Characterization of feeding behavior traits in steers with divergent residual feed intake consuming a high-concentrate diet

Abstract

The objective of this study was to examine the differences in feeding behavior patterns of steers with divergent phenotypes for residual feed intake (RFI). Three trials were conducted with 508 Angus-based composite crossbred steers (body weight [BW] = 309 ± 57 kg) fed a high-concentrate diet in pens equipped with electronic feed bunks (GrowSafe System). Initial and final carcass ultrasound measurements (intra-muscular fat, backfat depth, and rib-eye area) were collected on days 0 and 70, and BW measured at 14-d intervals. Individual dry matter intake (DMI) and feeding behavior traits were collected for 70 d, and RFI calculated as the residual from the regression of DMI on average daily gain (ADG) and mid-test BW^0.75. Steers were ranked by RFI and assigned to low-, medium-, and high-RFI classes based on ± 0.5 SD from the mean RFI within the trial. The feeding behavior traits evaluated in this study included frequency and duration of bunk visit (BV) and meal events, head-down (HD) duration, mean meal length, time-to-bunk interval, the maximum nonfeeding interval, and the day-to-day variation of these traits, defined as the root mean squared error (RMSE) from linear regression of each trait on the day of trial. Additionally, three ratio traits were evaluated: BV events per meal, HD duration per BV event, and HD duration per meal event. Low-RFI (feed-efficient) steers consumed 16% less (P < 0.01) DMI, while BW and ADG were not different compared with high-RFI steers. Low-RFI steers had 18% fewer and 21% shorter (P < 0.01) BV events, and 11% fewer and 13% shorter (P < 0.01) meal events per day compared with high-RFI steers. Furthermore, low-RFI steers exhibited less (P < 0.05) day-to-day variance in DMI, as well as in frequency and duration of BV and meal events and HD duration compared with high-RFI steers. Differences in feeding behavior traits due to RFI were minimally affected by covariate adjustment for DMI, indicating that steers with divergent RFI have distinct feeding behavior patterns that are largely independent of differences in DMI. These results suggest that feeding behavior traits may be useful biomarkers for the prediction of feed efficiency in beef cattle.

Introduction

Although feed intake is regulated by multiple, complex mechanisms not completely understood, the consumption of feed is inextricably associated with feeding behavior (Allen, 2014). Day-to-day variation in feed intake has been associated with an increased incidence of digestive disorders and reduced feed efficiency in feedlot cattle (Pritchard and Bruns, 2003). Previous studies have found that feed efficiency, measured using residual feed intake (RFI), is highly correlated with feeding behavior patterns (Hafla et al., 2013; Cantalapiedra-Hijar et al., 2018). However, studies to date have focused on the frequency and duration of bunk visit (BV) or feeding events, while few studies have examined the associations between RFI and day-to-day variation in feeding behavior patterns.

Selecting animals that have superior genetic merit for growth efficiency offers considerable economic and environmental benefits (Carstens and Kerley, 2009). Previous research found that variation in RFI was associated with distinctive differences in feeding behavior (Allen, 2014; Scanes and Hill, 2017). This indicates that feeding behavior may serve as a reliable indicator for many complex mechanisms controlling the efficiency of feed utilization. Understanding the relationship between feeding behavior and traditional metrics of feed efficiency will provide new variables to be utilized in selection indexes designed to aid in selection for animals with superior RFI phenotypes. The objectives of this study were to characterize feeding behavior patterns in steers with divergent RFI phenotypes and to explore the relationship between growth, feeding behavior, and feed efficiency.

Materials and Methods

Animals and experimental design

All animal care and use procedures were in accordance with the guidelines for the use of Animals in Agriculture Teaching and Research as approved by the Texas A&M University Institutional Animal Care and Use Committee.

