| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
ANIMAL GENETICS |
,1
* Department of Animal and Dairy Science, University of Georgia, Athens 30602-2771;
and
Norsvin, Pb 504, 2304 Hamar, Norway;
and
Departamento de Producción Animal, Universidad de León, León, 24071, Spain; and
Smithfield Premium Genetics Group, PO Box 668, Rose Hill, NC 28458
 |
Abstract |
|---|
 Top Abstract INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
Key Words: carcass weight • genetic parameter • heat stress • pig • random regression
|
INTRODUCTION |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
. In pig production, the economic losses caused by heat stress in sow production and growing-finishing pigs are estimated at $299 to $316 million per year in the United States. Of these losses, those for grow-finish pigs are estimated at $202 million. In North Carolina, one of the largest hog-producing states, the loss in body weight gain is estimated at about 2.9 kg/animal. Heat abatement might improve animal welfare and is practiced for sow production. In growing-finishing pigs the gains in productivity are nearly all negated by additional capital and operation costs.
Genetic differences of animals in response to high temperatures or other stressors have been reported in pigs (Collin et al., 2001), cattle (Ravagnolo and Misztal, 2002; Bohmanova et al., 2005; Petersson et al., 2005), and sheep (Pollott and Greeff, 2004). Exploitation of this genetic variability could improve production efficiency and performance.
A genetic evaluation that accounts for heat stress requires each record to be associated with some easily available measure of heat stress. In dairy cattle, Ravagnolo and Misztal (2002) used weather information from public weather stations. A daily temperature-humidity index (THI) was found to be a useful predictor of daily milk production 1 to 2 d later. In pigs, Zumbach et al. (2008) found that heat stress for weight in finisher pigs was best accounted for by heat load during the last 10 wk of life. Heat load was defined as mean daily THI exceeding the threshold of 18°C and compounded weekly. The objective of this study was to estimate genetic parameters for carcass weight as a function of heat load.
|
MATERIALS AND METHODS |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
Data
Warm carcass weight data from 23,556 terminal crossbred pigs [Duroc1 x (Landrace x Large White)] harvested from May 2005 through December 2006 were available (Smithfield Premium Genetics Group, Rose Hill, NC). Duroc1 corresponds to the Duroc nucleus line P1 (Zumbach et al., 2007). The animals were raised on 2 commercial 1,200-sow farrow-to-finish farms.
Farm 1 contributed 12,771 records, and farm 2 provided 10,785 records. Both farms were located in North Carolina less than 10 km apart. The average age at harvest was 200 d, and the average carcass weight was 89 kg. Typical and similar corn-soybean diets were used at both sites. Feeding was ad libitum for all animals.
As no inside temperature/humidity data were available, these were obtained from a public weather station located at an airport within a distance of 21 and 14 km from farm 1 and farm 2, respectively. Average daily temperature (T) in degrees Celsius (°C) and relative humidity (RH) based on hourly observations were available (State Climate Office of North Carolina, 2007). A THI (NOAA, 1976) for each day of the experimental period was calculated based on the formula
 |
where T is temperature in °C and RH is relative humidity.
Daily heat load was defined as the number of degrees exceeding the threshold of 18°C. The weekly heat load was defined as a mean of daily heat loads. The total heat load was defined as the sum of weekly heat loads over the last 10 wk of life. The threshold of 18°C and the period of 10 wk provided the best fit to the data (Zumbach et al., 2008).
There were a total of 208 purebred sires and 2,731 crossbred dams. Pedigrees were not available for dams. The pedigree file for sires contained 770 animals, of which 562 parents were without records.
Statistical Analyses
Variance components were estimated with 3 different models. As the dam and litter effects were confounded, only the litter effect was included in all models.
Model 1 was a single-trait model (UNI) disregarding heat stress and was defined as follows:
 |
where y = a vector of observations, and β = a vector of fixed effects, which included the contemporary group (farm-year-week) and sex; age at harvest standardized to 200 d was included as a linear covariate; s = a vector of additive genetic effects of the sire; l = a vector of birth litter effects, also containing dam additive and dam environmental effects; e = a vector of residual effects; X, Z, and W = appropriate incidence matrices. Relationships among sires were taken into account. All of the data were considered in the analysis.
