Copyright 2004 by the Genetics Society of America DOI: 10.1534/genetics.104.026401
Mapping of Multiple Quantitative Trait Loci Affecting Bovine Spongiform Encephalopathy Chi Zhang, Dirk-Jan de Koning, Jules Herna´ndez-Sa´nchez, Chris S. Haley, John L. Williams and Pamela Wiener1 Roslin Institute (Edinburgh), Roslin, Midlothian EH25 9PS, United Kingdom Manuscript received January 13, 2004 Accepted for publication April 21, 2004 ABSTRACT A whole-genome scan was conducted to map quantitative trait loci (QTL) for BSE resistance or susceptibility. Cows from four half-sib families were included and 173 microsatellite markers were used to construct a 2835-cM (Kosambi) linkage map covering 29 autosomes and the pseudoautosomal region of the sex chromosome. Interval mapping by linear regression was applied and extended to a multiple-QTL analysis approach that used identified QTL on other chromosomes as cofactors to increase mapping power. In the multiple-QTL analysis, two genome-wide significant QTL (BTA17 and X/Yps) and four genome-wide suggestive QTL (BTA1, 6, 13, and 19) were revealed. The QTL identified here using linkage analysis do not overlap with regions previously identified using TDT analysis. One factor that may explain the disparity between the results is that a more extensive data set was used in the present study. Furthermore, methodological differences between TDT and linkage analyses may affect the power of these approaches.
B
OVINE spongiform encephalopathy (BSE) is a chronic, degenerative disorder affecting the central nervous system of cattle (http://www.bse.org.uk). It was first described in 1987 (Wells et al. 1987) and confirmed to be a transmissible spongiform encephalopathy (TSE) in 1988 (http://www.bseinquiry.gov.uk). TSEs also include scrapie in sheep and Creutzfeldt-Jakob disease (CJD) and Kuru in humans. According to the aberrant protein agent hypothesis, a disease-specific isoform of the prion protein PrPSc interacts with normal host PrP, resulting in its conversion to PrPSc. PrPSc is resistant to digestion with protease and can be detected in peripheral organs following infection; it later accumulates in the brain as the disease progresses. Clinical onset of disease is associated with neuronal cell death resulting in the spongiform appearance of the brain and results in “nervousness,” kicking, abnormal gait, and pelvic limb ataxia in affected animals (Patterson and Painter 1999). BSE can be experimentally transmitted across species (Bruce et al. 1994) and a distinct variant of CJD (vCJD) in humans has been linked to BSE (Collinge et al. 1996a; Almond and Pattison 1997). A major feature of prion diseases is that, although they are not genetic, susceptibility of individuals to TSE diseases is influenced by their genetic makeup and in particular, their PrP genotype. Three codons in the sheep PrP gene, at positions 136, 154, and 171, have been strongly associated with different incubation peri-
1 Corresponding author: Roslin Institute (Edinburgh), Roslin, Midlothian EH25 9PS, United Kingdom. E-mail:
[email protected]
Genetics 167: 1863–1872 ( August 2004)
ods following experimental challenge of sheep with different sources of scrapie and BSE (see Hunter 1999 for review). In humans, the sporadic forms of CJD and vCJD both seem to be associated with certain PrP gene polymorphisms (Palmer et al. 1991; Collinge et al. 1996b). Although Neibergs et al. (1994) suggested an increased BSE incidence in families of cattle with a particular genotype within the octapeptide repeat region of the PrP gene, to date, no convincing association between polymorphisms at the PrP locus and incidence of BSE disease has been demonstrated for cattle, which may be due partly to the relatively limited variability of the bovine PrP gene (Hills et al. 2003). Studies of the progeny of cows that developed BSE have suggested that there may be some elements of genetic control of susceptibility in cattle as there is an increased risk of BSE in offspring of BSE-affected cows compared with unaffected controls (Donnelly et al. 1997; Ferguson et al. 1997; Wilesmith et al. 1997). However, it has been difficult to distinguish between increased susceptibility and maternal transmission of disease. Intriguingly, recent studies of TSE transmission in mice showed that genetic loci other than the PrP gene are likely to be involved in the genetic control of prion diseases (Baron 2002; Moreno et al. 2003). In an earlier study on cattle, we identified three marker loci on bovine chromosomes 5, 10, and 20 that were significantly associated with BSE and are not linked to the PrP locus (Herna´ndez-Sa´nchez et al. 2002). A prion disease can be considered as a continuous trait if, for example, incubation period is measured or as a binary trait, if we consider individuals as being either
