| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
ANIMAL GENETICS |
,2
* Departamento de Producción Animal, Universidad de León, León, 24071, Spain;
and
Animal and Dairy Science Department, University of Georgia, Athens 30602; and
Departamento de Ciencia Animal, Universidad Politécnica de Valencia, Valencia, 46022, Spain
 |
Abstract |
|---|
 Top Abstract INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: constitution of line • cumulative production • fertility • longevity • prolificacy • rabbit breeding
|
INTRODUCTION |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
). Selective breeding to increase the length of productive life could be an alternative to reduce costs attributable to replacements. Previously reported estimates of heritability of functional longevity were 0.16 and 0.18 (Piles et al., 2006b) and 0.1 (Sánchez et al., 2006b).
Standard rabbit breeding programs are based on selection within lines for prolificacy and growth. Size of the maternal lines ranges from 100 to 250 does (20 to 40 bucks) and the generational interval is approximately 9 mo (Baselga, 2004). This generational interval is achieved because the information used to evaluate the candidates comes from females’ own performance, collateral relatives, and ancestors, but no information from offspring is used. The within-line response for the number of young at weaning ranges from 0.076 (Tudela et al., 2003) to 0.085 (García and Baselga, 2002) young per parturition per generation; these values were considered to be lower than expected by the authors. Under this structure, it does not seem feasible to include longevity as another selection objective, because the evaluation accuracy is expected to be low because of its low heritability and lack of information.
Alternative approaches to within-line selection have been used to improve prolificacy traits. One of them (Cifre et al., 1998a,b) was inspired by the successful hyperprolificacy programs in pigs (Bichard and David, 1985; Sorensen and Vernersen, 1991; Herment et al., 1994; Noguera et al., 1997).
The aim of this study was to compare the performance for several maternal traits of a new line, founded by selecting does outstanding for longevity that had prolificacy near or above the population average, to a well-reputed maternal line of high prolificacy and standard longevity. The objective of this comparison was to obtain some indications of the new strain’s potential as a maternal line for rabbit meat production.
|
MATERIALS AND METHODS |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
All experimental procedures were approved by the Universidad Politécnica de Valencia Research Ethics Committee.
The 2 lines compared in this experiment were the V line and the LP line; the former was selected for litter size at weaning for 31 nonoverlapping generations. This line was founded in the early 1980s (Estany et al., 1989) from different synthetic populations. The LP line was recently founded by selecting females from commercial farms (H_LP does) that showed extremely high productive lives (measured as the number of parturitions), but with prolificacy (measured as the mean number of young born alive per parturition) near or above the average of the Spanish commercial rabbit population. The average value for prolificacy in the Spanish commercial rabbit population is approximately 9 young per parturition, and the average value for number of parturitions during the whole life of a doe is approximately 8 (Ramón and Rafel, 2002).
Foundation of the line took place in 3 steps. Table 1 shows the summary statistics of the selected does in each step for the traits of interest. In the first step, the selected H_LP does were mated to V bucks; in the second step, the male offspring from the previous step were mated to a second batch of H_LP females, and in the third step, the male offspring from the second step were mated to a third batch of H_LP does. The offspring of this step were randomly mated to produce the experimental population of the LP line.
|
On farms I and II, mating was by AI with semen from another paternal line, but on farm III, the mating was natural, using bucks from the respective lines. The first mating took place when the does were 4.5 to 5.5 mo old, and the females were mated again 25 d after parturition. On all of the farms, a pregnancy test was carried out by abdominal palpation 12 d after mating. On farms I and II, the nonpregnant does were remated 28 d after the previous insemination. On farm III, the does that did not accept the buck were presented to the male 1 wk later and the does that accepted the male, but did not become pregnant, were remated 21 d after the previous attempt. Hence, although no AI was used on farm III, the reproductive management, in terms of volunteer waiting periods, was similar on the 3 farms. A count of the number of live young was conducted 28 d after parturition and was considered as the number of young at weaning.
The same culling policies were implemented on all 3 farms. Does with low prolificacy were never culled. In addition to the common practice of culling because of evident pathological problems (i.e., snuffles, mastitis, sore hocks, diarrhea, etc.), does with 3 consecutive non-fertile matings (on all 3 farms), or with 6 consecutive refusals of the buck (on farm III) were culled. The does that did not have any young alive at weaning after 2 consecutive pregnancies were also culled. All of these reasons for culling are considered indicators of reproductive problems but not indicators of poor production of healthy animals.