This study was comprised of three trials that were conducted with Angus-based composite steers (N = 508), with an initial body weight (BW) of 309 ± 56 kg and age of 290 ± 16 d. Upon arrival at the Texas A&M AgriLife McGregor Research Center (McGregor, TX), steers were vaccinated, dewormed, and fitted with passive, half-duplex radio frequency identification (RFID) ear tags (Allflex USA Inc., Dallas, TX). Steers were randomly assigned to one of two pens (46 × 58 m, 85 steers per pen) equipped with 10 electronic feed bunks (GrowSafe Systems LTD., Airdrie, AB, Canada) to measure the feed intake and feeding behavior traits. Steers were adapted to a high-concentrate diet (Table 1) for 28 d, after which ad libitum feed intake and feeding behavior data were collected for 70 d. The GrowSafe system consisted of feed bunks equipped with load bars to measure feed disappearance and an antenna within each bunk to record the animal presence by detection of the animal’s unique RFID tag during feeding events.

Table 1.

Ingredient and chemical composition of experimental diets

Item	1	2	3
Ingredient composition, % as-fed
Dry rolled corn	72.7	73.7	74.3
Brome hay	5.5	6.0	5.4
Cottonseed meal	8.0	6.0	7.8
Cottonseed hulls	5.5	6.0	5.4
Molasses	5.0	5.0	6.0
Mineral premix1	2.5	2.5	2.5
Urea	0.8	0.8	0.7
Chemical analysis, % DM
DM, %	88	90.2	88
CP, %	11	12.6	14.9
NDF, %	17.9	20.3	20.8
NEm, Mcal/kg	1.59	1.74	1.70
NEg, Mcal/kg	1.06	1.16	1.08

¹Mineral premix contained minimum 15.5% Ca, 2,800 ppm Zn, 1,200 ppm Mn, 12 ppm Se, 14 ppm Co, 30 ppm I, 45.4 KIU/kg Vit-A, 2.3 KIU/kg Vit-D, 726 IU/kg Vit-E, 1,200 Monensin, and 400 ppm Tylan.

Computation of traits and statistical analysis

Individual feed intake was computed using a subroutine of the GrowSafe 6000E software (Process Feed Intakes) based on continuous recordings of feed disappearance during feeding events. Assigned feed disappearance (AFD) rates were computed daily for each feed bunk to assess data quality. Similar to other feeding behavior studies (Mendes et al., 2011; Kayser and Hill, 2013), data were excluded by pen when the AFD of an individual bunk in a pen was less than 90% or the average AFD for a pen was less than 95%. For pens 1 and 2, 15 and 11 d were excluded in trial 1, 22 and 4 d were excluded in trial 2, and 12 and 15 d were excluded in trial 3, respectively. The average AFD for the days included in the analysis was 98.3%, 98.8%, and 97.3% for trials 1, 2, and 3, respectively. While not optimal, the data collection exceeded the 35 d recommended in the literature for the optimum duration of feed intake testing (Archer et al., 1997).

Feeding behavior traits evaluated in this study were based on the frequency and duration of BV events. BV frequency was defined as the number of BV events recorded during a 24-h period, regardless of whether feed was consumed, with BV duration defined as the sum of the lengths of BV events recorded during a 24-h period. A BV event commenced when the RFID tag of an animal was first detected at a feed bunk and ended when either the duration of time between the last two consecutive RFID recordings exceeded 100 s (a parameter setting in the GrowSafe 6000E software [GrowSafe Systems Ltd.]), the RFID tag was detected in another bunk, or the RFID tag of another animal was detected at the same bunk (Mendes et al., 2011). The intervals between BV events were defined as the nonfeeding intervals (NFI), with maximal NFI defined by the longest NFI within each day. Head-down (HD) duration was computed as the sum of the number of times the RFID ear tag for an animal was detected each day multiplied by the scan rate of the GrowSafe system. Time to bunk (TTB) was defined as latency between feed delivery and an animal’s first BV event each day (Table 2).

Table 2.

Definitions of feeding behavior traits evaluated in the study

Item	Definition
BV frequency, events/d	Number of BV events recorded each day
BV duration, min/d	Sum of the durations of BV events recorded each day
HD duration, min/d	Number of RFID recordings each day multiplied by the scan rate of the GrowSafe system
Meal frequency, events/d	Number of meal events recorded each day
Meal duration, min/d	Sum of the durations of meal events recorded each day
Mean meal length, min/meal	Average length of meal events
Max NFI	Maximum nonfeeding interval each day
TTB, min	Length of interval from feed delivery to the first BV event each day
HD duration per meal duration	Ratio of HD duration to meal duration
HD duration per BV duration	Ratio of HD duration to BV duration
BV events per meal event	Ratio of number of BV per meal event

