Implications

Research to date has clearly shown that feed efficiency is a complex multifaceted trait, under the control of many biological processes. Thus, a thorough understanding of the biochemical mechanisms regulating appetite, intestinal absorption, digestion and nutrient partitioning, amongst other key processes, underpinning the trait is warranted to expedite selection for feed efficient cattle, necessary for the continued economic and environmental sustainability of beef production.

Measuring and calculation of residual feed intake

As variation amongst individual animals in feed intake cannot be estimated from knowledge of BW and level of production alone, accurate measurement of feed intake remains a necessary requirement in national cattle evaluation systems (Nielsen et al., Reference Nielsen, MacNeil, Dekkers, Crews, Rathje, Enns and Weaber2013). In order to aid the standardisation of the regimen, criteria for measuring, recording and assessing feed efficiency have been established (Beef Improvement Federation, 2010). This includes a period of feed intake measurement of at least 70 days duration, preceded by an acclimatisation period of at least 21 days, with live weight recorded on 2 consecutive days at the beginning and end and periodic intervals throughout. Notwithstanding this, recent studies have attempted to reduce the test duration further (Culbertson et al., Reference Culbertson, Speidel, Peel, Cockrum, Thomas and Enns2015; Cassady et al., Reference Cassady, Felix, Beever and Shike2016). Indeed, where BW was recorded for 63 (Wang et al., Reference Wang, Nkrumah, Li, Basarab, Goonewardene, Okine, Crews and Moore2006) and 84 (Manafiazar et al., Reference Manafiazar, Basarab, McKeown, Stewart-Smith, Baron, MacNeil and Plastow2017) days, apparently adequate feed intake test durations as short as 35 to 42 days, respectively have been reported. Although shortening the duration of the feed intake test period resulted in a loss in accuracy (reduction in Spearman’s correlation coefficient) of 5% to 7%, such an approach would reduce the cost of feed intake recording and increase annual animal throughput (Manafiazar et al., Reference Manafiazar, Basarab, McKeown, Stewart-Smith, Baron, MacNeil and Plastow2017). The accuracy of shorter test durations, however, is likely to be dependent on prevailing diet composition and animal growth rate (Goonewardene et al., Reference Goonewardene, Okine, Wang, Spaner, Mir, Mir and Marx2004).

Residual feed intake is the difference between observed and predicted feed intake (dry matter or energy) and is usually calculated as the residuals from a multiple regression model of dry matter intake (DMI) on selected sources of significant energy expenditure or sinks such as maintenance, growth and activity (Savietto et al., Reference Savietto, Berry and Friggens2014). This computation generally forces RFI to be mathematically independent of the traits used to predict DMI at a phenotypic level but does not necessarily ensure genetic independence (Berry and Crowley, Reference Berry and Crowley2013). The conventional basic multiple regression model used to predict DMI in many studies includes metabolic live weight and average daily gain (ADG) but other potential sources of variation, such as measures of body composition (see later), can also be included. The coefficient of determination (R ²) of this regression model predominantly quantifies the relative cumulative contribution of various energy-demanding processes, included in the model, to variation in DMI, and, by extension RFI, but also contains measurement error. Despite much research on the topic and assessment of the potential contribution of a variety of traits and physiological processes, very little of the unknown residual component of the model has been explained to-date.

The majority of published studies that have evaluated RFI in growing (finishing) beef cattle pertain to animals offered energy-dense diets. An appraisal of 14 such publications (e.g. Kelly et al., Reference Kelly, McGee, Crews, Fahey, Wylie and Kenny2010a; Fitzsimons et al., Reference Fitzsimons, Kenny and McGee2014a) indicates a mean R ² of 0.70 for the ‘base’ model used to predict DMI. The corresponding value from eight other such studies where growing cattle were offered mainly forage diets (e.g. Shaffer et al., Reference Shaffer, Turk, Wagner and Felton2011; Lawrence et al., Reference Lawrence, Kenny, Earley and McGee2012) was 0.61. The lower degree of variation explained for forage-based diets is not surprising considering their poorer intake characteristics and slower rate of passage through the rumen (Forbes, Reference Forbes2005). Compared with concentrate-based diets, feeding high-forage diets may limit voluntary feed intake capacity thus reducing the expression of inherent DMI potential.

