Quantitative trait locus affecting birth weight on bovine chromosome 5 in a F2 Gyr x Holstein population
http://www.100md.com
《遗传学和分子生物学》
IUniversidade Federal de So Carlos, Programa de Pos-Graduao em Genetica e Evoluo, So Carlos, SP, Brazil
IIUniversidade de So Paulo, Departamento de Produo Animal, Piracicaba, SP, Brazil
IIIEmbrapa Gado de Leite, Juiz de Fora, MG, Brazil
IVEmbrapa Pecuaria Sudeste, So Carlos, SP, Brazil
VCNPq fellowship
ABSTRACT
Segregation between a genetic marker and a locus influencing a quantitative trait in a well delineated population is the basis for success in mapping quantitative trait loci (QTL). To detect bovine chromosome 5 (BTA5) birth weight QTL we genotyped 294 F2 Gyr (Bos indicus) x Holstein (Bos taurus) crossbreed cattle for five microsatellite markers. A linkage map was constructed for the markers and an interval analysis for the presence of QTL was performed. The linkage map indicated differences in the order of two markers relative to the reference map (http://www.marc.usda.gov). Interval analysis detected a QTL controlling birth weight (p < 0.01) at 69 centimorgans (cM) from the most centromeric marker with an effect of 0.32 phenotypic standard-error. These results support other studies with crossbred Bos taurus x Bos indicus populations.
Key words: BTA5, birth weight, cattle, QTL, microsatellite markers.
Introduction
The use of crossbred cattle has been an alternative for the intensification of bovine milk production, making possible the exploration of the genetic differences between breeds and the benefits of heterosis. The need for bovine genetic resources adapted to tropical conditions has lead to the use of crosses between Holstein (Bos taurus) cattle, selected for milk production for more than 2.000 years (Friend and Bishop, 1978), and Zebu (Bos indicus) cattle, adapted to tropical conditions (Madalena et al. 1990). Most economically important traits in dairy cattle are quantitative traits, which have been under selection for several generations, with quite favorable results.
The development of saturated genetic maps (Barendse et al., 1997; Kappes et al., 1997; Ihara et al., 2004) has allowed the identification of quantitative trait loci (QTL) affecting economically important traits. MacNeil and Grosz (2002) have stated that the identification of QTL has the potential to significantly increase the genetic improvement rate through the implementation of marker-assisted selection (MAS). For traits difficult or expensive to determine, with low heritability, late expression or measured only after slaughter, MAS can substantially increase the rate of response compared to selection based exclusively on estimates of performance values (Davis and DeNise, 1998). The use of MAS also allows the opportunity for more efficient breakage of antagonistic genetic correlations between characters (Grosz and MacNeil, 2001). For instance, estimates of the genetic correlation between direct effects on birth weight and yearling weight are approximately 0.5 in all bovine breeds (Koots et al., 1994) and as result selection based on mature weight or growth rate to mature weight can significantly increase birth weight potentially increasing the incidence and severity of calving difficulties. The opposite is also true, selection for smaller birth weight will reduce adult weight (Grosz and MacNeil, 2001). Identification of genomic regions affecting birth weight with no effect on mature weight would be a powerful tool for the manipulation of pre and post birth growth rate.
Quantitative trait loci have been detected in experimental and commercial bovine, swine and ovine populations (Davis and DeNise, 1998). There are several described bovine QTL affecting many characteristics. Davis et al. (1998) detected and mapped five QTL for birth weight on chromosomes 5, 6, 14, 18 and 21 in three families of paternal half-sib Charolais x Brahman (B. taurus x B. indicus) and Elo et al. (1999) detected a QTL for adult live weight on chromosome 23. Stone et al. (1999) reported significant evidence for the presence of QTL on chromosomes 1, 2, 5 and 13 and suggestive evidence for QTL on chromosomes 7, 11, 14, 18 and 26 affecting carcass and growth traits in a family of paternal half-sib Brahman cattle. Casas et al. (2003) suggested the existence of QTL segregating on chromosomes 5, 6, 7, 13, 14, 17, 19, 22, 27 and 29 for carcass composition and growth in families of Piedmontese and Belgian Blue cattle, both breeds from the B. taurus group, while Kim et al. (2003) found a total of 35 QTL (five significant and 30 with evidence suggesting linkage) in 19 chromosomes of an experimental F2 Angus x Brahman population.
Various studies have produced highly significant evidence for the presence of QTL on chromosome 5 affecting different characteristics such as ovulation rate (Kirkpatrick et al., 2000; Arias and Kirkpatrick, 2004), carcass traits (Stone et al., 1999; Casas et al., 2000; Kim et al., 2003) and birth weight (Davis et al., 1998; Casas et al., 2002; Li et al., 2002; Machado et al., 2003b; Kim et al., 2003). The insulin-like growth factor 1 gene (igf-1) acting on growth and metabolism maps to chromosome 5. The QTL for growth traits can be attributed to igf-1 itself or to one or more surrounding genes such as the high growth and myf5 genes, a hypothesis reinforced by the work of Machado et al. (2003b) and Kim et al. (2003).
QTL fine mapping, which means reducing marker interval near the QTL, is necessary to provide useful reference for future searches for candidate genes which eventually may lead to the identification of the nucleotide substitution(s) underlying the phenotypic variation. This kind of approach usually requires very large and well designed experimental populations appropriate for mapping QTL (Li, 2002) and this is especially true when mapping small effect QTL, although obtaining cattle populations such as these is expensive and time consuming for commercial lines. Even so, since 1995 the Embrapa (Empresa Brasileira de Pesquisa Agropecuaria, the Brazilian national agricultural company) Dairy Cattle Research Center has developed a second filial (F2) generation Gyr x Holstein cattle population using an experimental design with the main aim of creating a population for QTL mapping. The F2 design is very suitable for mapping economically important loci because not only a considerable amount of phenotypic variability is generated when the first filial (F1) generation is intercrossed to produce the F2 generation but linkage disequilibrium between markers and QTL is also promoted. The objective of the work presented in this paper was to use an interval mapping approach to map chromosome 5 for birth weight QTL in the F2 Gyr x Holstein cattle population.
Materials and Methods
Population
A bovine Gyr x Holstein population of 400 F2 animals is under development using multiple ovulation with embryo transfer at the Embrapa Dairy Cattle Research Center, Juiz de Fora, Minas Gerais, Brazil. Five F1 sires and 59 F1 dams, obtained from crosses between four Holstein sires and 28 Gyr dams, were intercrossed to obtain the F2 generation. Each F2 generation animal was weighted within 24 h of birth. For the present study, birth weight data from 294 F2 animals were available.