1864
C. Zhang et al.
affected or unaffected. A binary trait can be assumed to be controlled by an underlying continuous variable (Falconer and Mackay 1996) where the values of the variable above a certain threshold level are designated as affected (1) and those below it are designated as unaffected (0). Under this assumption, binary data can be analyzed for QTL with the same methods used for quantitative traits, although power and accuracy may be reduced (Xu and Atchley 1996; Kadarmideen et al. 2000). Visscher et al. (1996a) studied QTL for binary traits in backcross and F2 populations using linear regression (LRG) and a generalized linear model (GLM) and concluded that both LRG and GLM gave similar results when the errors within each genotype class were normally distributed. Kadarmideen et al. (2000) compared GLM with LRG to map QTL for binary traits in a multi-family half-sib design and also reported that they had similar power to detect the QTL and gave similar estimates of QTL location, effects, and variances. In cattle, QTL mapping is commonly performed in a collection of paternal half-sib families that are produced by artificial insemination. Neimann-Sorensen and Robertson (1961) and Weller et al. (1990) proposed, respectively, the daughter and granddaughter designs to analyze linkage between a single marker and a QTL when data are structured in half-sib families. The drawback of these methods is that they use information from a single marker at a time. Interval mapping, which uses information from multiple markers simultaneously, has been used by Georges et al. (1995) and Knott et al. (1996) for analyzing data from half-sib families. de Koning et al. (2001) developed a cofactor analysis strategy for simultaneous analysis of multiple chromosomes to increase the power and the precision of QTL mapping in outbred populations, structured in half-sib families. In this study, a genome-wide scan for QTL associated with BSE resistance or susceptibility was performed by applying interval mapping and cofactor analysis methods in paternal half-sib Holstein cattle families. The goodness of fit of a two-QTL model was also tested to search for multiple QTL within the same linkage group. The results were compared with previous findings obtained using transmission-disequilibrium tests (TDT; Herna´ndez-Sa´nchez et al. 2002). MATERIALS AND METHODS
Samples and genotyping: Four half-sib Holstein sire families with 360 cows, of which 268 were BSE affected and 92 BSE unaffected, were analyzed (Table 1). These families were selected from the same sample of animals as used in the previous study (Herna´ndez-Sa´nchez et al. 2002). Study animals contracted BSE from the environment during the United Kingdom’s epidemic of the late 1980s and 1990s. They were diagnosed by veterinarians using standard clinical criteria and BSE was confirmed
TABLE 1 Numbers of BSE-affected and BSE-unaffected offspring per sire in the study Sire 1 2 3 4 Total
Case
Control
Total
83 70 72 43
30 22 18 22
113 93 90 65
268
92
360
by histopathology. No other information on their disease progression was available. Samples were collected by the U.K. Veterinary Field Service at the time of slaughter of BSE suspects. Their sires were commonly used as artificial insemination donors and were not known to have an increased frequency of BSE-affected offspring. Control (unaffected) animals and BSE cases were sampled at approximately the same time from the same farms and were sire matched and age matched; hence, we make the assumption that both sets of animals were exposed to the same level of infectious agent. However, there is no way to confirm that they were exposed to the same degree as the affected animals. None of the controls subsequently appeared in the BSE case database of the British Ministry of Agriculture. Microsatellite markers were chosen from two published bovine linkage maps, the International Bovine Reference Panel (IBRP) map (http://www.cgd.csiro.au) and the U.S. Meat Animal Research Center map (http://www.marc.usda.gov), to cover all autosomes (Bos taurus chromosomes BTA1–29) and the pseudoautosomal region of the sex chromosomes (BTAX/Yps). A panel of 173 microsatellite markers was genotyped on all samples using an ABI373 DNA sequencer. The genotypes of the sires were inferred from those of their daughters while the genotypes of dams were unknown. For more details about the samples and genotyping see Herna´ndez-Sa´nchez et al. (2002). Markers and map construction: Marker linkage maps were constructed using Cri-map software (Green et al. 1990) after setting all identifiable genotyping errors to unknown genotypes. Various Cri-map options were used to determine linkage groups, marker orders, and marker distances within linkage groups. If the map distance of two neighboring markers was ⱖ100 cM, that linkage group was divided into two unlinked groups. Thirty-three linkage groups, covering 2835 cM of the bovine genome across all autosomes and the pseudoautosomal region of the sex chromosomes, were produced. The complete set of maps can be viewed as supplementary material (supplement 1 at http://www.genetics.org/supplemental/). QTL analysis: Linear regression methods were used for QTL analysis of the binary data (Visscher et al. 1996a; Kadarmideen et al. 2000). Initial QTL mapping was per-
QTL Affecting BSE