Some studies have reported a positive, though weak, phenotypic correlation between RFI and measures of body fat content (see later) and as a result, many studies include an adjustment for body fat in the statistical model for computing RFI. This adjustment can increase the R² from less than one (Basarab et al., Reference Basarab, Colazo, Ambrose, Novak, McCartney and Baron2011) to up to five percentage units (Fontoura et al., Reference Fontoura, Montanholi, Diel de Amorim, Foster, Chenier and Miller2015), although the significance and contribution of body composition per se to the accuracy of the model is not always stated (McGee et al., Reference McGee, Ramirez, Carstens, Price, Hall and Hill2014). In nine studies using pregnant beef females, the mean reported R ² for the prediction of DMI was only 0.37, even where measures of body composition were included. This poorer relationship may be partially attributed to the fact that the diets offered were forage-only but also to the relatively low or near zero growth rates (i.e. conceptus-adjusted ADG) of the pregnant females used in those studies (Lawrence et al., Reference Lawrence, Kenny, Earley and McGee2013). Although variation in tissue accretion and depletion between the conceptus and dam exists during pregnancy, adjusting live weight data for conceptus growth should largely account for the confounding effects of stage of gestation and estimated foetal size. In addition, factors such as maturity (parity, age) and the energy requirements pertaining to stage of gestation and colostrogenesis may influence this.

Few studies have calculated RFI for lactating beef females. Black et al. (Reference Black, Bischoff, Mercadante, Marquezini, DiLorenzo, Chase, Coleman, Maddock and Lamb2013), using ADG, energy-corrected milk yield and change in back fat thickness, reported an R ² of 0.60 for the prediction of DMI of lactating beef cows, offered a forage-based diet. Of note is that BW was excluded from this prediction model, suggesting that this variable did not account for a statistically significant proportion of the variation in DMI observed. Similarly, it is recognised that measuring RFI in lactating dairy cows is much more complicated than for growing cattle exhibiting a linear growth trajectory (Connor, Reference Connor2015). Collectively, these studies suggest that the regression models used to predict RFI in fast-growing cattle may not be appropriate for pregnant and lactating beef cows, as the majority of the phenotypic variance in DMI remains unexplained, or alternatively, that the error in estimation of weight and weight gain is too high relative to that of DMI.

Sources of biological variation in phenotypic residual feed intake

There is a substantial inherent inter-animal variation for feed efficiency where cattle are offered feed to appetite. Phenotypic differences in DMI between the most feed efficient and inefficient terciles of up to 15% in young growing cattle (Lawrence et al., Reference Lawrence, Kenny, Earley and McGee2012; Fitzsimons et al., Reference Fitzsimons, Kenny, Fahey and McGee2013) and 25% in pregnant beef cows (Lawrence et al., Reference Lawrence, Kenny, Earley and McGee2013) have been reported. In agreement, significant genetic variance for the trait has been reported (Berry and Crowley, Reference Berry and Crowley2013; Savietto et al., Reference Savietto, Berry and Friggens2014). In addition, it must be noted that the inter-animal variance in DMI will be greater than this. Given the existence of such variation, potential biological sources of variation are discussed below.

Digestion, rumen fermentation and microbiome

Increasing consumption of feed usually decreases diet digestibility, mainly as a result of a reduction in ruminal residency time. Consequently, a reduction in apparent digestibility per se would be expected in cattle classified as high- compared with low-RFI, but the literature does not support this speculation (Table 1). Nevertheless, several studies (including some of those that did not detect a difference in dry matter digestibility (DMD) between RFI classifications), reported that diet DMD was negatively correlated with RFI. It is unclear, however, whether the apparently improved digestive ability of more feed efficient animals is inherent, or simply a function of a slower passage rate of digesta through the rumen due to lower DMI. In some instances, the absence of differences in DMD between cattle of varying RFI phenotype may be related to the nature of the diets offered, as the effect of feed intake on digestion is less with forage than concentrate-based diets. In addition, in relation to the indirect marker methodologies employed in the majority of studies, these technologies may not be sufficiently sensitive to detect differences, where they exist and their accuracy may also be affected by the nature and homogeneity of the diet offered (Herd et al., Reference Herd, Arthur and Archer2004).

Table 1

The effect of residual feed intake (RFI) rank on apparent dry matter digestibility in beef cattleFootnote ¹

HC=high concentrate; TFC=total faecal collection, AIA=acid insoluble ash; TMR=total mixed ration; DMD=dry matter digestibility.

¹ Reference list for this table provided in Supplementary Material S1.

² r=correlation between RFI and DMD; †P<0.10, *P<0.05 and **P<0.01.

³ TMR (70 : 30 corn silage:concentrate on a dry matter basis).

⁴ An interaction was reported in this study whereby low-RFI heifers had greater DMD than their high-RFI contemporaries when consuming a grass silage diet but this difference was not observed when the same animals grazed pasture or were offered a TMR indoors.