Marker selection and genotyping
Marker data were obtained from the MARC-ARS-USDA (Meat Animal Research Center-Agriculture Research Service-United States Department of Agriculture) database at http://www.marc.usda.gov. We chose five microsatellite markers covering the entire chromosome 5 with an average spacing between them of approximately 20 centimorgans (cM), the markers being chosen based on allele number, polymorphism information content, allele range, annealing temperature and primer availability in the Animal Biotechnology laboratory of Embrapa Southeast Cattle (So Carlos-SP, Brazil). An important criterion for locus selection was that the F1 bulls should be heterozygous for the chosen markers.
All microsatellites were amplified in a final reaction volume of 12.5 mL, consisting of 50 ng of genomic DNA, 50 mM KCl, 1.5 mM MgCl2, 10 mM Tris-HCl pH 8.4, 0.2 mM of each nucleotide, 0.5 units of Taq DNA polymerase, and 0.1 mM of each primer. The forward primers for each locus were marked with a fluorescent dye, primer identification codes, annealing temperature, map position and fluorescent dye are summarized in Table 1. Thermocycling conditions consisted of an initial denaturation at 94 °C for 2 min followed by 20 cycles of denaturation at 94 °C for 30 s and primer annealing for 30 s at 10 °C above the annealing temperature of each primer pair which was reduced by 0.5 °C at each cycle (i.e. a 'touchdown cycle profile designed to increase the specificity of the PCR reaction). The thermocycles were concluded with 10 cycles of denaturation at 94 °C, primer annealing temperature and primer extension, each for 30 s. Amplification products were submitted to a final extension step at 72 °C for 45 min. Polymerase chain reactions (PCR) were performed in a Mastercycler Gradient thermocycler (Eppendorf). At the end of the amplifications, products were analyzed in an ABI Prism 3100 Avant sequencer (Applied Biosystems) containing four capillaries and which has the capacity to analyze four fluorescent spectra simultaneously. The results were analyzed using the GeneScan and Genotyper software that calculate the size of the amplified fragments based on the internal size standard GeneScan 500, with fragments between 50 and 500 base pairs.
Statistical analysis
Allelic frequencies and Hardy-Weinberg equilibrium were calculated for the population as a whole and also for each of the generations in the study using the Cervus 2.0 software (Marshall et al., 1998).
A chromosome 5 linkage map was built using the BUILD and ALL functions of the CRIMAP program (Green et al., 1990) and the parental contribution of each locus was determined using the CHROMPIC function which, starting from the genotypes of a three generation pedigree (parental, F1 and F2), allows the determination of the linkage phase of the markers and haplotype identification in the F2 generation. The map was derived from the observed recombination fraction for each marker interval using Kosambis mapping function to transform recombination into distance.
This map was used for QTL analysis by the method of multiple interval mapping for F2 families (Haley et al., 1994) using the QTL Express program (Seaton et al., 2002) at http://qtl.cap.ed.ac.uk. The F-statistic was calculated to test the hypothesis of QTL segregation at 1 cM intervals using a model that included the fixed effects of year and season of birth, sex of the calf and additive and dominance effects of the QTL. We also applied permutation tests for the threshold determination (Churchill and Doerge, 1994) and the bootstrap technique for determination of the confidence interval (CI) of the presence of a possible QTL (Visscher et al., 1996). In this study 10.000 permutations were adopted for the value of a = 0.01 to obtain stable estimates (Churchill and Doerge, 1994).
The resampling bootstrap method proposed by Visscher et al. (1996) to determine the confidence interval removes samples (individuals) of the population that contain information about genotype and phenotype, generating a new population with sample replacement in which statistical analysis is used to identify QTL. After a pre-established resampling number and QTL analysis of the new populations the 95% CI of the QTL is determined by ordering the generated estimates, with 2.5% of the values representing the superior and inferior ends of the distribution. The width of the confidence interval depends on the size of the population and on the QTL effect, although variation between marker spacing does not result in very different confidence intervals (Visscher et al., 1996).
Results and Discussion
Genotypic analysis of parental and F1 bulls
The purpose of this analysis was to select the most informative markers from the initial panel based on F1 heterozygosity and chromosome coverage. The microsatellite markers chosen showed polymorphism, covering a segment of about 122 cM, according to the results obtained with the CRIMAP software.
The allelic frequencies of the five markers chosen for the Holstein bulls and Gyr dams are given in Table 2. Seven alleles were identified for markers BM6026 and BMS490, three for BM321, and eight for BMS1617 and BMS1248. It is interesting to observe that many alleles were breed-specific although these data should not be considered representative of the allelic distributions in the Holstein and Gyr populations due to the reduced sample size. They were only estimated with the purpose of verifying the occurrence of allelic divergence among the animals crossed to generate the F2 progeny.
The heterozygosity of a locus is defined as the probability that an individual is heterozygous for that locus in a population (Liu, 1998). This information is crucial for determining the number of markers needed to expand the linkage map to a desired level of coverage and to determine the number of animals or markers needed to search for economic trait loci (Bishop et al., 1994). As defined by Ott (1992), a locus is considered polymorphic if its heterozygosity (H) is greater than 0.1 and is considered highly polymorphic if H = 0.7. This definition implies that a marker is considered polymorphic when its most frequent allele has a frequency less than 0.95, and highly polymorphic when its most frequent allele has a frequency less than 0.55. Liu (1998) has shown that a sample size as small as 15 specimens can, with reasonable probability, reveal about 70% of the expected heterozygosity of a locus for all the allelic frequency distributions, although a large sample size is needed if the goal is to detect 95% of the heterozygosity.
It is also common to use polymorphism information content (PIC) to quantify marker polymorphism (Botstein et al., 1980), PIC being an estimate of the probability of obtaining informative crosses. Polymorphism information content approximately equals the heterozygosity when the locus has a large number of alleles and, as the number of alleles increases, both PIC and heterozygosity also increase (Liu, 1998). Our values for PIC, F1 heterozygosity and probability of Hardy-Weinberg equilibrium are summarized in Table 3.
