from these BW measures (National Swine Improvement Federation, 2006). At each of these time points, measures of fat-free total lean (FFTOLN), total body fat tissue (TOFAT), empty body protein (EBPRO), and empty body lipid (EBLIPID) were calculated by using equations similar to those used by Wagner et al. (1999).
Whole blood was collected from all F0, F1, and F2 animals for DNA isolation. White blood cells were separated and frozen for subsequent DNA extraction. A total of 206 dinucleotide microsatellite genetic markers were considered for genotyping, and 128 informative markers were chosen. Markers were selected from the published pig genome linkage map (http://www.marc.usda.gov/genome/swine/swine.html; USDA, 2005), and the initial set of markers was tested for informativeness (markers segregating between F0 breeds and estimated to have 3 or more alleles) by using fluorescent primers distributed by the US Pig Genome Coordinator (supported by the USDA-Cooperative State Research, Education, and Extension Service through National Research Support Project 8).
Genotypes for the 128 markers were determined for 510 F2 animals, their parents, and grandparents at a commercial laboratory (GeneSeek Inc., Lincoln, NE). These 510 animals were sampled across all farrowing groups from 61 entire litters and represented all F1 sires with at least 100 grand progeny from each F0 sire. Fifteen of the 16 F0 dams had a son or daughter as a parent that produced multiple litters of the selected F2 pigs, with the remaining F0 dam represented by a single F1 daughter with 1 litter in this group. Mendelian inheritance of alleles could not be verified within F2 animals for SW1349 on SSC 9, SW2067 on SSC 10, SW1632 on SSC 11, or S0229 on SSC 12, and these markers were removed from the data set. The remaining 124 markers were used for a whole-genome scan of approximately 20-cM spacing across the 18 autosomes and X chromosome. Among these 124 markers, some individual animal genotypes could not be verified and were removed. Markers used in this genome scan are listed in Table 1, along with the number of alleles segregating for each marker, the relative position of each marker, and the number of verified genotypes for each marker. Genetic linkage maps were constructed for each of the 18 autosomes and the X chromosome by using Crimap version 2.4 software (Green et al., 1990). To handle nonpseudoautosomal, X-linked loci, a dummy allele was created and assigned as the second allele for all males in the pedigree.
|
Serial BW and ultrasound estimates from 10 to 22 wk of age were used to generate random regression equations to model pig BW, BF10, LMA, LRF, FFTOLN, TOFAT, EBPRO, and EBLIPID on age at measurement for individual animals. Random regression accounts for the correlation structure between serial measurements on the same animal, and thus may lead to discovery of different QTL than those at the single time point measurements, for which the correlation structure is not taken into account. Age at measurement was modeled as week on-test, calculated as age in weeks minus 10 (i.e., 0, 3, 6, 9, and 12 as distinct covariate values used in the analysis). A random intercept for each animal and a linear regression on age for each animal were included in each model. Main effects and interactions that contributed significantly to the increase of the log likelihood were kept in the model. Table 2 lists the polynomial order of week of age and interactions used in models for these 8 traits as determined by log likelihood tests of significance. The following model was used:
|
 |
in which Yijklmno is the record on the oth pig within the jth sex, kth finisher, lth group, mth pen, and nth litter regressed on the 
th polynomial week i; µ is the overall mean of the trait; weeki is the fixed regression coefficient for polynomial terms (1 to 4) of week i; sexj is the fixed effect of the sex of animal j (barrow or gilt); fink is the fixed effect of finisher k (modified open front or test station building); grp(fin)kl is the random effect of farrowing group l (1 to 11) nested within finisher, where grp ~N(0, I
grp 2); pen(grp)lm is the random effect of pen m (1 to 25) nested within farrowing group, where pen ~N(0, Igrp2); litn is the random effect of litter n (1 to 142), where lit ~N(0, Ilit2); go is the random intercept for animal o, 
o is the random linear regression coefficient on age for animal o, Zi is the week on test as a covariate, and eijklmno is the random error.
