J. Anim Sci. 2007. 85:1865-1872. doi:10.2527/jas.2006-665
© 2007 American Society of Animal Science
Association of single nucleotide polymorphisms in the leptin gene with body weight and backfat growth curve parameters for beef cattle1
J. L. Lusk2
Department of Agricultural Economics, Oklahoma State University, Stillwater 74078
 |
Abstract
|
|---|
Previous research has identified differences in carcass characteristics across SNP in the bovine leptin gene at slaughter, but before feedlot operators implement selection and sorting strategies, more information is needed to determine how carcass characteristics change over time. The objective of this study was to investigate the effect of 2 leptin SNP on growth curve parameters for BW and backfat. Two SNP (UASMS2 and R25C) were genotyped on 1,653 cross-bred steers and heifers in a commercial feedlot. Up to 4 serial measures of BW and ultrasound estimates of backfat thickness were taken for each animal from the time of placement on feed to slaughter. The measures were used to estimate growth models that describe changes in BW and backfat thickness as a function of days on feed. Data analysis was carried out by estimating nonlinear mixed models to determine the individual and joint effect of each SNP on growth curve parameters. Brody growth curves were fit to the BW data. Variations in the R25C SNP did not significantly affect growth parameters individually or in combination with the UASMS2 SNP. Variations in the UASMS2 SNP were significant in Brody growth curve parameters for BW growth (P < 0.001). The genotype UASMS2-CC was the heaviest at the beginning of the feeding period and exhibited the largest asymptotic mature BW, but UASMS2-TT cattle exhibited the fastest rate of BW growth. A modified power function was fit to the serial ultrasound backfat measures. Models that included the combined effect of the R25C and UASMS2 SNP provided the best fit to the data. Genotypes differed significantly in power function parameters for backfat growth (P < 0.001). The R25C-CC/UASMS2-TT cattle had the smallest backfat thickness at placement. The genotype R25C-CC/UASMS2-TT exhibited the fastest backfat growth rate, whereas backfat in R25C-CC/UASMS2-CC cattle grew at the slowest rate. The association between leptin genotype and growth in BW and backfat presents opportunities to identify genetically distinct cattle and to differentially optimize feeding times accordingly.
Key Words: backfat • beef cattle • growth curve • leptin gene • single nucleotide polymorphism
|
INTRODUCTION
|
|---|
Leptin is a protein that regulates appetite and metabolism. The economic importance of factors related to appetite, such as feed intake, BW gain, and fat deposition has led to studies of the effect of leptin on carcass characteristics in beef cattle. Serum concentrations of leptin in the bloodstream obtained 24 h before slaughter have been found to be significantly correlated with marbling, yield grade, and dressing percentages in beef cattle (Geary et al., 2003
). Differences in leptin in the bloodstream are due, in part, to genetics. Research has shown genetic diversity in the ability of animals to produce and recognize leptin and has led to the identification of the obese gene, also known as the leptin gene (Friedman and Halaas, 1998; Buchanan et al., 2002).
A number of SNP have been reported in the bovine leptin gene (Fitzsimmons et al., 1998; Buchanan et al., 2002; Nkrumah et al., 2005). Polymorphisms in the locus of the leptin gene have been associated with carcass fat measurements in beef bulls (Fitzsimmons et al., 1998) and with lean yield and tenderness (Schenkel et al., 2005), quality and yield grade (Kononoff et al., 2005), and feed intake and average daily gain in beef cattle (Nkrumah et al., 2005), among other factors.
To date, previous studies have focused on measures such as BW, lean yield, marbling, and total feed intake at the time of slaughter. Before such genetic information is utilized by feedlots to improve management and marketing decisions, work is needed to identify whether and how the leptin genotype is associated with changes in carcass characteristics across time. Such growth models are necessary to optimally sort animals into feeding and marketing groups.
The objective of this study was to investigate the effect of 2 leptin SNP on growth curve parameters for BW and backfat estimated from a relatively large sample of commercially fed cattle.
|
MATERIALS AND METHODS
|
|---|
Animals and Genotyping
Animal Care and Use Committee approval was not obtained for this study because the data were obtained from an existing database of performance records.