We found that most of the microsatellite markers in the population studied were highly polymorphic, as expected by the loci pre-selection based on the MARC database, although this was not the case for marker BM321 which presented only three alleles in our research population. Machado et al. (2003b) found that in a 5/8 Charolais 3/8 Zebu crossbred population heterozygosity was between 0.468 and 0.755 (average 0.646) for the four markers used in their research. In our study heterozygosity was between 0.578 and 1.0 (average 0.865) (Table 3), this value being higher than the average heterozygosity obtained with nine microsatellites for the Gyr (0.305) and Holstein (0.339) populations by Machado et al. (2003a). In a study to incorporate more markers in the bovine genetic linkage map, Bishop et al. (1994) found an average heterozygosity of 0.747 in Bos taurus X Bos indicus F1 crosses.
We found that the PIC values of the markers ranged from 0.402 to 0.754, with average of 0.665. According to Botstein et al. (1980), a marker can be considered highly informative in a mapping population if it has an expected PIC value greater than 0.5, with values of between 0.5 and 0.25 considered to be reasonably informative. Our data suggest that the parental populations of Gyr and Holstein cattle were sufficiently divergent with respect to the analyzed markers to generate a highly informative F2 generation.
Of the five markers used to genotype the chromosome 5 only the BM6026 marker showed allelic frequencies for the F1 generation under Hardy-Weinberg equilibrium (Table 3). The absence of equilibrium was the expected result since the F1 generation was formed from two different breeds with the purpose of creating new associations between genotype and phenotype in the F2 generation. The observation of Hardy-Weinberg equilibrium for the BM6026 marker can result from the absence of differences among the allelic frequencies verified for the two breeds, probably due to sample size.
Linkage map construction
The chromosome 5 linkage map constructed by us (Figure 1) showed differences in the marker order of two microsatellites (BMS1617 and BMS490) in relation to the MARC reference map (http://www.marc.usda.gov). According to Liu (1998), this could happen due to the reduced sample size used to build the map, because the correct ordering of very close loci implies great number of informative meiosis events. Another hypothesis which could explain the variation in marker order could be the fact that in the MARC reference map two of these markers (BM321 and BMS1248) showed a great number of informative meiosis events while the other markers had a low number of informative meiosis events. The number of informative meiosis events in our map was very homogeneous, except for the BM321 marker for which only three alleles were identified in the population studied. Differences in sample size and the number of informative meiosis events may also explain the fact that the distance (121.7 cM) between the first and the fifth marker in our map was larger than the distance shown in the MARC reference map (81,7 cM). The number of informative meiosis events for the five markers is summarized in Table 4.
QTL mapping
A highly significant QTL (p < 0.01) for birth weight was detected in chromosome 5, with an additive effect of -1.6 kg corresponding to approximately 0.32 of the phenotypic standard-error and 5% of the trait average. The highest F-statistics value was at 69 cM from the most centromeric marker between the BMS1617 and BMS490 markers and only 4.6 cM from the BMS490 marker (Figure 2). Bootstrap analysis determined a confidence interval of 40 cM, between 39 and 79 cM, a value considered acceptable for the number of F2 animals used (Visscher et al. 1996). It is important to emphasize that this analysis was performed with 10.000 permutations between genotypes and phenotypes, a much higher value than the value considered by Churchill and Doerge (1994) as being statistically reliable. This QTL effect can be considered as strictly additive, since the estimated dominance deviation (0.735 ± 0.63 kg) was not significant (p > 0.05). Machado et al. (2003b), in a study with a Canchim population (5/8 Charolais, 3/8 Zebu), detected the presence of a QTL affecting birth weight (p < 0.05) located at 82.9 cM from the most centromeric marker and 9.9 cM from the igf-1 gene towards the telomeric end. However, the identified confidence intervals in Machado et al. (2003b) and in the present study do not allow us to conclude that the same QTL is segregating in the two populations. Li et al. (2002), using chromosome 5 haplotypes for association studies with growth traits in commercial populations of Bos taurus identified three chromosomic regions (0 to 30 cM, 55 to 70 cM, and 70 to 80 cM) that showed significant association with birth weight. However, the association of the igf-1 gene with birth weight was refuted by Machado et al. (2003b) and by Li et al. (2004), the latter authors having performed candidate gene analysis between the single nucleotide polymorphisms (SNPs) of chromosome 5 and growth and carcass traits. According to Li et al. (2004) SNPs detected in the igf-1 gene should not be responsible, nor even be tightly linked to, the mutation that affects growth and carcass traits in that beef cattle population.
Analysis for the presence of chromosome 5 QTL and their effect on weaning weight (WW) in this F2 population did not reveal any significant QTL effect (data not shown). This result indicates a tendency towards breaking the high correlation between birth weight and weaning weight, even though a slightly smaller number (N = 290) of animals was used in the weaning weight analysis. This could imply that the selection for this QTL could contribute to minimize the distocia effects since birth weight is the most significant factor that affects the distocia, which in turn affects several traits such as calve and dam mortality, increased susceptibility to disease and smaller calve weaning weight (Grosz and MacNeil, 2001). According to Grosz and MacNeil (2001), the annual losses due to distocia can reach 83.4 million dollars for dairy cattle and 142.5 million dollars for beef cattle in the United States.
The results presented in this paper are consistant with the presence of a QTL for birth weight in bovine chromosome 5 segregating in the F2 experimental population studied. This QTL has shown to be strictly additive and with no influence on weaning weight, which makes it appropriate for selection in cattle. Further investigation of this chromosome region may allow the identification of favorable haplotypes to be used in marker assisted selection. Analysis of this QTL in the parental Gyr and Holstein populations will also be relevant for the successful incorporation of this information in breeding programs.
Acknowledgments
Research supported by Embrapa-Prodetab. LL Coutinho, LCA Regitano and ML Martinez received a research fellowship from the Brazilian National Research Council (CNPq).
References
Arias J and Kirkpatrick B (2004) Mapping of bovine ovulation rate QTL: An analytical approach for three generation pedigree. Anim Genet 35:7-13.
Barendse W, Vaiman D, Kemp SJ, Sugimoto Y, Armitage SM, Willians JL, Sun HS, Eggen A, Agaba M, Aleyasin SA, Band M, Bishop MD, Buitkamp K, Byrne K, Collins F, Cooper L, Coppertiers W, Denys B, Drinkwater RD, Easterday K, Elduque C, Ennis S, Erhardt G, Ferretti L, Flavin N, Gao Q, Georges M, Gurung R, Harlizius B, Hawkins G, Hetzel J, Hirano T, Hulme D, Jorgensen C, Kessler M, Kirkpatrick BW, Konfortov B, Kostia S, Kuhn C, Lenstra JA, Leveziel H, Lewin HA, Leyhe B, Lil L, Martin-Burriel I, McGraw RA, Miller JR, Moody DE, Moore SS, Nakane S, Nijman IJ, Olsaker I, Pomp D, Rando A, Ron M, Shalom A, Reale AJ, Thieven U, Urquhart BGD, Vage DI, Van De Weghe A, Varvio S, Velmala R, Vilkki J, Weikard R, Woodside C, Womack JE, Zanotti M and Zaragoza P (1997) A medium-density genetic linkage map of the bovine genome. Mamm Genome 8:29-36.