The distributional assumptions on g = {go} and = {o} were such that
 |
where g2 is the intercept genetic variance for the individuals, 2 is the linear age by animal genetic variance, and g is the genetic covariance between the intercept and linear term for each animal. To account for residual variances across serial measurement and the relationship between time points, the e = {eijklmno} was specified as normally distributed with a general (co)variance structure calculated from the data and specified within and across weeks, with 5 variance and 10 covariance terms. These analyses were performed by using the ASREML software package (Gilmour et al., 1999).
QTL Analysis
The genetic linkage maps constructed for each chromosome were used in an F2, least squares, interval mapping framework for QTL analysis (Haley et al., 1994). Briefly, this method proceeded in 2 steps. In the first step, marker positions and genotypes were used to calculate the probability that an individual inherited 0, 1, or 2 alleles from each of the 2 founder lines. Second, phenotypic data were regressed on coefficients derived from these probabilities. The F2 analysis option of the QTL Express software (Seaton et al., 2002) was used to search for single QTL with additive, dominance, or imprinting effects on the 18 autosomes and additive effects on the X chromosome. For traits listed in Table 3, the model used included the fixed effects of sex of animal and litter. The model for 10- to 22-wk ADG also included a covariate of 10-wk BW to account for differences in BW when pigs were placed into the finishers.
|
. Analyses that included the animal intercept and linear regression terms from the random regression models were also carried out with QTL Express, but only for additive effects. During the course of estimation of the animal intercept and linear terms, dominance effects were confounded with common environment (as stated in Hill, 1999). Because these traits have already been adjusted for fixed and random effects in random regression models, no further effects were included in their QTL analyses.
Additive effects were defined as half the difference between Pietrain and Duroc genotypes at the QTL. Dominance effects were defined as the difference between the heterozygote genotype and the average of the homozygote genotypes. Imprinting effects were defined as the difference between the heterozygous genotypes when the Duroc allele was inherited from the sire vs. from the dam. Haley et al. (1994) and Knott et al. (1998) describe further details on the parameterization of these effects. Significance thresholds of 5 and 1% at the chromosome-wise level and 5 and 1% at the genome-wise level were determined through the use of permutation tests (Churchill and Doerge, 1994) in QTL Express. These thresholds were determined for genome-wise levels and for each chromosome based on appropriate models for each chromosome and 30,000 permutations. For each QTL determined to be significant at the 5% genome-wise level, confidence intervals of the QTL position were determined by using a bootstrap method with 5,000 permutations in QTL Express (Visscher et al., 1996).
|
RESULTS AND DISCUSSION |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
Trait means and SD are listed in Table 3, and are similar to those measured in other resource populations for similar traits (e.g., Malek et al., 2001b; Stearns et al., 2005). Table 1 contains a list of the 124 markers used in this analysis along with their linkage map positions on each chromosome, the number of alleles segregating in this population, and the number of genotypes determined for F2 animals. The average distance between markers was 24 cM, although the distance between some markers was larger because of the difficulty in identifying informative markers in those regions. Overall, our map generally agreed with or was slightly longer than the USDA-ARS map (Rohrer et al., 1996; http://www.marc.usda.gov/genome/swine/swine.html, USDA, 2005).
The permutation test results to determine significance thresholds for the F-statistics for the model testing additive, dominance, and imprinting effects ranged from 3.44 to 4.35 across chromosomes for the 5% chromosome-wise thresholds and from 4.71 to 5.64 for the 1% chromosome-wise thresholds. Genome-wise threshold levels were 6.34 and 7.63 for the 5 and 1% thresholds, respectively. Significance threshold levels for the additive effect only models ranged from 5.39 to 7.64 and 8.68 to 10.99 for the 5 and 1% chromosome-wise levels, respectively, and were 12.86 and 16.05 for the 5 and 1% genome-wise levels, respectively.