formed by using the web-based software package QTL Express (Seaton et al. 2002), which implements the multi-marker linear regression method (Knott et al. 1996). In short, a conditional probability of inheriting a particular haplotype from the sire was inferred from the marker genotypes in all half-sib offspring. Then the phenotypic value (i.e., affected or unaffected) was regressed on the probability that a particular QTL allele (associated with susceptibility or resistance) was inherited from the sire. The genetic model was Yij ⫽ i ⫹ biXij ⫹ eij , where Yij is the phenotype of animal j, offspring of sire i; i is the mean of sire family i; bi is the allele substitution effect of the QTL within family i; Xij is the probability that animal j inherited the (arbitrarily assigned) first haplotype of sire i; and eij is the residual effect. For every linkage group, F-statistics were calculated by comparing the pooled mean squares obtained from regression within families to the residual mean square. Numerator degrees of freedom was S and denominator degrees of freedom was N ⫺ 2S, where S is the number of sires and N is the total number of progeny. This process was repeated at 1-cM intervals along each linkage group with the maximum F-value indicating the most likely position of a QTL. A multiple (cofactor) analysis can partly account for the variance generated by other segregating QTL and substantially increases both power to detect a QTL and precision of estimating the QTL position (Lynch and Walsh 1998). The strategy adopted was the simultaneous analysis of multiple chromosomes in an outbred half-sib design (de Koning et al. 2001) where all positions within a linkage group could be included in the analysis of any family. Every round of analysis in this procedure involved two steps. Chromosome positions showing significant associations with the trait from the preliminary analysis were first selected as cofactors. Then all linkage groups were analyzed by standard interval mapping including these cofactors as covariates, using the following genetic model, Yij ⫽ i ⫹ biXij ⫹
n
兺 bikXijk ⫹ eij ,
k⫽1
where the variables are as given above, and the summation term reflects the contribution of cofactors, QTL identified on other chromosomes. Cofactors on the chromosome under analysis were omitted (n is the number of linkage groups minus 1). QTL exceeding 5% chromosome-wide threshold levels were selected as cofactors in the next round and this analysis was repeated until no new QTL were revealed. Cofactors were dropped from the analysis if the corresponding significance level was lower than the threshold calculated at every round for each linkage group. To distinguish between the presence of one QTL with a large effect and two linked QTL with smaller effects, a two-dimensional QTL search was carried out for those
1865
linkage groups where significant evidence for a single QTL was detected by the cofactor analysis. A 1-cM grid search was performed in QTL Express to estimate the effects of two QTL at separate positions within the same linkage group simultaneously, examining all possible pairs of locations, and to test whether the two-QTL model explained significantly more variation than the best QTL from the single-QTL analysis. Cofactors from other linkage groups identified in the cofactor analysis were also fitted. The genetic model was Yij ⫽ i ⫹ bilXijl ⫹ bimXijm ⫹
n
兺 bikXijk ⫹ eij ,
k⫽1
where the variables are as described above and l and m refer to the two putative QTL positions on the chromosome. The F-statistics were calculated by comparing mean squares from the two-QTL model with that of the best one-QTL model with degrees of freedom equal to S (numerator) and N ⫺ S(3 ⫹ C) (denominator), where S is the number of sires, C is the number of cofactors, and N is the total number of offspring. Significance thresholds and confidence intervals: Searching for QTL across an entire genome involves a large number of statistical tests for marker-trait associations. The use of an appropriate significance threshold for each test is necessary to keep the number of false positives to an acceptable level (Lynch and Walsh 1998). Significance thresholds were determined empirically by permutation tests to account for missing genotypes and differences in marker density (Churchill and Doerge 1994). Chromosome-wide significance levels (Pchromosome) were obtained for each linkage group using 10,000 permutations. Two genome-wide thresholds were set, one for “suggestive” and one for “significant” QTL. For genomewide suggestive linkage, one false positive is expected (Lander and Kruglyak 1995). The threshold for this level for a specific linkage group corresponds to that for a P-value equal to the contribution (R) of that linkage group to the total genome length, which was obtained by dividing the length of a linkage group by the total length of the genome. The second threshold, for genome-wide significant linkage, was set to that corresponding to 5% significance (i.e., one false positive expected in 20 genome scans; Lander and Kruglyak 1995), and it was obtained from the chromosome-wide significance levels using the following Bonferroni correction (de Koning et al. 1999): Pgenome ⫽ 1 ⫺ (1 ⫺ Pchromosome)1/R. Confidence intervals (C.I.) for the location of possible QTL were constructed by bootstrapping the samples 1000 times (Visscher et al. 1996b). Taking the top and bottom 2.5% of resampled positional estimates gave the estimated 95% confidence interval for each QTL. In the two-QTL model, an F-test of two QTL vs. one QTL was performed for linkage groups harboring sug-
1866
C. Zhang et al. TABLE 2 Distribution of markers, marker intervals, and information content across chromosomes
BTA
Markers
Length (cM)
1 2 3 4 5 6 7A 7B 8 9 10 11A 11B 12 13 14 15 16 17 18 19 20 21B 22 23 24 25 26 27 28 29 X/Yps
11 6 7 5 12 6 3 3 8 6 8 2 2 6 7 5 4 5 10 6 6 6 4 4 5 4 3 6 3 2 3 4
185 134 147 121 182 126 30 11 113 102 157 11 2 121 151 81 65 66 164 86 104 96 73 81 77 67 52 84 19 10 29 88
Total
172
2835
Average marker interval (cM)
Contribution (R )
Information content
16.8 22.3 21.0 24.2 15.2 21.0 10.0 3.7 14.1 17.0 19.6 5.5 1.0 20.2 21.6 16.2 16.3 13.2 16.4 14.3 17.3 16.0 18.3 20.3 15.4 16.8 17.3 14.0 6.3 5.0 9.7 22.0
0.065 0.047 0.052 0.043 0.064 0.044 0.011 0.004 0.040 0.036 0.055 0.004 0.001 0.043 0.053 0.029 0.023 0.023 0.058 0.030 0.037 0.034 0.026 0.029 0.027 0.024 0.018 0.030 0.007 0.004 0.010 0.031
0.52 0.65 0.58 0.50 0.62 0.63 0.49 0.61 0.61 0.63 0.48 0.28 0.57 0.52 0.48 0.57 0.50 0.67 0.51 0.62 0.57 0.57 0.45 0.55 0.50 0.47 0.53 0.57 0.54 0.62 0.68 0.36
16.5 (average)
1.000
0.55 (average)
Contribution was calculated by dividing the length of a linkage group by the total length. Information content measured as the ratio of the variance of QTL conditional probability to the maximum variance was averaged over every centimorgan along a linkage group. Chromosomes 7, 11, and 21 were each divided into two linkage groups, 7A/7B, 11A/11B, and 21A/21B (21A contains only one marker and is not shown), because the map distance between two adjacent markers on each chromosome was ⬎100 cM.
