| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
,2
* Area de Producció Animal, Centre UdL-IRTA, Universitat de Lleida (UdL)—Institut de Recerca i Tecnologia Agroalimentàries (IRTA);
and
Departament de Producció Animal, UdL, Rovira Roure 177, 25198 Lleida, Spain; and
and
Centre de Control Porcí, IRTA, Veïnat de Sies s/n, 17121 Monells, Girona, Spain
2 Correspondence:
phone: 34 973 70 28 93; fax: 34 973 70 28 74; E-mail:
jestany{at}prodan.udl.es.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
Key Words: Correlated Responses • Fats • Feeding Behavior • Growth • Litter Size • Pigs
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
In order to study the responses to selection for number of liveborn piglets, a large-scale selection experiment was conducted from 1993 to 1998 (Noguera et al., 2002b). The results of this study, which successfully improved litter size in all parities, indicated that correlated responses in performance tested traits (BW and backfat thickness at the end of test) were negligible.
These results, however, are not sufficient to preclude the existence of an underlying pattern of correlated changes in production traits. On one hand, the posterior probabilities associated with undesirable correlated responses in performance tested traits were large enough (higher than 90%) to further examine this hypothesis. Moreover, to gain more insight into this point, extending the analyses to ages other than the age at the end of the test and to different environments is desirable. Also, feeding behavior traits, as well as carcass and meat quality traits, were not analyzed in the aforementioned study.
The objective of this article was to investigate the correlated effects of selection for litter size on production traits. In particular, using the same lines described in Noguera et al. (2002b), the correlated responses to selection for litter size in growth, fat deposition, and feeding behavior patterns during the test period were analyzed. A subsequent paper will examine the responses in carcass, meat, and fat quality traits.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
The pigs used in this study were from a selection experiment with Landrace pigs conducted at Nova Genètica breeding farm in Solsona, Spain. Two lines were maintained: line H, selected for litter size, and line C, the control line. Details of the establishment of lines H and C, as well as the difference for litter size at different parities between both lines, were given in Noguera et al. (2002b). Briefly, line H was founded after one cycle of intense selection for the number of piglets born alive per litter, while the foundation stock for line C was constituted with a random sample of animals before selection took place. First generation of line H showed an average genetic superiority of 0.46 piglets born alive compared with contemporary sows from line C. As line C was formed and maintained as a true control line the difference between both lines is an unbiased estimator of the genetic change caused by the selection process carried out to improve litter size.
The data used in the analyses were obtained from a random sample of 115 and 72 pigs from the progeny of the first generation of lines H and C, respectively. Pigs, intact males and females, were from litters produced by 48 sows and 18 boars for line H and 18 sows and 7 boars for line C. After weaning, at 27.2 d of age (SD 5.6), pigs were weighed and delivered in three batches to the early-weaning unit associated with the IRTA swine test station in Monells, Spain. At a weight of about 25 kg they were moved to a unit equipped with the IVOG automatic feeding system (De Haer et al., 1992). After a period of adaptation, the pigs were individually put on test at an average weight of 29.3 kg (SD 7.8). Pigs were allocated in pens with 12 individuals of the same sex and were given ad libitum access to a pelleted finishing diet (12.9 MJ/kg of ME, 17.5% CP). Voluntary feed intake, number of visits to the feeder, and feeding time were measured on a daily basis (Fernández-Capo, 2001), and BW was recorded nine times during the test period. From 80 kg to slaughter weight, ultrasonic measures of midback (UMB) and loin (ULB) backfat were recorded weekly at 4 cm off the midline on both sides of the pig. The pigs were slaughtered at a weight of 102.0 kg (SD 9.1) in a commercial slaughterhouse. A muscle sample of each carcass was used to determine the halothane genotype at Universitat Autònoma de Barcelona, following the method described by Otsu et al. (1992). Only homozygous stress-resistant and stress-carrier pigs were identified in the animals used in this study. The number of records in each line and sex is shown in Table 1
.
|
Polynomial models with random regression coefficients were used to describe the data. Individual performances for BW, UMB, and ULB were analyzed as a third degree polynomial on age (d). Let Yid denote the observation of animal i at age d. The model was then:
 | [1] |
where bline,j, bsex,j, bhal,j, and bbatch,j are, respectively, the fixed effects of line (H, C), sex (boar, gilt), halothane genotype (NN, Nn), batch (1 to 3); ai,j and pi,j, are, respectively, the random additive genetic value and the random permanent non-genetic effect of animal i associated to the degree j of the polynomial, and eid the residual term. Additive genetic values were multivariate normally distributed N(0, A 
Va), in which A is the numerator relationship matrix and Va is the variance-covariance matrix of (ai,0, ai,1, ai,2, ai,3). A three-generation pedigree was considered to calculate A. It was also assumed that permanent nongenetic effects were N(0, I Vp), with Vp being the variance-covariance matrix of pi,j effects for animal i. Additive genetic values, permanent effects, and residuals were assumed to be mutually independent. Thus, variation between individuals is explained in terms of ai,j and pi,j, whereas variation within individuals is done in terms of eid. A common litter effect was not considered because previous analyses performed with a univariate mixed model did not show important changes in the results when this effect was included in the model. Also, very low contributions to variation were observed in on-farm performance traits (Noguera et al., 2002a).
