Search
Author
Title
Vol.
Issue
Year
1st Page

Journal of Animal Science - Animal Genetics

Genotype × environment interaction as it relates to egg production in turkeys (Meleagris gallopavo)

 

This article in JAS

  1. Vol. 88 No. 6, p. 1957-1966
     
    Received: Apr 01, 2009
    Accepted: Feb 11, 2010
    Published: December 4, 2014


    1 Corresponding author(s): lcase@uoguelph.ca
 View
 Download
 Share

doi:10.2527/jas.2009-2004
  1. L. A. Case 1,
  2. M. J. Kelly*,
  3. S. P. Miller* and
  4. B. J. Wood*†
  1. Department of Animal and Poultry Science, University of Guelph, Guelph, Ontario, Canada; and
    Hybrid Turkeys, Suite C, 650 Riverbend Drive, Kitchener, Ontario, Canada

ABSTRACT

Genotype × environment (G×E) interactions can reduce the accuracy of a model to predict the performance of an animal and have an undesirable influence if not accounted for when estimating breeding values. Consequently, identification of these G×E is necessary when considering a turkey breeding program. Reranking based on the genetic prediction of turkey egg production, fertility, and hatchability in different seasons was indicative of a potential G×E interaction. Quantification of the G×E interactions was based on the genetic correlation estimated when traits were expressed in different seasons. Egg production was expressed as the percentage of days with an egg produced; fertility represented the proportion of hatched eggs that contained a fertile embryo; and hatchability was defined as the percentage of fertile eggs that produced a live bird. Variance components and heritability for egg production, fertility, and hatchability were estimated using ASReml. The heritability (h2) of egg production was calculated to be 0.32 for both lines with the phenotypic and genetic variance, 141.3 and 45.58 (percent days with egg produced)2 and 118.3 and 38.35 (percent days with egg produced)2 for female and male lines, respectively. The h2 estimates for fertility were 0.08 in both lines with and of 293.3%2 and 24.03%2, and 576.9%2 and 48.43%2 for female and male lines, respectively. The hatchability h2, and estimates were 0.09, 267.1%2, and 24.44%2, respectively, for the female line and 0.15, 582.2%2, and 90.01%2 for the male line, respectively. Based on an animal model, the variance components were used to calculate estimated breeding values for each trait. The annual fluctuation in estimated breeding values resulted in the need to evaluate egg number, fertility, and hatchability as 2 traits, summer and winter lay. The correlation between the 2 traits was less than unity (female line: regg production = 0.76, rfertility = −0.20, rhatchability = 0.75 and male line: regg production = 0.86, rfertility = 0.19, rhatchability = 0.68) suggesting a G×E interaction, and animals will significantly rerank in genetic predictions for these reproductive phenotypes in different seasons of lay. Egg production, fertility, and hatchability in turkeys could be considered as 2 distinct traits in an animal model based on season of lay.



INTRODUCTION

Environmental sensitivity refers to the different phenotypic expression of a genotype in response to different environments (Kolmodin et al., 2003). Reproductive traits can fluctuate annually (Horn and Perenyi, 1974; Tona et al., 2007), and this may be due to phenotype regulation by different genetic controls in different seasons. If environmental sensitivity is not consistent for all individuals a genotype × environmental interaction (G×E) will result. Due to the varying magnitude of phenotypic change between environments there is a reranking of individuals for trait expression and, consequently, selection for superior performance in one environment will not necessarily result in enhanced performance under other sets of environmental conditions (Falconer and Mackay, 1996).

Genotype × environment interactions have been detected for egg production in layer hens and BW traits in broiler chickens (Mathur and Horst, 1994; Settar et al., 1999). High performance genotypes in the spring were associated with reduced EBV under the heat stresses of summer. As seen with BW, seasonal environmental changes may be producing the G×E observed for egg production in layer chickens. Currently, there is no documented literature investigating seasonal G×E interactions for turkey reproductive traits.