gestive QTL detected by the single-QTL model. The significance of the test statistics was determined from a standard F-table with degrees of freedom as given above. Segregation distortion: The TDT explores segregation distortion within the affected individuals from a heterozygous sire, significant distortion being seen as evidence for the presence of a disease-related gene. To compare our results to those from a previous study (Herna´ndez-Sa´nchez et al. 2002), we tested for segregation distortion within affected and unaffected individuals for all linkage groups. Segregation distortion was studied by testing whether the average probability that a particular QTL allele inherited from the sire deviated from 0.5. The inheritance
probabilities along each linkage group at 1-cM intervals were collected for every offspring in a family. The mean probability (Pmean) and standard error (SE) for each family were calculated and t-statistics were obtained from the formula: T⫽
|Pmean ⫺ 0.5| . SE
Thresholds for genome-wide significant segregation distortion were set at 5%, adjusted using the Bonferroni correction for multiple testing (number of tests ⫽ 688 ⫽ 172 markers ⫻ 4 families). This corresponds to a tabulated P-value of 0.000071. Degrees of freedom were determined by the number of animals in the smallest fam-
QTL Affecting BSE
1867
TABLE 3 The result of cofactor analysis by fitting five cofactors Effect (standard error) by family
BTA
Position (cM)
95% C.I. (cM)
F
Pchromosome
1 6 13 17 19 X/Yps
125 60 55 144 97 58
0–171 11–110 0–106 91–164 0–104 50–69
3.9 4.5 3.4 6.2 4.9 6.1
0.0308 0.0121 0.0447 0.0015 0.0057 0.0004
1 ⫺0.18 0.23 ⫺0.26 ⫺0.01 ⫺0.12 0.07
2
(0.100) (0.083) (0.135) (0.100) (0.094) (0.237)
0.26 ⫺0.14 ⫺0.27 0.20 ⫺0.04 0.67
(0.096) (0.082) (0.144) (0.093) (0.087) (0.144)
3 ⫺0.08 0.08 0.06 0.12 0.33 0.41
(0.153) (0.084) (0.130) (0.110) (0.091) (0.145)
4 ⫺0.43 ⫺0.26 ⫺0.31 ⫺0.84 0.30 ⫺0.01
(0.181) (0.112) (0.120) (0.186) (0.117) (0.392)
Two QTL, BTA17 and BTAX/Yps , are 5% genome-wide significant. Others are 5% chromosome-wide significant. Significance threshold were determined by 10,000 permutations.
ily (43 for affecteds, 18 for unaffecteds), making this a conservative test. RESULTS
Genetic map: A total of 173 microsatellite markers were used to construct a 2835 cM (Kosambi) genetic linkage map of the bovine genome that included all autosomes and the pseudoautosomal region of the sex chromosomes (Table 2). In total, 32 linkage groups were analyzed, excluding one, that contained a single marker, because interval mapping would not be very powerful for such a case. The average marker interval per linkage group ranged from 1 to 24.2 cM with a genome-wide mean of 16.5 cM. Most of the linkage groups were moderately informative with average information content (Knott et al. 1998) ranging from 0.28 to 0.68. The average information content at genome level was 0.55. The mapping results confirmed that the X/Yps markers underwent recombination in the sire and therefore were located in the pseudoautosomal region of the sex chromosomes. Ten linkage groups were not in agreement with one or both of the two published maps. The majority of the differences were minor flips between closely linked markers. BTA13 was the only chromosome for which there was a major difference in the orientation of markers; DIK93 was found at the centromeric end of the chromosome in our study and at the telomeric end of the IBRP map (most of the BTA13 markers used in this study were not found on the MARC map). Chance may explain the minor differences between maps, but it is not clear why the BTA13 maps were so different. As the average number of informative meioses per marker was 167 in this study and all maps were well supported by the Cri-map “FLIPS” option, we used this map rather than published map positions. QTL analysis: Six genome-wide suggestive QTL (Pchromosome ⬍ R) were identified using single-QTL mapping (BTA6, 13, 17, 19, 22, and X/Yps). The most striking was on the BTAX/Yps region, which also exceeded the
5% genome-wide significance threshold (Pgenome ⵑ0.01). The estimated QTL position for the X/Yps linkage group is at 55 cM with a 95% C.I. between 52 and 65 cM. The F-ratios for all linkage groups can be viewed as the “no cofactor” points in Figure 2. Cofactor analysis was then applied, where the cofactors were the QTL exceeding the 5% chromosome-wide significance threshold found in the previous round. The QTL analysis was repeated including the cofactors until no new QTL were revealed. By the final round, five cofactors in total were fixed and six QTL were obtained (one of the QTL did not feature as a cofactor). The numbers of cofactors fitted in successive rounds were zero, seven, three, seven, four, six, and five. Two linkage groups (BTA17 and X/Yps) showed genome-wide significant evidence and four linkage groups (BTA1, 6, 13, and 19) showed genome-wide suggestive evidence for a single QTL (Table 3 and Figure 1). The significant QTL on BTAX/Yps was mapped to 58 cM (Pgenome ⵑ0.01), near marker TGLA325, with 95% C.I. between 50–69 cM. Families 2 and 3 accounted for most of this effect. The genome-wide significant QTL on BTA17 was at marker INRA25, located at 144 cM (Pgenome ⵑ0.02), with 91–164 cM 95% C.I. Families 2 and 4 explained most of the effect of this QTL. To calculate the proportion of variance explained by the joint QTL, we estimated for each of the six QTL the ratio between the sum of squares explained by the multiple-QTL model and the total sum of squares. The models including QTL plus four cofactors (BTA1, 6, 17, 19, and X) gave an estimate of 0.23 for the proportion of variance explained, while the model including QTL plus five cofactors (BTA13) gave a slightly higher estimate (0.25). Compared with the six QTL obtained without cofactors, one QTL on BTA22 was removed and one new QTL on BTA1 was added. Among the remaining five QTL, the estimated positions of three QTL, on BTA6, 19, and X/Yps, moved 2–5 cM after the inclusion of cofactors, and the F-statistics for individual QTL were either decreased or increased, with changes ranging