Meal events represent clusters of BV events that are differentiated from the next meal by an NFI that is longer than the NFI within a meal (Yeates et al., 2001). The longest NFI considered to be a part of a meal is defined as the meal criterion and can be estimated by fitting a two-pool bimodal probability density function to log₁₀-transformed NFI data for individual animals (Bailey et al., 2012). For this study, meal criterion was computed using a two-pool Gaussian–Weibull probability density function using the Meal Criterion Calculation Software (MCC; ver. 1.7.6836.33854; http://nutritionmodels.tamu.edu), which was then used to transform BV-event data into meal traits (Table 2).

Day-to-day variation of the frequency and duration of BV and meal events, HD duration, Max NFI, TTB, Mean meal length, and DMI were computed for each animal as the root mean squared error (RMSE) of the residuals from the linear regression of each of the feeding behavior trait on the day of trial (Putz et al., 2019) using the Standard Least Squares procedure of JMP (SAS Inst. Inc., Cary, NC). Additionally, to further assess the intensity of feeding behavior patterns, the following ratios of feeding behavior traits were also evaluated: BV frequency per meal event, HD duration per meal event, and HD duration per BV event.

During the 70-d trials, BW was measured at 14-d intervals, and ultrasound measurements of backfat depth, intramuscular fat, and longissimus muscle (LM) area obtained on days 0 and 70 by a certified technician using an Aloka 500-V instrument with a 17-cm, 3.5-MHz transducer (Corometrics Medical Systems Inc., Wallington, CT). Diet samples were collected weekly and composited by weight at the end of each trial. Diet dry matter percentage was measured by drying samples in a forced air oven for 48 h at 105 °C. Samples were sent to Cumberland Valley Analytical Services Inc. (Maugansville, MD) for analysis of ash (535 C; AOAC 2000; method 942.05), acid detergent fiber (AOAC, 2000; method 973.18), neutral detergent fiber (NDF) with heat-stable ∞-amylase and sodium sulfite (Van Soest et al., 1991), and CP (N × 6.25; AOAC 2000; method 990.03, Leco FP528 Nitrogen Analyzer, Lec, St. Joseph, MI).

Growth rates of individual steers were calculated by linear regression of serial BW on the day of trial using the PROC GLM procedure of SAS, and the regression coefficients used to compute average daily gain (ADG), and mid-test BW^0.75 (Koch et al., 1963). Moisture analysis from weekly samples was used to adjust feed intake measurements to determine daily dry matter intake (DMI). Estimates for missing feed intake data were derived from linear regression of the feed intake on the day of trial as described in Hebart et al. (2004).

RFI was computed as the residual from the linear regression of DMI on ADG and mid-test BW^0.75 (Koch et al., 1963), with trial included as a fixed effect. Similarly, residual gain (RG) was computed as the residual from the linear regression of ADG on DMI and mid-test BW^0.75 (Koch et al., 1963), with trial included as a fixed effect. Steers were ranked by RFI and classified into one of three RFI phenotypic groups: low, medium, or high RFI based on ± 0.5 SD from mean RFI within trial (Lancaster et al., 2009; Crowley et al., 2010). To evaluate the effect of RFI classification on performance, feed intake, carcass ultrasound, and feeding behavior traits, a mixed model (SAS Inst. Inc., Cary, NC) was used that included the fixed effect of RFI classification and the random effect of trial. To determine if the effects of RFI on feeding behavior traits were due to the differences in DMI, a mixed model was used that included the fixed effect of RFI classification, DMI as a covariate, RFI × DMI interaction, and random effect of trial class (Table 6). The RFI × DMI covariate interaction was not significant and thus removed from the final model. Tukey–Kramer post hoc test was used to evaluate the differences among RFI class means. To generate phenotypic correlation coefficients, performance, feed efficiency, and feeding behavior traits were adjusted for the random effect of trial, then used in the multivariate platform of JMP (SAS Inst. Inc., Cary, NC) to obtain phenotypic correlations (Table 5).

Table 6.

