Association of Twelve Immune-Related Genes with Performance of ...

Association of Twelve Immune-Related Genes with Performance of Three Broiler Lines in Two Different Hygiene Environments X. Ye,* S. Avendano,† J. C. M. Dekkers,* and S. J. Lamont*1 *Department of Animal Science and Center for Integrated Animal Genomics, Iowa State University, Ames 50011; and †Aviagen Limited, Newbridge, Midlothian, EH28 8SZ, Scotland, UK ABSTRACT Elite populations of farm animals under genetic selection are often maintained in high hygiene conditions, yet the commercial populations may be raised in, and are expected to perform well in, environments of varied hygiene levels. This presents special challenges to genetically improve those traits for which genotype by environment interactions are important. Twelve immunerelated genes were studied for associations with general mortality and other performance traits in 3 elite commercial broiler chicken lines raised in high and low hygiene environments. The genes were toll-like receptor 4, MD-2 (accessory protein of TLR4), interferon-γ, transforming growth factor-β 3, inducible nitric oxide synthase, macrophage migration inhibitory factor, interleukin-2, caspase1, inhibitor of apoptosis protein-1, tumor necrosis factorrelated apoptosis-inducing ligand, chicken B-cell marker, and bone morphogenetic protein-7. From a total of 56 identified single-nucleotide polymorphisms (SNP) in 12 genes, 14 SNP that had moderate allelic frequencies in at least 2 of the 3 lines were typed in about 100 progenytested sires from each of 3 elite commercial broiler chicken lines using restriction fragment length polymorphism techniques and then used in association analysis. The traits measured on the progeny (total progeny = 145,467)

were: mortality from hatching to 14 d and from 14 to 40 d of age, BW at 7 and 40 d of age, feed conversion, ultrasound breast depth, percentage of breast, eviscerated carcass weight, twisted legs or evident tibial dyschondroplasia, x-ray-inspection–based subclinical or incipient development of tibial dyschondroplasia, curly or crooked toes or bowed legs, oxygen content of blood, and female’s antibody titer to infectious bursal disease virus at 27 wk. Association analyses were conducted with allele and haplotype substitution effect models using progeny mean data adjusted for fixed and mate effects as sire trait records. Ten of the 12 genes had SNP associations with at least 1 trait. Most detected effects were with mortality and growth traits. Most gene–SNP trait associations varied by genetic line or with environment. These results indicate that associations of candidate genes with important broiler traits can be identified in multiple environments, and they offer a potential for the implementation of marker-assisted selection for traits expressed in the environment in which the commercial broiler needs to perform. The effects of these immune-related candidate genes, however, are complex and affected by genetic background and environment.

Key words: immune-related gene, single nucleotide polymorphism, environment, trait, broiler chicken 2006 Poultry Science 85:1555–1569

INTRODUCTION The phenotype of an individual represents the complex sum of the effects of genotype and environment. Many studies on interactions of genetic factors and environments (G × E) have been reported for chickens (Sheridan, 1990; Leenstra and Cahaner, 1991; Settar et al., 1999; Ali et al., 2001; Deeb and Cahaner, 2001a,b, 2002; Tixier-Boichard, 2002; Mathur, 2003; Fulton, 2004). Chickens raised in intensive vs. extensive rearing systems show large differences in live BW, feed intake, and carcass weight (Ali

©2006 Poultry Science Association Inc. Received January 23, 2006. Accepted March 13, 2006. 1 Corresponding author: [email protected]

et al., 2001). Settar et al. (1999) found highly significant interaction effects between seasons (spring and summer) and genotypes for BW in broilers, and the correlation of BW between spring broilers and summer broilers was low, although the broilers were from the same stock. Such effects of genotype and environment complicate the detection of genes or DNA markers that are associated with traits of interest, especially for genes with small effects or that are strongly affected by environment (i.e., traits with lower heritability and G × E). Genetic selection generally takes place in high-hygiene environments (HH), but commercial production may be in contrasting hygiene environments. Differences between selection environments and production environments can affect phenotypic expression of traits. Kolmodin and Bijma (2004) showed that the optimum selection environment for a trait depends not only on the