Given the central nature of ruminal digestion to the profile of nutrients available for post-absorptive processes, it is surprising that few obvious differences in the primary rumen fermentation variables measured are evident between high- or low-RFI cattle and, in cases where variance was observed, results were not in agreement (Table 2). Consistent with studies showing no association between RFI and production and composition of volatile fatty acid (VFA), Kong et al. (Reference Kong, Liang, Chen, Stothard and Guan2016) conducted transcriptome profiling of the rumen epithelium of steers differing in RFI and reported no differences in the expression of genes involved in VFA metabolism, however, concentration and absorption of ruminal VFA were not measured. Nevertheless, there is evidence for associations between the rumen microbiome, VFA and RFI phenotypes in growing beef cattle (Carberry et al., Reference Carberry, Kenny, Han, McCabe and Waters2012; Myer et al., Reference Myer, Smith, Wells, Kuehn and Freetly2015; Shabat et al., Reference Shabat, Sasson, Doron-Faigenboim, Durman, Yaacoby, Berg Miller, White, Shterzer and Mizrahi2016). For example, the association between RFI ranking and bacterial profiles was more pronounced when a forage-only (grass silage), as opposed to a cereal-based diet, was offered to beef heifers (Carberry et al., Reference Carberry, Kenny, Han, McCabe and Waters2012). Furthermore, Prevotella, one of the most dominant bacterial genera within the rumen microbiome, was more abundant in inefficient cattle (Carberry, et al., Reference Carberry, Kenny, Han, McCabe and Waters2012; McCann et al., Reference McCann, Wiley, Forbes, Rouquette and Tedeschi2014; Myer et al., Reference Myer, Smith, Wells, Kuehn and Freetly2015).

Table 2

Rumen fermentation traits and residual feed intake (RFI) in beef cattleFootnote ¹

NH₃=ammonia; VFA=volatile fatty acid; HC=high concentrate; H=high RFI; L=low RFI; A:P=acetate:propionate ratio.

^a,b,c,d,eLeast squares means within a column without a common superscript letter differ (a, b: P<0.05; d, e: P<0.10).

¹ Reference list for this table provided in Supplementary Material S1.

² Acetate, propionate and butyrate reported as mmol/mol of volatile fatty acid.

³ NH₃ reported as millimolar (mM).

Although enteric methane (CH₄) production is an integral component of rumen fermentation, it constitutes what can be a significant loss of energy to the host animal (Pickering et al., Reference Pickering, Oddy, Basarab, Cammack, Hayes, Hegarty, Lassen, McEwan, Miller, Pinares-Patiño and de Haas2015). The well-documented strong positive relationship between DMI and ruminal methane production in cattle suggests that low-RFI cattle should have lower CH₄ emissions (g/day), at least proportionate to their lower feed intake, however, the published literature does not support this expectation (Table 3). For example, CH₄ emissions (g/day) were found to be lower for low- compared with high-RFI cattle, when offered unrestricted access to feed (Hegarty et al., Reference Hegarty, Goopy, Herd and McCorkell2007; Fitzsimons et al., Reference Fitzsimons, Kenny, Fahey and McGee2013), but also where animals were offered an equal, though restricted, feed allowance (Nkrumah et al., Reference Nkrumah, Okine, Mathison, Schmid, Li, Basarab, Price, Wang and Moore2006). In contrast, McDonnell et al. (Reference McDonnell, Hart, Boland, Kelly, McGee and Kenny2016) found no effect of RFI status on CH₄ emissions (g/day) of beef heifers offered a grass silage diet followed by a high-starch cereal-based diet. These findings suggest that there is little evidence of a direct effect of RFI per se on ruminal CH₄ emissions (g/day) and that differences observed are likely a reflection of the variance in DMI between animals. This appears to be true regardless of whether cattle were the result of divergent selection for RFI (Hegarty et al., Reference Hegarty, Goopy, Herd and McCorkell2007; Jones et al., Reference Jones, Phillips, Naylor and Mercer2011) or not (Nkrumah et al., Reference Nkrumah, Okine, Mathison, Schmid, Li, Basarab, Price, Wang and Moore2006; Fitzsimons et al., Reference Fitzsimons, Kenny, Fahey and McGee2013; McDonnell et al., Reference McDonnell, Hart, Boland, Kelly, McGee and Kenny2016). In addition, other work from our laboratory (Carberry et al., Reference Carberry, Waters, Kenny and Creevey2014) shows that with the exception of various genotypes of Methanobrevibacter smithii found to be more abundant in cattle of high- compared with low-RFI when compared across a number of contrasting diet types, overall methanogen abundance in rumen digesta was not affected by host feed efficiency status.

Table 3

Methane emissions and residual feed intake (RFI) in beef cattleFootnote ¹

CH₄, methane; DM=dry matter; SF₆=sulphur hexafluoride tracer gas technique; OPFTIR=open path Fourier transform IR spectrophotometer technique; TMR=total mixed ration.

¹ Reference list for this table provided in Supplementary Material S1.

² Animal model=phenotype study – cattle used were from a random population; selection lines – the progeny of cattle divergently bred and selected for RFI.