One thousand six hundred sixty-eight crossbred steers and heifers fed in a commercial feedlot in western Kansas were used in the study. All animals were placed on feed between August and November 2004. The average BW at the beginning of the feeding period was 313.3 kg (SD = 45.6). At the beginning of the feeding period at the feedlot, cattle were weighed, and ultrasound readings were taken to estimate backfat thickness using procedures that utilize image analysis software (Brethour, 1994). The cattle were then placed in pens. Cattle were reweighed, and additional backfat ultrasound measures were taken up to 3 additional times before slaughter. All cattle were fed the same corn-based ration during the feeding period, with estimated NEm and CP contents of 2.1 Mcal/kg and 13.2%, respectively, on a DM basis. Cattle were fed for an average of 139 d (SD = 22) and were slaughtered at an average BW of 552.3 kg (SD = 59.5). The final slaughter date for each pen was determined by feedlot managers attempting to maximize profit based on the initial ultrasound readings, the cost of gain, and visual inspection.
To analyze genetic variation in the leptin gene, a hair was plucked from each animal. Based on findings in Schenkel et al. (2005) and Woodward et al. (2005), 2 particular SNP in the bovine leptin gene locus were expected to be predictive of the characteristics of interest in this study; that is, AB070368.2:g.528C>T and R25C. For the sake of consistency with previous literature and for ease of exposition, the SNP AB070368.2:g.528C>T is referred to as UASMS2 for the remainder of the paper. Note, further, that the SNP R25C has previously been referred to as EXON2FB or as R4C in previous studies (Liefers et al., 2003; Nkrumah et al., 2005; Schenkel et al., 2005). Genotyping was carried out by Igenity (a business unit of Merial Ltd., Atlanta, GA) at a commercial genotyping facility. Primers and probes were designed according to procedures previously described elsewhere (Buchanan et al., 2002; Nkrumah et al., 2004, 2005). Both SNP contain 2 alleles (C and T) and were homozygous (CC or TT) or heterozygous (CT).
In this study, the independent effect of each SNP and the interaction between the 2 SNP on growth curve parameters were investigated. For the UASMS2 SNP, the majority of the cattle were genotype CC (n = 827), whereas 41% of the sample was heterozygous (n = 690) and the remainder was genotype TT (n = 136). The sample differed by the R25C SNP as follows: genotype CC (n = 522), genotype CT (n = 808), and genotype TT (n = 323). In the combined analysis, there were 3 possible outcomes for each SNP (CC, CT, or TT), theoretically making 9 possible genotypes available for analysis. However, when the combined effects of both SNP were investigated, 3 genetic combinations occurred with very low frequencies in the population, and cattle with these combinations were simply pooled in an "other" category. Thus, the combined analysis focused on the following 7 genotypes R25C-CC/UASMS2-CC (n = 131), R25C-CC/UASMS2-CT (n = 226), R25C-CC/UASMS2-TT (n = 125), R25C-CT/UASMS2-CC (n = 392), R25C-CT/UASMS2-CT (n = 406), R25C-TT/UASMS2-CC (n = 304), and all other R25C/UASMS2 combinations (n = 29). The Lewontin’s D' measure of linkage disequilibrium was 0.8695, and the R2 measure of linkage disequilibrium was 0.4917.
Statistical Methods
Body weight growth was modeled using the Brody growth curve (Kaps et al., 1999). The general model is
 | [1] |
where Wit is BW at time t for animal i, A represents the asymptotic mature BW, W0 represents BW at t = 0, and k is the maturing rate index representing the ratio of maximum growth rate to mature size. For estimation purposes, equation [1] was modified to account for the facts that: a) the best-fit values for k were close to zero (Staniar et al., 2004), b) several observations were obtained for each animal i, c) there was likely to be variation in A and W0 within a genotype, and d) A and W0 were likely to be correlated for each animal. It is also possible that k might vary within genotype, but the results indicate that the variance of k was not significantly different from zero when a random effect was estimated for this term. Thus, equation [2] was fit for each genotype:
 | [2] |
where ai 
N(0, 
2a) and wi N(0, 2w) are individual-specific random effects with Cov(ai, wi) = 
aw, 
it N(0,s2) is an independently distributed error term, and k = exp(k').