The aim of the current study was to estimate reproductive trait genetic parameters in male and female turkey lines and to investigate if a G×E interaction existed for egg production, fertility, and hatchability in different seasons. Identification of any interactions could lead to the consideration of distinct seasonal traits in selection decisions to more effectively manage the selection for egg production.


MATERIALS AND METHODS

All data utilized was from an existing database of information collected from industry sources; therefore, animal care and use committee approval was not required for the analysis.

Study Population and Management

The populations used in this study were a commercial Large White female (ntotal = 430,975; nsire = 2,195; ndam = 10,471) and male (ntotal = 423,159; nsire = 1,772; ndam = 15,949) line with a breeding objective balanced between commercial and reproductive traits. Full pedigree information was available for each line. A stronger selection emphasis was placed on reproductive traits in the female line compared with the male line. Traits were recorded between 1995 and 2007 and trait values for egg production (nfemale line = 8,809; nmale line = 13,875); fertility (nfemale line = 8,787; nmale line = 15,735), and hatchability nfemale line = 8,765; nmale line = 15,631) were recorded. The male to female ratio in the female and male lines was 1 to 5 and 1 to 10, respectively. Artificial insemination was used for all matings with 203 flocks across the 13-yr period, and a consistent number of hens started lay throughout the year. All eggs were stored for between 4 and 14 d.

Rearing up until 20 wk of age was under a standard commercial production environment and feeding regimen. After 20 wk of age hens were placed onto a commercial parent stock BW-restriction diet. Light conditions were controlled in dark-out barns, and lighting consisted of equal 12-h periods of light and dark with a light intensity of approximately 107 lx. From 16 to 30 wk, birds were exposed to 6-h periods of dark and 18 h of light with an intensity of approximately 88 lx. Light intensities varied with flock requirement, temperament, and other management factors. From 30 wk of age the light conditions consisted of 16 h of light with an intensity of approximately 107 lx and 8-h periods of dark.

Egg production was evaluated as percentage days with egg produced between 210 and 420 d of age. Fertility was measured as the proportion of eggs that were candled fertile. Hatchability was calculated as the percentage of fertile eggs that produced a live poult.

Statistical Analysis

A multiple-trait analysis was conducted by treating traits across 2 seasons as independent. The seasonal partition of data was winter lay (January to May, G×E winter) and summer lay (July to November, G×E summer). June and December were considered transitional periods. Partitioning of these periods was based on observed EBV transition patterns from the more consistent values associated with the summer lay and winter lay seasons. Genetic correlations between the 2 seasons were then used to determine if a G×E influenced the reproductive traits. The following animal model was used to estimate additive genetic effects:where y was the evaluated trait (egg production, fertility, and hatchability) and µ the average trait value of all animals. Flock was a fixed classification factor incorporated to account for the temporary environmental effects influencing each group of hens laying in a barn at the same time, and the age regression, where b represents the linear coefficient, accounts for the age of the hen within the flock contemporary group. The a and e represent the random animal genetic effect and residual random effect, respectively. The ANOVA was used to determine the significance of fixed effects. The random animal and residual effects were assumed to be normally distributed with a mean of zero and a variance ofwhere A represents the relationship matrix, and I is an identity matrix. Variance components and heritability for egg production, fertility, and hatchability were estimated with ASReml (Gilmour et al., 2002). Values were preliminarily estimated for each reproductive trait in a single trait model. A bivariate model was then used to generate parameter estimates (Tables 1, 2, and 3).

Table 1.

Please see the pdf to view this table.

 
Table 2.

Please see the pdf to view this table.

 
Table 3.

Please see the pdf to view this table.

 

To account for genetic trends in the EBV estimated by BLUP for egg production, fertility, and hatchability as the result of selection over time, monthly average EBV from the bivariate model were adjusted aswhere EBV equals EBVall, EBVG×EWinter, or EBVG×ESummer, and µ was the average breeding value for all animals. Hatch_year was a fixed classification effect representing the hatch year of the turkey, and hatch_month was a fixed class effect representing the month a poult was hatched. These parameters account for genetic trends in the EBV estimated by BLUP for egg production, fertility, and hatchability as the result of selection over time. These adjusted values were used to estimate the correlation between summer and winter lay seasons and also to calculate annual EBV fluctuations as the difference between the greatest monthly EBV value and the least monthly EBV value. A genetic correlation between summer and winter lay of less than 1 would indicate the presence of a G×E interaction.


