| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
ANIMAL GENETICS |
Istituto Zootecnico e Caseario per la Sardegna, 07040 Olmedo, Italy
 |
Abstract |
|---|
 Top Abstract INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
Key Words: milk yield • prion protein (PrP) locus • Sardinian sheep • udder morphology
|
INTRODUCTION |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
). Primarily, 5 alleles out of the possible combinations are involved: A136R154R171, and thus, AHQ, ARH, ARQ and VRQ. The ARR allele is associated with resistance to natural and experimental infections with TSE (Jeffrey et al., 2001). The VRQ allele has the greatest susceptibility to scrapie, whereas ARQ and AHQ are associated with different levels of susceptibility depending on the breed and country of origin (Elsen et al., 1999; Baylis and Goldmann, 2004).
In 2003, the European Community introduced a regulation to launch breeding programs to eradicate scrapie in small ruminants (European Union, 2003). This was because bovine spongiform encephalopathy may have entered the sheep population and it was not obviously distinguishable from classical scrapie by clinical signs or current rapid tests for TSE (EFSA, 2003, 2005). As a result, selection for scrapie resistance has been recently introduced into several European sheep breeding schemes. In the Sardinian breed, selection for scrapie resistance began in 2002. This has meant that breeders now need to optimize the selection scheme to increase scrapie resistance without slowing down genetic improvements of production and functional traits.
Breeding for scrapie resistance may affect the other traits of interest because of 1) a direct effect of the PrP or a closely linked gene; 2) an effect of a gene not closely linked to the PrP gene; or 3) the loss of selection pressure due to the introduction of a new breeding goal.
The aim of this study was to investigate the existence of these effects on milk yield and udder morphology traits in Sardinian sheep.
|
MATERIALS AND METHODS |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
Data
From 2001 to 2005, 150,264 hair bulb or blood samples were collected from ewes born between 2001 and 2004, adult rams, and young males without progeny registered in the herd book. Finally, 23,077 animals (10,029 males) were then genotyped according to their impact on the selection scheme (high EBV or pedigree index for milk yield). Genotyping was performed using real-time PCR with the Taqman method or the PCR-RFLP technique. The latter technique does not discriminate between Q and H alleles at the 171 codon. Therefore, the ARH allele, which had a frequency of 0.05%, was considered jointly with ARQ (together, these were labeled ARQ*).
Milk yield (MY), expressed as mature-ewe equivalent (Carta et al., 1998), and 4 udder morphology traits: teat position (TP), degree of udder suspension (SU), udder depth (UD), and degree of separation of 2 udder halves (DS) were analyzed. Udder traits were scored with 9-point linear scales (Casu et al., 2006). Both milk yield and udder morphology records came from the national genetic evaluations of 2005 performed within the framework of the Sardinian breed selection program. Descriptive statistics of the data sets used for genetic evaluation are reported in Table 1.
|
Descriptive statistics of the data sets used for each analysis are reported in Table 2. Different approaches were applied depending on the nature of the investigated effect.
|
In the female model (GF), the dependent variable was animal performance deviation (PD), calculated as the sum of the animal additive genetic, permanent environmental and residual effects obtained from national genetic evaluations. The permanent environmental effect was included to take into account nonadditive effects of the PrP locus. In the male model (GMR), the dependent variable was the daughter yield deviation (DYD) of the sire; that is, the weighted average of the lactation records of the sires’ daughters adjusted for the fixed effects, and half of the breeding value of the sires’ mates obtained from genetic evaluations.
The applied model was
 |
where yijk was the PD k of the ewe j or the DYD of sire j, gi was the fixed PrP genotype effect, aj was the random animal additive genetic effect including relatives without records [a ~ N(0,A
), where A is the relationship matrix], and eijk was the random residual effect [e ~ N(0,I
)]. The analyzed genotypes were AHQ/AHQ, AHQ/ARQ*, ARQ*/ARQ*, ARQ*/VRQ, ARR/AHQ, ARR/ARQ* ARR/ARR, and ARR/VRQ. In males, the VRQ allele was not detected, and thus genotypes carrying VRQ were not present in the analysis.
Breeding values and additive genetic and error variances were estimated using the REML algorithm provided by ASREML (Gilmour et al., 2002). The significance of the PrP genotype factor and of the pairwise comparisons between genotypes was assessed by using the F-test and the Tukey-Kramer multiple-comparison procedure. The variance used to derive the error term of the statistical tests was the total individual variance (s2tiv), calculated as the variance of the sum of the additive genetic, permanent environmental, and residual effects estimates of animals with records. The degrees of freedom of the F-test and the Tukey-Kramer multiple-comparison procedure were calculated taking into account only the fixed part of the models.