Trait Definitions and Statistical Models
Functional Longevity. Longevity was defined as the time in days from the first positive pregnancy test until the female left the herd by death or culling. This trait can be considered as a measurement of functional longevity, because no culling based on production was carried out. All females alive at the end of the experiment had a right censored record for this trait.
Longevity was analyzed by using a semiparametric log-normal animal frailty model under a Bayesian Markov chain Monte Carlo (MCMC) framework (Sánchez et al., 2006b). The following factors were included in the model:
13 born alive; and ninth included all of the does before the first parturition). Changes in level for this effect took place at parturition. The physiological status had 4 levels (first, pregnant; second, lactating; third, unknown; and fourth, empty). A doe was considered to be pregnant after a positive pregnancy test, lactating after the parturition if she had at least 1 offspring born alive, unknown when the doe was mated and had not been tested for pregnancy, and empty after parturition with zero born alive or after a negative pregnancy test. 
2a, where A is the numerator relationship matrix.
Uniform unbounded priors were assumed for the systematic effects (nongenetic factors). A probability mass was assigned at point 0.2 for the additive genetic variance. This figure can be considered as an average value for the genetic variance estimates previously reported in the literature (Piles et al., 2006b, Sánchez et al., 2006b).
Results of the survival analysis can be expressed in units of time to better understand what the log-hazard or relative risks imply. This can be achieved by using mean life estimates, so the survival functions of 6 prototype animals were computed, each one belonging to a different level of the combination of farm and genetic type. For these animals, the same reproductive performance was assumed; it was considered that they had 11 parturitions out of a total of 13 AI, and an average of 9.8 kits born alive. These animals were assumed to live for 636 d. Afterward, the mean life (restricted to 636 d) based on these predicted survival functions was computed, following the methods of Klein and Moeschberger (1997).
Prolificacy Traits.
Three prolificacy traits were studied: total born (TB), number born alive (BA), and number of young alive at weaning (28 d, NW). They were studied by using a single-trait repeatability animal model, with a Bayesian MCMC approach for the estimation (Sorensen and Gianola, 2002), and the factors included were as follows:
eighth). 2a. 2p. 2e.
Uniform unbounded priors were assumed for the systematic effects (nongenetic factors). The variance components were fixed to values producing those heritabilities and ratios of the nonadditive genetic plus permanent environmental variances to the phenotypic variances previously estimated by García and Baselga (2002) for line V.
Fertility.
Fertility was studied by using the results of the successive matings or AI as a binary trait (success or failure). When parturition records were available, this information was used to confirm the result provided by the pregnancy test; otherwise, the results from the pregnancy test were used. A threshold model (Wright, 1934) was used, which was implemented by using Bayesian MCMC estimation techniques (Sorensen et al., 1995).
The following factors were included in the model on the liability scale:
2a. 2p. 2e.
Uniform unbounded priors were assumed for the systematic effects. A probability mass was assigned at points 0.324 and 0.084 for the nonadditive genetic plus permanent environmental and additive genetic variances. These values were previously reported by Piles et al. (2005) for other rabbit populations. In general, in the threshold model 2 constraints are needed. When the variable is binary, one of them has to be imposed on the threshold, and in our case it was constrained to zero. The other constraint has to be imposed either on a linear function of the variances or on one of the variance components itself; we constrained the residual variance to 1, as in Sorensen et al. (1995).
Cumulative Production.
Cumulative production was defined as the total number of young weaned by a doe during her lifetime (from the first positive pregnancy test until death or culling or censoring date). With this definition, the trait is subjected to right censoring, because in the case of does alive at the end of the experiment, we only know that the real value of this trait was greater than the one recorded until that time. Hence, we implemented the Bayesian MCMC approach described by Sorensen et al. (1998) to fit normally distributed traits in the presence of right-censored records.
The model for this analysis included the following:
2a. 2e.
Uniform unbounded priors were assumed for the systematic effects (nongenetic factors). A probability mass was assigned at points 2,401.0 and 186.8 for the residual and additive genetic variances. These values were previously estimated with a large data set of the V line described by Sánchez et al. (2006b).
For testing the significance of the genotype x environment interaction (G x E), reduced models for all of the traits that did not include the interactions between farm and genetic type and between cycle and genetic type were fitted. Both models were compared by using deviance information criterion (DIC; Sorensen and Gianola, 2002; Spiegelhalter et al., 2002).
