J. Anim. Sci. 2005. 83:2289-2296
© 2005 American Society of Animal Science
Large-scale, multibreed, multitrait analyses of quantitative trait loci experiments: The case of porcine X chromosome1
M. Pérez-Enciso*,
,2,
A. Mercadé,
J. P. Bidanel
,
H. Geldermann
,
S. Cepica#,
H. Bartenschlager,
L. Varona¶,
D. Milan|| and
J. M. Folch
* Institut Català de Recerca i Estudis Avançats, Lluis Companys 23, Barcelona 08010, Spain;
and
Departament de Ciència Animal i dels Aliments, Facultat de Veterinària, Universitat Autònoma de Barcelona, Bellaterra, 08193, Spain;
and
Station de Génétique Quantitative et Appliquée, INRA, 78352 Jouy-en-Josas, France;
and
Department of Animal Breeding and Biotechnology, University of Hohenheim, D-70593 Stuttgart, Germany;
and
# Institute of Animal Physiology and Genetics, Academy of Sciences of the Czech Republic, 277 21 Libechov, Czech Republic;
and
¶ Àrea de Producció Animal, Centre UdL-IRTA, Alcalde Rovira Roure 177, Lleida, 25198, Spain; and
and
|| Laboratoire de Génétique Cellulaire, INRA, 31326 Castanet-Tolosan, France
 |
Abstract
|
|---|
A QTL analysis of multibreed experiments (i.e., crossed populations involving more than two founder breeds) offers clear advantages over classical two-breed crosses, among them increased power and a more comprehensive coverage of the total genetic variability in the species. An alternative to designed multibreed crosses is to reanalyze jointly several experiments involving different breeds. We report a multibreed, multitrait QTL analysis of SSCX that involves five different crosses, six breeds, and almost 3,000 genotyped individuals using a truly multibreed strategy to allow for any number of founder breed origins. Traits analyzed were growth, fat thickness, carcass length, and shoulder and ham weights. Generally, the joint analysis resulted in more significant QTL than the single-experiment analyses. We show that the QTL for fatness, which is highly significant (nominal P < 10–43), is of Asiatic origin (Meishan). The next most significant QTL (nominal P < 10–15) affected ham weight and seems to be segregating only between Large White and the rest of the breeds. A multitrait, multi-QTL analysis suggests that these are two distinct loci. Additionally, a locus segregating only between Iberian and Landrace affects live weight. The advantages of joint, multibreed analyses clearly outweigh their potential risks.
Key Words: Fatness • Growth • Pig • Quantitative Trait Locus • Sex Chromosome
|
Introduction
|
|---|
Traditionally, QTL analyses in crosses have involved an experimental design with two parental lines. Logically, the risk exists that the QTL has the same allele fixed in both lines (or more generally, that allelic frequencies are the same in the two lines), and thus, a potentially important region may remain undetected. To minimize this risk, phenotypically extreme lines are usually chosen as founders. The drawback to this approach is that only a portion of the total genetic variability in that species will be detected, restricted only to the differences between the two particular lines chosen. In principle, multibreed crosses should allow for uncovering a much larger fraction of the total genetic variation than two parental crosses. A multibreed experiment, if it uses all available information in the different crosses, should be more powerful than single cross studies. The problems now are 1) that such an experiment will be more expensive and difficult to set up than traditional designs, 2) that specific QTL statistical methodology has not been fully developed, and 3) that computing constraints exist. One logical alternative that addresses the first concern is jointly reanalyzing data from several experiments carried out at different times and/or by different researchers (Walling et al., 2000
; Kim et al., 2005).
Here we report a multibreed, multitrait QTL analysis of SSCX that involves five different crosses and almost 3,000 genotyped individuals. We develop a multibreed approach to allow for any number of founder breed origins that uses an efficient computing strategy. In addition, we also performed multi-QTL multitrait studies. In so doing our goals were 1) to have a more accurate picture of how many QTL are segregating in this chromosome and 2) to choose between the single pleiotropic QTL and the two-linked QTL hypotheses. We have focused on growth, fatness, and carcass traits because these are characteristics usually measured in all QTL experiments and in comparable conditions.
|
Materials and Methods
|
|---|
Materials
Three experiments (ES, FR, and D) comprising five crosses were analyzed. These involved a total of six pig breeds, including Wild Boar: Iberian x Landrace (ES; IB x LD); Large White x Meishan (FR; LW x MS); MS x Pietrain (PI) (D1; MS x PI); Wild Boar (WB) x MS (D2; WB x MS); and WB x PI (D3). The experiments and traits recorded are fully described elsewhere (Bidanel et al., 2001; Milan et al., 2002; Pérez-Enciso et al., 2002; Geldermann et al., 2003), and short summary statistics are presented in Table 1. Regarding the IB x LD cross, only the QTL analysis of intramuscular fat and pigment content in the X chromosome have been reported (Pérez-Enciso et al., 2002).
