A genome scan for QTL influencing milk production and health traits in dairy cattle
D. W. HEYEN1,
J. I. WELLER2,
M. RON2,
M. BAND1,
J. E. BEEVER1,
E. FELDMESSER2,
Y. DA1,
G. R. WIGGANS3,
P. M. VANRADEN3 and
H. A. LEWIN1
1 Department of Animal Sciences, University of Illinois at Urbana-Champaign, Urbana, Illinois 61801
2 Institute of Animal Sciences, Agricultural Research Organization, the Volcani Center, Bet Dagan 50250, Israel
3 United States Department of Agriculture-Agricultural Research Service, Animal Improvement Programs Laboratory, Beltsville, Maryland 207052350
 |
ABSTRACT
|
---|
Heyen, D. W., J. I. Weller, M. Ron, M. Band, J. E. Beever, E. Feldmesser, Y. Da, G. R. Wiggans, P. M. VanRaden, and H. A. Lewin. A genome scan for QTL influencing milk production and health traits in dairy cattle. Physiol. Genomics 1: 165175, 1999.A genome scan was conducted in the North American Holstein-Friesian population for quantitative trait loci (QTL) affecting production and health traits using the granddaughter design. Resource families consisted of 1,068 sons of eight elite sires. Genome coverage was estimated to be 2,551 cM (85%) for 174 genotyped markers. Each marker was tested for effects on milk yield, fat yield, protein yield, fat percentage, protein percentage, somatic cell score, and productive herd life using analysis of variance. Joint analysis of all families identified marker effects on 11 chromosomes that exceeded the genomewide, suggestive, or nominal significance threshold for QTL effects. Large marker effects on fat percentage were found on chromosomes 3 and 14, and multimarker regression analysis was used to refine the position of these QTL. Half-sibling families from Israeli Holstein dairy herds were used in a daughter design to confirm the presence of the QTL for fat percentage on chromosome 14. The QTL identified in this study may be useful for marker-assisted selection and for selection of a refined set of candidate genes affecting these traits.
genomics; lactation; mastitis
 |
INTRODUCTION
|
---|
THE GENETIC DISSECTION of complex traits is one of the most challenging problems in contemporary biology (37, 42). Complex traits, such as disease resistance and lactation, are under the control of many genes, each with a different phenotypic contribution. Most complex traits have a continuous underlying distribution of measurement, and the genes responsible for genetic variation in these traits have been termed quantitative trait loci (QTL; 16). Detection of QTL in outbred populations may be complicated by environmental effects and epistasis. To deal with this complexity, a number of experimental strategies and statistical tools have been developed that maximize the probability of detecting QTL (30). When the trait is known to be heritable, QTL can be identified by conducting a genome scan. This involves selection of an appropriate resource population segregating the trait(s) of interest and genotyping that population with a panel of polymorphic genetic markers that span all possible chromosome segments. Statistical methods such as analysis of variance or maximum likelihood are then used to determine if a marker or marker interval is associated with an effect on the trait (8, 27). Although the entire field is presently stymied by the absence of appropriate tools for understanding the molecular basis of QTL effects, the identification of marker intervals bracketing QTL is a prerequisite to progress in this area of scientific inquiry. Moreover, knowledge of the relative position of QTL may itself be useful as a tool for the genetic improvement of plant and livestock species using marker-assisted breeding schemes (24, 34, 46).
The "daughter design" and "granddaughter design" are well-accepted schemes to identify marker-QTL associations in commercial dairy populations (35, 47, 57). The daughter design uses marker information and phenotypic data from daughters having a common sire (half-sibs) for detection of QTL effects. The granddaughter design employs marker information and sire breeding values for quantitative traits that are estimated from the production and health records of their daughters. Markers are scored among half-sib sons and tested for allele effects using one or more statistical methods (33, 49, 57). The effects detected among the sons indicate that the sons' sire was heterozygous for both the marker and the QTL. In general, the allele effect detected in the sons is one-half the true effect, because the sons' breeding values are estimated from the phenotypic records of the elite sire's granddaughters (half of the granddaughters will not have any contribution from their paternal grandsire at a particular locus). This strategy is feasible because of the widespread use of artificial insemination and selective breeding in the dairy cattle industry. Identification of QTL contributing to the genetic variation of milk production and composition will enhance our understanding of the genetics and physiology of these traits while accruing practical benefits to the dairy industry by facilitating genetic improvement using marker-assisted selection (14, 24, 34). In addition, mapping QTL may lead to pharmacological and transgenic approaches that will enhance an animal's production potential or disease resistance.
The current genetic linkage map of cattle consists of more than 1,500 genetic markers. The majority of these markers are polymorphic microsatellites with known map locations that are well-suited for genome scans. Microsatellites have been used to perform scans for QTL affecting milk production (1, 3, 5, 12, 17, 18, 29, 32, 40, 49, 53, 60), health traits (4, 5, 38, 60), and body conformation traits (2, 3). Both consistent and conflicting results have been obtained for particular markers, which may be due to differences in the size, composition, and geographical location of the populations studied and methods of analysis employed.
The objective of the present study was to identify QTL affecting economically important traits in North American Holstein-Friesian cattle using the granddaughter design. QTL were identified with large effects on milk production, milk composition, and health traits. Our results will provide a basis for genetic improvement using marker-assisted selection and an understanding of the molecular mechanisms of complex physiological traits.
 |
MATERIAL AND METHODS
|
---|
Resource Populations and Description of Phenotypic Data
The Dairy Bull DNA Repository (DBDR) was established in 1992 at the University of Illinois as a DNA resource for mapping QTL in elite North American Holstein-Friesian families (13). The collection currently consists of semen from 3,562 sons of 35 elite sires contributed by nine North American artificial insemination organizations. Eight of the largest half-sib families consisting of 1,068 bulls were selected for QTL detection using the granddaughter design (57). These bulls are not a random sample of grandsires' sons because semen availability is based on the results of progeny testing programs. As a result, the statistical power of the experiment is lower because of a reduction in the average allele substitution effect (17). An independent population of 3,264 daughters of nine Israeli Holstein sires was collected for confirmation of the QTL on BTA14 using the daughter design. Although several grandsires in the DBDR are related [e.g., families 1 and 8 are half-sibs (13)], sires were treated in the analysis as unrelated on the basis of simulations showing only minor effects on the power of QTL detection in a granddaughter design when taking family relationships into account (20).