The effects of RFI classification on feeding behavior traits in steers fed a high-concentrate diet (model included DMI as a covariate)

Trait1	Low RFI	Medium RFI	High RFI	SE	P-value
					RFI	DMI2
BV traits:
BV frequency, events/d	42.8a	49.2b	54.7c	6.24	<0.001	0.025
BV duration, min/d	58.2a	62.3b	69.7c	3.07	<0.001	0.001
Meal traits:
Meal frequency, events/d	5.30a	6.20b	6.50b	0.75	<0.001	0.037
Meal duration, min/d	120a	121a	131b	5	<0.001	0.068
Intensity traits:
HD duration, min/d	38.6a	44.0b	53.2c	1.3	<0.001	0.015
TTB, min	38.1	35.8	35.3	7.7	0.33	0.058

¹RMSE = day-to-day variation of the traits during the trial.

² P-values for the DMI covariate term.

^a–cMeans within row with different superscripts differ (P < 0.05).

Table 5.

Pearson correlations between performance and feed efficiency and feeding behavior traits in steers fed a high-concentrate diet

Item1	DMI	ADG	G:F	RG	RFI
DMI	---	0.45*	−0.39*	0.00	0.75*
ADG	0.45*	---	0.63*	0.82*	0.00
BV frequency	0.17*	−0.05	−0.18*	−0.05	0.38*
BV duration	0.42*	0.07	−0.30*	−0.02	0.55*
Head down (HD) duration	0.41*	0.04	−0.31*	−0.06	0.57*
Max NFI	−0.12*	−0.13	−0.04	−0.16*	−0.07
TTB	−0.10*	−0.21*	−0.14*	−0.14*	−0.15*
Meal frequency	−0.02	−0.04	−0.01	−0.04	0.11*
Meal duration	0.20*	0.19*	0.03	0.13*	0.25*
Meal length	0.14*	0.13	0.00	0.08	0.08
HD duration per BV duration	0.30*	−0.02	−0.29*	−0.12*	0.47*
HD duration per meal duration	0.30*	−0.09	−0.34*	−0.16*	0.42*
BV events per meal event	0.15*	0.01	−0.12*	−0.01	0.15*
Day-to-day variation (RMSE) of feeding behavior
BV frequency RMSE	0.23*	−0.33*	−0.38*	−0.08	−0.21*
BV duration RMSE	0.16*	−0.22*	−0.32	−0.09	0.25*
HD duration RMSE	0.24*	−0.16*	−0.31*	−0.12*	0.38*
TTB RMSE	−0.15*	0.01	0.11*	−0.04	−0.14*
Meal frequency RMSE	0.03	−0.13*	−0.14*	0.01	0.10*
Meal duration RMSE	0.16*	0.02	−0.09*	0.00	0.17*

¹RMSE = day-to-day variation during the trial.

*Correlations are different from zero at P < 0.05.

Results and Discussion

Growth and performance

Performance, feed efficiency, and ultrasound least squared means are presented in Table 3. The initial age of steers at the start of the trials averaged 290 ± 16 d and ranged from 280 to 313 d. The means and SD for ADG and DMI were 1.71 ± 0.26 and 10.1 ± 1.1 kg/d, respectively, which are consistent with growth patterns expected from steers of this breed, weight, and age class. In this study, variation in ADG and mid-test BW^0.75 only accounted for 45% of the variation in DMI, which is less than in previous studies (Lancaster et al., 2009; Hafla et al., 2013) that reported R² ranging from 0.61 to 0.80. Means and SD for RG and RFI were 0.00 ± 0.19 and 0.00 ± 0.78 kg/d, with RG ranging from −0.55 to 0.57 kg/d and RFI ranging from −3.38 to 2.30 kg/d, respectively. The effect of RFI classification did not affect initial age, BW, or hip height. Low-RFI steers consumed 16.4% less DMI and had 20.6% higher G:F than high-RFI steers, with no differences observed in ADG or final BW. These results agree with other studies that have reported 13% to 20% less feed consumption by low-RFI compared with high-RFI cattle (Lancaster et al., 2009; Baldassini et al., 2018). As expected, low-RFI steers exhibited greater (P < 0.01) RG compared with high-RFI steers, which is similar to the results previously reported by Crowley et al. (2010).

Table 3.