1868
C. Zhang et al.
Figure 1.—Profile of the F-statistics for four genome-wide suggestive QTL (BTA1, 6, 13, 19) and two genome-wide significant QTL (BTA17 and X/Yps) identified by cofactor analysis. Horizontal dashed line indicates genomewide suggestive threshold and horizontal dashed-and-dotted line indicates 5% genome-wide significance threshold. Vertical dotted lines mark 95% confidence intervals. Where no dotted line is shown, the confidence limit is the end of the linkage group. Markers in parentheses are at the same position as their left adjacent marker.
from ⫺0.1 (BTAX/Yps) to ⫹2.0 (BTA17). Figure 2 shows the development of the maximum F-statistic for each individual linkage group during the cofactor analysis process. Comparison of two-QTL vs. one-QTL models provided no support for additional QTL on any linkage group where genome-wide significant and suggestive QTL were found using cofactor analysis. Test for segregation distortion on chromosomes with significant QTL: The TDT analysis of Herna´ndez-Sa´nchez et al. (2002) found associations between BSE infection and loci on chromosomes 5, 10, and 20 (see below). In the current study, significant segregation distortion
within affected individuals was found for one or more families on the following chromosomes: BTA1, 5, 10, 13, 14, 17, 19, 21B, 26, and X/Yps. These include five of the six chromosomes with significant QTL (excluding BTA6), two of the three chromosomes with significant TDT results (BTA5 and 10; Herna´ndez-Sa´nchez et al. 2002), and three others. Significant segregation distortion within unaffected individuals was found on BTA1, 10, 13, 14, 17, 19, 21B, and X/Yps for one or more families. These include the same five chromosomes with significant QTL as seen for the affected individuals, one of the three chromosomes detected using TDT (BTA10), and two others (also seen for the affected
QTL Affecting BSE
1869 DISCUSSION
Figure 2.—Development of the F-statistics for 32 linkage groups during cofactor analysis of the first round (no cofactor), the second round (seven cofactors), the sixth round (six cofactors), and the final seventh round (five cofactors). The different symbols indicate the maximum F-statistic for the linkage group during the process of fitting cofactors and reanalyzing the data.
individuals). The direction of the segregation distortion (i.e., which allele is overrepresented) was different for the affected and unaffected offspring in a few cases. This pattern was seen for family 3 on BTA1, 19, and X, but not for the six other family/chromosome combinations where there was significant segregation distortion for both affected and unaffected offspring. For the chromosomes where linkage analysis identified QTL, the maximum t-statistics for segregation distortion in affected offspring within individual families and their positions are shown in Table 4. When the position of the maximum t-statistic is compared with that of the maximum F-statistic on the same linkage group, there is substantial overlap in some cases but not others. The greatest overlap in t-statistic and F-ratio positions is for the pseudoautosomal X chromosome where the maximum statistics were physically close for both families 2 and 3.
In this study, a genome-wide linkage analysis of QTL involved in susceptibility or resistance to BSE was conducted in a population of cattle consisting of four halfsib families. The QTL analysis, in which five cofactors were considered, identified four genome-wide “suggestive” and two genome-wide significant QTL. There was evidence for significant segregation distortion of sire alleles on five of the six chromosomes with QTL, but also on other chromosomes. The most consistent picture between the QTL and segregation distortion analyses was seen on BTA19 and X/Yps, where segregation distortion was in opposite directions in the affected and unaffected offspring of a family with a significant QTL. The positions of the QTL and the maximum level of segregation distortion overlapped for BTAX/Yps. Candidate genes: Although variation in the PrP gene has been associated with differences in susceptibility or incubation period of TSEs in other species, up to now polymorphisms in the bovine PrP gene have not been associated with variation in susceptibility to BSE. One suggestive QTL in this study was identified on BTA13, which harbors the prion gene (PRNP). However, the 95% C.I. for the QTL identified on BTA13 does not include the region where PRNP has been mapped (between markers HUJ616 and ABS10; Schla¨pfer et al. 2000); the distance between that QTL and PRNP is at least 74 cM. Because of inconsistencies between our map order and that of the IBRP map for BTA13, linkage analysis was also run using the subset of markers found on the IBRP map and the IBRP positions for this chromosome (results not shown). There was still a suggestive QTL and its position was again distant from PRNP. In mice, polymorphisms in PRNP are known to be associated with variation in incidence or onset of disease. It is therefore interesting that in a QTL study in a mouse model of BSE infection, Manolakou et al. (2001) also failed to detect any effect closely linked to PRNP, although a QTL was found on mouse chromo-
TABLE 4 The maximum t-statistics for segregation distortion within affected offspring for linkage groups where putative QTL were detected Chromosomal position of maximum t - statistic and maximum F - ratio (cM)
Maximum t - statistic
BTA1 BTA6 BTA13 BTA17 BTA19 BTAX
Family 1
Family 2
Family 3
Family 4
3.01 2.96 5.15 6.01 13.44 2.25
1.87 1.02 1.92 8.61 5.03 20.1
7.38 2.07 2.02 8.28 4.93 5.08
3.71 2.45 3.87 3.12 1.38 1.92
Family 1
Family 2
Family 3
Family 4
122 (0) 54–57 (52) 144 (164) 6 (3)
105–106 (144) 0–23 (105) 54 (58)
112–113 (7) 0 (100) 44–88 (68)
Values in italic were above the 5% genome-wide threshold. For cases of significant segregation distortion, positions of maximum t-statistics are shown with the maximum F-ratio position in parentheses.