The IB x LD cross consisted of 321 F2 animals, 85 backcross pigs (produced after mating four F2 boars and 22 Landrace gilts), and 87 F3 pigs (offspring of four F2 boars and 16 F2 gilts). The rest of the experiments were F2 crosses. The FR experiment used MS as maternal founder line, crossed to LW boars, and generated 1,084 F2. The German (D1, D2, D3) family material included three different F2 crosses, each originated with a single founder male and four to nine sows. All three crosses were connected because the WB was the same in crosses D2 and D3, as were two PI sows in crosses D1 and D3. The D1, D2, and D3 families encompassed 316, 335, and 315 F2 animals, respectively. The total data set analyzed consisted of 2,813 animals; 2,551 had phenotypic records.
The traits analyzed were live weight at or close to slaughter (LWT), backfat thickness in the neck (first rib; BF1) and at the last rib (BF2), carcass length (CL), shoulder weight (SW), and ham weight (HW). Measurements BF1 and BF2 coincide approximately with the thickest and thinnest fatness, respectively. Fatness measurements were obtained according to several techniques. In the ES experiment, BF1 and BF2 were measured 24 h postmortem in the slaughterhouse. Measurements of BF2 were obtained with ultrasound in the FR experiment. In this experiment, males and females were managed in different batches; weights and backfat thickness were recorded every 3 wk. Measurements of BF1 and BF2 were obtained as the mean of three measurements in the D1, D2, and D3 crosses (Geldermann et al., 2003).
Table 1
shows the main phenotypic statistics. On average, the fastest growing animals were IB and LD, and the slowest growing were those involving WB and MS. The fattest animals were WB x MS, and the leanest were WB x PI. Although some means were quite different between experiments, phenotypic variances were similar. As a consequence, the risk of false positives caused by heteroskedasticity can be assumed to be small. Nevertheless, we standardized the data within experiment before the QTL analyses. The P-values and QTL positions were very similar in the analyses to either standardized or raw data (results not presented).
There were 17 different markers. Only two of the markers (SW2456 and SW1943) were genotyped in all populations, but seven markers were genotyped in more than one cross, making comparisons between experiments easier. For QTL mapping purposes, SW259 and SW1994 were considered the same marker because they were located in identical positions. Three new markers were added in the ES experiment, SW2470, SW1943, and SW2059, compared with previously reported results (Pérez-Enciso et al., 2002). Given that the estimated distances between the two common microsatellites (SW2456 and SW1943) were similar across experiments and close to those published, we used the distances from the female map available at NCBI (http://www.ncbi.nlm.nih.gov/mapview/map_search.cgi?taxid=9823); these markers and positions are detailed in Table 2.
QTL Analysis
The general univariate model was
 | [1] |
where yij is observation i within cross j (j = 1, ..., 5), Pijk is the probability of individual ij having an allele of breed origin k (at the position analyzed), 
k is the breed allelic effect k, cross is the cross effect, ß is the covariate effect (nested within cross), batch is the slaughter batch, a is the infinitesimal genetic effect, and e is the residual. The covariate c was age at slaughter for live weight or live weight itself for the remaining traits. The covariate was hierarchized to cross, as we found that interactions of experiment x age or weight were very important. Batch and the infinitesimal genetic effect were treated as random, with covariance matrix I
and A
, respectively (A being the numerator relationship matrix). The probabilities P in [1] were obtained using a modification of the MCMC algorithm described by Pérez-Enciso et al. (2000). As described fully in that reference, the P terms are computed, allowing for the fact that females have two X chromosomes and males have one, and include a weighting factor to account for gene inactivation in females.
Note that in [1], we are implicitly assuming that all breeds have the same allele fixed. This may not be justified in all cases, either because the alleles are actually segregating or because there are alternative alleles fixed in different lines of the same breed. We could test whether this possibility or the most parsimonious model [1] is the most appropriate model by simply expanding model [1] and considering the same breed in distinct experiments as different alleles. However, this does increase the number of parameters with the consequent loss of power and decreases the advantage of joint analyses. In this work, model [1] is at least fully justified for the German (D1, D2, D3) experiment because some founder animals were shared between crosses.