The effects of RFI classification on performance, feed efficiency, and ultrasound traits in steers fed a high-concentrate diet

Trait1	Low RFI	Medium RFI	High RFI	SE	P-value
No. of animals	147	199	152
Performance traits:
Initial age, d	290	290	290	9	0.908
Initial BW, kg	308	310	310	35	0.827
Final BW, kg	428	430	429	28	0.864
ADG, kg/d	1.72	1.72	1.70	0.11	0.570
Initial hip height, cm	122	122	121	3	0.136
DMI, kg/d	9.2a	10.2b	11.0c	0.2	<0.001
DMI RMSE, kg/d	2.18a	2.28b	2.42c	0.08	0.033
Feed efficiency traits:
RFI, kg/d	0.911a	0.009b	0.880c	0.030	<0.001
RG, g/d	59.09a	9.76b	−68.29c	11.53	<0.001
G:F	0.187a	0.170b	0.155c	0.014	<0.001
Ultrasound traits:
Initial backfat depth, cm	0.370a	0.394ab	0.407b	0.109	0.013
Final backfat depth, cm	0.648a	0.734b	0.758b	0.085	<0.001
Initial intramuscular fat,%	2.91	2.82	2.82	0.26	0.198
Final intramuscular fat,%	3.05	3.18	3.21	0.23	0.054
Initial LM area, cm2	20.4	20.6	20.2	1.3	0.348
Final LM area, cm2	26.2	25.9	25.8	1.0	0.415

¹DMI RMSE = day-to-day variation in DMI during the trial.

^a–cMeans within row with different superscripts differ (P < 0.05).

Initial intramuscular fat and initial and final LM area were not different (P > 0.10) across divergent RFI phenotypes. However, low-RFI steers had decreased (P = 0.05) final intramuscular fat and reduced (P < 0.01) initial and final backfat depth compared with high-RFI steers. Similar results were reported by Lancaster et al. (2009) with bulls and by Lines et al. (2009) in heifers and are consistent with other studies that have found carcass fatness to be positively associated with RFI (Nkrumah et al., 2007; Shaffer et al., 2011; Hafla et al., 2013).

Feeding behavior traits

The least squared means for feeding behavior traits are presented in Table 4. Low-RFI steers had 18% fewer (P < 0.01) BV events and spent 21% less (P < 0.01) time at the bunk each day compared with high-RFI steers. These studies are in agreement with Nkrumah et al. (2007) who found that low-RFI heifers fed a high-concentrate diet had 14% fewer BV events and spent 24% less time at the bunk than high-RFI heifers. Likewise, other studies have reported a lower frequency of BV events by low-RFI growing steers and heifers fed high-concentrate diets (Golden et al., 2008; Montanholi et al., 2009), and cows consuming high-roughage diets compared with their respective high-RFI contemporaries (Fitzsimons et al., 2014). In Angus cattle divergently selected for RFI and a fed high-concentrate diet, Herd et al. (2019) reported that BV frequency was highly correlated in a positive manner with both genetic and phenotypic RFI. However, in several studies that utilized high-concentrate diets, BV frequency was not associated with variation in RFI (Basarab et al., 2007; Dobos and Herd, 2008). In agreement with our results, Kenny et al. (2018) conducted a nine-study meta-analysis with beef cattle fed high-concentrate diets and found that low-RFI cattle had a 10% shorter BV duration than high-RFI cattle. Some studies have not observed associations between BV duration and RFI when high-roughage diets were offered (Kelly et al., 2010; Basarab et al., 2011), suggesting that the type of diet offered may influence whether or not variation in BV duration is associated with RFI. The lack of consistency in BV feeding patterns due to RFI is also likely influenced by the type of electronic feed-intake measurement system (e.g., GrowSafe, Insetec system) and the stocking density of feed bunks used in the studies. Similar differences in frequency and duration of BV events have also been observed in sheep (Cammack et al., 2005) and swine (Young et al., 2011) with divergent RFI phenotypes, indicating that this phenomenon occurs across species.

Table 4.