1870
C. Zhang et al.
some 2 where PRNP is located. This could suggest that in both cattle and mice there is another locus on the same chromosome as PRNP that is involved in BSE resistance or susceptibility or, alternatively, that in both studies, shortcomings in the analyses misplace the QTL regions. Four previous studies that mapped QTL related to the incubation period of TSE diseases in mouse models revealed significant or suggestive evidence for QTL on mouse chromosomes 2, 4–12, 15, and 17–19 (Stephenson et al. 2000; Lloyd et al. 2001; Manolakou et al. 2001; Moreno et al. 2003). Chromosomes 2 and 11 showed significant evidence of a QTL in two of the studies. Other chromosomal regions were supported for association with TSE susceptibility either by significant evidence in only one study or by suggestive evidence in more than one study. Regions of the bovine genome showing conservation of synteny with the mouse chromosomes were deduced via Mouse Genome Informatics (http://www.informatics.jax.org/). Several of the chromosomes containing regions where QTL were located in this study (BTA1, 6, 13, 17, 19, and X/Yps) showed homology with mouse chromosomes with putative QTL. Due to the lack of resolution of the bovine map, however, it was difficult to align regions where the cattle QTL were located relative to the mouse map. A putative candidate gene, Nf1, falls within the 95% C.I. of a genome-wide suggestive QTL on BTA19 in the current study and also falls within a QTL region on mouse chromosome 11 described in two independent studies (Stephenson et al. 2000; Lloyd et al. 2001). Neurological and behavioral defects have been described in Nf1 homozygous mutant mice (Costa et al. 2001), suggesting that NF1 could play a role in the neurodegenerative process of BSE. Although none of the above-mentioned mouse studies identified QTL on the sex chromosomes, a sex effect on incubation period and survival time was noted by Manolakou et al. (2001) and Moreno et al. (2003), respectively. Manolakou et al. (2001) were able to dissect maternal effects from those due to the X chromosome in the F1 generation and concluded that some of the sex differences seen in that population could be explained by X chromosome differences. The magnitude of the BTAX/Yps effect in our study emphasizes the need for further investigation into the genetic basis of sex differences in TSE susceptibility. Comparison with previous study: Herna´ndez-Sa´nchez et al. (2002) conducted a genome-wide search for markers associated with BSE incidence using a subset of the data used here. In that study, transmission probabilities of the two sire alleles were assessed for the affected daughters at each marker. A number of daughter genotypes had to be excluded from the analysis either because the transmitted sire allele could not be identified (i.e., daughters with the same genotype as their
sire) or because their inclusion biased the estimation of segregation ratios in the absence of maternal genotypes (i.e., daughters with homozygous genotypes). Furthermore, sires homozygous for a marker were also excluded from the analysis. Using a very stringent significance threshold, Herna´ndez-Sa´nchez et al. (2002) found three significant markers associated with BSE on chromosomes BTA5, 10, and 20. Surprisingly, none of these three regions are in common with genome-wide suggestive or significant QTL found in this study, although significant segregation distortion within affected individuals was found on two of these chromosomes. When markers closely linked to the three significant associations were tested, only one, on BTA5, showed a significant association with disease status under TDT. None of the markers identified using TDT showed the reverse pattern of allele transmission in the unaffected animals. A number of factors could contribute to differences in the results between linkage and TDT analyses. Although the initial data set was the same, the samples used in the two analyses differed. The TDT analyzes individual loci separately and requires the sire to be heterozygous for the marker locus tested. In addition, the absence of maternal genotypes meant that the paternal allele inherited for a particular locus could not be determined for many of the progeny from heterozygous sires and so these also could not be used in the TDT. As a consequence, the data set used by Herna´ndez-Sa´nchez et al. (2002) for any particular locus included approximately half the data that was used in the linkage analysis. This would have reduced the statistical power of the TDT study (Cardon and Bell 2001). To test whether differences in results arose from different data sets used in each study, the same restrictions on genotypic data used in the TDT analysis were applied to BTA5, 10, and 20, where significant associations were previously identified by TDT (however, in this case, both affected and unaffected individuals were included in the analysis). These data were then analyzed using linkage methods. Results using the restricted data set still failed to find significant effects above the 5% chromosome-wide threshold although the maximum F-statistics on BTA5 and BTA10 moved closer to markers BM315 and INRA107, which showed the greatest association under the TDT (results not shown). Another possible explanation for the differences in results between the two approaches could be that TDT and linkage analysis via multiple regression are very different tests. The TDT examines distortion within affected individuals for each sire and locus and uses counts of segregating alleles as raw data. Linkage analysis measures mean phenotypic differences between genotype groups. Moreover, the power of TDT is affected by the recombination rate and the linkage disequilibrium (LD) in the population (composed of four families), whereas linkage analysis depends on the recombination rate and is not affected by population-level LD.