Each son's daughter yield deviations (DYDs) for seven traits were extracted from the November 1997 U.S. national Holstein evaluation and used for the analysis. The seven traits are DYDs for milk yield (MY), fat yield (FY), protein yield (PY), fat percentage (FP) and protein percentage (PP), somatic cell score (SCS), and productive herd life (HL). DYDs are unregressed weighted averages of daughter records adjusted for environmental effects and the additive genetic values of the daughters' dams (52). The DYD for percent fat (DYDPF) is calculated as follows: DYDPF = 100 x (MF + DYDF)/(MM + DYDM) - MPC, where DYDF and DYDM are DYD for fat and milk, respectively, and MF, MM, and MPC are mean fat yield, milk yield, and percent fat, respectively, for cows born during the base year, 1990. The DYD for percent protein is calculated similarly, substituting DYD for protein and mean protein yield into the formula. The underlying yield and percentage traits consist of averages for five or fewer lactations adjusted to a twice-a-day milking and a 305-day lactation record. Somatic cell score is an indicator trait for resistance to mastitis, calculated as 3 + log2(cells/100), where cells equals somatic cells per microliter of milk (45). Productive herd life is the number of months in milk (maximum of 10 mo per lactation) until the age of 84 mo (51). The number of individuals with phenotypic records in each family (F) ranged from 58 to 203, with an average of 133, 128, and 134 for the milk production traits, SCS, and HL, respectively (Table 1). The means, standard deviation (
P), and ranges for the DYDs of the sons included in the study are given in Table 2. Estimated cow breeding values for the Israeli sample were calculated from an animal model evaluation (56), base year 1990. For statistical analysis, the seven traits were considered as uncorrelated based upon a previous study that found only marginal differences in the number of significant effects using significance thresholds calculated on the basis of four canonical traits compared with seven independent traits (60). This approach is also supported by the results of Weller and co-workers (59), who found minimal differences in the empirical type I error rates for detecting QTL among the seven correlated traits evaluated and the theoretical type I error rates for seven uncorrelated traits.
Genotyping
DNA was extracted from semen, and PCR reactions were carried out using fluorescence-labeled primers or by incorporating [
-32P]dCTP into PCR products (22, 31, 39). Either frozen whole milk samples (1 µl per PCR reaction) or DNA isolated from whole blood was used as the template for genotyping daughters of the nine Israeli sires for the marker D14S31. For the genome scan, 174 microsatellite markers were selected at ~20-cM intervals on all 29 autosomes using published cattle linkage maps as a guide for intermarker distances (6, 7, 10, 23, 31). For the fluorescent markers, actual allele sizes were scored for all individuals. The radioactively labeled alleles were coded according to size. Assuming a 3,000-cM genome, these 174 markers provide coverage of greater than 85 percent of the cattle genome between flanking markers. The average number of markers per chromosome was six, with an average spacing of 20.8 cM (Table 3). The number of markers typed per grandsire ranged from 84 to 111 (
= 101, or 3.48 markers per chromosome), which corresponds to a grandsire-specific genome coverage of 5877%. The mean number of informative families for each marker was 4.5.
Statistical Methods
Analysis of variance.
Analysis of variance (ANOVA) for each marker and trait combination was performed using the PROC GLM function of SAS (SAS Institute, Cary, NC) employing the model Yijk = Gi + Mij + eijk,where Yijk is the DYD for son k of grandsire i that received microsatellite allele j, Gi is the effect of grandsire i, Mij is the effect of allele j in grandsire family i, and eijk is the random residual for each son's DYD. The DYDs were weighted by the reliabilities of their evaluations. Reliability, a measure of accuracy of an evaluation, is the squared correlation of an animal's predicted and true transmitting abilities (52). Sons with a reliability less than 0.5 for a trait were deleted from the data set. Analysis was performed for the combined data from all the families and for individual sire families for each marker and trait, testing for significance of the Mij effect. Under the null hypothesis of no segregating QTL, the ratio of the mean squares of the Mij effect to the residual mean squares should have a central F distribution. This is by definition a two-tailed test. Any segregating QTL should increase the F value regardless of the direction of effects. A normal distribution of residuals is assumed, but this assumption is not completely true because the sires with semen are a selected sample. However, F tests are robust to minor deviations from normality, and this was confirmed for our sample by permutation. The empirical distribution of the F statistic was very close to the theoretical distribution.
Genome scanning was performed in two stages. In the initial screen, a subset of 70 sons were genotyped in F1F7, inclusive; all sons were genotyped in F8. If a single-family contrast with P < 0.05 was obtained, the remaining sons in those families were genotyped. For the daughter design analysis, Yijk is the breeding value for daughter k of sire i that received microsatellite allele j, Gi is the effect of sire i, Mij is the effect of allele j in sire family i, and eijk is the random residual for each daughter breeding value.
Multimarker regression analysis.
A linkage map based on the DBDR families was constructed using the BUILD option of CRIMAP 2.4 (19). Genome coverage was calculated as the sum of all marker intervals measured in Kosambi centimorgans (Kosambi cM; 26). Linkage groups were constructed for all 29 autosomes with a total map length of 2,551 Kosambi cM (Table 3). Chromosomes containing putative QTL from the ANOVA (P < 0.01) were analyzed for effects in all families as well as in individual families using a multimarker regression program (49). The contribution of each son was weighted as described by Spelman et al. (49) according to the number of daughters contributing to the DYD. Heritabilities of 0.25 for yield traits, 0.5 for percentage traits, and 0.1 for SCS (45) and HL (51) were used.
Calculation of critical values.
The statistical power of detecting segregating QTL is affected by sample size, the genetic distance between the markers, and the QTL and the QTL effect size (57). Because neither QTL location nor effect size are known a priori, rejecting putative QTL at too stringent a level defeats the purpose of a preliminary genome scan (48). Therefore, three significance thresholds are reported for the ANOVA: genomewide, suggestive, and nominal. For genomewide significance and suggestive threshold levels, the type I error rate was calculated using the equation suggested by Lander and Kruglyak (28), assuming a 3,000-cM male genome. To account for the analysis of multiple traits, the obtained
values were divided by 7 (the number of traits). Weller et al. (59) used a permutation analysis to show that the empirical genomewide error rate for the seven traits analyzed in this study was similar to the genomewide error rate computed for seven independent traits. Therefore, the number of independent traits could not be reduced, even though there are strong correlations among the studied traits. Using the formula of Lander and Kruglyak (28), a marker-trait association with a genomewide error rate of 0.05 must exceed P = 8.1 x 10-6 to be significant. The type I error rate for a suggestive association, which is the expected number of type I errors in the experiment when the null hypothesis of no QTL segregating is true, must exceed P = 2.7 x 10-4. The values can be adjusted for the analysis of marker effects within families by dividing
by 56 (8 families x 7 traits). For within-family analysis, the threshold increases to 8.9 x 10-6 for a significant association and 3.4 x 10-5 for a suggestive association. For completeness, marker effects with unadjusted P values of P < 0.01 for the joint analysis and P < 0.001 for individual families are also reported. Additional marker effects with higher type I error rate (P < 0.10 for multiple families and P < 0.05 for individual families) can be found on the World Wide Web at http://cagst.animal.uiuc.edu.