Bishop MD, Kappes SM, Keele JW, Stone RT, Sunden SLF, Hawkins GA, Toldo SS, Fries R, Grosz MD, Yoo JY and Beattie CW (1994) A genetic linkage map for cattle. Genetics 136:619-639.
Botstein D, White RL, Skolnick M and Davis RW (1980) Construction of a genetic linkage map in man using restriction fragment length polymorphisms. Am J Hum Genet 32:314-331.
Casas E, Shackelford SD, Keele JW, Koohmaraie M, Smith TPL and Stone RT (2003) Detection of quantitative trait loci for growth and carcass composition in cattle. J Anim Sci 81:2976-2983.
Casas E, Keele JW, Shackelford SD, Stone RT and Koomaraie M (2002) Quantitative trait loci for birth weight longissimus muscle area and marbling on bovine chromosome 5. Abstracts 28th International Conference on Animal Genetics Guttingen Germany, Guttingen, International Socyety for Animal Genetics pp 163.
Casas E, Shackelford SD, Keele JW, Stone RT, Kappes SM and Koomaraie M (2000) Quantitative trait loci affecting growth and carcass composition of cattle segregating alternate forms of myostatin. J Anim Sci 78:560-569.
Churchill GA and Doerge RW (1994) Empirical threshold values for quantitative trait loci mapping. Genetics 138:963-971.
Davis GP and DeNise SK (1998) The impact of genetic markers on selection. J Anim Sci 76:2331-2339.
Davis SK, Hetzel DJS, Corbet NJ, Scacheri S, Lowden S, Renaud J, Mayne C, Stevenson R, Moore SS and Byrne K (1998) The mapping of quantitative trait loci for birth weight in a tropical beef herd. Proceedings 6th World Congress on Genetics Applied to Livestock Production Armidale Australia Papers 26:441-444.
Elo KTL, Vilkki J, De Koning DJ, Velmala RJ and Maki-Tanila AV (1999) A quantitative trait locus for live weight maps to bovine chromosome 23. Mamm Genome 10:831-835.
Friend J and Bishop D (1978) Cattle of the World. Blandford Press, Poole, 198 pp.
Green P, Falls K and Croocks S (1990) CRI-MAP version 2.4. St. Louis, University School of Medicine.
Grosz MD and MacNeil MD (2001) Putative quantitative trait locus affecting birth weight on bovine chromosome 2. J Anim Sci 79:68-72.
Haley CS, Knott SA and Elsen JM (1994) Mapping quantitative trait loci in crosses between outbred lines using least squares. Genetics 136:1195-1207.
Ihara N, Takasuga A, Mizoshita K, Takeda H, Sugimoto M, Mizoguchi Y, Hirano T, Itoh T, Watanabe T, Reed KM, Snelling WM, Kappes SM, Beattie CW, Bennett GL and Sugimoto Y (2004). A comprehensive genetic map of the cattle genome based on 3802 microsatellites. Genome Res 14:1987-1998.
Kappes SM, Keele JW, Stone RT, McGraw RA, Sonstergard TS, Smith TP, Lopez-Corrales NL and Beattie CW (1997) A second-generation linkage map of the bovine genome. Genome Res 7:235-249.
Kim JJ, Farnir F, Savell J and Taylor JF (2003) Detection of quantitative trait loci for growth and beef carcass fatness traits in a cross between Bos taurus (Angus) and Bos indicus (Brahman) cattle. J Anim Sci 81:1933-1942.
Kirkpatrick BW, Becky MB and Gregoy KE (2000) Mapping quantitative trait loci for bovine ovulation rate. Mamm Genome 11:136-139.
Koots KR, Gibson JP and Wilton JW (1994) Analyses of published genetic parameter estimates for beef production traits. 2 Phenotypic and genetic correlations. Anim Breed Abstr 62:825-853.
Li C, Basarab J, Snelling WM, Benkel B, Murdoch B, Hansen C and Moore SS (2004) Assessment of positional candidate genes myf5 and igf1 for growth on bovine chromosome 5 in commercial lines of Bos taurus. J Anim Sci 82:1-7.
Li C, Basarab J, Snelling WM, Benkel B, Murdoch B and Moore SS (2002) The identification of common haplotypes on bovine chromosome 5 within commercial lines of Bos taurus and their associations with growth traits. J Anim Sci 80:1187-1194.
Liu BH (1998) Statistical Genomics: Linkage Mapping and QTL Analysis CRC Press, Boca Ranton, FL, 611 pp.
Machado MA, Schuster I, Martinez ML and Campos AL (2003a). Genetic diversity of four cattle breeds using microsatellite markers. R Bras Zootec 32:93-98.
Machado MBB, Alencar MM, Pereira AP, Oliveira HN, Casas E, Coutinho LL and Regitano LCA (2003b) QTL affecting body weight in a candidate region of cattle chromosome 5. Genet Mol Biol 26:259-265.
MacNeil MD and Grosz MD (2002) Genome-wide scans for QTL affecting carcass traits in Hereford x composite double backcross populations. J Anim Sci 80:2316-2324.
Madalena FE, Teodoro RL, Lemos AM, Monteiro JBN and Barbosa RT (1990) Evaluation of strategies for crossbreeding of dairy cattle in Brazil. J Dairy Sci 73:1887-1901.
Marshall TC, Slate J, Kruuk LEB and Pemberton JM (1998) Statistical confidence for likelihood-based paternity inference in natural populations. Mol Ecol 7:639-655.
Ott J (1992) Strategies for characterizing highly polymorphic markers in human gene mapping. Am J Hum Genet 51:283-290.
Seaton G, Haley CS and Knott SA (2002) QTL Express: http://qtl.cap.ed.ac.uk.
Stone RT, Keele JW, Shackelford SD, Kappes SM and Koomaraie M (1999) A primary screen of the bovine genome for quantitative trait loci affecting carcass and growth traits. J Anim Sci 77:1379-1384.