Estimates of positions and F-ratios of QTL significant at the 5% chromosome-wise level for traits listed in Table 3 are listed in Table 4. These results were derived from single QTL models on each chromosome. Information in the table is sorted by chromosome and position within each chromosome. Additionally, the additive, dominance, and imprinting effect at each QTL along with their SE are listed in Table 4. A total of 43 QTL for traits in Table 3 were found to be significant at the 5% chromosome-wise levels. Of these 43 QTL, 20 were significant at the 1% chromosome-wise significance threshold, 14 of these 20 were also significant at the 5% genome-wise significance threshold, and 10 of these 14 were also significant at the 1% genome-wise significance threshold. No significant QTL were detected in this population for 3-, 6-, 10-, 13-, or 16-wk BW, 13-wk LMA, or 22-wk FFTOLN.
|
. In addition, the additive effect and SE at each QTL are listed in Table 5. A total of 22 QTL for 13 of the 16 animal random regression terms were found to be significant at the 5% chromosome-wise level. Of these 22 QTL, 6 were significant at the 1% chromosome-wise significance threshold, 4 of these 6 were also significant at the 5% genome-wise significance threshold, and 3 of these 4 were also significant at the 1% genome-wise significance threshold.
|
, but a few cryptic alleles exist that act in an opposite direction to the general trend for the overall breed effects. Body Weight
Four QTL influenced BW growth at different ages at the 5% chromosome-wise level (Table 4). A QTL for birth weight was discovered on SSC 5 that has not been reported before in other pig resource populations (Hu et al., 2005). No QTL for growth between 3 and 16 wk of age were detected in this analysis. However, a QTL on SSC 16 was significant for BW at 19 and 22 wk of age and influenced 10- to 22-wk ADG. This was a cryptic allele, because the Duroc alleles caused lower BW and 10- to 22-wk ADG, which contrasted with results from Edwards et al. (2006), who reported faster growth from Durocsired vs. Pietrain-sired progeny. The 1-cM F-ratio tests for these QTL are illustrated for SSC 16 in Figure 1, with the F-ratio plotted vs. relative marker position. Another QTL affecting 10- to 22-wk ADG was found on SSC 7. Other studies have reported a QTL for ADG on SSC 7, but for differing time periods of development than in this study (Knott et al., 1998; Nezer et al., 2002) or in different positions on the chromosome (Bidanel et al., 2001). One QTL was found that influenced the animal random regression linear rate of BW gain from 10- to 22-wk of age (Table 5). This QTL, on SSC 7, had larger additive effects from Duroc alleles that increased the rate of BW gain. No linear rate QTL, or traits with similar definitions, was reported on SSC 7 in a summary of other pig QTL studies (Hu et al., 2005). In addition, a QTL for days to 105 kg was discovered on SSC 9. Although ADG from 10 to 22 wk of age and days to 105 kg are related traits, because differences in growth from birth to 10 wk of age are included in the calculation of days to 105 kg, these traits may be influenced by different combinations of genes; thus, QTL for the 2 traits were identified on different chromosomes.