We used a maximum likelihood technique as implemented in the Qxpak package (Pérez-Enciso and Misztal, 2004) modified to estimate a different P for each breed. Briefly, this is a two-step strategy, First, the identity by descent probabilities are computed via a Monte Carlo Markov Chain; then maximum likelihood estimates are obtained at successive chromosome positions (here every cM). Maximum likelihood estimates were obtained using an accelerated expectation maximization algorithm, taking advantage of sparse matrix techniques (Misztal et al., 2002). This strategy allows rather large data sets to be analyzed with reasonable computing time and memory requirements. For instance, obtaining the estimates at each chromosome position took approximately 2.5 min on a Pentium 4 (Intel, Santa Clara, CA) personal computer with Linux in the bivariate model (approximately 6,000 equations).
Because 
k Pijk always equals one, absolute allelic effects are not estimable (i.e., here only the allelic differences between IB and LD and among MS, PI, LW, and WB can be estimated.) Nominal P-values were obtained using the usual 
2 approximation to minus twice the likelihood ratio against a model without the QTL effect. Note that permutation techniques cannot be used to compute chromosomewise P with model [1] because the pedigree structure is broken. We have argued elsewhere (Mercadé et al., 2005) that a nominal 10–3 P-value is equivalent to a 1% chromosomewise level, approximately. An alternative would be to simulate the same structure data set under the null hypothesis, but then one is forced to assume that the non-QTL effects such as heritability are known, and it becomes even more complicated with multitrait and multi-QTL tests.
We also carried out a series of multitrait and multi-QTL analyses to refine the genetic architecture of this chromosome. The model used was the same as in [1], although in the multitrait case the distribution of infinitesimal genetic effects is now (e.g., in a bivariate setting):
with the subindices 1 and 2 referring to Traits 1 and 2, respectively, and 
denoting the Kronecker product. In the case of multi-QTL analyses, different P and terms are fitted per QTL. We distinguished pleiotropy from linkage by comparing the likelihood ratios of the model containing two linked loci vs. that of the model with a single pleiotropic QTL, with the difference in degrees of freedom as one, approximately, because of the extra position of the second QTL (Knott and Haley, 2000).
|
Results
|
|---|
Table 3 shows the results of the single experiment analyses. The number of significant results (defined as nominal P < 10–3) varied according to the experiment. Overall, crosses involving MS produced the largest percentage of significant P-values. By far, the most significant QTL was found in the French LW x MS experiment for backfat, a QTL located between markers SW1994 and SW1943. In fact, this also was the most significant result obtained in the FR cross for all traits measured across all chromosomes (Bidanel et al., 2001; Milan et al., 2002). This QTL was confirmed in the remaining crosses involving MS (D1 and D2) in similar positions. The next most significant results were for ham weight in the same cross (FR), growth in the IB x LD cross, and carcass length in the MS x PI cross. All QTL were located between positions 56 (HW; LW x MS) and 74 cM (LWT; IB x LD) (i.e., bracketed between SW2456 and SW1943 markers).
A more global and comprehensive picture than the single experiment studies was obtained in the joint experiment analysis (Table 4; Figure 1). Again, the significance of the BF2 QTL stood out and was even increased with respect to the results in Table 3. The BF2 QTL dramatically dwarfed all remaining results (Figure 1), even if non-negligible significance results were found for other traits, in particular for HW in a location 14 cM away from the BF2 QTL, but also for SW and CL. The significance of HW, CL, and SW QTL also were (notably) increased compared with those in the single experiment analyses, clearly illustrating the advantage of joint analyses. The exceptions were LWT and BF1, although for different reasons. In the case of LWT, it seems that this QTL was segregating only in the IB x LD cross. Therefore, including noninformative extra data and extra parameters (the additional allelic effects) decreased power. Regarding BF1, the fact that this measurement was not recorded in the FR experiment might have contributed to the decreased significance of this QTL.
View larger version (16K):
[in this window]
[in a new window]
|
Figure 1. P-value profile of the joint experiment analysis. The null model contains identical effects as the full model except for the QTL. Traits are live weight before slaughter (LWT), fat thickness at the loin (BF2), carcass length (CL), shoulder weight (SW), and ham weight (HW).
|
|
The sign of the QTL effects were generally in the expected direction; MS alleles increased fatness and decreased CL. The IB allele decreased growth but curiously tended to increase CL, whereas we have observed that the IB allele on the SSC4 QTL has a strong effect in decreasing CL (Mercadé et al., 2005). Traditionally, a long CL has been associated with a lean animal. The PI allele had a strong influence on the shape of the animal, increasing SW relative to all breeds and decreasing HW with respect to LW. The WB alleles affected mostly fatness but did not affect the other traits much, suggesting that modern selection has penalized mainly alleles increasing fat deposition.