³ No RFI×diet interaction was reported for this study.

⁴ TMR (70 : 30 corn silage:concentrate on a DM basis).

Size of and metabolic processes within the visceral organs

Due to the high metabolic cost associated with the gastrointestinal tract and liver, it is likely that inter-animal variation in the size and functionality of these organs may influence energy requirements for basal metabolism. However, the published literature that has examined variation in visceral organ size amongst animals of divergent feed efficiency status is inconsistent (Table 4). Likewise, in terms of energy expenditure of visceral organs, there are a number of recent molecular-based studies, such as that of Paradis et al. (Reference Paradis, Yue, Grant, Stothard, Basarab and Fitzsimmons2015), that have demonstrated inconsistencies in the association between RFI phenotype and transcript abundance for genes involved in metabolic processes within gastrointestinal tissues. For a more in-depth discussion on this topic, the reader is referred to the recent review of Fitzsimons et al. (Reference Fitzsimons, McGee, Keogh, Waters and Kenny2017).

Table 4

Visceral organ weights and residual feed intake (RFI) in beef cattleFootnote ¹

H, high RFI; L, low RFI; GIT, gastrointestinal tract.

^a,bLeast squares means within a column without a common superscript letter differ (P<0.05).

¹ Reference list for this table provided in Supplementary Material S1.

² Entire stomach complex reported.

Nutrient partitioning: protein and fat deposition

In addition to its central importance to the value of beef cattle, body content of both muscle and fat tissues make a significant contribution to overall energy status. The potential contribution of differences in energy utilisation relating to composition, maintenance and metabolic processes within muscle and adipose tissue depots to inter-animal variation for the RFI trait, has been reviewed by Fitzsimons et al. (Reference Fitzsimons, McGee, Keogh, Waters and Kenny2017). As highlighted earlier, there is currently much equivocation in the published literature on body compositional differences between cattle of divergent feed efficiency status and this is consistent for both growing and pregnant beef cattle. For example, where the base model (DMI=βBW+βADG+(e=RFI) has been used to calculate RFI, studies have reported both positive (Hafla et al., Reference Hafla, Carstens, Forbes, Tedeschi, Bailey, Walter and Johnson2013) and negative (Lawrence et al., Reference Lawrence, Kenny, Earley, Crews and McGee2011) associations between RFI status and ultrasonically measured longissimus muscle size. Similarly, positive (Arthur et al., Reference Arthur, Archer, Johnston, Herd, Richardson and Parnell2001; Basarab et al., Reference Basarab, McCartney, Okine and Baron2007; Berry and Crowley, Reference Berry and Crowley2013), though sometimes weak and close to zero (Mao et al., Reference Mao, Chen, Vinsky, Okine, Wang, Basarab, Crews and Li2013), or no association (Fitzsimons et al., Reference Fitzsimons, Kenny and McGee2014a) between RFI status and ultrasonically measured fat depth in the live animal and carcass fatness traits have been reported. Discrepancies between studies may partly be due to variation in fat deposition in different breeds, differences in the site and technique for measurement of the traits between operators and also disparities in carcass classification methodologies that differ between countries. Similarly, inconsistencies in the literature exist for systemic metabolic indicator traits for body composition such as creatinine (negative association, Lawrence et al., Reference Lawrence, Kenny, Earley and McGee2012; no association, Fitzsimons et al., Reference Fitzsimons, Kenny and McGee2014a) and leptin (no association, Kelly et al., Reference Kelly, McGee, Crews, Fahey, Wylie and Kenny2010a). Reports on the effects of insulin and IGF-I, which are also indicators of body composition and overall metabolic status, on RFI status are equally in disagreement. In terms of circulating concentrations of IGF-I, some studies have reported higher (Nascimento et al., Reference Nascimento, Branco, Bonilha, Cyrillo, Negrão and Mercadante2015), others lower (Lancaster et al., Reference Lancaster, Carstens, Ribeiro, Davis, Lyons and Welsh2008) or no difference (Lawrence et al., Reference Lawrence, Kenny, Earley and McGee2012; Welch et al., Reference Welch, Thornton, Murdoch, Chapalamadugu, Schneider, Ahola, Hall, Price and Hill2013) in concentrations between low- and high-RFI cattle. Plasma concentrations of insulin at the end of a test period were found to be higher in the steer progeny of high- compared with low-RFI parents (Richardson et al., Reference Richardson, Herd, Archer and Arthur2004). Nevertheless, similar circulating concentrations of insulin (Kolath et al., Reference Kolath, Kerley, Golden and Keisler2006; Lawrence et al., Reference Lawrence, Kenny, Earley and McGee2012) and insulin response to a glucose tolerance test (Fitzsimons et al., Reference Fitzsimons, Kenny, Waters, Earley and McGee2014c) suggest no differences in insulin sensitivity or insulin-mediated body composition between cattle divergent for RFI.