The effects of RFI classification on feeding behavior traits in steers fed a high-concentrate diet

Trait1	Low RFI	Medium RFI	High RFI	SE	P-value
No. animals	147	199	152
BV traits:
BV frequency, events/d	43.9a	49.1b	53.7c	6.1	<0.001
BV duration, min/d	56.2a	62.3b	71.5c	2.8	<0.001
Max NFI, min/d	723	726	709	20	0.103
BV eating rate, g/min	170a	168a	159b	9	0.002
Meal traits:
Meal criterion, min	14.1a	11.9b	12.7a	2.1	0.019
Meal frequency, events/d	5.59a	6.22b	6.28b	0.72	0.013
Meal duration, min/d	117a	121a	134b	5	<0.001
Mean meal length, min/event	25.0ab	24.6a	27.4b	2.9	0.018
Mean meal size, kg/event	1.80a	1.85a	2.04b	0.19	0.001
Meal eating rate, g/min	81.2a	87.2b	85.7ab	4.6	0.006
Intensity traits:
HD duration, min/d	36.9a	44.1b	54.8c	0.9	<0.001
TTB, min	39.5a	35.8b	34.0b	7.7	0.002
HD duration per BV duration	0.649a	0.702b	0.762c	0.030	<0.001
HD duration per meal duration	0.321a	0.371b	0.420c	0.013	<0.001
BV events per meal event	8.30a	8.63a	9.48b	0.48	<0.001
Day-to-day variation of traits:
BV frequency RMSE, events/d	14.6a	16.3b	17.5c	1.5	<0.001
BV duration RMSE, min/d	17.8a	18.8b	20.5c	1.0	<0.001
Max NFI RMSE, min	184	182	182	9	0.752
Meal frequency RMSE, events/d	1.81a	2.12b	2.08b	0.28	0.009
Meal duration RMSE, min/d	32.2a	31.6a	34.7b	1.0	0.003
Mean meal length RMSE, min/event	8.49	8.51	9.12	0.80	0.222
HD duration RMSE, min/d	12.4a	13.9b	16.3c	0.6	<0.001
TTB RMSE, min	53.1a	50.7ab	48.2b	7.1	0.040

¹RMSE = day-to-day variation of the traits during the trial.

^a-cMeans within row with different superscripts differ (P < 0.05).

In the current study, BV eating rates were 7% higher (P < 0.01) for low- than high-RFI steers, which is in contrast to most studies that have reported slower BV eating rates for low-RFI cattle fed high-concentrate diets (Robinson and Oddy, 2004; Montanholi et al., 2010; Kenny et al., 2018). In addition to recording uninterrupted BV events, the feed-intake measurement system used in this study also records BV events when animals switch feed bunks or when animals are displaced temporarily from the same feed bunk by other animals (Mendes et al., 2011). Although frequency and duration of BV events are clearly associated with RFI, meal-behavior patterns may be more biologically relevant in assessing mechanisms that regulate appetite and feed efficiency as they are less subject to factors, such as social hierarchy, bunk competition, and environmental conditions (Tolkamp and Kyriazakis., 1999; Forbes and Gregorini, 2015). To date, most studies that have examined the effects of RFI on meal-feeding behavior have used a static meal criterion of 5 or 7 min to transform BV data into meal traits (Lancaster et al., 2009; Montanholi et al., 2009; Kayser and Hill, 2013; McGee et al., 2014). However, in these studies, associations between meal frequency and RFI were not observed. While Lancaster et al. (2009) and McGee et al. (2014) reported that low-RFI bulls had lower meal duration then high-RFI bulls, the other two studies found no associations between meal duration and RFI. Alternatively, meal criterion can be estimated on an individual-animal basis by fitting a two-pool bimodal probability density function to NFI data (Bailey et al., 2012). Computed in this manner, low-RFI steers had a 11% longer (P < 0.05) meal criterion, which resulted in fewer (P < 0.01) meal events per day than high-RFI steers. The average meal length of low-RFI steers did not differ from either medium- or high-RFI steers, but the average meal size was smaller (P < 0.001) in low- compared with medium- and high-RFI steers. Daily meal duration was 13% less (P < 0.001) in steers with low RFI than steers with high RFI. Thus, low-RFI steers exhibited fewer and shorter BV as well as meal events each day. Although meal eating rates of low-RFI steers were slower (P < 0.05) than medium-RFI steers, meal eating rates of high-RFI steers were intermediate and did not differ from low-RFI steers.