1555

1556

YE ET AL.

environment in which selection response is realized, but also on the degree of G × E, the correlation among the population average level in different environments, sensitivity to environmental change, and heritability of the trait. Mulder and Bijma (2005) showed that genetic gain in commercial conditions is reduced due to the G × E interaction between the selection environment and the production environment. Leenstra and Cahaner (1991) showed that when broilers were raised in different temperatures, broilers derived from Israeli strains (selected in hot environments) had higher weight gain and protein deposition at moderate to warm temperatures than those derived from Dutch strains (selected in temperate environments), whereas the Dutch chicks performed better at lower temperatures. Therefore, identification of gene-trait associations in different environments, including environments similar to that of the commercial production setting, is very important. Previous studies have reported associations of immune-related genes with immune response, bacterial burden, and growth performance in experimental crosses and commercial chicken lines (Kramer et al., 2003; Lamont et al., 2002; Liu et al., 2003; Malek and Lamont, 2003; Malek et al., 2004; Zhou et al., 2001; Zhou and Lamont, 2003a,b). However, immune-related genes may have pleiotropic effects, and their expression is affected by environment and genetic background. Immune responsiveness is hypothesized to be important in maintaining performance under challenging environmental conditions. In this context, the objective of the current experiment was to study associations of single-nucleotide polymorphisms (SNP) in 12 immunity-related genes with growth, mortality, yield, and support traits in 3 elite commercial broiler chicken lines raised in HH and low-hygiene (LH) environments.

MATERIALS AND METHODS Chickens and Traits Performance data on progeny from about 100 sires from each of 3 elite commercial broiler chicken lines (designated lines X, Y, and Z) were recorded. Each sire’s progeny consisted of 3 independent progeny groups, of which 1 was raised in HH and 2 in 2 LH conditions (LH1 and LH2). Traits recorded for all progeny groups were BW at 7 and 40 d of age (BW7 and BW40) and mortality prior to 14 d of age (EMORT) and between 14 and 42 d of age (LMORT). An additional 9 traits were recorded in the HH environment at different ages (Table 1), including breast depth measured by ultrasound (US), breast percentage of BW (BR), feed conversion ratio (FCR), eviscerated carcass weight (EV), 3 leg-defect related traits [twisted legs or evident tibial dyschondroplasia (LEG), x-ray inspection-based subclinical or incipient development of tibial dyschondroplasia detected by Lixiscope examination (LIXI), and curly or crooked toes or bowed legs (TOBO)]. Two further traits related to cardiovascular function and antibody response, the oxygen content of

blood measured by oximeter (OXI) and the antibody titer of hens to infectious bursal disease virus at 27 wk (IBD) were also recorded. The traits with numbers of sires and mean numbers of progeny per sire for each trait are given in Table 1.