Model 2 was a 2-trait model (MT2) with carcass weight under heat stress and non-heat stress months with the same fixed and random effects as in Model 1:
 |
where subscripts HS and NHS indicate harvest months where full heat stress and no heat stress were observed, respectively. August to October 2005 and July to October 2006 were considered as months of heat stress; May to July 2005 and January to June 2006, as well as December 2005 and 2006, were considered as months of non-heat stress. The months of November 2005 and 2006 were considered as intermediate and not considered in MT2. The number of sires and dams during periods of heat stress was 152 and 1,580, respectively; during the months of non-heat stress there were 161 sires and 1,865 dams.
The (co)variances were
where A = numerator relationship matrix and I = identity matrices of appropriate dimensions.
The third model (RR) was the same as UNI, except that the sire effect was converted to a random regression on H. The scale of H was adjusted to the range of 0 (no heat stress) to 5 (highest heat stress). Initially, the litter and residual effects were also made functions of H. However, analyses showed that residuals were homogeneous throughout H stages, and the effect of birth litter nested within H was not significant. The sire variance of RR was assumed to be H G H', where G = 2 x 2 additive genetic covariance matrix for linear RR. Heritability estimates were calculated assuming the additive variance was 4 times the sire variance. Analyses were by AIREMLF90 (Misztal et al., 2002).
|
RESULTS AND DISCUSSION |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
. The variances based on UNI, which did not account for H, were intermediate between those in MT2 under non-heat stress and heat stress. They were closer to those under no heat stress because the fraction of observations under no heat stress was greater than that under heat stress.
|
).
Whereas birth litter variances and residual variances under heat stress were similar to the variances estimated under non-heat-stress conditions, the sire variance doubled during heat stress months, as did the heritability estimate. The larger heritability estimate for carcass weight under heat stress indicated that selection for this trait under high THI would be efficient. Thus, selection under heat stress would be one method of selecting for heat tolerance. The genetic correlation between carcass weight during heat stress and non-heat stress was 0.42 (Table 1
), indicating that joint selection in hot and cold environments is possible.
Collin et al. (2001) investigated the effect of constant temperature ranging from 19 to 35°C. They found increased variability for BW gain and voluntary feed intake under highest temperatures, suggesting genetic differences among pigs from different litters. Pollott and Greeff (2004) investigated the level of genotype x environment interactions of Merino sheep for several traits in a range of flock environments in Australia. They found greater heritability for BW for flocks with lower contemporary group means.
The decrease in ADG under heat stress seems to be mainly the result of the decrease in voluntary feed intake (Huynh et al., 2005). Hermesch et al. (2006) found increasing heritability estimates for ADG of pigs under decreasing feeding levels: 0.05 for ad libitum, 0.13 for semiad libitum, and 0.28 for restricted feeding.
Random Regression Model
Estimates obtained with RR are in Table 2. The variances estimated with RR for non-sire effects were similar to the estimates from MT2. Estimates of the sire variances converted to the multiple-trait scale were:
|
The average H for the months considered in MT2 in which heat stress was observed and not observed was 3.4 and 0.5, respectively. The RR sire variances obtained for these values of H were very similar to the estimates obtained with MT2.
The genetic correlation between the intercept and slope was negative (–0.50). This negative correlation could have been due to increased sensitivity of genetically superior animals to heat stress (e.g., Kolmodin et al., 2002) or to the unbalanced data structure (Hermesch et al., 2006). Another possibility was that the RR with only a linear component was too simple and the estimates were inaccurate. However, attempts to make the RR model more complicated by extending the linear component to quadratic or linear splines with 2 knots were unsuccessful.