Multitrait studies add some important information concerning the genetic architecture of chromosome X (last rows of Table 4). The BF1-BF2 analysis shows that, given the dramatic effect of the BF2 QTL, adding a second trait (BF1) did not convey much extra information; the P-values and the QTL positions were the same as in the BF2-only study. The infinitesimal genetic correlation between BF1 and BF2 was 0.67. Perhaps the most remarkable result of the bivariate BF1-BF2 study compared with the univariate results is that it seems that there is an increased estimate of the IB allele relative to the LD allele for fatness (e.g., 
BF2 = 0.21 ± 0.09 vs. 0.10 ± 0.09 mm.). Nevertheless, there was a relatively large variability across multitrait analyses of the estimated effects; for instance, BF2 in the IB-LD contrast ranged from 0.01 to 0.21 (Table 4), suggesting that the amount of information about some parameters is limited.
Next, we analyzed the two traits exhibiting the most significant QTL, HW and BF2. In this case, we compared a single pleiotropic QTL vs. a two-QTL linkage model. Results are shown in the bottom rows of Table 4. The univariate estimates of the positions differed by 14 cM, and the allelic effects were clearly different for each trait; for BF2, there was a highly significant allele in the MS-PI contrast, whereas the most significant contrast for HW was LW-PI. As expected, the pleiotropic model was highly significant (i.e., there must exist at least one QTL; P < 1.9 x 10–29) but, importantly, it also is quite likely (P < 2.7 x 10–3) that there exist two linked QTL, one affecting HW (between SW2456 and SW259) and the other affecting fatness (SW2456-SLC25A5 interval). By breed, it is likely that the fatness QTL is segregating between MS and the rest of breeds, whereas the HW shows up in the IB x LD cross, in LW and WB, with no differences between PI and MS effects.
|
Discussion
|
|---|
The X chromosome seems to be particularly influential in the porcine species regarding fatness and growth as confirmed in numerous instances (Knott et al., 1998; Rohrer and Keele, 1998; Harlizius et al., 2000; Bidanel et al., 2001; Malek et al., 2001; Cepica et al., 2003a), although no causal mutation or candidate gene has been identified convincingly yet. Here we provide more detailed and compelling evidence regarding the effects of loci harbored by SSCX.
Regarding the single experiment analyses published using these data, the results in Table 3 should be similar to those already published (Bidanel et al., 2001; Milan et al., 2002; Cepica et al., 2003a), which were analyzed using a least squares approach (Haley et al., 1994). The main difference here is the inclusion of the infinitesimal genetic effect. Other differences should have a minor effect, such as slightly different marker positions or regression vs. maximum likelihood estimates. Overall, the agreement was very good (e.g., the QTL for fatness is reported in the FR, D1, and D2 crosses in similar positions and with high significance values). Nevertheless, the P-values obtained here tended to be larger (less significant) than those already published. For instance, Cepica et al. (2003a) reported a significant result (5%) for CL in the D3 cross, whereas the P-value obtained here was 0.09 (Table 3). This result occurred because we included an infinitesimal genetic effect. We have repeatedly observed that the QTL effect may be partially confounded with the infinitesimal genetic variation.
Fatness
Although the presence of several allelic effects cannot be ruled out, the SSCX fatness allele with the largest effect is certainly of Asiatic origin; it showed up predominantly in crosses involving MS (FR, D1, and D2; Table 3), and the largest effect reported was the MS minus PI contrast (Table 4). This hypothesis is reinforced by the fact that other experiments involving MS (Rohrer and Keele, 1998; Harlizius et al., 2000) found higher significance values than those crosses that did not include MS (Knott et al., 1998; Malek et al., 2001; Nezer et al., 2002). Interestingly, the effect is much larger when MS was used as a maternal rather than as a paternal line (i.e., compare crosses MS x PI [D1] vs. WB x MS [D2] or LW x MS [FR]).