Model (1) was also used to fit the observations for daily feed intake (DFI, g/d), number of visits per day (NVD), feeding time per day (FTD, min/d), and rate of feed intake (RFD, g/min). Rate of feed intake was obtained dividing DFI by FTD. A more detailed interpretation of Model [1] is given by Andersen and Pedersen (1996) and Jamrozik et al. (1997). The variance-covariance matrices were estimated by REML using the EM algorithm, as applied in the REMLF90 programs (Misztal, 1999). Iterations were performed until the criterion of convergence was less than 1 x 10-7. Estimated fixed effects and predicted random effects were obtained regarding the estimated variance-covariance parameters as the true parameters. Henderson’s mixed model equations were solved by direct inversion of the coefficient matrix.
The first derivatives of Eq. [1]
for BW, UMB, and ULB give the growth rate and the backfat deposition rate curves, and the integration of Eq. [1] for DFI, NVD, and FTD describes the accumulated curves for feed intake, number of visits, and feeding time. The accumulated curves can be used to derive efficiency curves. Thus, when weight gain is modeled as a function of accumulated feed intake, the first derivative gives the feed efficiency as a function of feed intake. Feeding time efficiency, as a function of cumulative feeding time, and efficiency of a visit, as a function of cumulative number of visits, were calculated similarly. On the other hand, when accumulated feed intake is modeled as a function of accumulated feeding time and accumulated number of visits, the first derivative represents, in terms of feed intake, the efficiency of a visit and the efficiency of feeding time, respectively. The efficiency of a visit, in terms of feeding time, can be modeled from the first derivative of the regression of feeding time on cumulative number of visits.
Estimation of Correlated Response
Genetic differences between the animals from line H and C were used to estimate correlated responses. The estimated genetic value of an animal is given by the line effect plus the additive genetic value within line. From these values, the line mean was estimated as the marginal mean that would be expected had the experimental design been balanced. Accordingly, estimation of the correlated response for trait k at age d was determined as follows:
 | [2] |
where nH and (n - nH) are the number of performance tested animals from line H and line C, respectively. Estimation of the standard error of line means and CRkd were calculated using the (co)variances of the estimated line effects and breeding values. These values were obtained from the inverse coefficient matrix from the mixed-model equations associated to [1]. Equation [2] was also used to estimate the correlated responses in traits modeled as a function of other variables different from age but substituting d by the fixed level of the corresponding independent variable. Tests of hypothesis were carried out using the t-test.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
and in the associated (co)variance components did not modify the main results of this study, even though a fixed model was used (Estany et al., 1998). The (co)variance components estimates used in the present analyses were within the range of expected values.
The estimated curve for growth rate for lines H and C is given in Figure 1. No significant differences were observed at weaning (P= 0.45) and at the end of the test (P = 0.18), even though line H showed higher (P< 0.05) BW during most of the test period (Table 2). Maximum relative difference for BW (29.88 kg vs 28.51 kg,P < 0.01, for lines H and C, respectively) was close to 5% near the beginning of the test period at 75 d. This occurred because the growth curves were different between lines, line H showing an earlier decline in growth rate than line C. The growth rate curve for line H is bent more than line C, as can be clearly seen in Figure 1 because the curves cross over. Thus, despite showing similar average growth rate on test, Table 2 reports that although line H grew faster at the start of the test (34 ± 11 g/d,P < 0.01), it grew more slowly by the end (-68 ± 27 g/d,P < 0.05). During the test period, the fat deposition rates were linear for both UMB and ULB and showed higher values for line H (0.017 ± 0.010 mm/d,P = 0.09 and 0.021 ± 0.010 mm/d,P < 0.05, for UMB and ULB, respectively). Therefore, line H had higher backfat depths than line C, and this difference increased with age (Table 3).
