www.elsevier.com / locate / livprodsci
Genetic variation in residual feed intake and its association
with other production traits in British Hereford cattle
a ,
*
bR.M. Herd
, S.C. Bishop
a
NSW Agriculture Beef Industry Centre, University of New England, Armidale, NSW 2351, Australia
b
Roslin Institute(Edinburgh), Roslin EH25 9PS, UK
Received 10 November 1998; received in revised form 18 May 1999; accepted 1 June 1999
Abstract
Variation in residual feed intake, that is, variation in feed intake in relation to liveweight (LW) and growth rate, was investigated using data from 540 progeny of 154 British Hereford sires, collected over ten 200-day postweaning performance tests conducted between 1979 and 1988. Residual feed intake (RFIReg) was calculated for each test as the difference between actual feed and expected feed intake predicted from a multiple regression of feed intake on metabolic mid-test LW and average daily gain (ADG). RFIRegwas heritable (0.16, S.E. 0.08) and phenotypically and genetically independent of size and
growth rate. RFIReg had favourable phenotypic and genetic correlations with feed conversion ratio (FCR) and estimated
maintenance energy expenditure. It was negatively correlated with estimated lean content of the carcase (LEAN) and
appeared to be genetically independent of mature cow LW (COWWT). RFIRegover the performance test was not affected by
differences in pre-test rearing treatments, in contrast to start-of-test LW and end-of-test LW, and in some years, ADG and FCR. Selection against RFI has the potential to increase the efficiency of beef production by reducing feed intake without
changing the growth rate of the young animal, or increasing mature cow size. 2000 Elsevier Science B.V. All rights
reserved.
Keywords: Beef cattle; Selection; Efficiency; Residual feed intake
1. Introduction production systems has been questioned, since in-creased mature size is a direct consequence, resulting The cost of feeding animals is a major determinant in an increased cost of maintaining breeding females of profitability in livestock production enterprises. (Barlow, 1984). Modelling work by Thompson and Traditionally selection for growth rate has received Barlow (1986) showed greater improvements in considerable emphasis in most breed improvement enterprise efficiency (lean output / feed input) would schemes. However its value to the improvement of result from improvement in feed conversion ef-enterprise efficiency and profitability of ruminant ficiency of the growing animal and reduction in feed intake of the mature dam, but at that time the evidence for genetic variation in either of these traits
*Corresponding author. Tel.: 161-2-6770-1808; fax: 1
61-2-was equivocal. In their review of genetic parameter
6770-1830.
E-mail address: [email protected] (R.M. Herd) estimates for beef cattle, Koots et al. (1994b)
ported that the weighted mean genetic correlations to be evaluated for selection on RFI measured over a for feed conversion ratio (FCR) with growth rate and postweaning performance test, then the influence of size where highly negative. These correlations indi- pre-test growth on RFI over the performance test cate that selection to reduce FCR and improve needs to be assessed.
efficiency would be accompanied by an increase in The purpose of this study was to establish whether growth rate, and presumably in mature cow size. there exists genetic variation in RFI in young British That this would indeed occur was demonstrated by Hereford bulls during postweaning performance test-Mrode et al. (1990b) who reported that selection to ing, and to determine the phenotypic and genetic reduce lean FCR resulted in a correlated increase in correlations of RFI, growth rate and FCR with some
cow size. key production traits, including mature cow size. A
Recent research has shown that there is genetic second objective was to investigate the effect of variation in the feed eaten by young growing beef pre-test rearing treatments on variation in RFI and cattle beyond that explained by the size and growth other production traits measured over the subsequent rate of the cattle (Archer et al., 1998). This variation performance test.
in feed intake net of expected requirements based on size and growth is measured as residual feed intake
(RFI). RFI is calculated as the difference between 2. Materials and methods
actual feed intake by an animal and its expected feed
intake based on its size and growth rate. Selection to 2.1. Source of data reduce RFI offers to reduce feed intake, without
ad libitum with a small quantity of hay to stimulate energy expenditure per kg of metabolic body weight rumination. Carcase lean content (LEAN) was pre- (MMBW) was calculated as MAINT divided by dicted at the end of test for each animal from backfat MBW.
