Detection of quantitative trait loci affecting non-return rate in French dairy cattle

Animals

In the QTL detection study, we used the extended mapping GDD available in PH, NO and MO breeds (Gautier et al. 2006b). They were composed of 78 half-sib families with more than 25 sons: 47 in PH, 18 in NO and 13 in MO. A total of 4993 progeny tested sires were available with an average of 70 sires per family (from 25 to 240). As shown in Table 1, for the QTL confirmation study, which required additional genotyping, a sub-sample of 41 families was chosen. As described previously (Gautier et al. 2006b), we tried to maximize the number of informative families for the QTL underlying the different traits initially analysed (FF but also production and disease resistance traits) while avoiding the overrepresentation of some closely related families. The 41 families included: 26 PH, 9 NO and 6 MO families. All 14 families (9 PH, 3 NO and 2 MO) from the first QTL programme were included to provide a link with the previous results.

Phenotypic data

For each individual, daughter yield deviation (DYD) and reliabilities were estimated (VanRaden & Wiggans 1991) and used as phenotypes. When reliability was lower than 0.3, the corresponding DYD was removed. DYD for NRR281 was used in the QTL detection study. To improve QTL characterization and detection, two additional NRR were analysed: NRR56 and NRR90. Measuring these two phenotypes should make it easier to distinguish between reproduction events that occur early in gestation and those, which continue to occur during the whole gestation period. The estimated correlations between NRR56 and NRR90 NRR56 and NRR281 and NRR90 and NRR281 were 0.91, 0.81 and 0.94, respectively (Guillaume, personal communication). The average DYD reliabilities for NRR56, NRR90 and NRR281 were 0.67, 0.66 and 0.70, respectively.

Genotyping data

For the QTL detection study, we used genotyping data from the 78 half-sib families already available from the French MAS programme (Boichard et al. 2006). As shown in Table 2, the data set consisted of genotypes for two to five microsatellites within 12 genomic regions (39 markers in total), which covered approximately 17% of the bovine genome. Based on QTL detection results, marker density on BTA01, BTA02 and BTA03 was further increased (see Table 3 for details on the different markers). For BTA01 and BTA02, respectively, nine and 11 microsatellites were chosen from the USDA database (http://www.marc.usda.gov/genome) for genotyping, in addition to the four and two markers already available for these two chromosomes from the MAS programme (Table 2). For BTA03, genotypes for four microsatellites were already available from the previous MAS programme (Table 2) and also for 14 microsatellites in PH (Guillaume et al. 2007). These latter markers were genotyped in the nine NO and six MO families. Since the region spanning the first 30 cM on BTA03, corresponding to the most likely QTL position (Guillaume et al. 2007), contained too few markers, seven additional microsatellite markers were selected for genotyping (BMS2904, DIK4193, INRA3001, INRA3012, INRA3013, INRA3015 and URB006).

Among these seven additional new markers, only INRA3001 and URB006 were genotyped in the three breeds as part of a first round of genotyping, the five other markers were subsequently developed and genotyped only in PH because no NRR QTL segregated in the two other breeds (see Table 3 and Results). The markers INRA3001, INRA3012, INRA3013 and INRA3015 were selected and developed from available bovine genome sequence data to target the region of interest according to the Btau_3.1 assembly (see ftp://ftp.hgsc.bcm.tmc.edu/pub/data/Btaurus/fasta/Btau20060815-freeze/ReadMeBovine.3.1.txt). Their expected positions on BTA03 were confirmed by radiation hybrid mapping using standard procedures (Gautier et al. 2003). Selected markers were then genotyped as previously described (Gautier et al. 2006a). Genotypes for 13, 13 and 26 (21 in NO and MO) markers, respectively, were available for BTA01, BTA02 and BTA03. Finally, for each of the three chromosomes, only individuals genotyped for at least 60% of the markers were considered in the analysis. A total of 2925 sons from the three breeds were included in the analysis (Table 1).

Linkage map construction

Linkage maps were constructed using crimap 2.4 software (Green et al. 1990). For each linkage group, the marker order obtained from known linkage maps and the bovine genome assembly (see above) was challenged using the flips option with a four-marker window. The Chrompic option was used to identify unlikely double crossing-overs. Final map distances were estimated using the Haldane mapping function.

QTL analysis

Linkage Analysis (LA) was performed using a multimarker regression approach fitting a one-QTL model (Haley & Knott 1992) with qtlmap software (Boichard et al. 2003) for QTL detection analysis and QTL-express (Seaton et al. 2002) for subsequent QTL confirmation analysis. Within each family, NRR DYD were regressed, at each centimorgan, on the probability of transmission of alternative paternal QTL alleles to their sons. The following model was used:

where Y_ij is the DYD of son j of sire i, s_i is the effect of sire i, p_ij is the probability of inheriting the first allele from sire i for son j, a_i is half of the substitution effect of the QTL carried by the sire i and e_ij is the residual. The three breeds were treated separately.