An additional interesting observation is that the P-levels were much more significant for BF2 (minimum or close to minimum backfat thickness) than at the neck (BF1, the thickest point). Several explanations are possible. The BF1 was not measured in the largest experiment (FR), making it more difficult to attain similar P-values. Nevertheless, the P-values were more significant for BF2 than for BF1 in experiments with MS, where both measurements were available (D1 and D2; Table 3). Thus, it rather seems that the pattern of fat deposition is different between MS and other European breeds. In fact, the infinitesimal genetic correlation between BF1 and BF2 was approximately 0.69, which is high but not close to 1.00, suggesting that the genetic basis for the two traits is partially different and thus controlled either by different QTL or the same QTL but with different allelic effects between traits. Different profiles of fat deposition between LW and MS have actually been reported (Legault et al., 1985); a simple glance at pictures from MS and European breeds (Porter, 1993) makes it easy for us to suspect that fat deposition is under different genetic controls along the pig body.
Once different studies have confirmed the presence of fatness-related QTL, the next logical step is to carry out a fine mapping study by adding new markers and/or identifying positional candidate genes. In a final step, an association study between a set of candidate polymorphisms and the trait(s) of interest is carried out, ideally in outbred populations, where the causal mutation is segregating because crosses are not a suitable material for fine mapping. Nevertheless, Figure 1 indicates that basically the whole of chromosome X is in high linkage disequilibrium with the BF2 QTL in the F2 cross, so one needs to be careful in interpreting the association signal. It is not completely unexpected, thus, that Gaboreanu et al. (2004) reported an association between an RFLP and fatness in a Berkshire x Yorkshire cross. The low reported significance values (P > 0.02) should not be interpreted necessarily as closeness to the causal mutation(s).
Shape
In addition to fatness and growth, modern intense selection has been accompanied by a change in the shape of the animal. Notably, a longer CL has been correlated with an increased leanness during the last century (Jonsson, 1975). Although the IB allele effect on CL was close to significance, the largest effect corresponded again to the Asiatic allele. Here, in contrast to fatness traits, significance was greater when MS was used as the paternal line vs. the maternal line (Table 3). With regard to the remaining shape measurements and HW and SW, the largest differences in allelic effects were found between LW and PI, where the effects were diametrical; an increased HW was matched by a comparable SW decrease. Note that HW and SW were corrected by total LWT; so in principle, we were measuring a change in how total weight is partitioned among the different body parts. Nonetheless, with a LW x PI cross, Nezer et al. (2002) did not report an effect on ham or shoulder percentage. Whether because of differences in statistical analyses or finding a false positive in the present study, this QTL needs to be confirmed in further studies. Nevertheless, it is interesting to observe that other chromosomes harbored loci affecting both traits in the same direction (e.g., SSC4 in the D1 and D2 crosses; Cepica et al., 2003b). In the IB x LD cross, the SCC4 effects also were in the same direction for the two traits but much more significant in SW than in HW (Mercadé et al., 2005). Thus, we can speculate that at least different alleles, if not different genes, have caused the changes in shape between breeds.
Advantages of Multibreed Analyses
The statistical approach used here offers several advantages over classical least squares techniques for crosses (Haley et al., 1994). In particular, it is a truly multibreed approach, estimating an effect per breed and using all pedigree information simultaneously. In contrast, the approach in Walling et al. (2000) considered an interaction between QTL and cross to identify whether the allelic effects are different. The approach used by Walling et al. (2000) does not allow estimation of the global breed effect across experiments. A further advantage of our methodology is that it accounts for common origins across experiments. For example, the German family material (D1, D2, and D3) used a single WB as well as several PI sows as parents in two crosses. Our methodology accounted for this fact. The advantage of multibreed analyses can be seen by comparing the P-values obtained in the single vs. multiple-experiment analyses. Despite the heterogeneity between management, more significant P-values were obtained in the latter than in the former (classical) studies for all traits except LWT and SW, where the minimum P-value and the global P-value were similar. Nonetheless, our methodology provides estimates for the QTL effect within each breed (i.e., those that are estimable) and thus provides clues as to which breeds are segregating the allelic effects. Finally, multitrait analyses also help to disentangle how many QTL are segregating in the population (Table 4).
One of the advantages of multibreed analyses is the possibility of estimating the number of different allelic effects. An interesting result in our analysis is that, for some traits, a continuum in the distribution of allelic effects can be observed (e.g., for BF2, allelic effects can be ranked as MS > WB > PI > LW), whereas other traits show quite discrete classes (e.g., HW or LWT weights). This result suggests that the alleles are at different frequencies between breeds or that different alleles exist. The latter can have important implications for a prospective fine mapping study, as the number of causal mutations to be found will vary from trait to trait. Pooling haplotypes from different breeds can be justified for some traits but not for others.