The false discovery rate (FDR) is the expected proportion of true null hypotheses within the class of rejected null hypotheses for each marker-trait association (9). The expected FDR (q) was calculated as q = mP(i)/i; where i is the number of the null hypothesis ranked by descending P and m is the total number of analyses [7 traits x 174 markers = 1,218 for multiple-family analyses and 7 traits x 793 typed families = 5,551 for individual family analyses (58)]. The fact that the sons are selected will reduce the magnitude of the effects detected, for those traits under selection, and therefore the power of the experiment.
Critical values for the multimarker regression analysis were calculated by permuting the phenotypic data as described by Churchill and Doerge (11). The DYDs with their weighting factors were shuffled 10,000 times among individuals within each sire family. Test statistics were calculated along the chromosome for each permutation using multimarker regression analysis. An empirical distribution of test statistics was created by ranking the test statistics obtained from each permutation in descending order. The 5% genomewide threshold values for analysis of all families jointly and for individual families were set equal to the 500th (10,000 x 0.05) largest test statistic after adjusting the test statistics with a correction factor equivalent to the Bonferroni correction for multiple testing (49). The genomewide critical values for the analysis of all families jointly account for testing of 7 traits and 29 chromosomes, whereas critical values used for analysis within families account for 7 traits, 29 chromosomes, and 8 families. Critical values were calculated independently for each chromosome to account for differing chromosome lengths, as performed by Spelman and co-workers (49). The type I error for a suggestive association was determined from the equation
= 1/203, where 203 equals the number of independent comparisons (29 chromosomes x 7 traits) and set equal to the 49th value of the 10,000 unadjusted ranked test statistics obtained from the permuted data sets (60). Suggestive significance thresholds were used only for the analysis of data from the joint analysis of all families.
Calculation of 95% confidence intervals.
Empirical confidence intervals (CI) for the location of QTL with genomewide or suggestive significance were determined by the bootstrapping method as described by Visscher et al. (54). For each bootstrap sample, 1,067 observations (consisting of both genotypic and phenotypic values) were drawn with replacement from the pool of original observations, keeping family size consistent in the analysis. Subsequently, multimarker regression analysis of 200 data sets was performed. The map position of the highest test statistic for each trait was retained from each data set. Empirical 90% and 95% CI were determined from the distribution of the retained map positions (top and bottom 5 and 2.5 percentiles).
 |
RESULTS
|
---|
Marker Effects on Milk Production Traits
Analysis of all families jointly using ANOVA identified marker effects on 7 chromosomes (BTA1, -2, -3, -5, -7, -14, and -29) that exceeded the genomewide, suggestive, or nominal threshold for QTL effects (Table 4). Putative QTL affecting MY and PY were found on BTA21 in a single family (q = 0.3886 and q = 0.222, respectively) but did not reach the nominal threshold in the joint analysis of all families (Table 5). The FDR was less than 0.36 for the 34 nominally significant marker-trait associations (P < 0.01), thus indicating that ~22 (34 x 0.64) of the reported effects are genuine (Table 4).
View this table:
[in this window]
[in a new window]
|
Table 4. Marker effects with genomewide, suggestive, and nominal significance levels (P<0.01) in the joint analysis of all families
|
|
View this table:
[in this window]
[in a new window]
|
Table 5. Markers with effects (P<0.001) in individual families that were not detected in the joint analysis of all families
|
|
Large effects on FP, exceeding the 5% genomewide or suggestive thresholds, were detected on BTA3 and BTA14 in both the ANOVA and multimarker regression analysis (Tables 4 and 6). The effect on FY associated with BTA14 exceeded the suggestive significance threshold in the ANOVA analysis (q = 0.0069) and the multimarker regression analysis. The suggestive significance threshold was also reached for the effects on BTA3 on PY and PP using the multimarker regression analysis.
View this table:
[in this window]
[in a new window]
|
Table 6. Effects exceeding genomewide or suggestive significance thresholds from the joint analysis of all families using multimarker regression analysis
|
|
The QTL on BTA3 and BTA14 were analyzed further within individual families using multimarker regression analysis. The QTL for MY, FY, FP, and PP were mapped between the markers D3S21 and D3S34 in F1, with the highest test statistic at 34 ± 1 cM (Fig. 1). The 95% CI for the QTL influencing FP spans 14 cM proximal to 26 cM distal of this map position. Analysis of F2 suggests the presence of a QTL affecting FY within the same marker interval as in F1 (Fig. 1), with a 95% CI ranging from the centromere to position 58 cM. In addition, analysis of F5 indicates a QTL affecting PY that is centromeric to one or more QTL affecting FP (Fig. 1). The 95% CI for this QTL overlaps the CIs obtained for effects on FY in F2 and for FP in F1.

View larger version (18K):
[in this window]
[in a new window]
|
Fig. 1. Multimarker regression analysis of marker effects on milk production traits for chromosomes 3 (BTA3; AC) and 14 (BTA14; D). A: analysis of F1 (BTA3) for milk yield (MY; thick solid line), fat yield (FY; +), protein yield (PY; ), fat percentage (FP; thin solid line), and protein percentage (PP; ). B: analysis of F2 (BTA3) for MY, FY, PY, FP, and PP. C: analysis of F5 (BTA3) for MY, FY, PY, FP, and PP. D: analysis of F4 and F5 combined (BTA14) for MY, FY, and FP. Arrows indicate location of markers; horizontal line represents the 5% genomewide threshold for the trait with the highest test statistic; vertical lines represent bounds of the 95% confidence intervals.
|
|
For comparison of effects associated with marker alleles in individual families, the DYD least squares means were computed among sons with alternatively inherited grandsire alleles (Table 7). The inheritance of D3S34 allele 3 in F1 is associated with an increase in the average sons' DYD of 112 kg MY and a 3.0-kg decrease in FY. These opposing effects cause a 0.07% (0.61 DYD within-family phenotypic SDs) decrease in the FP DYD among sons inheriting allele 3 compared with allele 2. Allele 3 of D3S34 is also associated with a 0.06% decrease in DYD for FP among sons in F8 (grandsires 1 and 8 are half-siblings). Alternatively, among the sons of grandsire 5, who is unrelated to grandsires 1 and 8 for at least four generations, the inheritance of D3S34 allele 3 is associated with a decrease of 146 kg MY (data not shown).