Four models were estimated to determine the relationship between each SNP and the growth curve parameters. As an initial investigation, equation [2] was fit separately for the 3 genotypes (CC, CT, and TT) in the UASMS2 SNP using the nonlinear MIXED procedure (SAS Inst. Inc., Cary, NC). This procedure was then repeated for the 3 genotypes (CC, CT, and TT) in the R25C SNP. In addition to the investigations into the independent effects of each SNP, the growth equation shown in [2] was fit for each of the combined 7 genotype categories. Fitting a separate model for each of the 7 genotypic categories permitted the most general investigation possible because it allowed growth curve parameters to differ for each of the 7 UASMS2, R25C combinations. To investigate whether any of these specifications exhibited significant explanatory power, all genotypes were pooled and equation [2] was estimated using all cattle in the data set.
Likelihood ratio tests were used to determine whether the growth models differed by genotype and to determine whether the interaction between genotypes was statistically significant. To first test whether each SNP independently influenced the growth curve parameters, equation [2] was estimated for each of the 3 genotypes in the UASMS2 SNP as follows. First, denote the sum of the maximum log-likelihood function values across the 3 models associated with the UASMS2 SNP as LLUASMS2 = LLUASMS2-CC + LLUASMS2-CT + LLUASMS2-TT. Similarly, let LLR25C represent the sum of the maximum likelihood function values for each of the 3 models associated with the genotypes in the R25C SNP, where LLR25C = LLR25C-CC + LLR25C-CT + LLR25C-TT. Further, let the maximum log-likelihood function value for the pooled model be LLP. The null hypothesis that the growth model parameters are the same across each of the 3 UASMS2 genotypes (CC, CT, TT) can then be tested by calculating the likelihood ratio value –2[LLP – LLUASMS2], which is distributed as 
2 with 14 degrees of freedom. The null hypothesis that the growth model parameters are the same across each of the 3 R25C genotypes (CC, CT, TT) can be tested by calculating the likelihood ratio value –2[LLP – LLR25C], which is also distributed as 2 with 14 degrees of freedom.
In addition to these models, equation [2] was estimated for each of the 7 combinations of the UASMS2, R25C SNP. The null hypothesis that the growth model parameters are the same across the 7 genotype categories can then be tested by calculating the likelihood ratio value 
which is distributed as 2 with (g-1)c degrees of freedom, where g is the number of combined genotypic categories and c is the number of model parameters, with c = 7 in this case. As a final inquiry, it is prudent to ask whether the most general specification where a model is estimated for each of the 7 UASMS2, R25C combinations provides a significantly better fit to the data than does the analysis of a single SNP. To test the null hypothesis that the growth model parameters provided by analyzing only the 3 variations in the UASMS2 SNP are not significantly different from the growth model parameters obtained by investigating the 7 UASMS2, R25C combinations, the likelihood ratio value 
can be compared against the critical 2 value with 28 degrees of freedom.
Because equation [2] is nonlinear in the random components, ai and wi, the expected BW at a given day t cannot be determined by simply inserting the estimated parameters into equation [1] and solving. However, given the estimated growth models, it is instructive to compare the expected BW of each genotype at different days on feed. Expected BW at day t is given by
 | [3] |
The integral in equation [3] was evaluated via Monte Carlo integration using 1,000 "intelligent", quasi-random Halton draws (Morokoff and Caflisch, 1995; Sloan and Wozniakowski, 1998).
The modified power function was used to determine how backfat changed with days on feed (Brethour, 2000). The general model is
 | [4] |
where bfatit is the projected backfat as measured by ultrasound at t days on feed for animal i, bfat0 is backfat thickness at the beginning of the feeding period, and 
represents the rate of increase in backfat. For estimation purposes, equation [4] was modified to account for a) the best-fit values for were close to zero, b) several observations were obtained for each animal i, and c) there was likely to be variation in bfat0 and within a genotype.