QTL Affecting BSE
It is known that TDT is more powerful than linkage if there is strong LD in the sample and the marker map is dense (Risch and Merikangas 1996). Therefore, linkage may have had insufficient power to detect the regions detected by TDT for the markers that were physically close to the genes affecting the trait. Conversely, over larger genomic distances (and thus higher recombination rates), the linkage approach may be more successful than TDT (Kolbehdari and Jansen 2003). It is likely that population-wide distortion in marker alleles caused by factors other than BSE is present in cattle populations that are subjected to intensive breeding selection. This distortion could be detected by TDT, giving misleading results. This is consistent with the fact that TDT found no significant effects when comparing cases vs. controls by testing interactions in the transmission of alleles (Herna´ndez-Sa´nchez et al. 2002) while one would have expected different alleles to be overrepresented in the affected and unaffected classes. A similar pattern was seen in our survey of segregation distortion across the genome, although there were a few cases where the direction of the distortion differed for the affected and the unaffected classes. On the other hand, the linkage profiles using the same reduced data set as used for TDT (as mentioned above) reached their maximum at the positions where TDT found significant effects, although the results from the linkage analysis did not exceed the threshold values. Herna´ndez-Sa´nchez et al. (2002) noted that TDT could not be used to test the X/Yps chromosome because of the real segregation distortion of this region caused by linkage to X combined with the fact that all samples were from females (i.e., it is expected that alleles from the sire’s X chromosome-linked pseudoautosomal segment will be overrepresented in his daughters relative to alleles from his Y chromosome-linked pseudoautosomal segment). Our preliminary exploration indicated that linkage analysis is robust to segregation distortion of this kind (see supplement 2 at http://www.genetics.org/supple mental/). In summary, both analyses may have detected actual effects and thus provide complementary evidence for loci involved in susceptibility or resistance. The fact that QTL and cofactors identified in the linkage analysis explained only a quarter of the variation allows for the possibility that other regions, such as those previously identified by Herna´ndez-Sa´nchez et al. (2002), could also be important. Additional work to explore the regions identified by this and the previous study will be required to confirm these results. Conclusions: Six putative QTL, two genome-wide significant and four genome-wide suggestive QTL, were found associated with BSE incidence using linkage analysis. Lack of correspondence with previous results obtained using a TDT approach highlights differences between linkage and association methodologies and emphasizes
1871
the potential advantages of using different methods to explore the data and the need for independent verification of results. We thank the Biotechnology and Biological Sciences Research Council for additional funding, D. Matthews and J. Wilesmith for assistance in obtaining samples, and I. Maclean and D. Pomp (GeneSeek) for technical assistance in genotyping. Data for this work were collected through funding from the U.K. Department for Environment, Food and Rural Affairs (DEFRA, project SE 1744) and the European Commission (project CT97 3311).
LITERATURE CITED Almond, G., and J. Pattison, 1997 Human BSE. Nature 389: 437– 438. Baron, T., 2002 Mouse models of prion disease transmission. Trends Mol. Med. 8: 495–500. Bruce, M., A. Chree, I. McConnell, J. Foster, G. Pearson et al., 1994 Transmission of bovine spongiform encephalopathy and scrapie to mice: strain variation and the species barrier. Philos. Trans. R. Soc. Lond. B Biol. Sci. 343: 405–411. Cardon, L. R., and J. I. Bell, 2001 Association study designs for complex diseases. Nat. Rev. Genet. 2: 91–99. Churchill, G. A., and R. W. Doerge, 1994 Empirical threshold values for quantitative trait mapping. Genetics 138: 963–971. Collinge, J., K. C. L. Sidle, J. Meads, J. Ironside and A. F. Hill, 1996a Molecular analysis of prion strain variation and the aetiology of ‘new variant’ CJD. Nature 383: 685–690. Collinge, J., J. Beck, T. Campbell, K. Estibeiro and R. G. Will, 1996b Prion protein gene analysis in new variant cases of Creutzfeldt-Jakob disease. Lancet 348: 56. Costa, R. M., T. Yang, D. P. Huynh, S. M. Pulst, D. H. Viskochil et al., 2001 Learning deficits, but normal development and tumor predisposition, in mice lacking exon 23a of Nf1. Nat. Genet. 27: 399–405. de Koning, D. J., L. L. Janss, A. P. Rattink, P. A. van Oers, B. J. de Vries et al., 1999 Detection of quantitative trait loci for backfat thickness and intramuscular fat content in pigs (Sus scrofa). Genetics 152: 1679–1690. de Koning, D. J., N. F. Schulman, K. Elo, S. Moisio, R. Kinos et al., 2001 Mapping of multiple quantitative trait loci by simple regression in half-sib designs. J. Anim. Sci. 79: 616–622. Donnelly, C. A., N. M. Ferguson, A. C. Ghani, J. W. Wilesmith and R. M. Anderson, 1997 Analysis of dam-calf pairs of BSE cases: confirmation of a maternal risk enhancement. Proc. R. Soc. Lond. B Biol. Sci. 264: 1647–1656. Falconer, D. S., and T. F. C. Mackay, 1996 Introduction to Quantitative Genetics, Ed. 4. Addison-Wesley Longman, Harlow, UK. Ferguson, N. M., C. A. Donnelly, M. E. J. Woolhouse and R. M. Anderson, 1997 A genetic interpretation of heightened risk of BSE in offspring of affected dams. Proc. R. Soc. Lond. B Biol. Sci. 264: 1445–1455. Georges, M., D. Nielsen, M. Mackinnon, A. Mishra, R. Okimoto et al., 1995 Mapping quantitative trait loci controlling milk production in dairy cattle by exploiting progeny testing. Genetics 139: 907–920. Green, P., K. Falls and S. Crooks, 1990 Cri-map, Version 2.4. Washington University School of Medicine, St. Louis. Herna´ndez-Sa´nchez, J., D. Waddington, P. Wiener, C. S. Haley and J. L. Williams, 2002 Genome-wide search for markers associated with bovine spongiform encephalopathy. Mamm. Genome 13: 164–168. Hills, D., J. Schlaepfer, S. Comincini, I. Maclean, G. Dolf et al., 2003 Sequence variation in the bovine and ovine PRNP genes. Anim. Genet. 34: 183–190. Hunter, N., 1999 Molecular biology and genetics of bovine spongiform encephalopathy, pp. 229–246 in The Genetics of Cattle, edited by R. Fries and A. Ruvinsky. CAB International, Wallingford, UK. Kadarmideen, H., L. G. Janss and J. C. M. Dekkers, 2000 Power of quantitative trait locus mapping for polygenic binary traits