thickness measured by ultrasound scanning at the Residual feed intake for each animal was calcu-10th and 13th ribs and the third lumbar vertebrae. lated as the difference between actual feed intake LEAN was standardised each year to have a mean of over the 200-day performance test, less their ex-0.60 and a coefficient of variation of 0.04. Detailed pected feed intake over the test. For each of the 10 descriptions of the husbandry and selection methods performance tests, FI was regressed on MBW and are given by Simm (1983) and Mrode et al. (1990a). ADG (Proc REG; SAS Institute, 1989), and RFIReg
In the selection experiment, the traits selected for for each animal was calculated as the residual from were lean growth rate (LGR; growth rate to 400 the multiple regression. The distributions of RFIReg
days3predicted carcase lean content) and lean feed and its component traits (FI, MBW and ADG) for conversion ratio [LFCR; feed intake /(weight gain3 each performance test were checked for normality predicted carcase lean content] from 200 to 400 (Proc UNIVARIATE; SAS Institute, 1989). Feed days. Both LGR and LFCR were scaled by a intakes by calves born in year 9 of the experiment constant predicted killing out proportion of 0.577, were not normally distributed (P,0.05). The FI by estimated from bulls slaughtered in the initial years two bull calves were judged to be abnormally low, of the experiment. Responses to selection on these being more than three standard deviations below the traits, as well as the components traits of birth mean FI for their test group. Data for these two weight, predicted LEAN, growth rate to 400 days, animals were not subsequently used. Results for 540 feed intake and FCR have been presented by Mrode bull calves, from 154 sires, are presented in this
et al. (1990a,b). report, compared to 542 calves by Bishop (1992).
Although RFIReg is phenotypically independent of
2.2. Derivation of traits size and growth rate, Kennedy et al. (1993) showed
that under some circumstances it may be genetically The performance test data for LW and feed intake correlated with these same traits. Residual feed measurements were precorrected to 200 days of age intake was therefore also calculated from the pheno-at the start of the test and 400 days of age pheno-at the end typic (RFIPhen) or genetic (RFIGen) variances and of the test. This was not done by Mrode et al. covariances obtained from multivariate analyses for (1990a,b). The traits of LGR, LFCR, FCR, feed FI, MBW and ADG. These calculations assumed intake (FI), 200-day weight (W200), 400-day weight constant relationships between the component traits (W400) and 200 to 400-day daily weight gain across years whereas the calculation of RFIReg
(ADG) were then calculated for each animal. Also allowed these relationships to vary from year to year. calculated was average metabolic weight (MBW) as To investigate if there was association between the mean of W200 and W400, raised to 0.75. postweaning test performance and cow size, the LW Bishop (1992) also derived traits to describe the of the dam at just over 4.5 years of age (COWWT) energy required for the deposition of fat and protein was used as an estimate of her mature size. This in the body of the growing animal, for maintenance weight was taken after all the four-year-old cows had energy expenditure, and for maintenance energy weaned their calves. Weight records were available expenditure per unit metabolic LW. Maintenance for 331 cows, some of whom may have had more energy expenditure (MAINT) of growing animals than one bull-calf performance tested. The mean was defined as the difference between total weight of the cows was 498659 kg (SD).
metabolisable energy (ME) intake and the amount of
ME required for growth. The ME required for 2.3. Data analysis growth (DEP) was calculated using allometric
equa-tions describing the costs of protein and fat accre- Heritabilities, and phenotypic (r ) and geneticp
tion, and assumed standard efficiencies for fat and correlations (r ) for all traits were estimated withing
model along with the fixed effects of birth year (10 lines. This meant that the design was unbalanced and levels), rearing treatment (three levels), age of dam results were therefore calculated as least-squares (10 levels) and selection line (three levels: two means.