In the QTL confirmation analysis, the F-ratio test statistic and marker informativity were computed every centimorgan along the chromosome. Both 95 and 99% chromosome-wise significance thresholds were computed based on 10 000 permutations (Churchill & Doerge 1994). Confidence intervals (CI) were estimated using bootstrapping over 10 000 iterations. For each sire, the p-value associated to the null hypothesis of being homozygous for the QTL was computed based on the T-test value at the maximum peak position, assuming it is normally distributed. The rejection threshold was set to 5%. The average allele substitution effect (a) for heterozygous sires was estimated under a bi-allelic QTL and used to approximate the proportion of genetic variance explained by QTL (Var_QTL) according to the formula: Var_QTL = 2pqa² (Falconer 1996), where p and q are the two QTL allele frequencies and 2pq is the frequency of heterozygous sires. As p and q are unknown and assuming Hardy–Weinberg equilibrium (HWE) in the sire population, we approximate 2pq by H: the proportion of heterozygous sires. The proportion of genetic variance explained by the QTL then becomes:

QTL detection study

In the extended PH design (47 half-sib families), we detected three significant (p < 0.01) QTL underlying NRR281 and mapping respectively to BTA01, BTA02 and BTA03 chromosomes and three suggestive (p < 0.05) QTL mapping to BTA06, BTA20 and BTA23. In the NO extended pedigree (18 half-sib families), one significant QTL mapping to BTA01 and one suggestive QTL mapping to the BTA21 chromosome were detected. No QTL was identified in the extended MO pedigree (Table 2). Among the three QTL initially detected in the French QTL programme, which was based on 14 families (Boichard et al. 2003), the QTL on BTA01 was confirmed in the NO and PH breeds. The QTL mapping to BTA21 was only suggestive in NO while the QTL mapping to BTA07 could not be confirmed in any of the three breeds considered. Since only a few markers were genotyped for each chromosome (Table 2), we were not able to provide precise mapping intervals on the whole chromosome.

QTL confirmation study

As detailed in Table 3, on average 1152 informative meioses (from 188 for MB101 to 1888 for DIK2719) were available in PH to construct the linkage maps for BTA01 (13 markers), BTA02 (13 markers) and BTA03 (26 markers). Marker order and distances were in agreement with published data for markers mapping to BTA01 and BTA02 (http://www.marc.usda.gov/genome/). However, for BTA03, the marker order between RM019 and DIK2101 was different from published data (http://www.marc.usda.gov/genome/) but in agreement with Btau_4.0 assembly (ftp://ftp.hgsc.bcm.tmc.edu/pub/data/Btaurus/fasta/Btau20070913-freeze/README.Btau20070913.txt) (Table 3).

In the PH breed, at least one QTL underlying one of the three NRR traits was detected at a 5% significance threshold. No significant QTL was identified for the three different NRR traits in NO and MO. Results for each chromosome are detailed in the following paragraphs.

Chromosome BTA01

One significant QTL underlying NRR90 was detected in PH (p < 0.009). The peak position (at 97 cM on our genetic map) was located near marker BM864 (Figure 1). When considering NRR281, the QTL was confirmed at a slightly lower significance level (p < 0.019) and the peak position was approximately the same as that for NRR90. Based on the T-test value at the maximum peak position, four sires (3442, 3481, 3498 and 7210) were heterozygous (p < 0.05) for both NRR281 and NRR90 while sire 6865 was heterozygous (p < 0.05) only for the QTL underlying NRR90. The average QTL allele substitution effect was estimated as 4.86% NRR90 and the QTL explained 18% of the total genetic variance (Table 4).

Chromosome BTA02

One suggestive QTL (p < 0.047) underlying NRR56 was detected in PH. The peak position was located near BMS778 and RM041 at 70 cM on our genetic map. Although they showed a similar trend, Linkage Analysis (LA) curves obtained for NRR90 and NRR281 did not reach the 5% significance threshold (Figure 2). Six sires were heterozygous (p < 0.05) for the QTL: 3468, 3481, 3502, 5245, 6109 and 6865 (Table 4). The average QTL allele substitution effect was estimated as 3.8% NRR90 and the QTL variance explained 11.5% of the total genetic variance.

A QTL underlying both NRR90 (p < 0.048) and NRR281 (p < 0.034) was detected on BTA02 at a more telomeric position. The peak position was situated near BM2113 at position 115 cM according to our genetic map (Figure 2). Sires 3480, 3514 and 6865 were heterozygous (p < 0.05) for both NRR90 and NRR281 data while sire 3487 heterozygous (p < 0.05) only for NRR90. The average QTL allele substitution effect was estimated as 4.84% NRR90 and the QTL explained 14% of the total genetic variance (Table 4).

Chromosome BTA03

A QTL underlying NRR90 (p < 0.015) was detected on BTA03 in the PH breed. The peak position (25 cM on our linkage map) was near ILSTS096 (Figure 3). When considering NRR281 as a trait, the LA curve displayed a similar trend but did not reach the 5% significance threshold (p < 0.057). Sires 3514, 4442, 4729, 6145 and 6151 were heterozygous (p < 0.05) for this QTL. The 90% bootstrap CI spanned 32 cM between markers BMS871 and BMS482 which is half the size of that obtained using only the markers included in Guillaume et al. (2007). The average QTL allele substitution effect of the identified QTL was estimated as 4.72% NRR90 and the QTL explained 14% of the total genetic variance (Table 4).