The Studied Genes, Primer Design, and Identification of SNP Twelve immune-related genes were studied. They were selected to represent a variety of immune mechanisms and to be genes for which sequence variation in the chicken had been reported to be evaluated by PCR-RFLP. These genes were interferon-γ (IFN-γ; Zhou et al., 2001), inhibitor of apoptosis protein-1 (IAP1; Lamont et al., 2002; Liu and Lamont, 2003; Zhou and Lamont, 2003b), and chicken B-cell marker (CHB6; Zhou and Lamont, 2003b) on chromosome 1; MD-2 (accessory protein of the tolllike receptor 4; Malek et al., 2004) on chromosome 2; interleukin-2 (IL-2; Kramer et al., 2003) and tumor necrosis factor-related apoptosis-inducing ligand (TRAIL; Malek and Lamont, 2003) on chromosome 4; transforming growth factor-β3 (TGF-β3; Zhou and Lamont, 2003a) on chromosome 5; macrophage migration inhibitory factor (MIF; Malek et al., 2004) on chromosome 15; toll-like receptor 4 (TLR4; Malek et al., 2004) on chromosome 17; inducible nitric oxide synthase (iNOS; Kramer et al., 2003; Malek and Lamont, 2003) and caspase-1 (CASP1; Liu and Lamont, 2003; Zhou and Lamont, 2003b) on chromosome 19; and bone morphogenetic protein-7 (BMP7) on chromosome 20. Primer Design and Identification of SNP. The SNP were identified for each studied gene by designing primers to amplify specific regions, performing PCR, sequencing the PCR products to verify their identity, and aligning the PCR product sequences. Primers were designed based on the known DNA and cDNA sequences of the genes using Primer3 public software (http://frodo.wi.mit.edu/ cgi-bin/primer3/primer3_www.cgi). The PCR products were sequenced at the Iowa State University DNA Sequence and Synthesis Facility using the ABI model 377 (Applied Biosystems, Foster City, CA). The PCR product sequences were aligned, and SNP were identified using the software Sequencher 4.2 (demo version, Gene Codes Corporation, Ann Arbor, MI). Most of the primers used in this study were designed and used in previous research (Kramer et al., 2003; Lamont et al., 2002; Liu et al., 2003; Malek and Lamont, 2003; Malek et al., 2004; Zhou et al., 2001; Zhou and Lamont, 2003a,b). New primers were designed for genes BMP7, TLR4, TGF-β3, and iNOS (Table 2). The PCR conditions for the primers were basically the same as those used in previous studies, except the annealing temperature was 62 to 65°C for CASP1, 56°C for IL-2, and 53 to 56°C for MIF, and the magnesium concentration was 1.825 mM for BMP7, IL-2, iNOS, MD-2, MIF, and TLR4, and 2.25 mM for TGF-β3.

1557

IMMUNE-RELATED GENES AND BROILER PERFORMANCE Table 1. Trait abbreviations and definitions, numbers of sires genotyped, and progeny evaluated

Progeny group1

Trait Abbreviation

Mean number of progeny per sire

Number of sires per line

Description

X

Y

Z

X

Y

Z

LH1 BW7-LH1 BW40-LH1 EMORT-LH1 LMORT-LH1

BW at 7 d (g) BW at 40 d (decag) Mortality prior to 14 d (%) Mortality 14 to 42 d (%)

82 82 82 82

69 69 69 69

92 92 92 92

94 94 116 116

67 68 92 92

41 41 50 50

BW7-LH2 BW40-LH2 EMORT-LH2 LMORT-LH2

BW at 7 d (g) BW at 40 d (decag) Mortality prior to 14 d (%) Mortality 14 to 42 d (%)

102 102 102 102

99 99 99 99

113 113 113 100

93 93 105 105

65 65 77 77

66 70 78 21

BW7-HH BW40-HH EMORT-HH LMORT-HH US-HH BR-HH EV-HH FCR-HH IBD-HH LEG -HH LIXI-HH

BW at 7 d (g), BW at 40 d (decag) Mortality prior to 14 d (%) Mortality 14 to 42 d (%), Ultrasound breast depth at 5 wk (mm) Breast yield at 6 wk (%) Eviscerated yield at 6 wk (%) Feed conversion ratio between 6 and 8 wk Hen antibody titer to infectious bursal disease at 27 wk Twisted legs or evident tibial dyschondroplasia (TD) at 5 wk (%) X-ray-inspection-based subclinical or incipient development of TD (Lixiscope) at 5 wk (%) Curly or crooked toes, bowed legs, or both at 5 wk (%) Oximeter estimate of blood oxygen at 5 wk (%)

101 101 101 101 —2 92 92 92 70 101 93

100 100 100 100 99 92 92 94 — 100 96

113 113 113 113 — 110 110 110 74 113 102

379 383 437 437 — 45 35 28 19 383 31

309 310 335 335 88 41 40 29 — 310 31

210 211 248 248 — 55 56 61 17 211 26

101 91

100 —

113 114

383 28

310 —

211 118

LH2

HH

TOBO-HH OXI-HH 1

LH = low hygiene; HH = high hygiene. Trait not measured in this line.