In all analyses, burn-in periods of 200,000 rounds were discarded, after which 1 of each 200 sampled values was retained. Convergence of all Markov chains (total length 2,000,000) was assessed by means of the Geweke test for contrasts of interest and by visual inspection of the trace plots.
|
RESULTS |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Table 2 shows summary statistics and relative frequencies of censored and uncensored records by genetic type. The chi-squared test, to determine whether the distribution of censored and uncensored records for both lines was the same, had a value of 5.41, which under the null hypothesis (the same distribution) follows a chi-squared distribution with 1 degree of freedom. The P-value that corresponded to this statistic was 0.020; therefore, the null hypothesis was rejected.
|
), restricted to 636 d, which was the maximum uncensored record, were 439.8 (13.8) and 405.7 (14.1) d for the LP and V lines, respectively. The median lives were 574 and 479 d for the LP and V lines, respectively. The differences between mean and median length of life estimates are partially because survival times are right censored; therefore, the estimates of the mean and median lifetimes are biased downward, but this bias is greater for the mean than for the median (Klein and Moeschberger, 1997). The log-rank statistic (Klein and Moeschberger, 1997) was also computed with the crude data. Its value was 4.88, which under the null hypothesis (the same probability of death between lines) follows a chi-squared distribution with 1 degree of freedom. The P-value that corresponds to this statistic was 0.027; therefore, the null hypothesis was rejected.
The number and proportion of animals alive at the end of each cycle (defined by the positive pregnancy tests), by genetic type, are shown in Table 3. The period in which most females died or were culled was the interval between the first and second positive pregnancy test (first cycle), 9.13% in the LP line and 10.99% in the V line. After this cycle, the proportion of dead or culled females was lower in both lines.
|
. The high prolificacy demonstrated by both lines should be noted. The values for the V line were particularly large, and were similar to those previously reported for this line (Piles et al. 2006a). The average pregnancy rates in both lines were within the range of variation for commercial farms in Spain (Ramón and Rafel, 2002).
|
shows the results for the contrasts between the genetic type effects within cycle number for longevity, prolificacy, and fertility traits. With regard to longevity, the cycles in which clear differences in the log-hazard ratios were found in favor of the LP line were the fourth, seventh, and eighth, where the probabilities of a V female leaving the herd were 2.40, 2.03, and 1.72 times higher than for an LP female; the probabilities of these ratios being greater than 1 were 0.01, 0.04, and 0.03. In the fifth cycle, the log-hazard ratio between the LP and V lines was 0.78, which means that the probability of an LP female leaving the herd at that moment was 2.4 times higher than for a V female. Regarding prolificacy traits, strong evidence was observed of differences between lines in favor of the V line during the first 3 parturitions for TB, measured as the probability of values greater than 0 for the contrast; the values of these contrasts were 0.68, 0.79, and 0.75. For BA, it was in the second and third parturitions in which strong evidence was observed; in this case, the contrast values were 0.92 and 0.88. With regard to NW, the estimated values for the contrast between lines in the first and second parturitions were 0.55 and 0.58; the probabilities of these values being greater than 0 were approximately 0.975. It should be noted that for parturitions after the fourth or fifth, the differences between lines changed sign, but in these cases the evidence for these contrasts being different from 0 was weaker. None of the ratios between the probabilities of obtaining a positive pregnancy in the LP and V lines showed strong evidence of being different from 1, because on one hand, the 95% highest posterior density intervals were symmetric compared with the mean and always included the value 1, and on the other, because the probability of values for these ratios being greater than 1 never reached extreme values; the more extreme observed values were 0.06 and 0.93 for the second and the sixth parturition, respectively.
|
. In this case, the contrast for cumulative production was also included. Regarding longevity, it was 1.24 times more likely for a V doe to leave the herd than for an LP doe; the probability of this contrast being greater than 1 was 0.08. The average differences between lines for TB, BA, and NW were 0.21, 0.21, and 0.14, in favor of the V line, but the probabilities of these contrasts being greater than 0 were approximately 0.2, which is a non-extreme value. Nor was any evidence observed of different performance between lines for fertility (cumulative production of weaned kits), because the probability of the contrast being greater than 1 (0) never reached extreme values; this probability was 0.26 (0.89).