Visscher PM, Thompson R and Haley CS (1996) Confidence intervals in QTL mapping by bootstrapping. Genetics 143:1013-1020.(Gustavo Gasparin; Marcelo)
IIUniversidade de So Paulo, Departamento de Produo Animal, Piracicaba, SP, Brazil
IIIEmbrapa Gado de Leite, Juiz de Fora, MG, Brazil
IVEmbrapa Pecuaria Sudeste, So Carlos, SP, Brazil
VCNPq fellowship
ABSTRACT
Segregation between a genetic marker and a locus influencing a quantitative trait in a well delineated population is the basis for success in mapping quantitative trait loci (QTL). To detect bovine chromosome 5 (BTA5) birth weight QTL we genotyped 294 F2 Gyr (Bos indicus) x Holstein (Bos taurus) crossbreed cattle for five microsatellite markers. A linkage map was constructed for the markers and an interval analysis for the presence of QTL was performed. The linkage map indicated differences in the order of two markers relative to the reference map (http://www.marc.usda.gov). Interval analysis detected a QTL controlling birth weight (p < 0.01) at 69 centimorgans (cM) from the most centromeric marker with an effect of 0.32 phenotypic standard-error. These results support other studies with crossbred Bos taurus x Bos indicus populations.
Key words: BTA5, birth weight, cattle, QTL, microsatellite markers.
Introduction
The use of crossbred cattle has been an alternative for the intensification of bovine milk production, making possible the exploration of the genetic differences between breeds and the benefits of heterosis. The need for bovine genetic resources adapted to tropical conditions has lead to the use of crosses between Holstein (Bos taurus) cattle, selected for milk production for more than 2.000 years (Friend and Bishop, 1978), and Zebu (Bos indicus) cattle, adapted to tropical conditions (Madalena et al. 1990). Most economically important traits in dairy cattle are quantitative traits, which have been under selection for several generations, with quite favorable results.
The development of saturated genetic maps (Barendse et al., 1997; Kappes et al., 1997; Ihara et al., 2004) has allowed the identification of quantitative trait loci (QTL) affecting economically important traits. MacNeil and Grosz (2002) have stated that the identification of QTL has the potential to significantly increase the genetic improvement rate through the implementation of marker-assisted selection (MAS). For traits difficult or expensive to determine, with low heritability, late expression or measured only after slaughter, MAS can substantially increase the rate of response compared to selection based exclusively on estimates of performance values (Davis and DeNise, 1998). The use of MAS also allows the opportunity for more efficient breakage of antagonistic genetic correlations between characters (Grosz and MacNeil, 2001). For instance, estimates of the genetic correlation between direct effects on birth weight and yearling weight are approximately 0.5 in all bovine breeds (Koots et al., 1994) and as result selection based on mature weight or growth rate to mature weight can significantly increase birth weight potentially increasing the incidence and severity of calving difficulties. The opposite is also true, selection for smaller birth weight will reduce adult weight (Grosz and MacNeil, 2001). Identification of genomic regions affecting birth weight with no effect on mature weight would be a powerful tool for the manipulation of pre and post birth growth rate.
Quantitative trait loci have been detected in experimental and commercial bovine, swine and ovine populations (Davis and DeNise, 1998). There are several described bovine QTL affecting many characteristics. Davis et al. (1998) detected and mapped five QTL for birth weight on chromosomes 5, 6, 14, 18 and 21 in three families of paternal half-sib Charolais x Brahman (B. taurus x B. indicus) and Elo et al. (1999) detected a QTL for adult live weight on chromosome 23. Stone et al. (1999) reported significant evidence for the presence of QTL on chromosomes 1, 2, 5 and 13 and suggestive evidence for QTL on chromosomes 7, 11, 14, 18 and 26 affecting carcass and growth traits in a family of paternal half-sib Brahman cattle. Casas et al. (2003) suggested the existence of QTL segregating on chromosomes 5, 6, 7, 13, 14, 17, 19, 22, 27 and 29 for carcass composition and growth in families of Piedmontese and Belgian Blue cattle, both breeds from the B. taurus group, while Kim et al. (2003) found a total of 35 QTL (five significant and 30 with evidence suggesting linkage) in 19 chromosomes of an experimental F2 Angus x Brahman population.
Various studies have produced highly significant evidence for the presence of QTL on chromosome 5 affecting different characteristics such as ovulation rate (Kirkpatrick et al., 2000; Arias and Kirkpatrick, 2004), carcass traits (Stone et al., 1999; Casas et al., 2000; Kim et al., 2003) and birth weight (Davis et al., 1998; Casas et al., 2002; Li et al., 2002; Machado et al., 2003b; Kim et al., 2003). The insulin-like growth factor 1 gene (igf-1) acting on growth and metabolism maps to chromosome 5. The QTL for growth traits can be attributed to igf-1 itself or to one or more surrounding genes such as the high growth and myf5 genes, a hypothesis reinforced by the work of Machado et al. (2003b) and Kim et al. (2003).
QTL fine mapping, which means reducing marker interval near the QTL, is necessary to provide useful reference for future searches for candidate genes which eventually may lead to the identification of the nucleotide substitution(s) underlying the phenotypic variation. This kind of approach usually requires very large and well designed experimental populations appropriate for mapping QTL (Li, 2002) and this is especially true when mapping small effect QTL, although obtaining cattle populations such as these is expensive and time consuming for commercial lines. Even so, since 1995 the Embrapa (Empresa Brasileira de Pesquisa Agropecuaria, the Brazilian national agricultural company) Dairy Cattle Research Center has developed a second filial (F2) generation Gyr x Holstein cattle population using an experimental design with the main aim of creating a population for QTL mapping. The F2 design is very suitable for mapping economically important loci because not only a considerable amount of phenotypic variability is generated when the first filial (F1) generation is intercrossed to produce the F2 generation but linkage disequilibrium between markers and QTL is also promoted. The objective of the work presented in this paper was to use an interval mapping approach to map chromosome 5 for birth weight QTL in the F2 Gyr x Holstein cattle population.
Materials and Methods
Population
A bovine Gyr x Holstein population of 400 F2 animals is under development using multiple ovulation with embryo transfer at the Embrapa Dairy Cattle Research Center, Juiz de Fora, Minas Gerais, Brazil. Five F1 sires and 59 F1 dams, obtained from crosses between four Holstein sires and 28 Gyr dams, were intercrossed to obtain the F2 generation. Each F2 generation animal was weighted within 24 h of birth. For the present study, birth weight data from 294 F2 animals were available.