1872
C. Zhang et al.
using generalized and regression interval mapping in multi-family half-sib designs. Genet. Res. 76: 305–317. Knott, S. A., J. M. Elsen and C. S. Haley, 1996 Methods for multiple-marker mapping of quantitative trait loci in half-sib populations. Theor. Appl. Genet. 93: 71–80. Knott, S. A., L. Marklund, C. S. Haley, K. Andersson, W. Davies et al., 1998 Multiple marker mapping of quantitative trait loci in a cross between outbred wild boar and Large White pigs. Genetics 149: 1069–1080. Kolbehdari, D., and G. B. Jansen, 2003 Simulation of transmission disequilibrium tests for QTL detection in outbred livestock populations. Book of Abstracts of the 54th Annual Meeting of the European Association for Animal Production, 31 August–2 September, Rome. Lander, E. S., and L. Kruglyak, 1995 Genetic dissection of complex traits: guidelines for interpreting and reporting linkage results. Nat. Genet. 11: 241–247. Lloyd, S. E., O. N. Onwuazor, J. A. Beck, G. Mallinson, M. Farrall et al., 2001 Identification of multiple quantitative trait loci linked to prion disease incubation period in mice. Proc. Natl. Acad. Sci. USA 98: 6279–6283. Lynch, M., and B. Walsh, 1998 Genetics and Analysis of Quantitative Traits. Sinauer Associates, Sunderland, MA. Manolakou, K., J. Beaton, I. McConnell, C. Farquar, J. Manson et al., 2001 Genetic and environmental factors modify bovine spongiform encephalopathy incubation period in mice. Proc. Natl. Acad. Sci. USA 98: 7402–7407. Moreno, C. R., F. Lantier, I. Lantier, P. Sarradin and J.-M. Elsen, 2003 Detection of new quantitative trait loci for susceptibility to transmissible spongiform encephalopathies in mice. Genetics 165: 2085–2091. Neibergs, H. L., A. M. Ryan, J. E. Womack, R. L. Spooner and J. L. Williams, 1994 Polymorphism analysis of the prion gene in BSE-affected and unaffected cattle. Anim. Genet. 25: 313–317. Neimann-Sorensen, A., and A. Robertson, 1961 The association between blood groups and several production characteristics in three Danish cattle breeds. Acta Agric. Scand. 11: 163–196. Palmer, M. S., A. J. Dryden, J. T. Hughes and J. Collinge, 1991 Ho-
mozygous prion protein genotype predisposes to sporadic Crentzfeldt-Jakob disease. Nature 352: 340–342. Patterson, W. J., and M. J. Painter, 1999 Bovine spongiform encephalopathy and new variant Creutzfeldt-Jakob disease: an overview. Commun. Dis. Public Health 2: 5–13. Risch, N., and K. Merikangas, 1996 The future of genetic studies of complex human diseases. Science 273: 1516–1517. Schla¨pfer, J., N. Stahlberger-Saitbekova, J. Kuffer and G. Dolf, 2000 Genetic mapping of the prion protein gene (PRNP) on bovine chromosome 13. J. Anim. Breed. Genet. 117: 211–216. Seaton, G., C. S. Haley, S. A. Knott, M. Kearsey and P. M. Visscher, 2002 QTL Express: user-friendly software to map quantitative trait loci in outbred populations. Bioinformatics 18: 339–340. Stephenson, D. A., K. Chiotti, C. Ebeling, D. Groth, S. J. DeArmond et al., 2000 Quantitative trait loci affecting prion incubation time in mice. Genomics 69: 47–53. Visscher, P. M., C. S. Haley and S. A. Knott, 1996a Mapping QTLs for binary traits in backcross and F2 populations. Genet. Res. 68: 55–63. Visscher, P. M., R. Thompson and C. S. Haley, 1996b Confidence intervals in QTL mapping by bootstrapping. Genetics 143: 1013– 1020. Weller, J. I., Y. Kashi and M. Soller, 1990 Power of daughter and granddaughter designs for determining linkage between marker loci and quantitative trait loci in dairy cattle. J. Dairy Sci. 73: 2525–2537. Wells, G. A. H., A. C. Scott, C. T. Johnson, R. F. Gunning, R. D. Hancock et al., 1987 A novel progressive spongiform encephalopathy in cattle. Vet. Rec. 121: 419–420. Wilesmith, J. W., G. A. H. Wells, J. Ryan, D. Gravier-Widen and M. Simmons, 1997 A cohort study to examine maternally-associated risk factors for bovine spongiform encephalopathy. Vet. Rec. 141: 239–243. Xu, S., and W. R. Atchley, 1996 Mapping quantitative trait loci for complex binary diseases using line crosses. Genetics 143: 1417–1424. Communicating editor: J. B. Walsh