Another QTL with very large effect on FP was mapped to the centromeric end of BTA14. Results of the multimarker regression analysis of combined data for F4 and F5 suggest that the most likely position of the FP QTL is at chromosome position 2 cM, between the markers D14S55 and D14S31 (Fig. 1). The bootstrap 95% CI estimated from F4 and F5 data is 11 cM. Inheritance of D14S55 allele 2 correlates with an increase in the average sons' DYD of 8.6 kg FY in F4 (Table 7). The magnitude of this effect on FY in F4 is 0.81 DYD
P, which corresponds to 1.06 DYD
P for FP. The effect size in F5 is +5.8 kg FY (0.59 DYD
P) and -127 kg MY (0.46 DYD
P). Effects associated with the marker D14S26, which maps more than 88 cM from D14S31, give support for a second QTL influencing MY on BTA14 in F2 and F6 (Table 4).
To confirm the QTL for FP on BTA14, 3,264 daughters of nine Israeli Holstein sires were genotyped for the marker D14S31 (data not shown). Of these, 2,802 (86%) daughters with genetic evaluations were informative for the marker. The number of informative daughters per family ranged from 130 to 493. In the joint analysis of nine families, the effect on FP was significant at P < 10-6. The effects on MY and FY were significant at P < 0.05 and P < 0.01, respectively. Only one Israeli Holstein (family 2) had significant allele contrasts for FP (P < 0.0001), MY (P < 0.004), and FY (P < 0.002). The sire of this family had 392 daughters with informative genotypes; the magnitude of the contrast was 0.081 percent fat, at the level of the cows' breeding values (data not shown). The most likely location of the QTL was at D14S31 as determined using single-locus maximum likelihood analysis. However, the 95% CI was greater than 20 cM.
Marker Effects on SCS
No effects reaching genomewide or suggestive significance thresholds were found for SCS using ANOVA or multiple marker regression analysis. However, putative QTL (P < 0.005) were identified on chromosomes 5, 7, 21, 22, 23 and 26 (Tables 4 and 5). At position 26 cM on BTA23, the test statistic obtained using multimarker regression analysis (3.52) was nearest to the suggestive significance threshold (3.77). D23S5, the marker associated with the most significant effect on this chromosome in the ANOVA analysis (q = 0.2494), maps near the cattle major histocompatibility complex (MHC). Both D22S8 and D22S15, located 41 cM apart on BTA22, were associated with effects on SCS (q = 0.2155 and 0.2030, respectively). However, the marker D22S19, located between D22S8 and D22S15, failed to show any association with SCS among the seven families. The effects on SCS range from 0.07 to 0.13 DYD at the marker locations and are equivalent to a 5.09.4% change in the number of somatic cells present in the milk of sons' daughters.
Marker Effects on HL
Putative QTL influencing HL were identified on BTA4, BTA17, and BTA21 using ANOVA (Tables 4 and 5). Effects were found primarily in single families and did not exceed suggestive or genomewide significance thresholds in the ANOVA or the multimarker regression analysis.
 |
DISCUSSION
|
---|
A genome scan was conducted for the purpose of detecting QTL associated with milk production, milk composition, and health traits in the North American Holstein-Friesian population. The resource population consisted of 1,068 sons of eight genetically elite sires. Using the granddaughter design, breeding values estimated from the production and health records of the large granddaughter population were compared among the genotyped sons. Results presented herein represent the first genome scan of the eight largest DBDR families, with more than 85% of the genome saturated at ~20-cM intervals by 174 selected markers. Results of another large study of North American Holstein sires were recently reported (60), but it is unknown whether the grandsire families employed are different from the ones used in the present study. Ashwell and co-workers (25) have used many of the same DBDR families in their studies as reported here, but only a small fraction of the genome was evaluated for QTL and the threshold values were less stringent. Other studies have been performed with different resource populations, some Holstein-Friesian and others belonging to different breeds (1, 12, 18, 29, 32, 49). Despite the diversity of resource populations, experimental approach, and methods of analysis employed, several QTL reported by others were confirmed in the present study (discussed below). Other putative QTL have not previously been detected, such as the effects on SCS associated with D5S22, D22S8-D22S15, and D26S7, the effect on HL associated with D4S22, the effect on PP associated with D2S11, the effects on PY and MY associated with D29S13, and the effect on PY associated with D2S122. All of these QTL were detected at the nominal significance threshold so these results must be interpreted with caution.
A comprehensive approach was implemented for the analysis of our granddaughter design data. First, ANOVA was used for the primary genome scan, using a stepwise genotyping procedure to minimize the amount of genotyping required. Several chromosomes with markers exceeding the suggestive level of statistical significance according to the criteria established by Lander and Kruglyak (28) were then studied further by multimarker regression analysis (49). The FDR (58) was also used as a guide for interpreting results of the ANOVA. In reporting our results (Table 4), we chose a relaxed significance threshold because many effects detected that were below the recommended significance threshold levels will actually represent genuine QTL. For the FDR we can conclude that ~22 of the 34 marker-trait associations are genuine, despite the fact that only 9 exceeded the suggestive threshold for significance. These putative QTL may be confirmed in future studies, so we feel that it is important to report such effects even if they do not meet the most stringent statistical criteria. It is necessary to have larger resource populations and/or better marker coverage to confirm results for markers with effects that do not reach conservative thresholds for statistical significance.