|
|
|
shows changes in DFI and RFD with age for each line, and in Figure 3 similar curves for NVD and FTD are depicted. Both DFI and RFD increased consistently with age; thus, their values at d 165 were 2.6 times those observed at d 75. However, there were maximums for FTD and NVD near 100 d of age. Similar to growth rate, the bending of the DFI curve with age was more marked in line H (Figure 2). Thus, although DFI was almost identical in both lines at the start and at the end of the test period, it tended to be higher in line H throughout the test period, with a maximum difference at 120 d of age (79.2 ± 43.6 g/d, P= 0.09). However, this differential was not large enough to result in a significant (P = 0.10) line effect on total feed intake over the test period (Table 4). Only total feed intake between 105 and 135 d of age was slightly higher in line H compared with line C (2.28 ± 1.25 kg,P = 0.07). Although line H tended to spend more time in feeding (Figure 3b), no significant (P= 0.70) differences were observed between lines with respect to the total feeding time (Table 4) and RFD (Figure 2). In contrast, results from Table 4 show that along the test period line H visited the feeder around 150 times less than line C. The difference between lines was maintained over the whole test period, but decreased proportionally with fewer visits to the feeder made by older animals (Figure 3a).
|
|
|
that the weight gain during the test did not differ between lines. However, when weight gain is expressed in terms of cumulative feed intake (Table 5), small differences in feed efficiency were found throughout the test period. In particular, line H was more efficient transforming feed into body tissues up to approximately 100 d of age, whereas line C performed better from this age until the end of the test period at 165 d of age. Nonetheless, differences in feed efficiency never were higher than 18 ± 6 g/kg. Because of this, total feed efficiency throughout the test period was slightly higher in line C (1.37 ± 0.77 kg of weight gain after eating 185 kg of feed, P= 0.08).
|
shows the differences between lines in feeding time efficiency, where efficiency was expressed as weight gain and feed intake per unit of time a pig spent in consuming it. In Table 7, efficiency was defined by weight gain, feed intake, and feeding time per visit to the feeder. Feeding time efficiency and visit efficiency steadily increased throughout the growing period with the exception of feeding time per visit. Thus, at the end of the growing period, efficiency in weight gain and feed intake were from two to three times higher than at the beginning of either per unit of time or per visit. Even though feeding time efficiency only presented minor differences between lines (Table 6), line H was more efficient per visit to the feeder throughout the test period (Table 7). Thus, line H ate from 30 to 40 g of feed more than line C in each visit to the feeder. Lines H and C had distinct feeding patterns with regard to eating frequency. Pigs from line H ate less frequently (Table 4), but instead they spent more time and ate more per visit (Table 7). During the test period, pigs from line H gained per visit around 10 g of live weight more than those from line C.
|
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
Statistical methods used in this study were chosen for making inferences about line means at different points rather than about curve parameters. Thus, despite polynomial models having no straightforward biological interpretation, they were chosen because of their goodness of fit and flexibility, particularly if random regression coefficients are used (Andersen and Pedersen, 1996; Huisman et al., 2002). The flexibility of polynomial models was used to describe data from traits with unknown trajectory such as NVD, and to calculate cumulative curves when not all the daily values were available. Eissen et al. (1999) showed that the use of such functions to estimate average daily feed intake is a good strategy to handle missing data on DFI brought about by the malfunction of automatic feeding systems. For this reason, accumulated values for feeding traits were not calculated directly but by integrating the polynomial function fitted to daily data (DFI, NVD, and FTD).
The method used for estimation of polynomial regressions and subsequently to calculate CRkd in Eq. [2] did not take into account uncertainty about variance components, which were assumed to be known. Therefore, estimators of CRkd and associated sampling distributions were only approximate. Varona et al. (1997) presented within the Bayesian setting a general solution to this problem, but the application of this methodology to our data was not straightforward because the sampler procedure did not always work for higher order polynomials. Nonetheless, in those analyses where convergence was achieved, the results were comparable and did not modify the conclusions, similarly to other studies (Su et al., 1997).
Selection for litter size did not affect average daily gain on test for either farm (Noguera et al., 2002b) or test station data. The same result was obtained by Herment et al. (1994). However, in light of our results, correlated effects for growth need a more careful interpretation. Only looking at average values can be misleading because they do not necessarily reflect the underlying correlated effects in the growth pattern. In particular, our results indicate that line H and line C had different growth patterns, so that correlated responses for growth within the test period are age dependent. This might explain the opposite tendency between the correlated response found here and the results in Noguera et al. (2002b) using on-farm data (-0.66 kg with a 95% probability interval equal to (-1.38; 0.05)). In this latter case, pigs were tested around 2 wk later, a period within which pigs from line C are expected to grow faster than pigs from line H.
The same argument, in conjunction with the ample variety of criteria used for measuring growth, can be used to understand some of the inconsistencies revealed when comparing results from literature. For instance, results show positive relationships between litter size and average daily gain on test (e.g., Kerr and Cameron, 1996), no relationship for growth rate from birth to the end of the test (Rydhmer et al., 1995), and a negative relationship for age at 100 kg (Ducos and Bidanel, 1996; Crump et al., 1997). Our results are consistent with the results in Garnett and Rahnefeld (1976), where selection for average daily gain did not affect litter size and may help to interpret why Landrace pigs selected for weight at 70 d of age had a small positive impact on litter size but no effect when they were selected for weight at 200 d of age (Kuhlers and Jungst, 1992, 1993).