The rank correlations between breeding values generated using the 3 models are presented in Table 3. The greatest correlations (0.90 to 0.94) were between UNI, where heat stress and non-heat stress periods were averaged, and RR at H = 2.5 (intermediate), between UNI and MT2-non-heat stress, between MT2-heat stress and RR at H = 2.5, and between MT2-non-heat stress and RR at H = 0. The correlations between intermediate (RR, H = 2.5) or average heat stress (UNI) and the extremes (no heat stress or full heat stress) were moderate and ranged from 0.66 to 0.81. The same is true for the correlations between heat stress and non-heat stress of MT2. In the 2-trait analysis, data from periods of intermediate heat stress were included to provide a sufficient sample size and genetic link for the 2 traits, which could explain why the correlations between MT2 traits never dropped below 0.54. However, the correlation between the extremes of RR (H = 0; H = 5) was very low (0.18), which could be an artifact of using a simple model.
|
; van der Waaij, 2004). The present study showed that selection under non-heat stress conditions could lead to increased sensitivity to heat stress (e.g., negative correlation between slope and intercept). Thus, selection in heat-stress-affected areas should also consider performance under heat stress conditions to avoid or reduce losses due to heat stress.
If nucleus selection took place under environmental conditions with less heat stress compared with commercial production, the environmental sensitivity would be expected to increase. For the case of different environmental levels across farms, Knap (2005) proposed to integrate reaction norm parameters (slope and level) for environmental sensitivity into the breeding goal. These parameters were based on the contrast between performance on the nucleus and average commercial level. However, this approach requires extensive data collection and sophisticated data processing. In the present study the RR was on heat stress in a given commercial production environment. For the nucleus animals, effects of heat stress on finishing weight had been examined as in Zumbach et al. (2008) for the commercial farms; however, little heat stress could be detected. The genetic correlations of weight at finishing between "hot" and "cold" months were >0.9. The genetic correlations between weight at finishing of purebreds and carcass weight of the commercial crossbred animals were greatest (0.53) during the "cold" months (unpublished data).
In this study, nucleus farms had 25% more space available for animals and operated free of numerous disease challenges compared with their commercial counterparts. Whereas the nucleus farms were free of porcine reproductive and respiratory syndrome virus, Actinobacillus pleuropneumonia, and Mycoplasma hyopneumoniae, the commercial farms were not (Zumbach et al., 2007). Thus, on the commercial farms several factors (e.g., heat stress, stocking rate, and disease pressure) were acting simultaneously, resulting in decreased carcass weights during hot months. The effects of stocking density and disease challenge with regard to heat stress were reported by Kerr et al. (2003, 2005).
If genetic selection were performed at the nucleus level with good management, with some assumptions, the selection would use breeding values comparable to MT2-NHS. In this study, purebreds were evaluated based on crossbred data from commercial farms, in which case selection would use breeding values similar to UNI. Breeding values of UNI had the greatest correlations with RR at H = 2.5, or an average of cold and hot breeding values. Thus, selection using the crossbred information would account for an average heat stress at the crossbred farms. The relatively low genetic correlation between purebreds and crossbreds (see above) could justify the additional costs involved in including crossbred data in selection strategies (Zumbach et al., 2007).
Tables 4 and 5 quantify possible gains per commercial pig in response to various types of selection, assuming either MT2 or RR. These tables show average EPD of top 10 boars selected by different criteria for a period of 1 yr. Selection on MT2-NHS can be regarded as approximately equivalent to the current selection on purebreds, where no signs of heat stress could be detected (unpublished data). When the selection is on MT2-HS, the gain in carcass weight over MT2-NHS is 0.07 kg. Selection based on an average of MT2-NHS and MT2-HS results in a gain over MT2-NHS of 0.61 kg. When the evaluation procedure ignores the heat stress (UNI), the gain is 0.45 kg, which is better than selection using only MT2-NHS, but not as good as using the average. When the evaluation is by RR (Table 5), the relative advantage of using the average cold and hot performance is almost identical to the average with MT2. However, using UNI is only slightly better than using EPD of RR with no heat stress. All of the gains given above should be treated with caution, because other aspects such as the cost of data recording on commercial farms, reduced data quality, and use of unproven sires are not taken into account.
|
|
1 Corresponding author: Birgit.zumbach{at}norsvin.no
Received for publication May 18, 2007. Accepted for publication April 21, 2008.
|
LITERATURE CITED |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