In theory, the analysis of individual families should reveal QTL not detected in a joint analysis of all families because heterozygosity at QTL may be low because of selection (17). However, significance thresholds become extremely high for the analysis of individual families because it is necessary to account for the additional comparisons. This problem is one of the paradoxes in QTL mapping. In the present study, marker effects were analyzed by ANOVA in all families jointly as well as in individual families. Although the results were generally consistent for both analyses, additional putative QTL not reaching the nominal significance threshold in the joint analysis were detected on chromosomes 17, 21, and 26 in individual families (Table 5). As for the joint analysis, additional sire families or direct examination of daughter records will be necessary to determine if these are genuine QTL or type I errors. For those effects that exceeded the nominal threshold level in the joint analysis, low grandsire-specific marker coverage prevented the use of the multimarker regression analysis in individual families that contributed significantly to the model. However, for chromosomes with adequate numbers of markers, such as BTA3 and BTA14, the multimarker regression analysis within families was useful for interpreting data that may be confounded in the joint analysis by low grandsire-specific marker coverage or QTL genotype.
A QTL with a very large effect on FP was linked to the marker D14S31 using the granddaughter design and confirmed in a daughter design study in an independent population. These results support our preliminary analysis of the DBDR families and daughter design data (41). In two informative families, the FP QTL was mapped to a position 2 cM distal to D14S55, with a bootstrap 95% CI of 11 cM (Fig. 1). There was evidence for a second QTL affecting MY at the telomeric end of BTA14, but significance did not exceed the threshold level for suggestive linkage. In further support of our findings, two recent studies have confirmed a major QTL for FP near the centromere of BTA14 (12, 60). Combining the results of these studies would be extremely valuable for fine mapping this QTL, thus facilitating selection/identification of candidate genes for FY and MY and a reasonably sized interval for physical mapping of the region.
Significant marker-effects for several milk production traits were found on BTA3 (Table 4). The peak test statistic was reached at position 22 cM for FP, whereas the peak statistics for the effects on PY and PP were reached at positions 16 cM and 3 cM, respectively. Within-family multimarker regression analysis suggests that more than one QTL may be present on BTA3, although F1, F2, and F5 all indicate QTL for MY or FY at similar locations (Fig. 1). If these effects represent multiple QTL then they are most likely to be tightly linked because the peak test statistics map within a 10-cM region. Zhang and co-workers (60) and Lipkin et al. (29) also reported a QTL for PP on the centromeric end of BTA3. The study by Zhang and co-workers found an effect on PP using the same marker (D3S34 or TGLA263). However, in that study D3S34 was the terminal marker in the linkage group, whereas in our study it was located 41 cM from the terminus of the linkage group, thus accounting for the difference in reported chromosome location. These data add important new positional information for this QTL.
A more detailed analysis of allele effects of markers on BTA3 in F1, F2, and F5 suggests that there are three QTL located in the region flanked by D3S32 and D3S34, one each for MY, FY, and PY. In grandsire 1, the QTL alleles for MY and FY appear to be in repulsion, i.e., the high MY QTL allele is on the same haplotype with the low FY QTL allele and vice versa (Table 7, data not shown). This produces the very large effect on FP DYDs observed among the sons and a numerical (but nonsignificant) difference in PP. Further evidence for three QTL is observed in F2 and F5. In F2, the grandsire appears heterozygous for the FY QTL only (Table 7). Grandsire 5 appears heterozygous for the PY QTL, the MY QTL and the FY QTL, but the direction of the effects is the same, resulting in a nonsignificant contrast by ANOVA for FP (data not shown). By comparison, there appears to be only one QTL in the vicinity of D14S55 and D14S31 that affects FY (but not MY), thus contributing to the significant effect on FP. Examination of the QTL affecting percentage traits on BTA3 and BTA14 dramatically demonstrates the importance of fully understanding the nature of the traits under investigation. The percentage traits are derived from the product of MY and the yield of a solid. Thus, as observed for the QTL on BTA3, when a QTL allele associated with a positive effect on MY is in linkage phase with a QTL allele associated with a negative effect on a milk solid, a dramatic effect on percentage can result, even if the two QTL have small effects. Moreover, the mapping of QTL for percentage traits may not reflect the true positions of the underlying QTL that influence the solids or the water content of milk. An alternative explanation for our data is that there is one QTL with multiple alleles, each affecting one or more traits. Given that the molecular mechanisms regulating water content of milk and secretion of fat or protein are likely to be different, we feel that pleiotropy is less likely as an explanation for our observations.
SCS, an indicator trait for mastitis resistance, was implemented into sire evaluations in 1994 (43). Identification of QTL affecting SCS could be of great practical significance because mastitis is the disease responsible for the largest economic loss in the dairy industry, with annual losses estimated at more than $2 billion in the United States alone (21). Fewer years of selection and the lower heritability of this trait make SCS an ideal target for marker-assisted selection or gene-based therapeutic strategies. Although we detected no marker effects that reached the genomewide or suggestive significance thresholds, putative QTL affecting SCS were identified on chromosomes 5, 7, 21, 22, 23, and 26. Multimarker regression and ANOVA suggest the presence of more than one QTL influencing SCS on BTA22, one centromeric to D22S8 and one telomeric to D22S15. The effect sizes on SCS ranged from 0.07 to 0.13 DYD SCS units, which corresponds to a 5.09.4% change in the number of somatic cells present in milk. In support of our findings, several groups have detected putative QTL influencing mastitis or SCS within or near the MHC on BTA23 (15, 25, 36, 38, 44, 50, 55). Ashwell et al. (4) also detected a QTL for SCS linked to microsatellite markers on BTA23 in the DBDR families. Our results and those of Ashwell and co-workers offer a rare example of confirmation in the same resource population. Such results are important because they rule out experimental errors, such as sample switching and typing errors, as sources of the observed differences. Explanations for discrepancies between the studies of Ashwell and ours include differences in significance threshold levels and differences in which DBDR families were studied. Along these lines, Ashwell and co-workers (4, 5) reported markers on BTA21 to be associated with MY and PY. In the present study, D21S27 is associated with an effect on MY in F3 (Table 5). Similar results for SCS were obtained within individual families for markers on BTA26 [(5), see Table 5].
Marker effects exceeding genomewide or suggestive significance were not found for the HL trait. This result is not surprising given that the heritability for HL is quite low and the trait is a composite of a number of production, conformation, and health traits. However, a putative QTL for HL was found at a similar position on chromosome 17 as that reported by Zhang et al. (60) in North American Holstein-Friesians. The FDR for each of the marker effects on HL detected by ANOVA also suggests that more extensive studies may be warranted.