With respect to backfat measurements, the estimates of correlated response showed the same trend reported by Noguera et al. (2002b) using field data (0.20 with a 95% probability interval equal to (-0.02; 0.40)). In both analyses, selection for litter size tended to slightly increase fatness. This agrees with recent maximum likelihood estimates of the genetic correlation between litter size and backfat depths (e.g., Ducos and Bidanel, 1996; Crump et al., 1997) and with the result reported by Herment et al. (1994), who found that backfat thickness was from 0.2 to 0.6 mm higher in animals whose parents were "hyperprolific" in comparison with contemporary animals. However, in the case of Hemment et al. (1994) there was not a real control population, so part of the difference could be attributed to the genetic lag of hyperprolific animals. There are also studies, though nonsignificant in some cases, showing that selection against fat can have detrimental effects on litter size (Clutter and Brascamp, 1998).
There is some evidence that selection for litter size affected the feed efficiency pattern with age, albeit total feed efficiency during the fixed test period was only slightly changed. Our results indicate that selection for litter size advances the age at which maximum feed efficiency is achieved. Using data from the French hyperprolific scheme Herment et al. (1994) found that selection increased feed conversion ratio by 0.02 to 0.04, comparable to 0.06 in the present study. This result is also compatible with the indications that selection for feed efficiency decreased litter size, both in pigs (Kerr and Cameron, 1995) and in mice (Holder et al., 1999).
Residual feed intake has been proposed as an alternative to measure energy for maintenance, as it accounts for the difference between daily feed intake and predicted feed intake from metabolic body weight, body weight gain, and its composition. The analyses undertaken applying the model described in De Haer et al. (1993a) to our data did not show that residual feed intake was lower in line H (results not shown). These authors observed that a low residual feed intake was associated with short daily eating time and a low eating frequency. Even though our results indicate that pigs from line H ate less frequently than pigs from line C, residual feed intake was not affected by NVD. The reason for this might be that we have not found differences between lines in the time spent in feeding, the feeding trait with the highest correlation with residual feed intake (De Haer et al., 1993a). In our case the correlation of residual feed intake with NVD (P= 0.71) and FTD (P= 0.60)was not significantly different from zero. The observed feeding behavior is consistent with the fact that breeds with lower lean growth tend to eat less frequently and more at a time (De Haer and De Vries, 1993; Labroue et al., 1994), especially when they are mixed with pigs from other genotypes in the same pen (Labroue et al., 1994).
The pattern of correlated changes observed in this study indicates that fatness showed the most change. An increase in feed intake seems the most plausible way by which the surplus of ME needed for the extra fat production is satisfied but not the only one. The results suggest that selected pigs might be more mature at the same age as well. Rauw et al. (1999) also found that growing mice intensely selected for litter size increased feed intake but not residual feed intake and that correlated changes were not simply a matter of scale. On the other hand, the pattern of correlated changes may differ with selection criteria, genetic merit of pigs for lean or fat deposition (Clutter and Brascamp, 1998; Cameron et al., 1999), or with external resources. Thus, according to which parity is emphasized in the selection criteria, the correlated response may differ, as can be inferred from the estimated genetic correlations between litter size and on-farm tested traits (Noguera et al., 2002a). Therefore, it would be advisable to assess the pattern of correlated changes beyond the slaughter age and, if possible, use different genotypes and selection criteria.
|
Implications |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
|
Footnotes |
|---|
Received for publication August 21, 2001. Accepted for publication May 30, 2002.
|
Literature Cited |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionImplicationsLiterature Cited |
|---|
This article has been cited by other articles:
|
|
 |
|
  L. Bunger, R. M Lewis, M. F Rothschild, A. Blasco, U. Renne, and G. Simm Relationships between quantitative and reproductive fitness traits in animals Phil Trans R Soc B, July 29, 2005; 360(1459): 1489 - 1502. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Serenius, M.-L. Sevon-Aimonen, A. Kause, E. A. Mantysaari, and A. Maki-Tanila Genetic associations of prolificacy with performance, carcass, meat quality, and leg conformation traits in the Finnish Landrace and Large White pig populations J Anim Sci, August 1, 2004; 82(8): 2301 - 2306. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. B. Petry, J. W. Holl, and R. K. Johnson Responses to 19 generations of litter size selection in the NE Index line. II. Growth and carcass responses estimated in pure line and crossbred litters J Anim Sci, July 1, 2004; 82(7): 1895 - 1902. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Estany, D. Villalba, M. Tor, D. Cubilo, and J. L. Noguera Correlated response to selection for litter size in pigs: II. Carcass, meat, and fat quality traits J Anim Sci, October 1, 2002; 80(10): 2566 - 2573. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