There are currently a number of projects involving large numbers of crosses between many founder lines with many traits recorded. In mice, the complex trait consortium is endorsing the Collaborative Cross, a panel of eight-way RI strains optimizing the contribution from each parental strain (Churchill et al., 2004). A similar, more ambitious initiative also is being developed in maize (Casstevens and Buckler, 2004). It is clear that classical two-breed methodologies will not be able to extract all available information from these experiments, and we need to resort to multibreed and multitrait methods such as the one presented here.
Certainly, joint analyses have some associated risks that are important to bear in mind. The most important one is that phenotypes recorded in different experiments do correspond to different traits. This risk is higher in plants than in animals, where management in intensive production systems tends to be rather constant across countries and farms. The second potential caveat is that alleles are fixed in most populations, causing a power reduction compared with analyzing experiments individually. Herein we have noted that this is likely the case for LWT, but this risk is not severe, as we have diagnostics to evaluate it (i.e., a model with common effects for several breeds can be tested vs. a model with different effects per breed.). Finally, a possibly problematic assumption in our method is that we assume a single allelic effect per breed across experiments, which may not be true if the QTL is segregating within breeds and/or the origin of the animals is very different, even when pertaining to the same breed. Again, however, we can use diagnostics to check these assumptions. We can treat each breed origin within experiment as different; although possibly at present, differences between breeds could not be estimable across experiments, diminishing the advantage of our approach. We also can allow for a variance within breeds, as we have done previously (Pérez-Enciso et al., 2002). Thus, the advantages of joint analyses clearly outweigh their potential risks. One potential problem is the presence of many founder breeds; in estimating an effect per breed, power will decrease because of the inflation in the number of parameters. This problem can be alleviated by treating each origin as random, or alternatively, a model with random regressors for the allelic effect terms can be employed (Verbeke and Molenberghs, 2000).
|
Implications
|
|---|
Our study confirms that swine chromosome X harbors loci with important effects on fatness and carcass composition. At least two distinct regions exist, one in the neighborhood of SW259/SW1994 markers, with an effect on ham weight and carcass length, and another one between markers SW2476 and SW1943, with primary effects on fatness and shoulder weight. The most interesting region lies between markers SW2476 and SW1943 (i.e., an interval of about 10 cM), which is quite a small region, considering that we are dealing mostly with F2 individuals and that mapping in the X chromosome is less precise than in autosomes. Compared with previous analyses involving large data sets and combined experiments, the statistical approach used here is more powerful in terms of modeling, allowing infinitesimal genetic effects, different models for each trait in multitrait analyses, and so on, and explicitly allowing for different allelic effects for each breed.
|
Footnotes
|
|---|
1 The following agencies funded these experiments: Spain, MCYT (AGF99-0284-CO2) and INIA (CPE03-010-03); France, European Union (Bridge and Biotech+ Programs), INRA (Department of Animal Genetics and AIP Structure des génomes animaux), and the Groupement de recherches et d’études sur les génomes; Germany, German Research Foundation (DFG, Mu616/6, Ge291/20), EC Programmes BRIDGE and Copernicus; and Czech Republic, Czech Science Foundation (523/04/0106). A. Mercadé was granted a Ph.D. fellowship from the Generalitat de Catalunya (Spain). We thank L. Silió and J. Estellé for comments and suggestions on the manuscript. Qxpak software, implementing the options described in this work, among others, is freely available at http://www.icrea.es/pag.asp?id=Miguel.Perez. 
2 Correspondence: Departament de Ciència Animal i dels Aliments, Facultat de Veterinària, Universitat Autònoma de Barcelona (phone: 34 93 581 4225; fax: 34 93 5812106; e-mail: miguel.perez{at}uab.es)
Received for publication March 30, 2005.
Accepted for publication June 22, 2005.
|
Literature Cited
|
|---|
Bidanel, J. P., D. Milan, N. Iannuccelli, Y. Amigues, M. Y. Boscher, F. Bourgeois, J. C. Caritez, J. Gruand, P. Le Roy, H. Lagant, R. Quintanilla, C. Renard, J. Gellin, L. Ollivier, and C. Chevalet. 2001. Detection of quantitative trait loci for growth and fatness in pigs. Genet. Sel. Evol. 33:289–309.[Medline]
Casstevens, T. M., and E. S. Buckler. 2004. GDPC: connecting researchers with multiple integrated data sources. Bioinformatics 20:2839–2840.[Abstract/Free Full Text]
Cepica, S., G. Reiner, H. Bartenschlager, G. Moser, and H. Geldermann. 2003a. Linkage and QTL mapping for Sus scrofa: Chromosome X. J. Anim. Breed. Genet. 120 (Suppl. 1):144–151.