2

SNP Genotyping

Statistical Analysis

Fourteen SNP with moderate allelic frequencies in at least 2 of the 3 elite commercial broiler chicken lines were chosen for individual genotyping and association analysis (Table 3). Individual sires were genotyped for those SNP using PCR-RFLP. Restriction sites caused by the SNP were identified using Webcutter 2.0 (http://rna.lund berg.gu.se/cutter2/index.html). Digestion reactions for the PCR products were performed for 4 h or overnight in 10 ␮L of reaction solution, including 5 ␮L of PCR product solution and 2 or 3 units of restriction enzymes. Other conditions for restriction enzyme digestion (e.g., temperature and buffer solution) followed manufacturer’s recommendations (New England BioLabs Inc. Ipswich, MA).

Associations of the gene SNP with the traits were analyzed using progeny means of sires, which were derived from progeny records that were each adjusted for nongenetic environmental effects and for mate effects, using solutions obtained from routine animal model genetic evaluation procedures that are used within these lines. An adjusted progeny record was equal to its phenotypic record minus estimates for the effects of sex, age, hatch, mating group, and half the estimated breeding value of the dam. The average of adjusted progeny records for each sire was used in further analyses. Association analyses were conducted using weighted least squares of SAS PROC GLM (SAS Inst. Inc., 2004), with weights equal to the number of progeny included

Table 2. Primer design for polymorphism identification of bone morphogenetic protein-7 (BMP7), toll-like receptor 4 (TLR4), transforming growth factor-β3 (TGF-β3), and inducible nitric oxide synthase (iNOS) genes Gene

GenBank accession #

BMP7

AF223970

TLR4

AY064697

TGF-β3

X60091

iNOS

AF537190

1

Annealing temperature.

Primer

Sequence

Forward Reverse Forward Reverse Forward Reverse Forward Reverse

5′-CGGGAATGTCTGGAAGAAGAAA-3′ 5′-AGAGGGGGGAATGCTGAAAT-3′ 5′-GTGCTGCTCAGTGGGTGTAA-3′ 5′-GGAGGAAGGCAATCATCAAA-3′ 5′-CTGGAAAAACAAATAGGTCTTCCTT-3′ 5′-GAAGCAGTAGTTGGTATCCAG-3′ 5′-CCAATAAAAGTAGAAGCGA-3′ 5′-GAGAGCAAGGATGTTTGAAAAGTAA-3′

Product size (bp)

Tm1 (°C)

1,216

56

676 ∼1,000 400

61–65 56 50.5

1558

YE ET AL. Table 3. Allele frequencies of single-nucleotide polymorphisms (SNP) for the candidate genes

Gene1

SNP

Restriction enzyme

MD-2 iNOS MIF

G/A T/C A/C G/A A/C C/A G/A T/C C/T A/G T/G A/G C/A T/C

Ase I Alu I Hinf I Bcg I Mnl I Xba I Tsp509 I Bcg I BstX I Fok I BstN I Cac8 I Bsr I Msp I

IL-2 TLR4 IFNγ CASP1 IAP1 TRAIL CHB6 BMP7 TGF-β3

Line X (n = 102)

Line Y (n = 103)

Line Z (n = 114)

Freq2

Freq

Freq

0.43 0.64** 0.93 0.37 0.18 0.13 0.29 0.48 0.99** 0.66 0.47 0.74 0.03 0.29

0.81 0.54 0.33** 0.52 0.65** 0.71 0.29* 0.68 0.71 0.49** 0.84 0.98* 0.35 0.51

0.96 0.27 0.28** 0.60** 0.08 0.33 0.46** 0.62 0.79 0.29** 0.18 0.76 0.13 0.52

1 MD-2 = accessory protein of toll-like receptor 4 (TLR4); iNOS = inducible nitric oxide synthase; MIF = macrophage migration inhibitory factor; IL-2 = interleukin 2; IFN-γ = interferon-γ; CASP1 = caspase-1; IAP1= inhibitor of apoptosis protein-1; TRAIL = tumor necrosis-factor-related apoptosis-inducing ligand; CHB6 = chicken B-cell marker; BMP7 = bone morphogenetic protein 7; TGF-β3 = transforming growth factor-β 3. 2 Frequency of allele 1, which is the allele before the “/” of an SNP (e.g., for SNP “G/A,” allele 1 = G). *P < 0.05 and **P < 0.01 for significance of a χ2 test for Hardy-Weinberg equilibrium.

in the mean to account for differences in variance of residuals. Only additive associations were evaluated because progeny means primarily reflect the additive effects of genes, being related to one-half of the sire’s breeding value. Association analyses were performed separately for single SNP and for haplotypes of SNP within a gene.