selected lines plus an unselected control line). Al-though the data had been collected on animals
previously selected on the basis of LGR and LFCR, 3. Results
Bishop (1992) showed this to produce little bias in
the genetic variances and covariances estimated for 3.1. Performance test results the traits he studied. For this reason, in this study it
was judged sufficient to include only LFCR in the The phenotypic and genetic correlations between trivariate analyses of the other traits. As only males FI and MBW (0.6760.03 (S.E.) and 0.8960.08, were performance tested, no animals had both per- respectively), and between FI and ADG (0.4760.04 formance and cow traits available. Calculation of and 0.7060.14) were medium to high, but less than phenotypic correlations for COWWT with perform- one, indicating that there was both phenotypic and ance test traits was therefore not possible. genetic variation in the relationship between FI and Differences in test performance, resulting as a growth performance. RFIReg had a heritability of consequence of the three different pre-test nutritional 0.1660.08 and was phenotypically independent of treatments (i.e., ages of weaning), were analysed size and growth (i.e., r with W200, W400, MBWp
using a general linear model (GLM) procedure (Proc and ADG were all zero; Table 1). RFIReg was GLM; SAS Institute, 1989). Data for 339 calves genetically independent of ADG, but the genetic from years 1 to 6 of the selection experiment was correlations with size (i.e., r with W200, W400 andg
used, as in later years all calves were weaned at the MBW) were not so close to zero, even though not same age (84 days). The traits analysed were W200 statistically different from it. The large standard as a measure of pre-test growth rate, and FI, ADG, errors were due to the small size of the dataset as W400, LEAN, FCR, LFCR and RFIRegmeasured for well as the low heritabilities of the component traits. the 200-day performance tests. The GLM model RFIReg was positively correlated with FCR and included the fixed effects of year (1 to 6), rearing LFCR, both phenotypically and genetically, such that treatment (birth, 84 or 168 days of age), and line lower RFIRegwas associated with improved FCR and (control, LGR, LFCR), fitted sequentially. The inter- LFCR (Table 1). RFIReg was negatively associated action of test year with rearing treatment was also with LEAN and LGR, implying that superior re-included in the model. In years 1 and 2, calves were sidual feed intake was accompanied by a greater not assigned to selection lines, and in year 3, all proportion of lean in the weight gain and final calves were assigned to either the LGR or LFCR carcase of the calves. RFIReg was phenotypically
Table 1
Means and heritabilities (h ) for performance test traits, and their phenotypic (r ) and genetic (r ) correlations with RFI2 p g Reg
W200 FI ADG MBW W400 LEAN LGR FCR LFCR DEP MAINT MMBW
0.75 0.75
(kg) (kg / 200 d) (kg / d) (kg ) (kg) (kg / kg) (kg / d) (kg / kg) (kg / kg) (MJ ME) (MJ ME) (kJ / kg / d)
Mean 166 1458 1.21 69.2 408 0.600 0.32 6.14 17.76 5327 9083 655
(SD) (30) (176) (0.18) (50.6) (41) (0.024) (0.04) (1.07) (3.24) (818) (1821) (118)
h2 0.23 0.31 0.38 0.36 0.42 0.49 0.47 0.17 0.26 0.36 0.23 0.14
(S.E.) (0.08) (0.08) (0.10) (0.09) (0.10) (0.11) (0.10) (0.09) (0.09) (0.10) (0.08) (0.08)
rp 0.00 0.70 20.01 20.01 20.01 20.22 20.33 0.61 0.63 0.06 0.78 0.91
(S.E.) (0.04) (0.02) (0.05) (0.04) (0.04) (0.04) (0.04) (0.03) (0.03) (0.04) (0.02) (0.01)
rg 0.34 0.64 0.09 0.22 0.15 20.43 20.47 0.70 0.72 0.27 0.77 0.93
independent of feed energy required for gain of lean ues, with median values of 8.5 and 19.7 kg / 200 d, and fat (DEP), although the genetic correlation was co-efficients of skewness equal to 20.25 and not so close to zero, even though not statistically 20.50, and Shapiro–Wilk statistics (Proc Uni-different from it. RFIReg was highly correlated, both variate; SAS Institute, 1989) of 0.98 and 0.97, phenotypically and genetically, with variation in feed indicative of non-normality (P,0.05 and P,0.01), energy attributed to maintenance (MAINT) and to respectively.