RESULTS

The fixed effects in the animal model were significant for all reproductive traits (P < 0.05), except for the effect of age on hatchability (P < 0.12 for the male line and P < 0.17 for the female line). Based on single-trait model analysis, phenotypic performance for turkey reproductive traits fluctuated annually as shown in Figure 1. The patterns of EBV, used to determine appropriate season definitions, showed that the EBV for June and December were not consistent with either summer (January to May) or winter (July to November) EBV. These 2 mo were, therefore, considered transitional and were not included in either seasonal trait definition. Because the rank correlation between seasons is the focus of the present study, the omission of these data in the bivariate model will not greatly affect the analysis of reranking to detect a G×E.

Figure 1.
Figure 1.

The effects of month of hatch on phenotypic performance of turkey egg production (a), fertility (b), and hatchability (c). Dashed lines represent the 95% confidence interval. Color version available in the online PDF.

 

Plotting of the egg production phenotype in both the female and male line resembled a normal distribution, with a slightly elongated tail toward lesser trait values. The annual breeding value fluctuations for egg production are shown in Figure 2a. Breeding value estimates were not consistent between the summer lay season and the winter lay season in either line, and these trends persisted when seasonal traits were plotted independently in Figure 3a. Analysis using a bivariate model resulted in greater heritability values for the seasonal traits, indicating that a larger proportion of phenotypic variance was explained by genetic factors (Table 1). The correlation between summer and winter lay traits was 0.76 and 0.86 for the female and male lines, respectively. These values depart from a correlation of unity (r = 1) and indicate that selection decisions made on one environment would decrease the accuracy of selection in the other environment due to ranking inaccuracies of selection candidates.

Figure 2.
Figure 2.

The effects of hatch month on egg production (a), fertility (b), and hatchability (c) EBV. Winter lay season is defined as January to May. Summer lay season is defined as July to November. Annual fluctuation in egg production EBV is 1.21% days with egg produced in the female line and 0.45% days with egg produced in the male line. Annual fluctuation of fertility is 0.33% in the female line and 0.64% in the male line. Annual fluctuation in hatchability EBV is 0.33% in the female line and 0.64% in the male line. Dashed lines represent the 95% confidence interval. Color version available in the online PDF.

 
Figure 3.
Figure 3.

The effects of hatch month on winter lay EBV and summer lay EBV for egg number (a), fertility (b), and hatchability (c). Annual fluctuations in the winter and summer lay egg production EBV were 1.49% days with egg produced and 1.72% days with egg produced, respectively, for the female line. Annual fluctuations in the winter and summer lay egg number EBV were 1.29% days with egg produced and 1.16% days with egg produced, respectively, for the male line. Annual fluctuations in fertility winter and summer lay EBV were 0.72 and 0.64%, respectively, for the female line. Annual fluctuations in fertility winter and summer lay EBV were 1.08 and 1.27%, respectively, for the male line. Annual fluctuations in the winter and summer lay EBV were 0.71 and 0.59%, respectively, for female line hatchability. Annual fluctuations in the winter and summer lay EBV were 1.39 and 1.27%, respectively, for the male line hatchability. Dashed lines represent the 95% confidence interval. Color version available in the online PDF.

 

Variance component and heritability estimates for fertility are shown in Table 2. Using a bivariate model did not change heritability estimates; however, incorporating a G×E interaction reduced the level of difference between variance components in the summer and winter traits. Breeding values were not consistent annually in both the single trait and multiple trait analysis, and EBV were greater in the winter season than in the summer season (Figures 2b and 3b). The average fertility value was greater and phenotypic SD smaller in the female line (µ = 85.7%, SD = 12.2%) than the male line (µ = 75.4%, SD = 24.7%). The increased mean value and reduced SD resulted in a non-normal distribution of fertility in the female line, as shown in Figure 4. The correlation between seasonal fertility traits was −0.20 for the female line and 0.43 for the male line. These values are less than 1, which is indicative of a G×E for fertility.