Analysis of Single SNP Three allele substitution effect models were used to determine associations between an SNP and a trait. Model 1 was for within-line data analysis, and models 2 and 3 were for analyses across lines. In each model, the effect of allele 1 was estimated relative to allele 2 for each SNP. Allele 1 was defined as the allele with the restriction site model 1: yi = ␮ + b fi + εi (fitted for each line separately), model 2: yij = Lj + bj fji + εji (fitted across lines, with separate effects per line), model 3: yij = Lj + bfji + εji (fitted across lines with a common effect), where yij = the adjusted progeny mean of the ith sire of line j; ␮ = a general mean; Lj = effect of the jth line (j = 1, 2, 3); fji = the number copies of allele 1 of the SNP in the ith sire; bj = the substitution effect for line j; and εji = 1 2 the residual for sire i with variance = σ (Ni = the Ni e number of progeny for sire i).

Analysis of Haplotypes Two genes (MIF and TGF-β3) had 2 SNP each. Because SNP within a gene may be in linkage disequilibrium,

the haplotypes formed by these SNP were also used for analysis. For these 2 genes, haplotype frequencies were estimated by maximum likelihood using the software Arlequin (version 2.000; Schneider et al., 2000), and the presence of linkage disequilibrium between the SNP was quantified by r2 (Hill and Robertson, 1968) and tested using a χ2 test. Haplotype frequencies were then used to assign haplotype probabilities for sires whose haplotypes could not be inferred with certainty. Models to analyze associations of haplotypes with the traits were the same as described above, but replacing SNP allele effects by n−1

∑ bjh fijh

h=1

where fijh = the number of copies of haplotypes h in sire i (or the sum of the probabilities that the first and second haplotype of the sire are h) and bjh = the substitution effect for haplotype h. In these models, the most frequent haplotype (n) was set equal to zero, such that the models are of full rank. As a result, effects bjh represent the effect of substituting a copy of haplotype n by a copy of haplotype h. In all analyses, residuals were assumed uncorrelated. Genetic relationships among sires were limited and were ignored in the analyses; the 102, 103, and 114 sires that were analyzed for lines X, Y, and Z originated from a total of 51, 50, and 66 sires and 71, 75, and 93 dams, respectively. Thus, half- and full-sib relationships among the evaluated sires were small, and ignoring them was not expected to bias results. In addition, an outlier analysis was performed. A total of 21 possible outlier data points were identified in the 14 traits by visual inspection of distributions of progeny means against numbers of progeny. Outliers generally represented instances of small progeny groups. Results

IMMUNE-RELATED GENES AND BROILER PERFORMANCE

1559

Table 4. Maximum likelihood estimates of frequencies of haplotypes of single nucleotide polymorphisms in macrophage migration inhibitory factor (MIF) and transforming growth factor-β 3 (TGF-β3) genes Frequency Gene MIF

1

TGF-β32

Haplotype

Line X

Line Y

Line Z

11 12 21 22 P(LD)3 11 12 21 22 P(LD)3

0.33 0.60 0.04 0.03 0.01 0.01 0.02 0.28 0.69 0.00

0.07 0.26 0.45 0.22 0.19** 0.01 0.34 0.51 0.14 0.53**

0.10 0.18 0.50 0.23 0.08** 0.01 0.12 0.51 0.36 0.12**

1 Haplotypes were constructed as “Hinf I + Bcg I.” For Hinf I, allele 1 is “A” and allele 2 is “C” For Bcg I, allele 1 is “G” and allele 2 is “A.” 2 Haplotypes were constructed as “Bsr I + Msp I.” For Bsr I, allele 1 is “C” and allele 2 is “A.” For Msp I (T/C), allele 1 is “T” and allele 2 is “C.” 3 P = P-value of χ2 test for linkage equilibrium with **P < 0.01. LD = amount of linkage disequilibrium, measured as r2.

from analyses with all data were compared with those with removal of the outliers. In general, the identified outliers had minor effects on SNP trait association results. The results reported in this paper are, therefore, based on all data.