Marker selection and genotyping
Marker data were obtained from the MARC-ARS-USDA (Meat Animal Research Center-Agriculture Research Service-United States Department of Agriculture) database at http://www.marc.usda.gov. We chose five microsatellite markers covering the entire chromosome 5 with an average spacing between them of approximately 20 centimorgans (cM), the markers being chosen based on allele number, polymorphism information content, allele range, annealing temperature and primer availability in the Animal Biotechnology laboratory of Embrapa Southeast Cattle (So Carlos-SP, Brazil). An important criterion for locus selection was that the F1 bulls should be heterozygous for the chosen markers.
All microsatellites were amplified in a final reaction volume of 12.5 mL, consisting of 50 ng of genomic DNA, 50 mM KCl, 1.5 mM MgCl2, 10 mM Tris-HCl pH 8.4, 0.2 mM of each nucleotide, 0.5 units of Taq DNA polymerase, and 0.1 mM of each primer. The forward primers for each locus were marked with a fluorescent dye, primer identification codes, annealing temperature, map position and fluorescent dye are summarized in Table 1. Thermocycling conditions consisted of an initial denaturation at 94 °C for 2 min followed by 20 cycles of denaturation at 94 °C for 30 s and primer annealing for 30 s at 10 °C above the annealing temperature of each primer pair which was reduced by 0.5 °C at each cycle (i.e. a 'touchdown cycle profile designed to increase the specificity of the PCR reaction). The thermocycles were concluded with 10 cycles of denaturation at 94 °C, primer annealing temperature and primer extension, each for 30 s. Amplification products were submitted to a final extension step at 72 °C for 45 min. Polymerase chain reactions (PCR) were performed in a Mastercycler Gradient thermocycler (Eppendorf). At the end of the amplifications, products were analyzed in an ABI Prism 3100 Avant sequencer (Applied Biosystems) containing four capillaries and which has the capacity to analyze four fluorescent spectra simultaneously. The results were analyzed using the GeneScan and Genotyper software that calculate the size of the amplified fragments based on the internal size standard GeneScan 500, with fragments between 50 and 500 base pairs.
Statistical analysis
Allelic frequencies and Hardy-Weinberg equilibrium were calculated for the population as a whole and also for each of the generations in the study using the Cervus 2.0 software (Marshall et al., 1998).
A chromosome 5 linkage map was built using the BUILD and ALL functions of the CRIMAP program (Green et al., 1990) and the parental contribution of each locus was determined using the CHROMPIC function which, starting from the genotypes of a three generation pedigree (parental, F1 and F2), allows the determination of the linkage phase of the markers and haplotype identification in the F2 generation. The map was derived from the observed recombination fraction for each marker interval using Kosambis mapping function to transform recombination into distance.
This map was used for QTL analysis by the method of multiple interval mapping for F2 families (Haley et al., 1994) using the QTL Express program (Seaton et al., 2002) at http://qtl.cap.ed.ac.uk. The F-statistic was calculated to test the hypothesis of QTL segregation at 1 cM intervals using a model that included the fixed effects of year and season of birth, sex of the calf and additive and dominance effects of the QTL. We also applied permutation tests for the threshold determination (Churchill and Doerge, 1994) and the bootstrap technique for determination of the confidence interval (CI) of the presence of a possible QTL (Visscher et al., 1996). In this study 10.000 permutations were adopted for the value of a = 0.01 to obtain stable estimates (Churchill and Doerge, 1994).
The resampling bootstrap method proposed by Visscher et al. (1996) to determine the confidence interval removes samples (individuals) of the population that contain information about genotype and phenotype, generating a new population with sample replacement in which statistical analysis is used to identify QTL. After a pre-established resampling number and QTL analysis of the new populations the 95% CI of the QTL is determined by ordering the generated estimates, with 2.5% of the values representing the superior and inferior ends of the distribution. The width of the confidence interval depends on the size of the population and on the QTL effect, although variation between marker spacing does not result in very different confidence intervals (Visscher et al., 1996).
Results and Discussion
Genotypic analysis of parental and F1 bulls
The purpose of this analysis was to select the most informative markers from the initial panel based on F1 heterozygosity and chromosome coverage. The microsatellite markers chosen showed polymorphism, covering a segment of about 122 cM, according to the results obtained with the CRIMAP software.
The allelic frequencies of the five markers chosen for the Holstein bulls and Gyr dams are given in Table 2. Seven alleles were identified for markers BM6026 and BMS490, three for BM321, and eight for BMS1617 and BMS1248. It is interesting to observe that many alleles were breed-specific although these data should not be considered representative of the allelic distributions in the Holstein and Gyr populations due to the reduced sample size. They were only estimated with the purpose of verifying the occurrence of allelic divergence among the animals crossed to generate the F2 progeny.
The heterozygosity of a locus is defined as the probability that an individual is heterozygous for that locus in a population (Liu, 1998). This information is crucial for determining the number of markers needed to expand the linkage map to a desired level of coverage and to determine the number of animals or markers needed to search for economic trait loci (Bishop et al., 1994). As defined by Ott (1992), a locus is considered polymorphic if its heterozygosity (H) is greater than 0.1 and is considered highly polymorphic if H = 0.7. This definition implies that a marker is considered polymorphic when its most frequent allele has a frequency less than 0.95, and highly polymorphic when its most frequent allele has a frequency less than 0.55. Liu (1998) has shown that a sample size as small as 15 specimens can, with reasonable probability, reveal about 70% of the expected heterozygosity of a locus for all the allelic frequency distributions, although a large sample size is needed if the goal is to detect 95% of the heterozygosity.
It is also common to use polymorphism information content (PIC) to quantify marker polymorphism (Botstein et al., 1980), PIC being an estimate of the probability of obtaining informative crosses. Polymorphism information content approximately equals the heterozygosity when the locus has a large number of alleles and, as the number of alleles increases, both PIC and heterozygosity also increase (Liu, 1998). Our values for PIC, F1 heterozygosity and probability of Hardy-Weinberg equilibrium are summarized in Table 3.
We found that most of the microsatellite markers in the population studied were highly polymorphic, as expected by the loci pre-selection based on the MARC database, although this was not the case for marker BM321 which presented only three alleles in our research population. Machado et al. (2003b) found that in a 5/8 Charolais 3/8 Zebu crossbred population heterozygosity was between 0.468 and 0.755 (average 0.646) for the four markers used in their research. In our study heterozygosity was between 0.578 and 1.0 (average 0.865) (Table 3), this value being higher than the average heterozygosity obtained with nine microsatellites for the Gyr (0.305) and Holstein (0.339) populations by Machado et al. (2003a). In a study to incorporate more markers in the bovine genetic linkage map, Bishop et al. (1994) found an average heterozygosity of 0.747 in Bos taurus X Bos indicus F1 crosses.