Taking these factors into consideration, equation [5] was fit for each genotype:
 | [5] |
where = exp('), bi N(0, 2b) and 
i N(0, 2) are individual-specific random effects with Cov(bi, i) = 0, and it N(0, 2) is an independently distributed error term. As with the BW growth models, likelihood ratio tests were used to determine whether the backfat models differed by genotype and to determine whether the interaction between genotypes was statistically significant. Equation [5] was estimated using data pooled from across all genotypes. In addition to this model specification, the independent effect of UASMS2 was investigated by estimating 3 models for each of the polymorphisms in the UASMS2 SNP. To investigate the independent effect of R25C, 3 models were estimated for each of the variations in the R25C SNP. Finally, to explore the combined effect of both genotypes, the 7 genotypic-specific models were estimated and the likelihood ratio test statistics were calculated as previously described. To determine the expected backfat at time t, the integral given by equation [6] was evaluated via Monte Carlo integration:
 | [6] |
|
RESULTS AND DISCUSSION
|
|---|
Maximum log-likelihood function values for each of 4 competing model specifications for BW growth are reported in Table 1. The maximum log-likelihood function value for the most restricted pooled model (model 4) was –26,237.7. Allowing the growth curve parameters to differ by variations in the UASMS2 SNP (model 1) significantly increased the likelihood function value to –26,216.6, implying that the null hypothesis that the growth curve parameters are equal across the 3 polymorphisms in the UASMS2 SNP can be rejected at the P < 0.001 level. In contrast, the hypothesis that BW growth parameters differ by the 3 polymorphisms in the R25C SNP (model 2) cannot be rejected (P = 0.999). Comparing the likelihood function value of the most general model specification that permits growth curve parameters to differ across the 7 R25C, UASMS2 combinations (model 3 with log-likelihood function value of –26,200.7) to the model specification where growth curve parameters to differ only by the UASMS2 SNP (model 1 with log-likelihood function value of –26,216.6) reveals that the more general specification is not supported by the data. That is, the Brody growth curve parameters estimated using the specification that includes the combined effects of the R25C, UASMS2 SNP is not statistically different (P = 0.283) from the specification that only includes variations in the UASMS2 SNP. This implies that model specification 1, where model growth curve parameters differ only by the variations in the UASMS2 SNP, is the most appropriate model.
Table 2 reports Brody growth curve parameters for this preferred model specification. The R2 values for the models fit to each of the 3 UASMS2 genotypes are in excess of 0.99, indicating the Brody growth curve provides a good fit to the data. The UASMS2-CC genotype had the greatest asymptotic mature BW, whereas UASMS2-TT exhibited had the lowest. The UASMS2-CC genotype had the greatest BW at the beginning of the feeding period at about 313 kg, whereas UASMS2-TT genotype had the lowest at about 295 kg. Estimates of the SD of asymptotic mature BW were approximately 78 kg (e.g., 6,0500.5 = 78), for all 3 genotypes. The UASMS2-CC genotype exhibited the largest variability in terms of BW at the beginning of the feeding period with an estimated standard deviation of 1,9600.5 = 44.3 kg. Estimates also reveal strong within-animal correlation between asymptotic mature BW and beginning BW. The UASMS2-TT genotype, while having the lowest asymptotic mature BW, exhibited the largest growth rate parameter.
Figure 1 plots the estimated growth curves for each of the 3 UASMS2 genotypes by reporting the mean estimated BW (obtained by substituting the estimates from Table 1
into equation [3]) for each day from 65 to 175 d on feed. The figure shows the growth curve for the UASMS2-TT genotype falls well below that for the other 2 genotypes over the 65- to 175-d time period. Body weight gain is a measure that is of direct relation to profitability. Data in Figure 1 can be used to show that average daily gain through the first 150 d on feed for UASMS2-CC is 1.742 kg/day, average daily gain for UASMS2-CC through 150 d is 1.728 kg/d, but average daily gain over the same time period for UASMS2-TT is only 1.695 kg/d. Thus, up to 150 d on feed genotype UASMS2-CC cattle gain an average of about 0.047 more kg per day than UASMS2-TT cattle (P < 0.001), a difference which amounts to a total of 7.05 kg over a 150-d time period. Given that BW market prices averaged about $1.89/kg (or $86/cwt) in 2006, this implies a $13.35/head revenue difference between UASMS2-CC and UASMS2-TT cattle. For cattle marketed on a grid, where heavy carcasses receive discounts as high as $0.55/kg (or $25/cwt), it should be apparent that UASMS2-CC cattle have a much greater chance of receiving overweight discounts than UASMS2-TT cattle. Estimates reveal it would take more than 1 mo (36 d to be exact) longer for the UASMS2-TT genotype to reach 600 kg, on average, than the cattle of the UASMS2-CC genotype.