Figure 1. Cumulative frequency distribution (black dots) of P-values for tests of single-nucleotide polymorphisms associations with progeny means by line. Total number of tests was 993. The 45° angle line represents the expected cumulative distribution if all tests followed the null hypothesis of no association.

RESULTS Significance Level for Claiming Trait Associations Significance of SNP trait associations for a given line was tested by an F-test of the model sum of squares for model 1 against a model without SNP or haplotype effects. To test whether effects were consistent across lines, an F-test of model sums of squares of model 2 against model 3 was used. To account for the large number of tests conducted, (994 tests in single-SNP trait association analyses), methods developed by Mosig et al. (2001) and Fernando et al. (2004) were used to control the rate of false positive results across all tests. The number of positive tests that is required for these methods was estimated following the procedure described by Nettleton and Hwang (2003) by comparing the frequency distribution of observed P-values of individual tests to the expected uniform distribution of P-values for tests under the null hypothesis of no association. For SNP trait associations in the complex experimental design that was used here, including repeated measures or measures on related groups of phenotypes, besides the P-values for the individual trait associations, the consistency of effects across genetic or environmental groups and across related traits can give important insights into the biological impact of the genes. Thus, results are presented and discussed for associations that showed consistency across genetic lines, across environments, or across traits, although some of the individual P-values may not be significant.

Identified SNP and Allele/Haplotype Frequencies Fourteen SNP with moderate frequencies of alleles in at least 2 lines were examined for associations with traits. These 14 SNP, their corresponding restriction enzymes, frequencies of allele 1, and P-values for the Hardy-Weinberg equilibrium test are presented in Table 3. Six of the 14 SNP were in Hardy-Weinberg equilibrium in all 3 lines and the remaining 8 SNP (iNOS-Alu I, MIF-Hinf I, MIFBcg I, IL-2-Mnl I, IFN-γ-Tsp509 I, IAP1-BstX I, TRAIL-Fok I, and BMP7-Cac8 I) were in Hardy-Weinberg disequilibrium (P < 0.05) in 1 or 2 lines, for a total of 11 out of 42 SNP by line combinations. For those cases, heterozygotes were more frequent than expected for iNOS-Alu I, MIFBcg I, IL-2-Mnl I, and IFN-γ-Tsp509 I and less frequent than expected for MIF-Hinf I and TRAIL-Fok I. For MIF and TGF-β3, which had 2 SNP each, haplotype frequencies, amount of linkage disequilibrium, and P-values for tests of linkage disequilibrium are given in Table 4. The 2 SNP in TGF-β3 were in significant linkage disequilibrium in all 3 lines (P < 0.001). The 2 SNP in MIF were in significant linkage disequilibrium only in line X (P < 0.001).

General Significance of SNP Trait Associations The cumulative frequency distribution of P-values of association tests between SNP and traits by line is shown

1560

Figure 2 (Continued).

YE ET AL.



1561

1562


YE ET AL.