To evaluate the relationship between RFI status and measures of body composition we conducted meta-analyses of studies that used growing beef cattle offered energy-dense diets. Although numerous individual studies have examined this relationship, the huge variation in reporting of results severely limited the number of studies that could be included in our meta-analyses. In relation to muscle accretion, we found no difference in either live animal (n=5; P=0.36) or carcass (n=8; P=0.39) measures between cattle of high- or low-RFI status. An additional meta-analysis was conducted to quantify the relationship between variation in RFI and ultrasonically measured back fat depth and again we failed to observe a difference (P=0.65) between growing high- and low-RFI cattle (references for these meta-analyses in Supplementary Material S1). This suggests that RFI rank in growing cattle is not associated with final muscle area, carcass muscle area and change in back fat depth during the linear phase of the growth curve, typical of RFI test periods in many studies. These findings are in contrast to those of Berry and Crowley (Reference Berry and Crowley2013) who reported a genetically based tendency for RFI status to be negatively correlated with muscularity and positively associated with body fat in the live animal or in the carcass. Equivocation amongst literature reports is undoubtedly contributed to by variation between studies in the breed, gender and stage of physiological maturity of the cattle employed. In addition, while an explicit relationship between RFI status and body composition could not be established in our meta-analyses, the well-established influence of body fatness, in particular on key reproductive events (i.e. onset of puberty and resumption of postpartum ovarian cyclicity; Diskin and Kenny, Reference Diskin and Kenny2014) must be borne in mind in any attempt to select animals for improved energetic efficiency. Indeed, the potential for antagonistic relationships amongst economically important traits is most appropriately catered for within the context of multi-trait economically weighted selection indices, the basis of beef cattle genetic improvement programmes, worldwide.

Maintenance requirements, mitochondrial function and stress physiology

Typically, total dietary energy intake required for body maintenance far exceeds 50% in adult cattle and in most cases is in excess of 40% in growing cattle consuming forage diets. The large energetic requirement of maintaining homeorhesis is contributed to by a number of physiological and biochemical processes, which may have implications for feed efficiency status, some of which are discussed below.

Maternal traits and fertility

Despite the fact that the greatest benefits of improved RFI may be realised in the cow herd when compared with growing cattle, there are relatively few studies that have examined the effect of RFI status on fertility and maternal productivity traits. Colostrum and milk yield are the principle factors influencing beef calf passive immunity and pre-weaning growth, respectively. In our own studies we have not established any association between RFI ranking and cow serum immunoglobulin concentration prepartum, colostrum yield or total Ig concentration in colostrum of beef cows (McGee and Drennan, Reference McGee and Drennan2006) or indeed subsequent measures of calf passive immunity (McGee and Drennan, Reference McGee and Drennan2006; Lawrence et al., Reference Lawrence, Kenny, Earley, Crews and McGee2011). Although Montanholi et al. (Reference Montanholi, Lam, Peripolli, Vander Voort and Miller2013b) reported a tendency for a positive effect of RFI status on colostrum specific gravity (an indicator of higher immunoglobulin), these authors failed to establish any relationship between RFI and colostrum protein, fat, lactose or total solids concentrations.

Residual feed intake ranking had no significant effect on milk yield of beef cows (McGee and Drennan, Reference McGee and Drennan2006; Lawrence et al., Reference Lawrence, Kenny, Earley, Crews and McGee2011; Morris et al., Reference Morris, Chan, Lopez-Villalobos, Kenyon, Garrick and Blair2014). In terms of milk composition, Montanholi et al. (Reference Montanholi, Lam, Peripolli, Vander Voort and Miller2013b) reported a negative relationship between RFI and milk lactose concentration (r=−0.29) in beef cows but no association with other milk constituents. Consistent with reported results on cow milk yield, calf pre-weaning growth was not associated with maternal status for RFI (Lawrence et al., Reference Lawrence, Kenny, Earley and McGee2013; Morris et al., Reference Morris, Chan, Lopez-Villalobos, Kenyon, Garrick and Blair2014). Given that maternal weaning weight is representative of dam milk yield and, at a genetic level, Crowley et al. (Reference Crowley, Evans, Mc Hugh, Kenny, McGee, Crews and Berry2011) found no relationship between maternal weaning weight and RFI of growing males, these findings corroborate the absence of a phenotypic association between cow RFI ranking and milk yield or progeny performance pre-weaning.