In addition to the marker effects described above, other QTL confirmed in this study include QTL for MY and PY on BTA21 (40) and possibly a QTL for PY [detected as PP (29)]. There are several other putative QTL detected in this study that have not been confirmed, as well as QTL detected in other studies not confirmed here. Follow-up studies are necessary to distinguish which effects are authentic. As discussed above, detection of QTL using the granddaughter design will depend in part on effect size, genetic distance between the marker and QTL, number of QTL alleles, and maternal contributions (57). Therefore, for most studies reported to date, marker density must be increased on a grandsire-by-grandsire basis. Fortunately, the total number of QTL for economically important traits detected in dairy cattle is not very large (<50). This will permit efficient strategies for confirmation and fine mapping in more recent generations of elite sires.
Our results provide another example of the power of the granddaughter design for the detection of QTL in dairy cattle. With a number of QTL in dairy cattle now confirmed it is important to gain a better understanding of how much these QTL contribute to the phenotypic variation at the population level. Evaluation of descendants of the DBDR grandsires in the extant population will permit the assessment of how useful the QTL reported here will be for marker-assisted selection. Future studies will focus on more precise mapping and identification of candidate genes for these QTL.
 |
ACKNOWLEDGMENTS
|
---|
We thank Cheryl Green, Shane Cary, and Robin Everts for technical assistance, Dr. Richard Spelman for his QTL analysis program, Dr. Steve Kappes at the United States Department of Agriculture (USDA) Meat Animal Research Center for marker information, and Pim Kuurman for assistance with computer programming. We also thank Atlantic Breeders Cooperative, CIAQ, Eastern AI, Excelsior Farms, Landmark Genetics, NOBA, Select Sires, TriState Breeders and 21st Century Genetics for contributions of semen to the DBDR.
This research was supported in part by the United States-Israel Binational Agricultural Research and Development Fund (BARD), project no. IS-238394C; the Israel Milk Marketing Board; USDA Regional Project NC209 (IL Project No. 35308), and USDA National Research Support Project-8.
Address for reprint requests and other correspondence: H. A. Lewin, Dept. of Animal Sciences, Univ. of Illinois at Urbana-Champaign, 206 Edward R. Madigan Laboratory, Urbana, IL 61801 (E-mail: h-lewin{at}ux1.cso.uiuc.edu).
 |
FOOTNOTES
|
---|
Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).
 |
REFERENCES
|
---|
-
Arranz, J. J., W. Coppieters, P. Berzi, N. Cambisano, B. Grisart, L. Karim, F. Marcq, L. Moreau, C. Mezer, J. Riquet, P. Simon, P. Vanmanshoven, D. Wagenaar, and M. Georges. A QTL affecting milk yield and composition maps to bovine chromosome 20: a confirmation. Anim. Genet. 29: 107115, 1998.[Medline]
-
Ashwell, M. S., Y. Da, P. M. VanRaden, C. E. Rexroad, Jr., and R. H. Miller. Detection of potential loci affecting conformation type traits in an elite U.S. Holstein population using microsatellite markers. J. Dairy Sci.81: 11201125, 1998.[Abstract/Free Full Text]
-
Ashwell, M. S., Y. Da, C. P. Van Tassell, P. M. VanRaden, R. H. Miller, and C. E. Rexroad, Jr. Detection of putative loci affecting milk production and composition, health, and type traits in a United States Holstein population. J. Dairy Sci. 81: 33093314, 1998.[Abstract/Free Full Text]
-
Ashwell, M. S., C. E. Rexroad, Jr., R. H. Miller, and P. M. VanRaden. Mapping economic trait loci for somatic cell score in Holstein cattle using microsatellite markers and selective genotyping. Anim. Genet. 27: 235242, 1996.[Medline]
-
Ashwell, M. S., C. E. Rexroad, Jr., R. H. Miller, P. M. VanRaden, and Y. Da. Detection of loci affecting milk production and health traits in an elite U.S. Holstein population using microsatellite markers. Anim. Genet. 28: 216222, 1997.
-
Barendse, W., S. M. Armitage, L. M. Kossarek, A. Shalom, B. W. Kirkpatrick, A. M. Ryan, D. Clayton, L. Li, H. L. Neibergs, N. Zhang, W. M. Grosse, J. Weiss, P. Creighton, F. McCarthy, M. Ron, A. J. Teale, R. Fries, R. A. McGraw, S. S. Moore, M. Georges, M. Soller, J. E. Womack, and D. J. S. Hetzel. A genetic linkage map of the bovine genome. Nat. Genet. 6: 227235, 1994.[Medline]
-
Barendse, W., D. Vaiman, S. J. Kemp, Y. Sugimoto, S. M. Armitage, J. L. Williams, H. S. Sun, A. Eggen, M. Agaba, S. A. Aleyasin, M. Band, M. D. Bishop, J. Buitkamp, K. Byrne, F. Collins, L. Cooper, W. Coppieters, B. Denys, R. D. Drinkwater, K. Easterday, C. Elduque, S. Ennis, G. Erhardt, L. Ferretti, N. Flavin, Q. Gao, M. Georges, R. Gurung, B. Harlizius, G. Hawkins, J. Hetzel, T. Hirano, D. Hulme, C. Jorgensen, M. Kessler, B. W. Kirkpatrick, B. Konfortov, S. Kostia, C. Kuhn, J. A. Lenstra, H. Leveziel, H. A. Lewin, B. Leyhe, L. Lil, I. M. Buriel, R. A. McGraw, R. J. Miller, D. E. Moody, S. S. Moore, S. Nakane, I. J. Nijman, I. Olsaker, D. Pomp, A. Rando, M. Ron, A. Shalom, A. J. Teale, U. Thieven, B. G. D. Urquhart, D. I. Vage, A. Van de Weghe, S. Varvio, R. Velmala, J. Vilkki, R. Weikard, C. Woodside, J. E. Womack, M. Zanotti, and P. Zaragoza. A medium density genetic linkage map of the bovine genome. Mamm. Genome. 8: 2128, 1997.[Medline]
-
Beckmann, J. S., and M. Soller. Detection of linkage between marker loci affecting quantitative traits in crosses between segregating populations. Theor. Appl. Genet. 76: 228236, 1988.
-
Benjamini, Y., and Y. Hochberg. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Statist. Soc. B. 57: 289300, 1995.