Table 3 reports maximum log-likelihood function values for each of 4 competing model specifications for backfat growth. Results reveal that the independent effect of the UASMS2 SNP (model 1) had a significant effect on backfat growth parameters (P < 0.001), but the independent effect of the R25C SNP (model 2) did not (P = 0.408). The most general model specification that included the combined effect of the R25C and UASMS2 SNP (model 3) exhibited more explanatory power than did either specification analyzing variations in only 1 SNP models (P < 0.001). Thus, the preferred model specification is the one that permits backfat growth curve parameters to vary across the 7 categories associated with combined polymorphisms in the R25C and UASMS2 SNP.
Table 4 reports parameter estimates for the modified power function for the backfat growth models for this statistically preferred model specification. The R2 values for each model are in excess of 0.95, indicating a good fit to the data. The R25C-CC/UASMS2-TT genotype had the lowest average backfat thickness at placement, and cattle in the other genotype category had the greatest average backfat thickness. The R25C-CT/UASMS2-CT genotype exhibited the greatest rate of growth with = 0.012, whereas the R25C-CC/UASMS2-CC genotype exhibited the lowest rate of growth with = 0.0085.
View this table:
[in this window]
[in a new window]
|
Table 4. Estimates of modified power function growth curve parameters for backfat for preferred model specification
|
|
Figure 2 shows the expected backfat thickness for each genotype from 75 to 175 d on feed. The R25C-CC/UASMS2-CC genotype had the lowest expected backfat over this time period. Figure 2 shows that although the R25C-CC/UASMS2-TT genotype had lowest expected backfat at the beginning of the feeding period, that by about 155 d on feed, it had the greatest expected backfat of all 7 genotypes. It is useful to utilize the data shown in Figure 2 to draw inferences about which genotype exhibits traits that impact profitability. Backfat relates directly to yield grade, and cattle that obtain a yield grade less than 3 (which typically occurs when backfat is below 10 mm) usually receive a premium when marketed on a grid. Using equation [5] and the approach used to estimate mean backfat levels shown in equation [6], the likelihood that an animal of a particular genotype has backfat less than 10 mm can be calculated given a particular number of days on feed. Results indicate that at 150 d on feed there is only a 46.9% chance that R25C-CC/UASMS2-CC cattle will exhibit backfat less than 10 mm; in contrast all other genotypes will have greater than a 70% chance that backfat will exceed 10 mm by 150 d. The R25C-CC/UASMS2-TT cattle deposit backfat most quickly, having an 86% chance of having more than 10 mm backfat by 150 d on feed. Another way of investigating the issue is to calculate the day at which a genotype is more likely than not to have a backfat that exceeds 10 mm. This would occur at d 154, 130, 124, 130, 122, 124, and 116 for genotypes R25C-CC/UASMS2-CC, R25C-CC/UASMS2-CT, R25C-CC/UASMS2-TT, R25C-CT/UASMS2-CC, R25C-CT/UASMS2-CT, R25C-TT/UASMS2-CC, and other, respectively. Thus, there is 30-d difference between when a feedlot might expect the R25C-CC/UASMS2-TT to stop receiving yield grade premiums vs. when R25C-CC/UASMS2-CC would stop receiving yield grade premiums.
In conclusion, this study revealed significant differences in BW and backfat growth parameters across leptin genotypes for a sample of commercially fed beef cattle. If these findings are validated with additional study, the potential exists to use genetic information in selection and in sorting cattle to optimize marketing decisions. That the analysis revealed significant differences in BW and backfat across genotype at the beginning of the feeding period suggests breeders may be able to utilize genetic information as a tool to aid in selection. Further, because of differences in growth rates of backfat and BW across genotype, feedlot managers may be able to determine which genotypes can reach yield grade targets without becoming too heavy in the shortest amount of time.
|
Footnotes
|
|---|
1 The author would like to thank Wade Brorsen, Donald Nkrumah, and Jim Tate for helpful comments on a previous version of this paper. 
2 Corresponding author: jayson.lusk{at}okstate.edu
Received for publication October 4, 2006.