Calving difficulty contributes heavily to production losses and labour costs on beef farms. However, calving difficulty score was not found to differ between cows divergent for RFI (Basarab et al., Reference Basarab, Colazo, Ambrose, Novak, McCartney and Baron2011; Lawrence et al., Reference Lawrence, Kenny, Earley, Crews and McGee2011 and Reference Lawrence, Kenny, Earley and McGee2013; Fitzsimons et al., Reference Fitzsimons, Kenny, Fahey and McGee2014b). There is some evidence, however, to indicate that perinatal calf mortality may be lower for more feed efficient cows (Basarab et al., Reference Basarab, McCartney, Okine and Baron2007 and Reference Basarab, Colazo, Ambrose, Novak, McCartney and Baron2011), though more work is required on the relationship between RFI status and animal health and immunocompetence.

In terms of reproductive performance, no differences were found between high and low-RFI beef females with regard to pregnancy, calving and/or weaning rates (Basarab et al., Reference Basarab, McCartney, Okine and Baron2007; Morris et al., Reference Morris, Chan, Lopez-Villalobos, Kenyon, Garrick and Blair2014; Jones et al., Reference Jones, Accioly, Copping, Deland, Graham, Hebart, Herd, Laurence, Lee, Speijers and Pitchford2016), although in other studies a lower pregnancy and calving rate (Basarab et al., Reference Basarab, Colazo, Ambrose, Novak, McCartney and Baron2011) and lower weaning rate (Copping et al., Reference Copping, Accioly, Deland, Edwards, Graham, Hebart, Herd, Jones, Laurence, Lee, Speijers and Pitchford2016; Hebart et al., Reference Hebart, Accioly, Copping, Deland, Herd, Jones, Laurence, Lee, Lines, Speijers, Walmsley and Pitchford2016) was observed for low-RFI females. Donoghue et al. (Reference Donoghue, Arthur, Wilkins and Herd2011) and Hebart et al. (Reference Hebart, Accioly, Copping, Deland, Herd, Jones, Laurence, Lee, Lines, Speijers, Walmsley and Pitchford2016) reported that low-RFI females calved later in the calving season than their high-RFI contemporaries; however, this was not evident in other studies (Morris et al., Reference Morris, Chan, Lopez-Villalobos, Kenyon, Garrick and Blair2014). Crowley et al. (Reference Crowley, Evans, Mc Hugh, Kenny, McGee, Crews and Berry2011) reported a negative, but not statistically significant, genetic correlation (r=−0.29) between age at first calving and RFI status in beef cattle. The later calving date of low-RFI females recorded in some studies could be attributable to a delay in the onset of puberty (Shaffer et al., Reference Shaffer, Turk, Wagner and Felton2011), although age at puberty was not always different between RFI classifications (Basarab et al., Reference Basarab, Colazo, Ambrose, Novak, McCartney and Baron2011; Donoghue et al., Reference Donoghue, Arthur, Wilkins and Herd2011). The positive association between body fatness and the timing of onset of puberty and postpartum ovarian cyclicity has been well documented and has been proposed as a reason for delayed calving date in more efficient and often leaner animals. In the study of Basarab et al. (Reference Basarab, Colazo, Ambrose, Novak, McCartney and Baron2011), when the DMI prediction model was adjusted for back fat thickness (and feeding event frequency), low-RFI heifers were found to be older at puberty than their less efficient contemporaries. Despite this, however, the above-mentioned adjustments negated the adverse effects of low-RFI on pregnancy rate, calving rate and the proportion of calves born in the first 28 days of the calving season. Indeed, Basarab et al. (Reference Basarab, Colazo, Ambrose, Novak, McCartney and Baron2011) suggested that on-going selection for low-RFI within cohorts of heifers of mixed pubertal status may negatively impact long-term fertility in low-RFI heifers by favouring later-maturing animals that have not incurred the additional energy expenditure associated with sexual activity.

Bull fertility has a central role in herd productivity and is an important trait to include in genetic selection programmes for beef cattle. Certain studies have reported an unfavourable relationship between RFI status and some (Wang et al. Reference Wang, Colazo, Basarab, Goonewardene, Ambrose, Marques, Plastow, Miller and Moore2012; Awda et al., Reference Awda, Miller, Montanholi, Vander Voort, Caldwell, Buhr and Swanson2013) but not all (Awda et al., Reference Awda, Miller, Montanholi, Vander Voort, Caldwell, Buhr and Swanson2013) estimates of semen quality. In addition, other studies have shown no effect of RFI rank on scrotal circumference, a measure of spermatogenic potential (Awda et al., Reference Awda, Miller, Montanholi, Vander Voort, Caldwell, Buhr and Swanson2013; Fontoura et al., Reference Fontoura, Montanholi, Diel de Amorim, Foster, Chenier and Miller2015, Kowalski et al., Reference Kowalski, Fernandes, DiLorenzo, Moletta, Rossi and de Freitas2017) or indeed systemic concentrations of testosterone (Kowalski et al., Reference Kowalski, Fernandes, DiLorenzo, Moletta, Rossi and de Freitas2017). Within the context of multi-sire groups on pasture, Wang et al. (Reference Wang, Colazo, Basarab, Goonewardene, Ambrose, Marques, Plastow, Miller and Moore2012) concluded that there was no evidence for a detrimental effect of selection for improved feed efficiency on the reproductive performance of beef bulls. Similar to heifers (discussed above) the influence of factors such as fatness and sexual activity must be considered in any interpretation of the relationship between RFI status and age at onset of puberty in bulls.