Cepica, S., A. Stratil, M. Kopecny, P. Blazkova, J. Schroffel, R. Davoli, L. Fontanesi, G. Reiner, H. Bartenschlager, G. Moser, and H. Geldermann. 2003b. Linkage and QTL mapping for Sus scrofa chromosome 4. J. Anim. Breed. Genet. 120 (Suppl. 1):28–37.
Churchill, G. A., D. C. Airey, H. Allayee, J. M. Angel, A. D. Attie, J. Beatty, W. D. Beavis, J. K. Belknap, B. Bennett, W. Berrettini, A. Bleich, M. Bogue, K. W. Broman, K. J. Buck, E. Buckler, M. Burmeister, E. J. Chesler, J. M. Cheverud, S. Clapcote, M. N. Cook, R. D. Cox, J. C. Crabbe, W. E. Crusio, A. Darvasi, C. F. Deschepper, R. W. Doerge, C. R. Farber, J. Forejt, D. Gaile, S. J. Garlow, H. Geiger, H. Gershenfeld, T. Gordon, J. Gu, W. Gu, G. de Haan, N. L. Hayes, C. Heller, H. Himmelbauer, R. Hitzemann, K. Hunter, H. C. Hsu, F. A. Iraqi, B. Ivandic, H. J. Jacob, R. C. Jansen, K. J. Jepsen, D. K. Johnson, T. E. Johnson, G. Kempermann, C. Kendziorski, M. Kotb, R. F. Kooy, B. Llamas, F. Lammert, J. M. Lassalle, P. R. Lowenstein, L. Lu, A. Lusis, K. F. Manly, R. Marcucio, D. Matthews, J. F. Medrano, D. R. Miller, G. Mittleman, B. A. Mock, J. S. Mogil, X. Montagutelli, G. Morahan, D. G. Morris, R. Mott, J. H. Nadeau, H. Nagase, R. S. Nowakowski, B. F. O’Hara, A. V. Osadchuk, G. P. Page, B. Paigen, K. Paigen, A. A. Palmer, H. J. Pan, L. Peltonen-Palotie, J. Peirce, D. Pomp, M. Pravenec, D. R. Prows, Z. Qi, R. H. Reeves, J. Roder, G. D. Rosen, E. E. Schadt, L. C. Schalkwyk, Z. Seltzer, K. Shimomura, S. Shou, M. J. Sillanpaa, L. D. Siracusa, H. W. Snoeck, J. L. Spearow, K. Svenson, L. M. Tarantino, D. Threadgill, L. A. Toth, W. Valdar, F. P. de Villena, C. Warden, S. Whatley, R. W. Williams, T. Wiltshire, N. Yi, D. Zhang, M. Zhang, and F. Zou. 2004. The Collaborative Cross, a community resource for the genetic analysis of complex traits. Nat. Genet. 36:1133–1137.[Medline]
Gaboreanu, A. M., L. Grapes, A. M. Ramos, J. J. Kim, and M. F. Rothschild. 2004. Characterization of an X-chromosome PCR-RFLP marker associated with fat deposition and growth in the pig. Anim. Genet. 35:401–403.[Medline]
Geldermann, H., E. Müller, G. Moser, G. Reiner, H. Bartenschlager, S. Cepica, A. Stratil, J. Kuryl, C. Moran, R. Davoli, and C. Brunsch. 2003. Genome-wide linkage and QTL mapping in porcine F2 families generated from Pietrain, Meishan and wild boar crosses. J. Anim. Breed. Genet. 120:363–393.
Haley, C. S., S. A. Knott, and J. M. Elsen. 1994. Mapping quantitative trait loci in crosses between outbred lines using least squares. Genetics 136:1195–1207.[Abstract]
Harlizius, B., A. P. Rattink, D. J. de Koning, M. Faivre, R. G. Joosten, J. A. van Arendonk, and M. A. Groenen. 2000. The X chromosome harbors quantitative trait loci for backfat thickness and intramuscular fat content in pigs. Mamm. Genome 11:800–802.[Medline]
Jonsson, P. 1975. Methods of pig improvement through breeding in the European countries: A review. Livest. Prod. Sci. 2:1–28.