-
Bishop, M. D., S. M. Kappes, J. W. Keele, R. T. Stone, S. L. F. Sunden, G. Hawkins, S. S. Toldo, R. Fries, M. D. Grosz, J. Yoo, and C. W. Beattie. A genetic linkage map for cattle. Genetics 136: 619639, 1994.[Abstract/Free Full Text]
-
Churchill, G. A., and R. W. Doerge. Empirical threshold values for quantitative trait mapping. Genetics 138: 963971, 1994.[Abstract/Free Full Text]
-
Coppieters, W., J. Riquet, J. Arranz, P. Berzi, N. Cambisano, B. Grisart, L. Karim, F. Marcq, L. Moreau, C. Nezer, P. Simon, P. Vanmanshoven, D. Wagenaar, and M. Georges. A QTL with major effect on milk yield and composition maps to bovine chromosome 14. Mamm. Genome. 9: 540544, 1998.[Medline]
-
Da, Y., M. Ron, A. Yanai, M. Band, J. I. Weller, G. R. Wiggans, and H. A. Lewin. The Dairy Bull DNA Repository: a resource for mapping quantative trait loci. Proc. 5th World Congr. Genet. Appl. Livest. Prod. 21: 229232, 1994.
-
de Koning, G. J., and J. I. Weller. Efficiency of direct selection on quantitative trait loci for a two-trait breeding objective. Theor. Appl. Genet. 88: 669677, 1994.
-
Dietz, A. B., N. D. Cohen, L. Timms, and M. E. Kehrli, Jr. Bovine lymphocyte antigen class II alleles as risk factors for high somatic cell counts in milk of lactating dairy cows. J. Dairy Sci. 80: 406412, 1997.[Abstract/Free Full Text]
-
Geldermann, H. Investigations on inheritance of quantitative characters in animals by gene markers. I. Methods. Theor. Appl. Genet. 46: 319330, 1975.
-
Georges, M., D. Nielsen, M. Mackinnon, A. Mishra, R. Okimoto, A. T. Pasquino, L. S. Sargeant, A. Sorensen, M. R. Steele, X. Zhao, J. E. Womack, and I. Hoeschele. Mapping quantitative trait loci controlling milk production in dairy cattle by exploiting progeny testing. Genetics 139: 907920, 1995.[Abstract/Free Full Text]
-
Gomez-Raya, L., H. Klungland, D. L. Vage, L. Olsaker, E. Fimland, G. Klemetsdal, K. Renningen, and S. Lien. Mapping QTL for milk production traits in Norwegian cattle. Proc. 6th World Cong. Genet. Appl. Livest. Prod. 26: 429432, 1998.
-
Green, P., K. Falls, and S. Crooks. Documentation for CRI-MAP (version 2.4). St. Louis, MO: Washington University School of Medicine, 1990.
-
Grignola, F. E., I. Hoeschele, Q. Zhang and G. Thaller. Mapping quantitative trait loci in outcross populations via residual maximum likelihood. II. A simulation study. Genet. Sel. Evol. 28: 491504, 1996.
-
Harmon, R. J. Physiology of mastitis and factors affecting somatic cell counts. J. Dairy Sci. 77: 21032112, 1993.
-
Heyen, D. W., J. E. Beever, Y. Da, R. E. Evert, C. Green, S. R. E. Bates, J. S. Ziegle, and H. A. Lewin. Exclusion probabilities of 22 bovine microsatellite markers in fluorescent multiplexes for semiautomated parentage testing. Anim. Genet. 28: 2127, 1997.[Medline]
-
Kappes, S. M., J. W. Keele, R. T. Stone, T. S. Sonstegard, T. P. Smith, N. L. Lopez-Corrales, and C. W. Beattie. A second-generation linkage map of the bovine genome. Genome Res. 7: 235249, 1997.[Abstract]
-
Kashi, Y. E., E. Hallerman, and M. Soller. Marker-assisted selection of candidate bulls for progeny testing programmes. Anim. Prod. 51: 6374, 1990.
-
Kelm, S. C., J. C. Detilleux, A. E. Freeman, M. E. Kehrli, Jr., A. B. Dietz, L. K. Fox, J. E. Butler, I. Kasckovics, and D. H. Kelley. Genetic association between parameters of inmate immunity and measures of mastitis in periparturient Holstein cattle. 80: J. Dairy Sci. 17671775, 1997.
-
Kosambi, D. D. The estimation of map distance from recombination values. Ann. Eugenics. 12: 172175, 1944.
-
Lander, E. S., and D. Botstein. Mapping mendelian factors underlying quantitative traits using RFLP linkage maps. Genetics 121: 185199, 1989.[Abstract/Free Full Text]
-
Lander, E., and L. Kruglyak. Genetic dissection of complex traits: guidelines for interpreting and reporting linkage results. Nat. Genet. 11: 241247, 1995.[Medline]
-
Lipkin, E., M. O. Mosig, A. Darvasi, E. Ezra, A. Shalom, A. Friedmann, and M. Soller. Quantitative trait locus mapping in dairy cattle by means of selective milk DNA pooling using dinucleotide microsatellite markersanalysis of milk protein percentage. Genetics 49: 15571567, 1998.
-
Liu, B. Computational tools for study of complex traits. In:Molecular Dissection of Complex Traits, edited by A. H. Paterson. Salem, MA: CRC, 1998.
-
Ma, R. Z., J. E. Beever, Y. Da, C. A. Green, I. Russ, C. Park, D. W. Heyen, R. E. Everts, S. R. Fisher, K. M. Overton, A. J. Teale, S. J. Kemp, H. C. Hines, G. Guérin, and H. A. Lewin. A male linkage map of the cattle (Bos taurus) genome. J. Hered. 87: 261271, 1996.[Abstract]
-
Maki-Tanila, A., D. J. de Koning, K. Elo, S. Moisio, R. Velmala, and J. Vilkki. Mapping of multiple quantitative trait loci by regression in half sib designs. Proc. 6th World Cong. Genet. Appl. Livest. Prod. 26: 269272, 1998.
-
Mackinnon, M. J., and J. I. Weller. Methodology and accuracy of estimation of quantitative trait loci parameters in a half-sib design using maximum likelihood. Genetics 141: 755770, 1995.[Abstract/Free Full Text]
-
Meuwissen, T. H. E., and J. A. M. van Arendonk. Potential improvements in rate of genetic gain from marker-assisted selection in dairy cattle. J. Dairy Sci. 75: 16511659, 1992.[Abstract/Free Full Text]
-
Neimann-Sorensen, A., and A. Robertson. The association between blood groups and several production characters in three Danish cattle breeds. Acta. Agr. Scand. 11: 163196, 1961.