Figure 4.
Figure 4.

Cumulative density of phenotypic performance of fertility in the female line.

 

The hatchability phenotype followed a normal distribution; however, there was a truncation of the upper tail at the maximum value of 100%. This truncation of the tail was not as severe as the pattern seen for fertility and was assumed to not have largely influenced the parameter estimates. Incorporating a G×E into the model increased heritability estimates for the seasonal traits in both lines, except for winter lay in the female line. The range in EBV for the female and male lines is shown for a single annual trait in Figure 2c and from the multiple trait model in Figure 3c. Breeding value estimates from both models showed elevated values in the summer lay season relative to the winter lay season. The correlation between hatchability expressed as a summer and winter lay trait (female line, r = 0.75 and male line, r = 0.68) indicates that a G×E interaction will result a ranking discrepancy for one seasonal trait when selection decisions are based on performance in the other season (Table 4).

Table 4.

Please see the pdf to view this table.

 

DISCUSSION

The impact of G×E interactions on a breeding program must be considered, because genetic correlations between environments that are less than unity can result in a reranking of selection candidates between environments (Falconer and Mackay, 1996). Robertson (1959) suggested that genetic correlations of less than 0.8 were indicative of a significant G×E. More recently, Smith and Banos (1991) and Mulder and Bijma (2006) investigated G×E interactions in dairy cattle populations and determined that if genetic correlations between environments are greater than 0.8 to 0.9, genetic gain can be increased by selection across environments. The calculated genetic correlations for reproductive traits in the turkey ranged from −0.20 to 0.86, indicating that single-trait selection is limiting the possible rate of genetic progress. This is a result of the reranking of performance levels in one environment when EBV are based on measurements taken in the other season. As a result, the structure of selection programs related to turkey reproductive traits should be considered to increase rates of genetic progress in both seasons.

To improve the response to selection, the structure of turkey breeding objectives can be altered to include 2 seasonal traits for the egg production, fertility, and hatchability phenotypes. Hens are routinely in production for between 24 and 30 wk, and as a consequence, many turkeys will lay across seasons, which necessitates the genetic improvement of both seasonal traits. When genetic correlations are greater than 0.61, a single breeding program that estimates breeding values in both environments and has a breeding goal to simultaneously improve genetic gain in both environmental traits is most effective (Mulder et al., 2006). The calculated correlations in the current analysis meet these criteria, with the exception of fertility. Incorporating reproductive traits as EBVG×EWinter and EBVG×ESummer into the standard selection index to evaluate the birds based on performance in both seasons could therefore increase the rate of genetic progress for the traits. Mulder et al. (2006) determined that for traits with correlations of less than 0.61, or with increased selection intensity, it can be more effective to run 2 separate breeding programs specific to each environment. This option, however, is unrealistic for turkey production systems because each evaluated animal can perform and contribute genetics to offspring performing in both production seasons. Accordingly, the EBV estimation and incorporation of season-specific reproductive traits into breeding objectives is the most feasible implementation of the detected G×E interaction into turkey breeding programs.

Determining the unique economic values for winter lay and summer lay traits in a breeding objective, based on financial significance to the industry, can be difficult when attempting to improve the rate of genetic gain using the multiple trait approach (Wood, 2009). One approach could involve equally weighting each seasonal trait. A second consideration could involve increasing the relative importance of the seasonal trait with decreased performance in an effort to increase performance consistency throughout the year. As a result, there are 2 potential ways to improve overall reproductive performance, either by improving both seasonal traits at the same rate or by placing additional selection emphasis on the decreased performance environment to also increase phenotypic consistency between seasons.