1563

Figure 2 (Continued). Estimates of effect of allele 1 relative to allele 2 on traits in 3 broiler lines (X, Y, and Z). Early = mortality prior to 14 d of age; Late = mortality from 14 to 42 d of age; −log (P-value) = the corresponding P-value for single-nucleotide polymorphism (SNP) trait association; bullet point = −log (P-value); bars = estimate of allele substitution effect expressed in additive genetic SD units and with positive estimates reflecting desirable effects on the trait; SE bars are also expressed in additive genetic SD units; horizontal line = the proportion of falsepositive associations (PFP) of 50% corresponding a P-value of 0.031 [−log (P-value = 1.5)]. In addition, the PFP of 20% corresponds a P-value of 0.0029 [−log (P-value = 2.54)]. MD-2 = accessory protein of toll-like receptor 4; BW7 = BW at 7 d; BW40 = BW at 40 d; FCR = feed conversion ratio; US = breast depth measured by ultrasound; BR = breast percentage of BW; EV = eviscerated carcass weight; IBD = antibody titer of hens to infectious bursal disease virus at 27 wk; LEG = twisted legs or evident tibial dyschondroplasia; LIXI = x-ray inspection-based subclinical or incipient development of tibial dyschondroplasia detected by Lixiscope examination; TOBO = curly or crooked toes or bowed legs; OXI = the oxygen content of blood measured by oximeter; iNOS = inducible nitric oxide synthase; MIF = macrophage migration inhibitory factor; IL-2 = interleukin 2; TLR4 = toll-like receptor-4; IFN-γ = interferon-γ; CASP1 = caspase-1; TGF-β3 = transforming growth factor-β 3; IAP1= inhibitor of apoptosis protein-1; TRAIL = tumor necrosis-factor-related apoptosis-inducing ligand; CHB6 = chicken B-cell marker; BMP7 = bone morphogenetic protein 7.

in Figures 1 and 2. If the null hypothesis of no association was true for all SNP trait associations, P-values of significance tests would be expected to follow a uniform distribution between 0 and 1, 5% of tests conducted would have a P-value less than 0.05, and the cumulative frequency distribution of P-values would follow the 45° straight line shown in Figure 1. The actual distribution of P-values showed an excess of low P-values, demonstrating that true associations were present in these data. The estimated number of tests that were true among the 994 tests conducted was 95. Using this number of true tests, Pvalues that met different levels of false discovery rates were calculated. Based on this, on average, 20% of tests with a comparison-wise P-value less than 0.0029 were expected to be false positives, and 50% of tests with a comparison-wise P-value less than 0.031 were expected to be false positives. The significance level used to claim a significant association was set to be at P < 0.031.

Trait Associations of SNP Results of associations of single SNP with traits, by line and environment, are summarized in Figure 2. Each graph in Figure 2 displays estimates of the effects of allele 1 for all traits for each SNP in the lines, regardless of the effects being significant, such that allele effects for a trait can be compared across environments, across lines, and with those for related traits. To facilitate comparisons across traits, estimates are given in units of genetic standard deviation, which were estimated using routine genetic evaluation procedures, by dividing estimates by 0.5 of the genetic standard deviation, noting that differences in progeny means represent half of the difference in breeding values of the sires. In addition, the direction of allele effect estimates was set such that the positive direction (bars above the horizontal axis) represents a desirable effect for all traits. Thus, for traits such as FCR,

1564

YE ET AL. Table 5. Estimates of allele effects across lines for single nucleotide polymorphisms (SNP) that were significant using model 3 and that did not show a significant (P > 0.05) interaction among lines Estimate2 Gene SNP

Trait1

Mean

SE

P-value

iNOS-Alu I MIF-Hinf I

BW7-LH2 BW40-HH LEG-HH BW40-HH TOBO-HH BW7-HH BW40-HH FCR-HH TOBO-HH OXI-HH FCR-HH BW40-LH1 BW40-LH2 LMORT-LH2 BW7-LH2 BW40-LH1 BW40-LH2 TOBO-HH LMORT-LH2 LMORT-HH BW40-LH1 LMORT-LH2

−0.73 1.01 −0.82 −0.93 −0.16 −1.13 −1.12 0.06 −0.2 0.76 0.09 1.51 1.36 −0.84 1.06 −1.6 −1.31 0.23 −0.64 −0.54 −1.75 0.46

0.36 0.39 0.3 0.42 0.08 0.31 0.47 0.03 0.1 0.33 0.02 0.69 0.67 0.21 0.51 0.72 0.65 0.09 0.3 0.26 0.73 0.23

0.04 0.01 0.01 0.03 0.044