-
Odgeirsson, O., S. P. Simpson, A. L. G. Morgan, D. S. Ross, and R. L. Spooner. Relationship between major histocompatibility complex (BoLA), erythrocyte markers and susceptibility to mastitis in Icelandic cattle. Anim. Genet. 19: 1116, 1988.[Medline]
-
Phillips, P. C. From complex traits to complex alleles. Trends Genet. 15: 68, 1999.[Medline]
-
Reinsch, N., N. Xu, H. Thomsen, C. Looft, E. Kalm, S. Grupe, C. Kuhn, M. Schwerin, B. Leyhe, S. Hiendleder, G. Erhard, I. Medjugorac, I. Russ, M. Forster, B. Brenig, R. Reents, and G. Averdunk. First results on somatic cell count loci from the ADR bovine mapping project. Proc. 6th World Cong. Genet. Appl. Livest. Prod. 26: 426428, 1998.
-
Ron, M., M. Band, A. Wyler, and J. I. Weller. Unequivocal determination of sire allele origin for multialleleic microsatellites when only sire and progeny are genotyped. Anim. Genet. 24: 171176, 1993.[Medline]
-
Ron, M., M. Band, A. Yanai, and J. I. Weller. Mapping quantitative trait loci with DNA microsatellites in a commercial dairy cattle population. Anim. Genet. 25: 259264, 1994.[Medline]
-
Ron, M., D. W. Heyen, J. I. Weller, M. Band, M. E. Feldmesser, H. Pasternak, Y. Da, G. R. Wiggans, P. M. VanRaden, E. Ezra, and H. A. Lewin. Detection and analysis of a locus affecting milk concentration in the U.S. and Israeli dairy cattle populations. Proc. 6th World Cong. Genet. Appl. Livest. Prod. Armidale, Australia 26: 422425, 1998.
-
Schafer, A. J., and J. R. Hawkins. DNA variation and the future of human genetics. Nat. Biotech. 16: 3339, 1998.[Medline]
-
Schutz, M. M. Genetic evaluation of somatic cell scores for United States dairy cattle. J. Dairy Sci. 77: 21132129, 1994.[Abstract/Free Full Text]
-
Sharif, S., B. A. Mallard, B. N. Wilkie, J. M. Sargeant, H. M. Scott, J. C. Dekkers, and K. E. Leslie. Associations of the bovine major histocompatibility complex DRB3 (BoLA-DRB3) alleles with occurrence of disease and milk somatic cell score in Canadian dairy cattle. Anim. Genet. 29: 185193, 1998.[Medline]
-
Shook, G. E., and M. M. Schutz. Selection on somatic cell score to improve resistance to mastitis in the United States. J. Dairy Sci. 77: 648658, 1994.[Abstract/Free Full Text]
-
Soller, M. The use of loci associated with quantitative effects in dairy cattle improvement. Anim. Prod. 27: 133139, 1978.
-
Soller, M., and A. Genizi. The efficiency of experimental designs for the detection of linkage between a marker locus and a locus affecting a quantitative trait in segregating populations. Biometrics 34: 4755, 1978.
-
Southey, B. R., and R. L. Fernando. Controlling the proportion of false positives among significant results in QTL detection. Proc. 6th World Cong. Genet. Appl. Livest. Prod. Armidale, Australia. 26: 221224, 1998.
-
Spelman, R. J., W. Coppieters, L. Karim, J. A. M. van Arendonk and H. Bovenhuis. Quantitative trait loci analysis for five milk production traits on chromosome six in the Dutch Holstein-Friesian population. Genetics 144: 17991808, 1996.[Abstract/Free Full Text]
-
Starkenburg, R. J., L. B. Hansen, M. E. Kehrli, Jr., and H. Chester-Jones. Frequencies and effects of alternative DRB3.2 alleles of bovine lymphocyte antigen for Holsteins in milk selection and control lines. J. Dairy Sci. 80: 34113419, 1997.[Abstract/Free Full Text]
-
VanRaden, P. M., and E. J. H. Klaaskate. Genetic evaluation of length of productive life including predicted longevity of live cows. J. Dairy Sci. 76: 27582764, 1993.[Abstract/Free Full Text]
-
VanRaden, P. M., and G. R. Wiggans. Derivation, calculation and use of National Animal Model Information. J. Dairy Sci. 74: 27372746, 1991.[Abstract/Free Full Text]
-
Vilkki, H. J., D. J. de Koning, K. Elo, R. Velmala, and A. Maki-Tanila. Multiple marker mapping of quantitative trait loci of Finnish dairy cattle by regression. J. Dairy Sci. 80: 198204, 1997.[Abstract/Free Full Text]
-
Visscher, P. M., R. Thompson, and C. S. Haley. Confidence intervals in QTL mapping by bootstrapping. Genetics. 143: 10131020, 1996.[Abstract/Free Full Text]
-
Weigel, K. A., A. E. Freeman, M. E. Kehrli, Jr., M. J. Stear, and D. H. Kelley. Association of class I bovine lymphocyte antigen complex alleles with health and production traits in dairy cattle. J. Dairy Sci. 73: 25382546, 1990.[Abstract/Free Full Text]
-
Weller, J. I., C. Israel, and E. Ezra. A simple procedure for obtaining approximate interim cow solutions from an animal model. J. Dairy Sci. 77: 11261131, 1994.[Abstract/Free Full Text]
-
Weller, J. I., Y. Kashi, and M. Soller. Power of daughter and granddaughter designs for determining linkage between marker loci and quantitative trait loci in dairy cattle. J. Dairy Sci. 73: 25252537, 1990.[Abstract/Free Full Text]
-
Weller, J. I., J. Z. Song, D. W. Heyen, H. A. Lewin, and M. Ron. A new approach to the problem of multiple comparisons in the genetic dissection of complex traits. Genetics 150: 16991706, 1998.[Abstract/Free Full Text]
-
Weller, J. I., J. Z. Song, Y. Ronin, and A. B. Korol. Experimental designs and solutions to multiple trait comparisons. Anim. Biotech. 8: 107122, 1997.
-
Zhang, Q., D. Boichard, I. Hoeschele, C. Ernst, A. Eggen, B. Murkve, M. Pfister-Genskow, L. A. Witte, F. Grignola, P. Uimari, G. Thaller, and M. D. Bishop. Mapping quantitative trait loci for milk production and health of dairy cattle in a large outbred pedigree. Genetics 149: 19591973, 1998.[Abstract/Free Full Text]