Kim, J.-J., M. F. Rothschild, J. Beever, S. Rodriguez-Zas, and J. C. M. Dekkers. 2005. Joint analysis of two breed cross populations in pigs to improve detection and characterization of quantitative trait loci. J. Anim. Sci. 83:1229–1240.[Abstract/Free Full Text]
Knott, S. A., and C. S. Haley. 2000. Multitrait least squares for quantitative trait loci detection. Genetics 156:899–911.[Abstract/Free Full Text]
Knott, S. A., L. Marklund, C. S. Haley, K. Andersson, W. Davies, H. Ellegren, M. Fredholm, I. Hansson, B. Hoyheim, K. Lundstrom, M. Moller, and L. Andersson. 1998. Multiple marker mapping of quantitative trait loci in a cross between outbred wild boar and Large White pigs. Genetics 149:1069–1080.[Abstract/Free Full Text]
Legault, C., P. Sellier, J. C. Caritez, P. Dando, and J. Gruand. 1985. Expérimentation sur le porc chinois en France. II-Performances de production en croisement avec les races européennes. Genet. Sel. Evol. 17:133–152.
Malek, M., J. C. Dekkers, H. K. Lee, T. J. Baas, and M. F. Rothschild. 2001. A molecular genome scan analysis to identify chromosomal regions influencing economic traits in the pig. I. Growth and body composition. Mamm. Genome 12:630–636.[Medline]
Mercadé, A., J. Estelle, J. L. Noguera, J. M. Folch, L. Varona, L. Silió, A. Sánchez, and M. Pérez-Enciso. 2005. On growth, fatness and form: A further look at porcine chromosome 4 in an Iberian x Landrace cross. Mamm. Genome 16:374–382.[Medline]
Milan, D., J. P. Bidanel, N. Iannuccelli, J. Riquet, Y. Amigues, J. Gruand, P. Le Roy, C. Renard, and C. Chevalet. 2002. Detection of quantitative trait loci for carcass composition traits in pigs. Genet. Sel. Evol. 34:705–728.[Medline]
Misztal, I., S. Tsuruta, T. Strabel, B. Auvray, T. Druet, and D. H. Lee. 2002. BLUPF90 and related programs (BGF90). Proc. 7th World Cong. Genet. Appl. Livest. Prod., CD-ROM Communication No. 28–07.
Nezer, C., L. Moreau, D. Wagenaar, and M. Georges. 2002. Results of a whole genome scan targeting QTL for growth and carcass traits in a Pietrain x Large White intercross. Genet. Sel. Evol. 34:371–387.[Medline]
Pérez-Enciso, M., A. Clop, J. M. Folch, A. Sanchez, M. A. Oliver, C. Ovilo, C. Barragan, L. Varona, and J. L. Noguera. 2002. Exploring alternative models for sex-linked quantitative trait loci in outbred populations: Application to an Iberian x Landrace pig intercross. Genetics 161:1625–1632.[Abstract/Free Full Text]
Pérez-Enciso, M., A. Clop, J. L. Noguera, C. Ovilo, A. Coll, J. M. Folch, D. Babot, J. Estany, M. A. Oliver, I. Diaz, and A. Sanchez. 2000. A QTL on pig chromosome 4 affects fatty acid metabolism: Evidence from an Iberian by Landrace intercross. J. Anim. Sci. 78:2525–2531.[Abstract/Free Full Text]
Pérez-Enciso, M., and I. Misztal. 2004. Qxpak: A versatile mixed model application for genetical genomics and QTL analyses. Bioinformatics 20:2792–2798.[Abstract/Free Full Text]
Porter, V. 1993. Pigs: A Handbook to the Breeds of the World. Helm Information Ltd., Mountfield, East Sussex, UK.
Rohrer, G. A., and J. W. Keele. 1998. Identification of quantitative trait loci affecting carcass composition in swine: I. Fat deposition traits. J. Anim. Sci. 76:2247–2254.[Abstract/Free Full Text]
Verbeke, G., and G. Molenberghs. 2000. Linear Mixed Models for Longitudinal Data. Springer Verlag, New York, NY.
Walling, G. A., P. M. Visscher, L. Andersson, M. F. Rothschild, L. Wang, G. Moser, M. A. Groenen, J. P. Bidanel, S. Cepica, A. L. Archibald, H. Geldermann, D. J. de Koning, D. Milan, and C. S. Haley. 2000. Combined analyses of data from quantitative trait loci mapping studies. Chromosome 4 effects on porcine growth and fatness. Genetics 155:1369–1378.[Abstract/Free Full Text]
Top
Abstract
Introduction