We found that the PIC values of the markers ranged from 0.402 to 0.754, with average of 0.665. According to Botstein et al. (1980), a marker can be considered highly informative in a mapping population if it has an expected PIC value greater than 0.5, with values of between 0.5 and 0.25 considered to be reasonably informative. Our data suggest that the parental populations of Gyr and Holstein cattle were sufficiently divergent with respect to the analyzed markers to generate a highly informative F2 generation.
Of the five markers used to genotype the chromosome 5 only the BM6026 marker showed allelic frequencies for the F1 generation under Hardy-Weinberg equilibrium (Table 3). The absence of equilibrium was the expected result since the F1 generation was formed from two different breeds with the purpose of creating new associations between genotype and phenotype in the F2 generation. The observation of Hardy-Weinberg equilibrium for the BM6026 marker can result from the absence of differences among the allelic frequencies verified for the two breeds, probably due to sample size.
Linkage map construction
The chromosome 5 linkage map constructed by us (Figure 1) showed differences in the marker order of two microsatellites (BMS1617 and BMS490) in relation to the MARC reference map (http://www.marc.usda.gov). According to Liu (1998), this could happen due to the reduced sample size used to build the map, because the correct ordering of very close loci implies great number of informative meiosis events. Another hypothesis which could explain the variation in marker order could be the fact that in the MARC reference map two of these markers (BM321 and BMS1248) showed a great number of informative meiosis events while the other markers had a low number of informative meiosis events. The number of informative meiosis events in our map was very homogeneous, except for the BM321 marker for which only three alleles were identified in the population studied. Differences in sample size and the number of informative meiosis events may also explain the fact that the distance (121.7 cM) between the first and the fifth marker in our map was larger than the distance shown in the MARC reference map (81,7 cM). The number of informative meiosis events for the five markers is summarized in Table 4.
QTL mapping
A highly significant QTL (p < 0.01) for birth weight was detected in chromosome 5, with an additive effect of -1.6 kg corresponding to approximately 0.32 of the phenotypic standard-error and 5% of the trait average. The highest F-statistics value was at 69 cM from the most centromeric marker between the BMS1617 and BMS490 markers and only 4.6 cM from the BMS490 marker (Figure 2). Bootstrap analysis determined a confidence interval of 40 cM, between 39 and 79 cM, a value considered acceptable for the number of F2 animals used (Visscher et al. 1996). It is important to emphasize that this analysis was performed with 10.000 permutations between genotypes and phenotypes, a much higher value than the value considered by Churchill and Doerge (1994) as being statistically reliable. This QTL effect can be considered as strictly additive, since the estimated dominance deviation (0.735 ± 0.63 kg) was not significant (p > 0.05). Machado et al. (2003b), in a study with a Canchim population (5/8 Charolais, 3/8 Zebu), detected the presence of a QTL affecting birth weight (p < 0.05) located at 82.9 cM from the most centromeric marker and 9.9 cM from the igf-1 gene towards the telomeric end. However, the identified confidence intervals in Machado et al. (2003b) and in the present study do not allow us to conclude that the same QTL is segregating in the two populations. Li et al. (2002), using chromosome 5 haplotypes for association studies with growth traits in commercial populations of Bos taurus identified three chromosomic regions (0 to 30 cM, 55 to 70 cM, and 70 to 80 cM) that showed significant association with birth weight. However, the association of the igf-1 gene with birth weight was refuted by Machado et al. (2003b) and by Li et al. (2004), the latter authors having performed candidate gene analysis between the single nucleotide polymorphisms (SNPs) of chromosome 5 and growth and carcass traits. According to Li et al. (2004) SNPs detected in the igf-1 gene should not be responsible, nor even be tightly linked to, the mutation that affects growth and carcass traits in that beef cattle population.
Analysis for the presence of chromosome 5 QTL and their effect on weaning weight (WW) in this F2 population did not reveal any significant QTL effect (data not shown). This result indicates a tendency towards breaking the high correlation between birth weight and weaning weight, even though a slightly smaller number (N = 290) of animals was used in the weaning weight analysis. This could imply that the selection for this QTL could contribute to minimize the distocia effects since birth weight is the most significant factor that affects the distocia, which in turn affects several traits such as calve and dam mortality, increased susceptibility to disease and smaller calve weaning weight (Grosz and MacNeil, 2001). According to Grosz and MacNeil (2001), the annual losses due to distocia can reach 83.4 million dollars for dairy cattle and 142.5 million dollars for beef cattle in the United States.
The results presented in this paper are consistant with the presence of a QTL for birth weight in bovine chromosome 5 segregating in the F2 experimental population studied. This QTL has shown to be strictly additive and with no influence on weaning weight, which makes it appropriate for selection in cattle. Further investigation of this chromosome region may allow the identification of favorable haplotypes to be used in marker assisted selection. Analysis of this QTL in the parental Gyr and Holstein populations will also be relevant for the successful incorporation of this information in breeding programs.
Acknowledgments
Research supported by Embrapa-Prodetab. LL Coutinho, LCA Regitano and ML Martinez received a research fellowship from the Brazilian National Research Council (CNPq).
References
Arias J and Kirkpatrick B (2004) Mapping of bovine ovulation rate QTL: An analytical approach for three generation pedigree. Anim Genet 35:7-13.
Barendse W, Vaiman D, Kemp SJ, Sugimoto Y, Armitage SM, Willians JL, Sun HS, Eggen A, Agaba M, Aleyasin SA, Band M, Bishop MD, Buitkamp K, Byrne K, Collins F, Cooper L, Coppertiers W, Denys B, Drinkwater RD, Easterday K, Elduque C, Ennis S, Erhardt G, Ferretti L, Flavin N, Gao Q, Georges M, Gurung R, Harlizius B, Hawkins G, Hetzel J, Hirano T, Hulme D, Jorgensen C, Kessler M, Kirkpatrick BW, Konfortov B, Kostia S, Kuhn C, Lenstra JA, Leveziel H, Lewin HA, Leyhe B, Lil L, Martin-Burriel I, McGraw RA, Miller JR, Moody DE, Moore SS, Nakane S, Nijman IJ, Olsaker I, Pomp D, Rando A, Ron M, Shalom A, Reale AJ, Thieven U, Urquhart BGD, Vage DI, Van De Weghe A, Varvio S, Velmala R, Vilkki J, Weikard R, Woodside C, Womack JE, Zanotti M and Zaragoza P (1997) A medium-density genetic linkage map of the bovine genome. Mamm Genome 8:29-36.