Repeatability and genotype×environment interaction for residual feed intake

Clearly, if RFI is to be included as an economically important trait worthy of consideration in selection programmes, an animal’s status for the trait must be repeatable across the various phases and physiological states of its productive life, as well as across different dietary regimens. High repeatability for a trait is also important in breeding animals where predictions of performance can only be made early in life. Furthermore, genotype×environment interactions are particularly relevant if estimates of genetic merit for improved productivity or feed efficiency are derived under conditions different from that under which progeny are reared (Berry and Crowley, Reference Berry and Crowley2013). Worldwide, performance testing of beef breed bulls is usually carried out using high-energy, concentrate-based diets, whereas the majority of beef cattle are largely produced on predominantly forage-based diets, often grazed pasture, which have very different intake characteristics (as discussed earlier).

Studies examining the repeatability of RFI in growing beef cattle offered the same diet across two ‘separated’ periods have found that RFI was moderately repeatable (r=0.62, Kelly et al., Reference Kelly, McGee, Crews, Sweeney, Boland and Kenny2010b; r=0.40, Gomes et al., Reference Gomes, Sainz, Silva, César, Bonin and Leme2012) and had a moderate rank correlation of 0.52 in cattle offered the same diet across two ‘consecutive’ periods (Durunna et al., Reference Durunna, Colazo, Ambrose, McCartney, Baron and Basarab2012). Similar findings were obtained by Herd et al. (Reference Herd, Dicker, Lee, Johnston, Hammond and Oddy2006) evaluating females post-weaning and subsequently as non-pregnant, non-lactating beef cows 4 to 4.5 years old. However, in commercial practice cattle are usually not offered the same diet throughout life and productive cows are usually pregnant and/or lactating. Nevertheless, moderate phenotypic correlations were reported between RFI measured in steers offered a grower diet and subsequently offered a finisher diet (Durunna et al., Reference Durunna, Mujibi, Goonewardene, Okine, Basarab, Wang and Moore2011; Cassady et al., Reference Cassady, Felix, Beever and Shike2016). Lawrence (Reference Lawrence2011) and Hafla et al. (Reference Hafla, Carstens, Forbes, Tedeschi, Bailey, Walter and Johnson2013) found that RFI status was correlated when measured in heifers offered a diet of forage and concentrates and subsequently in the same animals as cows offered a forage-only diet. Conversely, Black et al. (Reference Black, Bischoff, Mercadante, Marquezini, DiLorenzo, Chase, Coleman, Maddock and Lamb2013) using a similar animal model found no such relationship, though low-RFI weanlings did consume less feed as cows.

A number of studies have reported on the effect of RFI classification, when feed intake of female cattle was measured in confinement and subsequently, at pasture. Beef females previously ranked as divergent for RFI indoors offered a grass silage diet (Lawrence et al., Reference Lawrence, Kenny, Earley and McGee2012 and Reference Lawrence, Kenny, Earley and McGee2013) or hay (Meyer et al., Reference Meyer, Kerley and Kallenbach2008) did not differ in herbage intake when subsequently grazing pasture during first pregnancy (Lawrence et al., Reference Lawrence, Kenny, Earley and McGee2012), during gestation or late lactation (Meyer et al., Reference Meyer, Kerley and Kallenbach2008) or during lactation (Lawrence et al., Reference Lawrence, Kenny, Earley and McGee2013). Similarly, Morris et al. (Reference Morris, Chan, Lopez-Villalobos, Kenyon, Garrick and Blair2014) reported no difference in herbage intake in grazing lactating beef heifers from high- and low-RFI selection lines. In contrast, Manafiazar et al. (Reference Manafiazar, Basarab, Baron, McKeown, Doce, Swift, Undi, Wittenberg and Ominski2015) reported that heifers ranked as low-RFI in an outdoor drylot offered a barley silage-based diet had a lower intake of grazed herbage when subsequently measured as pregnant replacement heifers. Reasons for the discrepancies in DMI between the confined and grazing dietary phases in the majority of the aforementioned studies may be attributed to: (i) the re-ranking per se of animals for RFI over time (maturity); (ii) differences in diet type and thus, associated intake and digestion characteristics; (iii), changes in the physiological state of the animals and, finally, and perhaps most importantly (iv) the inherent difficulty in accurately quantifying herbage intake in grazing cattle (Lawrence et al., Reference Lawrence, Kenny, Earley and McGee2012).