|
Table 7 shows the results of the contrasts between line effects within farm. The only farm on which differences in the log-hazard ratio between lines were observed was on farm II; it was 1.65 times more likely on this farm for a V doe than for an LP doe to leave the herd; the probability of this contrast being greater than 1 was approximately 0.99. With regard to prolificacy, on farms II and III the probability of a value greater than 0 for the contrasts between lines for TB were 0.01 and 0.97, respectively; on farm II the difference between lines was almost 1 kit in favor of the V line; on farm III this difference was 0.5, but in favor of the LP line. For BA and NW the differences between lines on farm II were 1.01 and 0.66, respectively, also in favor of the V line. However, on farm III these differences were 0.5 in both cases, but in favor of the LP line. In all of these cases, the probabilities of contrast values being greater than 0 reached extreme figures, around 0.01 on farm II and 0.97 on farm III. The probability of success of the successive AI or matings was also a trait showing differential performance between lines across farms, because no differences in fertility were observed on farms II and III, but on farm I a ratio of 0.92 between the probabilities of obtaining a pregnancy in each line was observed. In this case, the probabilities were 0.73 and 0.80 for the LP and V lines, respectively. With regard to cumulative production, the probability of values greater than 0 for the differences between lines did not reach extreme values either on farm I or II (0.3 and 0.83), but on farm III, line LP females weaned on average 16.4 more young than line V females; the probability of this value being greater than 0 was 0.99.
|
). For longevity and fertility, the models without a G x E interaction showed lower DIC and so were statistically preferable; in these cases the differences between DIC were also greater than 3. For cumulative production, the differences in DIC between models were very low (2.2), indicating that both models were equally plausible given the data. The results of all of these tests are shown in Table 8. It should be noted that in both traits subject to censoring (longevity and cumulative production), the originally considered quality of fit by Spiegelhalter et al. (2002; the deviance) was lower in the models with a greater number of parameters (result not shown). This apparently contradictory result could be accepted if it was taken into account that this term already includes a penalty for model complexity, as indicated by Aitkin (1991).
|
. This table shows the same results, concerning longevity, as those previously provided, but now they are expressed in units that are easier to understand. The LP line showed a mean life 79 d longer than the V line on farm II. In contrast, on farm I, almost no difference was observed between the lines. Farm III showed an intermediate position with a difference of 21.48 d in favor of the LP line, but the probability of this difference being greater than 0 was only 0.78. The contrasts between farm II and farms I and III for V females were 144.62 and 106.42 d against farm II. These values for an LP female were 70.13 and 48.86 d, also against farm II. The probabilities of these contrasts being greater than 0 were always near 1. On the contrary, the probability of values greater than 0 for the differences between farms I and III did not reach extreme values for the V and LP females (0.86 and 0.73 for the V and LP lines).
|
|
DISCUSSION |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
) carried out for the overall differences between lines for longevity showed a positive result. Moreover, the weighted average of the differences between lines for mean life across farms was 31.63 d. This value is 51% of the average interval between parturitions in this experiment and 59% of the interval between parturitions reported by Ramón and Rafel (2002) for this trait in Spain, which was 53.6 d.
To our knowledge, the only selection experiment involving reproductive longevity carried out in prolific species was in mice (Farid et al., 2002). These authors selected 2 lines and kept a control population. The estimated responses in their experiment after 24 generations were 148 and 105 d within each line; these values corresponded to an increase in the reproductive life per generation equal to 18 and 22% of the interval between parities. In the French Holstein population, the selection of dairy bulls based on their EBV for functional longevity was implemented in 1997. Ducrocq (2005) reported an almost flat genetic trend, which is reasonable given the strong selection on milk production and its negative genetic correlation with traits determining functional longevity, such as fertility or resistance to diseases.
Although the observed differences in longevity between lines were generated mostly on farm II, which was the facility with the lowest average for this trait, the hypothesis of the longevity data being generated under the G x E interaction was not supported, because the model without this interaction had the lowest value of DIC. This result was obtained because almost no differences between lines were observed on farms I and III. The changing differences between lines across farms, despite the DIC results for model comparison, could be interpreted as a threshold response. When the environmental conditions for longevity are good (as on farms I and III), almost no differences between lines in survival ability are observed. However, when the animals are subjected to poor environmental conditions for longevity, differences in survival ability favored the LP line because of its greater robustness.