Within-family analyses were carried out to investigate the effects of genes not closely linked to PrP that could not be detected by the previous analyses (no linkage disequilibrium between the PrP and the gene not closely linked is expected at the population level). Daughter design (DD) and granddaughter (GDD) design (Soller and Genizi, 1978; Weller et al., 1990) were performed on the sires heterozygous for the PrP gene. Using the actual population structure (Table 3), stochastic simulations (Carta and Elsen, 1999) were carried out to quantify the detection power of the available experimental designs. The Type I error threshold was fixed at 5%.
|
 |
where yijk was the yield deviation (YD) of the daughter k of the sire i in DD or the DYD of the son k of the sire i in GDD, si was the sire or grandsire i, and alij was the PrP allele j within-family i. The YD was calculated adjusting the lactation record of the sires’ daughter for the fixed effects, the permanent environmental effect, and half of the breeding value of the sires’ mates obtained from genetic evaluation. The ram’s mate breeding value was set to zero when it was not known. The DYD was the weighted average of YD (at least 5 daughters per sire). Only heterozygous rams (AHQ/ARQ*, ARR/AHQ, or ARR/ARQ*) with informative offspring (at least 8 in DD or 7 in GDD) and at least 30% of them sharing the same paternal allele were retained.
A loss of genetic gain on MY and udder traits was expected because of the different selection pressures applied on resistant and susceptible genotype classes. Indeed, resistant rams were selected even if they showed low genetic merit for the selection traits, whereas susceptible rams were selected only if they showed high genetic merit for the selection traits. To estimate the difference in selection pressures on the resistant and susceptible PrP genotype classes, a male analysis was performed using the same model as GMR but neglecting the relationships matrix (GMNR). This approach avoided adjusting the genotype class means for the additive genetic value of the animals, so as to compare the average genetic merit of the genotype classes. Moreover, because selection is simply oriented to increase ARR frequency and against susceptible alleles, there were only 3 classes of PrP genotype: ARR homozygous rams (RR), ARR heterozygous rams (RS), and rams without the ARR allele (SS), where R = resistant and S = susceptible.
|
RESULTS AND DISCUSSION |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
). The ARQ and AHQ alleles have the greatest susceptibility in Italian sheep breeds. The cumulative frequency of ARQ* and AHQ alleles was 56%. The VRQ allele was very rare in females (0.1%) and was not found in males. Approximately 1 out of 6 rams was homozygous resistant and half were ARR carriers (Table 4). The selection scheme for the Sardinian breed is based on pyramidal management of the purebred population with the registered flocks at the top (250,000 ewes from 1,200 farms). In these flocks, AI and controlled natural mating as well as official milk recording and breeding value estimation are carried out in order to generate genetic progress. Genetic progress is subsequently transferred to commercial flocks (approximately 3,000,000 ewes in 15,000 farms) through the flow of breeding animals, which are mainly males. The main breeding goal is lactation milk yield. Currently, the genetic gain is around 2 kg/yr per head, which corresponds to approximately 12% of the genetic standard deviation (Carta et al., 2004). Selection for udder morphology was introduced in 2004 (Casu et al., 2006). The National Breeding Plan for scrapie resistance was officially launched in 2005, although selection of resistant rams in the herd book began in 2002. The allelic frequencies calculated in the commercial flocks were similar to those estimated in the herd book (Mura et al., 2004). This finding suggests that the results for the association between selection traits and PrP genotypes obtained in the herd book can be extended to the whole population.
|
demonstrated that including the fixed effect of the PrP genotype in the animal model applied to the data set for the genetic evaluations of the Sardinian breed led to false detections. These spurious results occurred because the non-genotyped animals were grouped in the same class and compared with the genotyped ones. The latter were chosen from the most recent generations and among those with the greatest genetic merit for the selected traits. Thus, the lower average genetic merit of the nongenotyped animals affected their mean estimate and caused an inappropriate erosion of s2tiv and significant pairwise contrasts, which should, thus, be considered as false positives. To counteract this, our analyses were carried out only on records of genotyped animals. The dependent variables used were the most accurate estimates available of the animals’ production ability because they derived from the solutions of national genetic evaluation models that took into account all known relationships between animals, and had the best adjustments for fixed effects. For sex-limited traits, DYD gives the most accurate estimate of the male’s production ability (VanRaden and Wiggans, 1991).
As far as the population-wide analysis was concerned, no significant effects were detected in either female or male models. The most significant contrasts (P-value = 0.06) were ARR/AHQ vs. ARR/ARQ* in GMR for DS and AHQ/AHQ vs. ARR/AHQ in GF for UD. In the literature, associations of PrP genotypes with meat traits have been detected in several breeds (Brandsma et al., 2004; Alexander et al., 2005). In contrast, no association has been detected for dairy traits (Barillet et al., 2002; De Vries et al., 2005; Álvarez et al., 2006). However, previous studies are hardly comparable because the models used differed in terms of analyzed response variable as well as sex and genetic merit of the involved animals.
As far as the detection of not-closely-linked gene effects is concerned, simulations of within-family designs showed that under the hypothesis of a gene coincident with the PrP locus, the GDD had 90% detection power for a PrP effect of 0.3 and 0.4 phenotypic standard deviation units for MY and udder traits respectively. Daughter design had the same detection power for a PrP effect of 0.4 and 0.5 phenotypic standard deviation units. For a not-closely-linked gene, 90% detection power was achieved with a gene effect equal to [a/(1 – 2r)], where a is the gene effect needed to reach 90% detection power for a gene coincident with the PrP and r is the recombination rate with the PrP gene. Even greater gene effects are needed to reach 90% detection power if some of the sires are homozygous at the linked gene. For instance, if only one-half of sires out of those available for the analysis were heterozygous at the PrP locus, 90% detection power was attained for an allelic substitution effect of at least 0.7 phenotypic standard deviations for MY with both DD and GDD.