|
Several regions of the genome contributed to fat tissue phenotypes at many ages of development. When considering the associated traits of BF10 and LRF, a region on SSC 6 influenced both of these traits at all measured time points of 10, 13, 16, 19, and 22 wk of age, with all of them significant at the 1% genome-wise level (Table 4). The estimate of the peak position of this QTL ranged from 134 to 143 cM from the origin of our SSC 6 linkage map (S0099), with overlapping 95% confidence intervals for the traits; therefore, this is likely to be the same pleiotropic QTL for all of these related traits. Figure 2 demonstrates the F-ratio curves plotted vs. relative marker positions on SSC 6 and illustrates the similar shapes of F-ratios for 22-wk BF10 and 22-wk LRF. The estimate of this QTL in this analysis indicated that Duroc alleles contributed to larger measures of BF10 and LRF. Although most backfat QTL on SSC 6 found in other populations occur more proximal to our 0-cM marker than those reported here, Malek et al. (2001a) and Ovilo et al. (2002) also reported putative QTL for backfat in the same region as those reported in this study. Two other regions were significant for multiple backfat phenotypes. One of these was on SSC 11, which influenced BF10 at 19 and 22 wk of age and LRF at 19 wk of age (Figure 3). The only other study that reported a backfat QTL on SSC 11 was Milan et al. (2002), but their QTL was more proximal to our 0-cM marker than the QTL reported here. Another region, at 64 cM on SSC 16, influenced BF10 at 19 wk of age and LRF at 22 wk of age with a larger additive effect characterized by Duroc alleles that caused less BF10 and LRF (Figure 1). Interestingly, only 1 QTL, for small intestine length (Hu et al., 2005), has previously been identified on SSC 16, but several traits in this study were influenced by QTL on SSC 16 (Tables 4 and 5). A QTL on SSC 18 at 54 cM was significantly related to BF10 at 10 wk of age and LRF at 16 wk of age. Additional locations of QTL affecting BF10 and LRF at different time points from 10 to 22 wk of age occur on SSC 3, 8, and 9 (Table 4). Random regression QTL affecting animal intercept and linear terms for fat tissue phenotypes were identified on SSC 4, 6, 8, 10, 11, 16, and 18 (Table 5). Within a region on SSC 6 136 to 144 cM away from the S0099 marker, BF10 intercept, LRF intercept, and LRF linear random regression terms had significant QTL (Figure 4). This region is more distal from S0099 than the RYR1 region. On SSC 8, a QTL for the LRF random regression linear term was significant in a region that also influenced other fat tissue phenotypes (Figure 5).
|
|
|
|
Numerous QTL controlling LMA in this population were detected, with most of them significant only at the 5% chromosome-wise level. One chromosome with 3 regions significant for LMA QTL was SSC 6 (Tables 4 and 5). One region that was significant for 10 and 19 wk of age LMA had a larger dominance effect, with Duroc alleles contributing to a larger LMA. Another region was significant for 16 wk of age LMA with a larger dominance effect, where Pietrain alleles contributed to a larger LMA. Last, Duroc alleles contributed significantly to a larger LMA intercept derived from random regression analyses. A QTL influencing lean percentage was reported in the same region in 2 other populations containing Duroc animals (Grindflek et al., 2001; Szyda et al., 2002; Stearns et al., 2005). Additionally, SSC 1 had an additive QTL region that affected 13 wk of age LMA; Pietrain alleles contributed to larger LMA. The SSC 4 had a QTL region that affected 22 wk of age LMA, where Pietrain alleles contributed to a larger LMA. Regions just distal (relative to our 0-cM marker) to those reported here that influence LMA have been identified in a few populations (Pérez-Enciso et al., 2000; Malek et al., 2001a; Varona et al., 2002), whereas 1 study identified a QTL for LMA in the same region as reported here (Wimmers et al., 2002). A region at the proximal end relative to our 0-cM marker of SSC 18 affected the linear random regression of LMA (Table 5). For this QTL, Duroc alleles caused a faster rate of increase in LMA.