maintenance energy expenditure per unit MBW
(MMBW). 3.2. Associations with cow size
RFIReg (calculated phenotypically for each test)
had a high phenotypic correlation with RFIPhen and There was genetic variation in estimated mature RFIGen (0.8860.01 and 0.7360.02 respectively), but cow size (COWWT) as evidenced by its heritability the correlations were less than unity implying that of 0.6960.11. Even though estimated from a small RFIReg was phenotypically a different trait than dataset this value is close to the weighted mean RFIPhen and RFIGen. The genetic correlation of heritability for mature cow weight of 0.50 calculated RFIReg with RFIPhen (0.7560.14) was also less than from 24 published estimates by Koots et al. (1994a). unity, although not statistically different from it. The Although estimated with a rather large standard genetic correlation of RFIReg with RFIGen error, COWWT appeared to be genetically indepen-(0.4760.24) was considerably less than unity, imply- dent of RFIReg measured during the postweaning ing that they were genetically different traits. This performance test (rg5 20.0960.26). The genetic was unexpected as our preliminary calculations correlations between growth traits (ADG, W400 and based on expectations from the (co)variance com- LGR) and COWWT were all positive (0.4060.18, ponents indicated all these correlations should have 0.4060.16 and 0.4360.16, respectively). The ge-been greater than 0.95. Two assumptions used in the netic correlations between measures of feed conver-calculation of RFIPhen and RFIGen were that the sion efficiency (FCR and LFCR) and COWWT were component traits (FI, MBW and ADG) were normal- less than zero, although not significantly different ly distributed, and that the regression coefficients for from it (20.2960.24 and 20.2360.22, respective-FI with MBW and ADG were constant across years. ly).
With respect to the first assumption, FI and MBW
for the 540 calves were normally distributed (P. 3.3. Effect of pre-test weaning treatments 0.05) but ADG was not (P,0.01). To check the
Table 2
Least-squares means (S.E.s) for performance test results of bull calves weaned at birth, 84 or 168 days of age during the first six years of the selection experiment
LEAN 0.606 (0.005) 0.603 (0.003) 0.601 (0.002) ns
a b a,b
LGR 0.307 (0.006) 0.322 (0.003) 0.321 (0.003) ns
FCR 6.66 (0.14) 6.77 (0.07) 6.75 (0.06) ***
LFCR 19.1 (0.5) 19.5 (0.2) 19.5 (0.2) ***
RFIReg 1 (22) 212 (11) 4 (10) ns
Means within a row with different superscripts differ (P,0.05). ***P,0.001; *P,0.05; ns P.0.05.
calves. The faster growth, and better FCR, of the increasing the size of the cow. This is an important artificially-reared calves in years 2 and 5 was advantage over selection for growth rate, LGR or evidence that compensatory gain in LW occurred LFCR. In this study these three traits were ge-during their 200-day test in these two years. Feed netically correlated with COWWT indicating that efficiency, as measured by RFIReg, was unaffected selection to improve these traits would be accom-each year (i.e., over accom-each test) by differences in panied by an increase in cow size. Selection for rearing treatment and pre-test growth rate. growth rate has been repeatedly associated with an increase in cow size and its benefit to whole herd productivity has been seriously questioned (Barlow,
4. Discussion 1984). Within the current selection experiment, selection for LGR and LFCR were both shown to This study has shown that in young British produce correlated increases in cow LW (Mrode et Hereford bulls there exists both phenotypic and al., 1990b).
genetic variation in feed intake that is independent of Our estimate of the heritability for RFIReg (0.16) size and growth rate. Therefore it should be possible is modest and similar to five other estimates reported to implement genetic selection to reduce feed intake in the review by Arthur et al. (1998), but low without compromising growth, and to thereby im- compared to the estimate of 0.46 reported for prove the profitability of beef production. The ge- British-breed cattle by Archer et al. (1998). In both netic correlations of performance test traits with studies, the phenotypic variances (V ) of two com-p
mature cow size had rather large standard errors ponents traits of RFIReg (ADG and MMBW) are of because of the small size of the dataset. However similar magnitude, both in absolute size and relative estimates of the genetic correlation of FCR with cow to mean for each trait (Table 3). The V of the thirdp
Table 3
Means and phenotypic variances (V ) for RFIp Regand its components traits measured over postweaning performance tests with British cattle
This study Archer et al. (1998)
Mean Vp V / meanp Mean Vp V / meanp
FI (kg) 1458 20 560 14.1 1296 14 205 11.0
MMBW (kg) 69.2 23.6 0.34 71.8 21.9 0.31
ADG (kg / d) 1.21 0.021 0.018 1.20 0.0196 0.016
a a
RFI (kg) 0 10 156 6.97 0 4743 3.66
a
Using the mean for FI.
and 2561 kg, or 5.9- and 2.0-times their mean feed composition. The full importance of variation in intakes, respectively). This is presumably due to the body composition to variation in RFI remains to be tighter relationships between FI, and MMBW and determined.