Considering the impact of current turkey breeding programs, genetic selection has led to major improvements in growth characteristics in modern commercial birds (Havenstein et al., 2007). This has been due to an industry focus primarily on improving meat yield and efficiency, and less selection emphasis has been placed on reproductive ability due to its decreased relative economic value compared with growth characteristics (Wood, 2009). Extensive research has been conducted on reproductive traits in broilers (Sharma et al., 1984; Tona et al., 2007), and because selection emphases and performance requirements are similar in broiler and turkey breeding programs, broiler variance component estimates and trait expression characteristics can be used as preliminary indicators of potential values in the turkey, where there have been fewer reported studies.

Heritability estimates for egg production in the turkey have ranged from 0.13 to 0.61 (McCartney et al., 1968; Nestor et al., 1996), and it has been shown that single trait selection for egg production in the turkey can result in significant increases in egg numbers within a small number of generations (Asmundson and Lloyd, 1935; Knox and Marsden, 1954). Small to moderate heritability estimates for fertility of 0.42, 0.18, and 0.34 have been estimated in the turkey hen (McCartney et al., 1968; Nestor et al., 1972; Dunnington et al., 1990). Similar to the fertility reproductive trait, reported heritability estimates for hatchability in the turkey of 0.16 and 0.17 are in the small range (McCartney et al., 1968; Nestor et al., 1972). The moderate heritability estimate for egg production and low estimated values for fertility and hatchability in the turkey population in the current analysis are in agreement with previously published estimates.

The heritability, genotypic, and phenotypic variances were not consistent between seasons for egg number, fertility, and hatchability without consideration of the G×E interaction and when modeled as 2 distinct traits. This indicates that seasonal environmental influences affect egg production, fertility, and hatchability phenotypes to a different extent suggesting a G×E, which was supported by the genetic correlation below 1.0 for each trait expressed in different seasons. It has been shown that environmental sensitivity decreases the heritability of traits as the phenotypic expression is dispersed over different environments (de Jong, 1990). This trend was observed in the current analysis, as heritability estimates were greater from the bivariate model for egg production and hatchability in both lines. The increased heritability estimates can be attributed to increasing genetic variance, when each reproductive phenotype was considered as 2 seasonal traits. This indicates that the bivariate model can help to remove a portion of the confounding effect caused by differential genetic control of performance across the 2 seasonal production environments.

A strong and negative correlation (−0.20) between fertility expressed in the summer and winter seasons was detected in the female line. Using an animal model to estimate variance components, the distribution of trait values was assumed to be normal. Because of selection for reproductive traits in the female line, fertility performance is nearing the biological maximum of 100% with an increased average value and decreased variance. This non-normal distribution of fertility in the female line in conjunction with a small range in trait values resulted in a restricted ability of fertility to change in response to environmental fluctuations as values remain increased under all conditions. In contrast, the genetic correlation in the male line is more representative of environmental influences seen in other traits in this study. In this population, the phenotypic average is less and the variance in phenotypic values is greater, allowing for greater freedom to alter the performance levels in response to climatic fluctuations. When the distributions of egg production and hatchability performance were plotted the phenotypic records were close to resembling a normal distribution, and the impact on non-normality was assumed to have a little impact on parameter estimations for these traits.

The genetic trends detect in this study are related to turkey reproductive physiology, and turkey populations are stimulated by environmental cues to come into season and lay. Critical day length must be exceeded together with a sufficient contrast in light intensity, to simulate night and day, for egg production to occur (Siopes, 1994). A fluctuating trend in reproductive performance was evident in the studied turkey population that could not be explained by changes in photoperiod or intensity because these were managed in lightproof barns with all the light controlled. Barn management cannot completely eliminate turkey exposure to seasonal changes in temperature and humidity, and as a consequence, the ambient environment is not consistent throughout the year. As a result, G×E interactions influenced by climate may be expressed in the turkey.