Finally, no linked gene effect was detected by the within-family analyses on MY and udder traits. This result agrees with the findings of Barillet et al. (2002), who applied a more powerful granddaughter design in French dairy breeds and Álvarez et al. (2006), who applied a daughter design in the Churra breed.
With respect to the effect of different selection pressures applied on different genotypes, the results of GMNR showed that RR rams were significantly worse for MY than RS and SS, whereas no significant difference was detected between RS and SS rams (Table 5). This is most likely due to the selection strategy applied of using even low-genetic-merit ARR homozygous rams and only high-genetic-merit ARR heterozygous and susceptible homozygous rams in the herd book. This condition is expected to produce a loss of genetic gain on MY in the next few years. To counteract this, ARR heterozygous rams should replace some of the ARR homozygous rams with the lowest genetic merit to reduce the loss of genetic gain on MY. The nonsignificant difference between RS and SS suggests that farmers are selecting these genotype classes with the same intensity. These results show that, to increase the frequency of the resistant allele in the population, a greater percentage of ARR heterozygous rams should be used instead of susceptible homozygous rams.
|
As a whole, great care should be taken in the approaches used to study the relationships between a quantitative trait and a specific locus. Most of the previous works investigated the effect of the PrP genotype or ARR allele at the population-wide level (Brandsma et al., 2004; Alexander et al., 2005; De Vries et al., 2005). In contrast, Barillet et al. (2002) and Álvarez et al. (2006) performed similar studies using within-family analyses. There were other differences in the choice of genotyped animals (randomly chosen or selected among those with greater productive genetic merit), in the class of animals included in the analysis (only those genotyped or all) as well as in the inclusion or not of the relationship matrix in the model (Álvarez et al., 2006; Brandsma et al., 2005; De Vries et al., 2005). Moreover, different phenotypes were used as dependent variables: performance records (De Vries et al., 2005; Isler et al., 2006; Vitezica et al., 2006), official EBV (Prokopová et al., 2002; Brandsma et al., 2004), and YD (Álvarez et al., 2006) or DYD (Barillet et al., 2002). All the above-mentioned differences have important consequences on the statistical inferences that can be drawn. The most important is the error variance used for the statistical test. The error variance is first determined by the choice of the dependent variable. For instance, if only EBV are analyzed, then the residual variance available for the statistical test can, at most, include the EBV variance, which is certainly lower than the overall individual variability. In this sense, analyzing the YD or DYD appears more adequate. In addition, once the dependent variable has been chosen, further differences can arise when considering (or not) the total individual variance in the error term. This consideration is especially important for highly hereditable traits for which the residual variance is relatively small. In any case the use of the solutions obtained from the genetic evaluations is certainly more precise than analyzing only the raw performances of the genotyped animals with a specific model, because of the better adjustments for fixed effects and deeper pedigrees.
For dairy sheep populations, a suitable strategy for investigating the relationship between the PrP locus and a quantitative trait would be to apply both population-wide and within-family analyses, based only on the genotyped animals and on dependent variables derived from the genetic evaluations. For sex-limited traits or progeny test-based breeding schemes, analysis of males is preferable, because their DYD results are more reliable than those for females. However, population-wide male analyses and within-family designs cannot be easy applied with sufficient power in most dairy sheep populations because of the number and size of the families involved. This is true even when AI and controlled natural mating are used. Nevertheless, it would be useful to carry out further work for developing models and software able to consider sire and daughter genotypes simultaneously and to perform the correct statistical tests.
This study showed that no PrP locus or linked gene effects exist in Sardinian sheep on udder morphology traits and MY. Nevertheless, the analysis of MY based on the DYD of sires and not including the relationship matrix showed that there are some differences between genotypes because of different selection pressures applied to the PrP genotype classes.
In summary, no negative correlated genetic response on the selection traits is expected from the selection for scrapie resistance in the Sardinian breed. However, a loss of genetic gain for the selection traits is likely to occur in the next few years due to the different selection pressures on resistant and susceptible sires. In order to limit the effect of this process and considering the average frequency of the resistant allele, the committee in charge of selection decisions for the Sardinian breed has decided that young rams homozygous for ARR without progeny test and with high pedigree value for milk yield can be used as elite sires in the herd book to produce young males for the progeny test. This strategy will increase the availability of resistant rams and is expected to reduce the loss of genetic gain for milk yield by shortening the generation interval on the sireson pathway.
|
Footnotes |
|---|
2 This work was financially supported by the EU project "Scrapiefreesheep" (QLRT-2000-01733). The udder scoring was carried out within the framework of the Community Initiative Programme IN-TERREG III A Italy-France Islands "Sardinia-Corsica-Tuscany."
3 Corresponding author: slsalaris{at}tiscali.it
Received for publication September 7, 2006. Accepted for publication May 22, 2007.
|
LITERATURE CITED |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