ADG, and hence lower range in residuals (i.e., Our assumption that the feed energy requirements RFIReg) in the Australian postweaning performance for protein and fat accretion (DEP) can be de-tests compared to those in this study. One explana- termined from allometric equations is in agreement tion for this could be less measurement error in the with Veerkamp and Emmans (1995) who could find more recently conducted Australian experiment. The no conclusive evidence in dairy cows for variation in
2
R values for multiple regressions for the seven the partial efficiencies for conversion of substrate to Australian tests are typically 70% or higher (un- product. In defining the maintenance energy expendi-published data) compared to a mean of 54% (range ture (MAINT) of the young animals in this study as
2
39% to 65%) for the R values over the 10 tests in feed energy intake surplus to the requirements for this study. Had the environmental variance for RFI DEP, we have included in MAINT sources of in this study been lower and similar to that in the variation in feed intake recognised by Veerkamp and Australian study, then the differences in heritabilities Emmans (1995) as being distinct from k : the partialm
would not have been so large. efficiency of maintenance. These include differences
In this study variation in RFI was associated with in mobilisation of tissues to meet energy require-variation in measures of body composition and of ments, differences in partitioning of substrate to lean maintenance energy expenditure. The phenotypic and and fat compartments, and, perhaps, differences in genetic correlations of RFI with LEAN and LGR activity. We calculated both RFI and MAINT as that were both low indicating that selection against RFI feed intake surplus to expected requirements based might slightly increase carcase leanness. This is in on regression with MBW and ADG (for RFIReg) or agreement with the observed small reduction in from allometric equations (MAINT) and it is, there-subcutaneous-fat thickness reported by Richardson et fore, not surprising that they are phenotypically and al. (1998) in response to a single generation of genetically very similar traits. Both traits would have selection against RFI. However, variation in LEAN included unexplained variation in feed intake due to explained little additional variation in FI beyond that these sources. The high phenotypic and genetic explained by MBW and ADG. In the GLM described correlations between RFIReg and MAINT and above that was used to examine the relationship MMBW suggest that selection against RFI should between FI, and MBW and ADG, over the 10 years favour those animals with lower maintenance energy of testing, the latter two traits and their interactions expenditures. This should lead to improvement in the with year explained 68% of the variation in FI. apparent efficiency of maintenance but without com-Adding LEAN to this model explained only an promise to growth performance.
additional 1.5% of the variation in FI. LEAN is a Residual feed intake calculated from phenotypic measure of carcase leanness, largely dependent on (co)variances (RFIPhen) should be similar to RFIReg
was supported by the recent results of Archer et al. ranking of animals. Thus there remains an argument (1998). Archer et al. (1998) also reported a very favouring standardisation of the pre-test preparation high correlation between RFIPhenand RFIGenand led (i.e., growth) of cattle to ensure a fairer phenotypic to the conclusion that selection for RFIGen (RFI comparison over the subsequent test. The lack of calculated to be genetically independent of product- effect of pre-test environment on RFI implies that it ion) would give similar results to selection for may be more robust across farms or environments RFIReg. However, in this study the genetic correla- than the other growth or efficiency traits measured. tion of RFIReg with RFIGen was considerably less This study has shown that selection against RFI than unity, implying that they were genetically has the potential to improve FCR in the young different traits. This may have been due to two of the growing animal, to improve the efficiency of mainte-assumptions used in the calculation of RFIGen: that nance energy expenditure, and to avoid increasing the component traits (FI, MBW and ADG) were the size of the cow. These are key responses to normally-distributed, and that the regression coeffi- improvement in enterprise efficiency (Thompson and cients for FI with MBW and ADG were constant Barlow, 1986). RFI appeared to be less influenced by across years, having been violated. The calculation pre-test environmental variation than was LW, of RFIReg allowed these regression relationships to growth rate and other measures of efficiency during vary across year. Whilst RFIReg may under some the postweaning performance tests on beef cattle. circumstances be genetically correlated with
pro-duction (Kennedy et al., 1993), use of RFIGen (and
assuming constant relationships across years) may be Acknowledgements
ignoring real variation in these relationships. An
alternative approach that could be investigated on We thank the Roslin Institute for providing the larger datasets may be to fix the genetic variances data and the Ministry of Agriculture, Food and and covariances and allow environmental covar- Fisheries for funding the original project, and S. iances to vary between years, and then use the Barwick and J. Van der Werf for reviewing the parameters to calculate RFI which differs between manuscript. David Nicholson extracted the maternal years. Clearly, the choice of traits to be included in a data from a historic database and Keryn Zirkler selection program to improve feed efficiency will prepared it for our analysis. R.H. thanks the
Aus-require further consideration. tralian Scientific and Technological Exchange
Pre-test environmental variation in the form of Scheme and the Cattle and Beef Industry CRC for different ages of weaning was shown to affect Meat Quality for their award of travel grants, NSW phenotypic performance during the subsequent post- Agriculture for granting study leave, and the Roslin weaning test. Rearing treatment affected pre-test Institute for hosting his visit.