The preliminary evidence for G×E interactions is change in performance levels between environments, and this has been evidenced in broilers where seasonal environmental differences affect reproductive performance. Changes in performance of varying magnitude as the ambient temperature and humidity fluctuate have resulted in a reranking of genotypes (Mathur and Horst, 1994; Yalcin et al., 1997). Evidence of the ambient environment creating a G×E was reported by Hartmann (1990), as the correlation between egg production, expressed in different geographical locations, was below unity. Egg production has been shown to decrease in response to heat stress in broilers (McDaniel et al., 1995). This trend was also reflected in a study of the variation in broiler egg production levels between seasons, with the least egg production levels expressed in the summer months (Torshizi et al., 2008). The environmental influence on broiler egg production supports the G×E detected in the current study of egg production in the turkey. Broiler fertility is least in the summer compared with other times of the year (Smyth and Leighton, 1953; Horn and Perenyi, 1974; Keirs, 1982). Similarly, hatchability of eggs drops in the heat of the summer season compared with cooler seasons (Horn and Perenyi, 1974; Tona et al., 2007). The impact of seasonal changes in light exposure, humidity, and temperature on fertility and hatchability in broilers substantiates the G×E observed for these reproductive traits in the turkey.

Genotype × environment interactions may not be consistent across genetic groups, and previous studies have determined that greater growth rate broiler chicken lines are more sensitive to environmental fluctuations (Settar et al., 1999). The amount of selective pressure on a trait could influence the ability of the animal to respond to changing environments, and this is reflected in the variation in sensitivity of broiler growth rate to environmental fluctuations, based on selective pressure. Similarly, a trend was seen in the present study for egg number and fertility. The female line, with greater selection pressure on reproductive traits, produced weaker correlations between the summer and winter EBV. It could be considered that increasing the selection pressure on a trait may increase susceptibility to changes in expression environment (Kolmodin et al., 2003). This indicates that performance traits in male line turkeys may be more sensitive to changes in performance environment, whereas greater sensitivity of reproductive traits could be expected in female line birds, based on the relative selection pressures in the breeding program.

The advantage of using a multiple trait model is the consideration of each environment as a distinct character state, and the bivariate model provides a starting point for G×E interaction studies to which alternative models can be compared. Multiple trait models thus enable the differences between seasonal environments to be accounted for directly. A key advantage to the multiple-trait approach is the straightforward implementation of the results into a breeding program, with the ability to assign different economic values to each trait. There are, however, alternative models to estimate environmental impact on trait expression, including random regression models, which can be considered to validate results from a multiple trait model.

Random regression models used to study environmental influence on performance detect a G×E as significant differences in reaction norms which indicate individual variability in environmental sensitivity (de Jong, 1990). This approach has been used to study changes in laying hen and turkey egg production performance throughout a single lay cycle (Anang et al., 2002; Kranis et al., 2007), although there are no reported studies on the effect of seasonal environment on total egg production for a lay cycle. Previous random regression analysis in the turkey showed that multiple-trait models have greater heritability values; however, the mean squared error is reduced using a random regression approach (Kranis et al., 2007). The primary author is interested in considering the use of a random regression model on the data set used in the present study to determine if both analyses yield the same conclusion. A random regression model can be used if there is interest in determining the characteristics of environmental sensitivity in the population; however, the bivariate approach can be considered most useful to analyze the G×E interactions for consideration in a breeding program.

In conclusion, egg production, fertility, and hatchability could each be considered as 2 separate seasonal traits in a turkey breeding program. Although commercial breeding operations attempt to maintain a constant internal environment, the ability of the turkey to respond to seasonal environmental factors is expressed in the G×E interaction. A new multiple-trait animal model incorporating egg production, fertility, and hatchability as winter lay and summer lay traits could increase the consistency of prediction for animal performance generated from a BLUP model. Because reproduction by a hen occurs in the summer and winter seasons, selection indexes should be adjusted to evaluate turkey egg production in these distinct environments. By considering discrete seasonal egg production EBV in the selection of parents, a superior combination of summer and winter lay genotypes could be transmitted to the next generation of offspring.

 

References

Footnotes


Comments
Be the first to comment.



Please log in to post a comment.
*Society members, certified professionals, and authors are permitted to comment.