Composition Traits
Estimates of body composition (FFTOLN, TOFAT, EBPRO, and EBLIPID) at 22 wk of age, as well as random regression analyses of serial data from 10 to 22 wk of age, were tested for QTL. The use of random regression terms as traits for QTL analysis has not been previously reported for other swine resource populations. Two QTL with larger dominance effects for FFTOLN at 22 wk of age were discovered, with 1 on SSC 6 and another on SSC 9 (Table 4). The Duroc alleles increased the amount of FFTOLN at 22 wk of age. A QTL on SSC 4 affected the random regression linear term of FFTOLN (Table 5). In this instance, the Pietrain alleles increased the FFTOLN intercept. Results for QTL for TOFAT indicated 3 QTL, with 1 each on SSC 6, 9, and 16 affecting TOFAT at 22 wk of age (Table 4). The random regression animal intercept for TOFAT was influenced by a QTL on SSC 6, whereas random regression animal linear TOFAT was influenced by QTL on SSC 8 and 16 (Table 5). The QTL on SSC 6 and 8 were co-located with several other fat tissue traits (Figures 4 and 5). Putative QTL with larger dominance effects were found on SSC 6 and 9, where Duroc alleles contributed to an increase in EBPRO. Both of the protein composition traits of FFTOLN and EBPRO had similar patterns for F-ratio curves on SSC 6 as indicated in Figure 2. Random regression animal intercept and linear terms for EBPRO were influenced by QTL on SSC 5, with an increase in both terms resulting from Pietrain alleles. Effects of QTL on SSC 6 and 11 were evident for EBLIPID at 22 wk of age, as well as in the same region of SSC 9 that contributed to differences in EBPRO (Table 4). Additionally, a region on SSC 6 also contributed to an increase in the random regression EBLIPID intercept when Duroc alleles were present. The F-ratios for TOFAT and EBLIPID on SSC 6 were generally concurrent (Figure 2), indicating that the same chromosomal regions influenced both traits. Traits related to body composition of FFTOLN, TOFAT, EBPRO, and EBLIPID were influenced by a QTL between the 2 most proximal markers (SW21 and SW983) relative to the origin of our SSC 9 map (Figure 6). No other studies have reported QTL for these or similar traits on SSC 9 (Hu et al., 2005).
|
For QTL that were significant at the 5% genome-wise level, 95% confidence intervals were estimated by using bootstrapping with resampling in QTL Express (Seaton et al., 2002). These confidence intervals are listed in Table 6. Confidence intervals were not calculated for QTL significant at the 5 or 1% chromosome-wise level because preliminary analyses indicated that many of these intervals tended to encompass the entire chromosome on which they reside.
|
Discovery of QTL in an F2 Duroc x Pietrain resource population revealed many regions of the genome that potentially influence a portion of the phenotypic differences in growth and composition traits. Not only were QTL found for measured traits, but also for the animal terms from random regression analysis. Using both measured and composite traits allowed unique characterization of growth and composition phenotypes from this population. The use of the Duroc and Pietrain breeds in this population will allow for identified QTL that are potentially already segregating in commercial pig populations to be incorporated into breeding schemes. Although further analyses and refining of QTL positions are in progress, results of this study are useful in discovering potential QTL regions and leading to further understanding of the role certain regions of the genome have in determining phenotypic differences among potential parents for selection of breeding animals.
|
Footnotes |
|---|
2 Corresponding author: batesr{at}msu.edu
Received for publication September 12, 2006. Accepted for publication October 24, 2007.
|
LITERATURE CITED |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
This article has been cited by other articles:
|
|
 |
|
  G. J. van Essen, J. C. M. Vernooij, J. A. P. Heesterbeek, D. Anjema, D. Merkus, and D. J. Duncker Does cardiovascular performance of modern fattening pigs obey allometric scaling laws? J Anim Sci, June 1, 2009; 87(6): 1991 - 1997. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. Gallardo, R. N. Pena, M. Amills, L. Varona, O. Ramirez, J. Reixach, I. Diaz, J. Tibau, J. Soler, J. M. Prat-Cuffi, et al. Mapping of quantitative trait loci for cholesterol, LDL, HDL, and triglyceride serum concentrations in pigs Physiol Genomics, November 12, 2008; 35(3): 199 - 209. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 F. F. Cardoso, G. J. M. Rosa, J. P. Steibel, C. W. Ernst, R. O. Bates, and R. J. Tempelman Selective Transcriptional Profiling and Data Analysis Strategies for Expression Quantitative Trait Loci Mapping in Outbred F2 Populations Genetics, November 1, 2008; 180(3): 1679 - 1690. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. B. Edwards, C. W. Ernst, N. E. Raney, M. E. Doumit, M. D. Hoge, and R. O. Bates Quantitative trait locus mapping in an F2 Duroc x Pietrain resource population: II. Carcass and meat quality traits J Anim Sci, February 1, 2008; 86(2): 254 - 266. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