Accepted for publication April 10, 2007.
|
LITERATURE CITED
|
|---|
Brethour, J. R. 1994. Estimating marbling score in live cattle from ultrasound images using pattern recognition and neural network procedures. J. Anim. Sci. 72:1425–1432.[Abstract]
Brethour, J. R. 2000. Using serial ultrasound measures to generate models of marbling and backfat thickness changes in feedlot cattle. J. Anim. Sci. 78:2055–2061.[Abstract/Free Full Text]
Buchanan, F. C., C. J. Fitzsimmons, A. G. Van Kessel, T. D. Thue, D. C. Winkelman-Sim, and S. M. Schmutz. 2002. Association of a missense mutation in the bovine leptin gene with carcass fat content and leptin mRNA levels. Genet. Sel. Evol. 34:105–116.[CrossRef][Medline]
Fitzsimmons, C. J., S. M. Schutz, R. D. Bergen, and J. J. McKinnon. 1998. A potential association between the BM 1500 microsatellite and fat deposition in beef cattle. Mamm. Gen. 9:432–434.[CrossRef][Medline]
Friedman, J. M., and J. L. Halaas. 1998. Leptin and the regulation of body weight in mammals. Nature 395:763–770.[CrossRef][Medline]
Geary, T. W., E. L. McFadin, M. D. Macneil, E. E. Grings, R. E. Short, R. N. Fuston, and D. H. Keisler. 2003. Leptin as a predictor of carcass composition in beef cattle. J. Anim. Sci. 81:1–8.[Abstract/Free Full Text]
Kaps, M., W. O. Herring, and W. R. Lamberson. 1999. Genetic and environmental parameters for mature weight in Angus cattle. J. Anim. Sci. 77:569–574.[Abstract/Free Full Text]
Kononoff, P. J., H. M. Deobald, E. L. Stewart, A. D. Laycock, and F. L. S. Marquess. 2005. The effect of a leptin single nucleotide polymorphism on quality grade, yield grade, and carcass weight of beef cattle. J. Anim. Sci. 83:927–932.[Abstract/Free Full Text]
Liefers, S. C., M. F. W. te Pas, R. F. Veerkamp, C. Delavaud, Y. Chilliard, and T. van der Lende. 2003. Association of leptin gene polymorphisms with serum leptin concentration in dairy cows. Mamm. Gen. 14:657–663.[CrossRef][Medline]
Morokoff, W., and R. Caflisch. 1995. Quasi-Monte Carlo integration. J. Comput. Phys. 122:218–230.[CrossRef]
Nkrumah, J. D., C. Li, J. A. Basarab, S. Guercio, Y. Meng, B. Murdoch, C. Hansen, and S. S. Moore. 2004. Association of a single nucleotide polymorphism in the bovine leptin gene with growth, feed intake, feed efficiency, feeding behaviour and carcass quality and composition. Can. J. Anim. Sci. 84:211–219.
Nkrumah, J. D., C. Li, J. Yu, C. Hansen, D. H. Keisler, and S. S. Moore. 2005. Polymorphisms in the bovine leptin promoter associated with serum leptin concentration, growth, feed intake, feeding behavior, and measures of carcass merit. J. Anim. Sci. 83:20–28.[Abstract/Free Full Text]
Schenkel, F. S., S. P. Miller, X. Ye, S. S. Moore, J. D. Nkrumah, C. Li, J. Yu, I. B. Mandell, J. W. Wilton, and J. L. Williams. 2005. Association of single nucleotide polymorphisms in the leptin gene with carcass and meat quality traits of beef cattle. J. Anim. Sci. 83:2009–2020.[Abstract/Free Full Text]
Sloan, J., and H. Wozniakowski. 1998. When are quasi-Monte Carlo algorithms efficient for high dimensional integrals? J. Complexity 14:1–33.[Medline]
Staniar, W. B., D. S. Kronfeld, K. H. Treiber, R. K. Splan, and P. A. Harris. 2004. Growth rate consists of baseline and systematic deviation components in Thoroughbreds. J. Anim. Sci. 82:1007–1015.[Abstract/Free Full Text]
Woodward, B. W., J. Li, Z. Zhang, R. L. Quaas, and E. J. Pollak. 2005. Associations between markers in the leptin gene and carcass traits in commercial feedlot steers and heifers. J. Anim. Sci. 83(Suppl. 1):327.
Top
Abstract
INTRODUCTION