Bishop MD, Kappes SM, Keele JW, Stone RT, Sunden SLF, Hawkins GA, Toldo SS, Fries R, Grosz MD, Yoo JY and Beattie CW (1994) A genetic linkage map for cattle. Genetics 136:619-639.
Botstein D, White RL, Skolnick M and Davis RW (1980) Construction of a genetic linkage map in man using restriction fragment length polymorphisms. Am J Hum Genet 32:314-331.
Casas E, Shackelford SD, Keele JW, Koohmaraie M, Smith TPL and Stone RT (2003) Detection of quantitative trait loci for growth and carcass composition in cattle. J Anim Sci 81:2976-2983.
Casas E, Keele JW, Shackelford SD, Stone RT and Koomaraie M (2002) Quantitative trait loci for birth weight longissimus muscle area and marbling on bovine chromosome 5. Abstracts 28th International Conference on Animal Genetics Guttingen Germany, Guttingen, International Socyety for Animal Genetics pp 163.
Casas E, Shackelford SD, Keele JW, Stone RT, Kappes SM and Koomaraie M (2000) Quantitative trait loci affecting growth and carcass composition of cattle segregating alternate forms of myostatin. J Anim Sci 78:560-569.
Churchill GA and Doerge RW (1994) Empirical threshold values for quantitative trait loci mapping. Genetics 138:963-971.
Davis GP and DeNise SK (1998) The impact of genetic markers on selection. J Anim Sci 76:2331-2339.
Davis SK, Hetzel DJS, Corbet NJ, Scacheri S, Lowden S, Renaud J, Mayne C, Stevenson R, Moore SS and Byrne K (1998) The mapping of quantitative trait loci for birth weight in a tropical beef herd. Proceedings 6th World Congress on Genetics Applied to Livestock Production Armidale Australia Papers 26:441-444.
Elo KTL, Vilkki J, De Koning DJ, Velmala RJ and Maki-Tanila AV (1999) A quantitative trait locus for live weight maps to bovine chromosome 23. Mamm Genome 10:831-835.
Friend J and Bishop D (1978) Cattle of the World. Blandford Press, Poole, 198 pp.
Green P, Falls K and Croocks S (1990) CRI-MAP version 2.4. St. Louis, University School of Medicine.
Grosz MD and MacNeil MD (2001) Putative quantitative trait locus affecting birth weight on bovine chromosome 2. J Anim Sci 79:68-72.
Haley CS, Knott SA and Elsen JM (1994) Mapping quantitative trait loci in crosses between outbred lines using least squares. Genetics 136:1195-1207.
Ihara N, Takasuga A, Mizoshita K, Takeda H, Sugimoto M, Mizoguchi Y, Hirano T, Itoh T, Watanabe T, Reed KM, Snelling WM, Kappes SM, Beattie CW, Bennett GL and Sugimoto Y (2004). A comprehensive genetic map of the cattle genome based on 3802 microsatellites. Genome Res 14:1987-1998.
Kappes SM, Keele JW, Stone RT, McGraw RA, Sonstergard TS, Smith TP, Lopez-Corrales NL and Beattie CW (1997) A second-generation linkage map of the bovine genome. Genome Res 7:235-249.
Kim JJ, Farnir F, Savell J and Taylor JF (2003) Detection of quantitative trait loci for growth and beef carcass fatness traits in a cross between Bos taurus (Angus) and Bos indicus (Brahman) cattle. J Anim Sci 81:1933-1942.
Kirkpatrick BW, Becky MB and Gregoy KE (2000) Mapping quantitative trait loci for bovine ovulation rate. Mamm Genome 11:136-139.
Koots KR, Gibson JP and Wilton JW (1994) Analyses of published genetic parameter estimates for beef production traits. 2 Phenotypic and genetic correlations. Anim Breed Abstr 62:825-853.
Li C, Basarab J, Snelling WM, Benkel B, Murdoch B, Hansen C and Moore SS (2004) Assessment of positional candidate genes myf5 and igf1 for growth on bovine chromosome 5 in commercial lines of Bos taurus. J Anim Sci 82:1-7.
Li C, Basarab J, Snelling WM, Benkel B, Murdoch B and Moore SS (2002) The identification of common haplotypes on bovine chromosome 5 within commercial lines of Bos taurus and their associations with growth traits. J Anim Sci 80:1187-1194.
Liu BH (1998) Statistical Genomics: Linkage Mapping and QTL Analysis CRC Press, Boca Ranton, FL, 611 pp.
Machado MA, Schuster I, Martinez ML and Campos AL (2003a). Genetic diversity of four cattle breeds using microsatellite markers. R Bras Zootec 32:93-98.
Machado MBB, Alencar MM, Pereira AP, Oliveira HN, Casas E, Coutinho LL and Regitano LCA (2003b) QTL affecting body weight in a candidate region of cattle chromosome 5. Genet Mol Biol 26:259-265.
MacNeil MD and Grosz MD (2002) Genome-wide scans for QTL affecting carcass traits in Hereford x composite double backcross populations. J Anim Sci 80:2316-2324.
Madalena FE, Teodoro RL, Lemos AM, Monteiro JBN and Barbosa RT (1990) Evaluation of strategies for crossbreeding of dairy cattle in Brazil. J Dairy Sci 73:1887-1901.
Marshall TC, Slate J, Kruuk LEB and Pemberton JM (1998) Statistical confidence for likelihood-based paternity inference in natural populations. Mol Ecol 7:639-655.
Ott J (1992) Strategies for characterizing highly polymorphic markers in human gene mapping. Am J Hum Genet 51:283-290.
Seaton G, Haley CS and Knott SA (2002) QTL Express: http://qtl.cap.ed.ac.uk.
Stone RT, Keele JW, Shackelford SD, Kappes SM and Koomaraie M (1999) A primary screen of the bovine genome for quantitative trait loci affecting carcass and growth traits. J Anim Sci 77:1379-1384.
Visscher PM, Thompson R and Haley CS (1996) Confidence intervals in QTL mapping by bootstrapping. Genetics 143:1013-1020.(Gustavo Gasparin; Marcelo)