Not finding conclusive overall better performance of line V over line LP for prolificacy traits was unexpected. The V line has been selected successfully for more than 30 generations for litter size at weaning. This strain is considered a top line for its maternal performance both in Spain and around the world (Garreau et al., 2004). By analyzing the results of prolificacy in depth across farms, one may observe that the nonexistence of differences in prolificacy is because differences in favor of the V line on farms I and II are compensated for by differences in favor of the LP line on farm III. Theilgaard et al. (2007) analyzed the data generated on farm III to study the relationship between prolificacy and the use of body reserves. They indicated that the unusually low performance of the V line was due to indeterminate poor environmental conditions, mainly related to the cumulative effect of cycles of severe feed restriction after weaning and before the next parturition. The negative effect of this restriction was greater in V than in LP females, because the latter were more robust and responded to this poor environment with mobilization of body reserves. As a consequence of this dramatic difference in the relative reproductive performance between lines across farms, all of the prolificacy traits can be said, based on the results of the DIC model comparison, to be subject to a G x E interaction. However, the detected G x E interaction could be considered unusual, because it was observed when one of the environments was particularly extreme (strong feed restriction). This, given the usual farming conditions, makes this finding irrelevant in terms of questioning the current selection procedures, which do not consider this interaction. On the other hand, results for the prolificacy traits are again clear evidence of the robustness of the LP line to poor environmental conditions.
The differences between lines throughout successive cycles (Table 5) clearly showed better survival ability for line LP at later cycles. This could be expected, because the selection procedure in the LP line was focused on late survival. However, we do not have a good explanation for the result that in the fourth cycle, the LP line showed better survival ability than the V line, whereas in the fifth cycle the opposite occurred. One speculative explanation could be a kind of purging of V females during the fourth cycle so that those V females remaining alive after this point had on average better short-term survival ability than LP females, because in the fifth cycle a difference in favor of the V line was observed.
During the initial parturitions, better prolificacy was observed for the V line, between 0.50 and 0.92, but during later parturitions (>third) the differences between lines, in general, changed in sign, although it needs to be noted that in these later parturitions the probabilities of the contrasts showing extreme values were as high as 0.83. Theilgaard et al. (2007), analyzing only data from farm III, found no differences in prolificacy between these 2 lines during the initial parturitions, although at later ages differences in favor of the LP line appeared.
All of these results (better longevity for the LP line and better early prolificacy for the V line) are within those predicted by the theories of the disposable soma and the antagonist pleiotropy explaining the aging process (Kirkwood and Rose, 1991; Kirkwood and Austad, 2000), which postulate an antagonism between reproductive functions (especially at early ages) and longevity.
With regard to fertility, although no overall differences between lines were found, differential relative performance of the lines was observed across farms. On farms II and III, no differential performance between lines was observed, but on farm I, which was the worst in terms of fertility, line V females showed better fertility than LP females. As for longevity, poor environmental conditions for fertility favor the expression of the differences between the lines.
It was unexpected that, for each of the traits, the farm showing the worst environmental effect would be different. Because culling policies on the 3 farms were basically the same, one speculative explanation for this result could be that the environmental correlations between traits were close to 0. Sánchez et al. (2006a) previously reported that the environmental correlations between longevity and prolificacy are close to 0.
Although differences between lines were observed for longevity, this was not the case for prolificacy, so the probability of the contrast for cumulative production between LP and V females being greater than zero was 0.89. This means there is evidence that the cumulative production for LP females is greater than that for V females, although it is not extreme.
The main purpose of this experiment was to compare the performance of the new LP line with that of another well-known and well-performing line to determine whether this new line could be considered as a candidate maternal line for inclusion in the current 3-way crossing production schema. Although some factors during the experiment could favor the LP line when comparing its performance with that of the V line (i.e., higher inbreeding in the V line and possible heterotic effects in the LP line because of its recent foundation), this experiment provided evidence of the successful selection procedure during the foundation of the LP line. Because the V line is widely used in Spanish rabbitries, the V line itself is relatively close, genetically, to the population from which the selected animals for founding the LP line were obtained. Thus, this line could be considered good animal material to be used as a control line for estimating the response to the selection procedure, particularly when no truly unselected control population was available.
Our results are the first obtained in characterizing the LP line; they are promising, because they indicate that the LP line is as productive as the V line. However, before making strong recommendations for using the LP line, more evidence supporting its utility and confirming our results must be recorded. For example, because the current production system is based on cross-bred females, it is necessary to estimate genetic parameters in the crosses between this new line and other lines already used in this crossing scheme.
Beyond the animal production scope, the new LP line could be a valuable resource for other biological sciences interested in the aging process. For example, it has already been used to study, from a physiological point of view, the putative antagonism between early reproductive performance and longevity (Theilgaard et al., 2007), and could also be useful in investigating the molecular genetic mechanisms involved in the aging process or in reproductive senescence in mammals.
|
Footnotes |
|---|
2 Corresponding author: jpsans{at}unileon.es
Received for publication April 17, 2007. Accepted for publication October 31, 2007.
|
LITERATURE CITED |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