growth as measured by W200, but it also affected LGR, FI and W400, and in some years also affected
LW gain over the test, LEAN, FCR and LFCR. RFI References
was unaffected by differences in rearing treatment
and may therefore be less influenced by pre-test Archer, J.A., Arthur, P.F., Herd, R.M., Richardson, E.C., 1998. Genetic variation in feed efficiency and its component traits.
environmental variation than the other performance
Proc. 6th World Congr. Genet. Appl. Livest. Prod. 25, 81–84.
traits routinely measured on beef cattle. Current
Arthur, P.F., Archer, J.A., Herd, R.M., Richardson, E.C., 1998. A
genetic evaluation schemes require that animals are review of variation in feed efficiency of beef cattle. Proc. 6th evaluated within contemporary groups and that the World Congr. Genet. Appl. Livest. Prod. 25, 85–88.
test environments are linked by common sires. This Barlow, R., 1984. Selection for growth and size in ruminants: Is it time for a moratorium. In: Hofmeyr, J.H., Meyer, E.H.H.
is to remove environmental variation due to
differ-(Eds.), Proc. 2nd World. Congr. Sheep Beef Cattle Breed,
ences in management of groups, and between tests.
Pretoria, pp. 421–432.
These procedures improve the accuracy of the ge- Bishop, S.C., 1992. Phenotypic and genetic variation in body netic evaluation and ranking of animals but these weight, food intake and energy utilization in Hereford cattle. I.
Gilmour, A.R., Thompson, R., Cullis, B.R., Welham, S.J., 1996. Richardson, E.C., Herd, R.M., Archer, J.A., Woodgate, R.T., ASREML. Biometrics, Bulletin 3, NSW Agriculture, Orange. Arthur, P.F., 1998. Steers bred for improved net feed efficiency Kennedy, B.W., Van der Werf, J.H.J., Meuwissen, T.H.E., 1993. eat less for the same feedlot performance. Anim. Prod. Aust.
Genetic and statistical properties of residual feed intake. J. 22, 213–216.
Anim. Sci. 71, 3239–3250. SAS Institute, 1989. SAS / STAT Users Guide, Version 6. 4th ed., Koots, K.R., Gibson, J.P., Smith, C., Wilton, J.W., 1994a. Analyses SAS Institute, Cary, NC.
of published genetic parameter estimates for beef production Simm, G., 1983. Selection of beef cattle for efficiency of lean traits. 1. Heritability. Anim. Breed. Abstr. 62, 311–338. growth. Ph.D. Thesis. University of Edinburgh, Edinburgh. Koots, K.R., Gibson, J.P., Wilton, J.W., 1994b. Analyses of Thompson, J.M., Barlow, R., 1986. The relationship between
published genetic parameter estimates for beef production feeding and growth parameters and biological efficiency in traits. 2. Phenotypic and genetic correlations. Anim. Breed. cattle and sheep. Proc. 3rd World Congr. Appl. Livest. Prod.
Abstr. 62, 825–853. 11, 271–282.
Mrode, R.A., Smith, C., Thompson, R., 1990a. Selection for rate Veerkamp, R.F., Emmans, G.C., 1995. Sources of genetic vari-and efficiency of lean gain in Hereford cattle. 1. Selection ation in energetic efficiency of dairy cows. Livest. Prod. Sci. pressure applied and direct responses. Anim. Prod. 51, 24–34. 44, 87–97.