Together, the results of these studies suggest that RFI is a moderately repeatable trait across time (maturity), stages of production and different diet types in beef cattle. However, it is evident that some animal re-ranking occurs, suggesting the existence of a genotype × environment interaction for the trait.

Genetics of residual feed intake

The main obstacles to widespread adoption of feed efficiency in cattle breeding programmes are the large cost and technical difficulty associated with measuring the trait (Nielsen et al., Reference Nielsen, MacNeil, Dekkers, Crews, Rathje, Enns and Weaber2013). The advent of genomically assisted selection approaches, where genomic information is employed to aid the prediction of the breeding merit of an animal, should increase selection accuracy and accelerate genetic improvement (Berry et al., Reference Berry, Garcia and Garrick2016). From their meta-analysis, Berry and Crowley (Reference Berry and Crowley2013) reported a pooled heritability for RFI in growing cattle of 0.33 (range of 0.07 to 0.62). Coupled with its considerable genetic variance (Crowley et al., Reference Crowley, McGee, Kenny, Crews, Evans and Berry2010), the RFI trait is likely to respond favourably to genomic selection. However, genomic prediction accuracy in beef cattle is still not sufficiently high to allow selection of candidates without an appropriate phenotypic measurement (Bolormaa et al., Reference Bolormaa, Pryce, Kemper, Savin, Hayes, Barendse, Zhang, Reich, Mason, Bunch, Harrison, Reverter, Herd, Tier, Graser and Goddard2013). The calculation of genomically informed estimated breeding values depends on the generation of a reference population where the trait of interest (i.e. feed efficiency) has already been measured and animals genotyped for appropriate genomic markers (Hayes et al., Reference Hayes, Bowman, Chamberlain and Goddard2009; Stothard et al., Reference Stothard, Liao, Arantes, De Pauw, Coros, Plastow, Sargolzaei, Crowley, Basarab, Schenkel, Moore and Miller2015; Seabury et al., Reference Seabury, Oldeschulte, Saatchi, Beever, Decker, Halley, Bhattarai, Molaei, Freetly, Hansen, Yampara-Iquise, Johnson, Kerley, Kim, Loy, Marques, Neibergs, Schnabel, Shike, Spangler, Weaber, Garrick and Taylor2017). Such a reference population does not currently exist in beef cattle (Fitzsimons et al., Reference Fitzsimons, McGee, Keogh, Waters and Kenny2017) and its assembly would have to overcome factors such as contrasting breeds, age and nutritional management of cohorts of cattle between, and even within, collaborating research groups.

At present, research on the genomic control of feed efficiency in cattle is focused on the identification of panels of genetic variants of biological significance to the trait (Lu et al., Reference Lu, Sargolzaei, Li, Abo-Ismail, Vander Voort, Wang, Plastow, Moore and Miller2013; Lindholm-Perry et al., Reference Lindholm-Perry, Kuehn, Freetly and Snelling2015; Fitzsimons et al., Reference Fitzsimons, McGee, Keogh, Waters and Kenny2017). However, if ultimately of benefit to industry, it is essential that these polymorphisms are sufficiently robust across breed, phase of development and dietary regimen. A recent genome-wide association study (Seabury et al., Reference Seabury, Oldeschulte, Saatchi, Beever, Decker, Halley, Bhattarai, Molaei, Freetly, Hansen, Yampara-Iquise, Johnson, Kerley, Kim, Loy, Marques, Neibergs, Schnabel, Shike, Spangler, Weaber, Garrick and Taylor2017) comparing quantitative trait loci (QTL) and utilising the Illumina Bovine HD (778K) and SNP50 assay platforms has reported QTLs associated with and influencing feed efficiency-related traits which could potentially be used for genomic selection. Furthermore, projects with the objective of combining international DNA sequence information, such as the Canadian Cattle Genome Project (Stothard et al., Reference Stothard, Liao, Arantes, De Pauw, Coros, Plastow, Sargolzaei, Crowley, Basarab, Schenkel, Moore and Miller2015), aim to develop genomics-based tools to enhance the efficiency and sustainability of beef production. The focus of such collaborative projects should be on the identification of functional variants supported by imputation, where necessary, so that the association between these variants and traits of economic importance such as feed efficiency and related traits can be determined (Taylor et al., Reference Taylor, Beever, Decker, Freetly, Garrick, Hansen, Johnson, Kerley, Loy, Neibergs, Saatchi, Schnabel, Seabury, Shike, Spangler and Weaber2017). Future success in breeding for improved feed efficiency in beef cattle will depend on the incorporation of such genetic information into national and international multi-trait genomic selection based breeding programmes.