Distinctive differences observed in the daily feeding durations and BV eating rates of animals with divergent RFI phenotypes may impact the heat increment associated with eating, as the rate of ingestion and duration of feeding have been reported as key factors influencing energy expenditures during the ingestion of food (Adam et al., 1984; Susenbeth et al., 1998). Thus, low-RFI animals may have a reduced heat increment associated with eating due to the reduced feed intake and lower BV and meal durations compared with high-RFI animals.

HD duration is computed as the number of times an animal’s RFID is read by the feed-intake measurement system (GrowSafe System) multiplied by the read-rate of the system. Low-RFI steers had 33% lower (P < 0.001) HD durations than high-RFI steers. In studies that have used the GrowSafe system to compare feeding behavior patterns of cattle with divergent RFI, HD duration has been reported to be consistently lower in low- than in high-RFI cattle (Nkrumah et al., 2007; Lancaster et al., 2009; Durunna et al., 2011; Kayser and Hill, 2013). HD duration per BV and meal duration were less (P < 0.001) in low- than high-RFI steers, which is in agreement with the results from Kayser and Hill (2013). These findings suggest that the intensity of feeding behavior patterns during consumption of feed was greater in steers with high-RFI phenotypes. Furthermore, low-RFI steers took 5.6 min longer (P < 0.01) each day to approach the feed bunk following feed delivery and exhibited fewer (P < 0.001) BV events per meal than their high-RFI contemporaries. Collectively, these results suggest that cattle with low-RFI phenotypes may have altered appetite signaling mechanisms that mitigate feeding behavior patterns (Perkins et al., 2014).

Feeding behavior is governed by both the external environment factors and innate biological mechanisms (Allen, 2014). External factors include feeding management protocols (Pritchard and Bruns, 2003), weather, and competition for feed bunk space (Pritchard and Bruns, 2003; Haskell et al., 2019). Innate biological signals from the distention of the rumen and gastrointestinal tract control hunger in animals on high-fiber diets, while the metabolism of digestion products and subsequent signals from fuel-sensing tissues likely limit appetite in animals consuming energy-rich diet (Scanes and Hill, 2017). It has been hypothesized that inter-animal variation in energy expended on feeding activity may explain a portion of the variation in feed efficiency (Adam et al., 1984; Fitzsimons et al., 2014; Asher et al., 2018). Richardson et al. (2004) found that feeding behavior explained approximately 12% of the variation in feed intake, while Lancaster et al. (2009) reported that meal duration, HD duration, and meal frequency increased the R² from 0.78 to 0.86 in RFI models. This would agree with Susenbeth et al. (1998), who found total energy expenditure associated with eating was driven by total time spent eating, not ingestion or eating rates. However, studies on grazing animals in poor-quality forage indicate that the energy cost associated with foraging activity only accounts for 11% of total heat production (Brosh et al., 2006, 2010) and is even smaller in higher quality forages. Given that behavior is driven by a complex interaction of metabolic and social signals (Allen, 2014), it is important to consider that feeding behavior is a function of biological phenotypes and the environment. Thus, more research is needed to understand the physiological state and environmental influence on feed efficiency.

Day-to-day variation in feeding behavior

Compared with low-RFI animals, high-RFI animals exhibited more (P < 0.05) day-to-day variation in DMI. Additionally, high-RFI steers had greater (P < 0.01) day-to-day variation in HD duration, and frequency and duration of both BV and meal events. Few studies have examined the associations between daily variation in DMI or feeding behavior patterns and performance or feed efficiency. Stock et al. (1995) reported reduced G:F in cattle exhibiting greater daily variation in feed intake. Likewise, G:F was reduced in cattle subjected to experimentally imposed disruptions in daily intake delivery (Galyean and Hubbert., 1995; Soto-Navarro et al., 2000). Although associations between feed efficiency and daily feed intake fluctuations have not always been observed (Schwartzkopf-Genswein et al., 2011), previous studies generally attribute reduced feed efficiency to increased metabolic disturbances such as subclinical acidosis (Gibb and McAllister, 1999; Soto-Navarro et al., 2000; Pritchard and Bruns, 2003; Schwartzkopf-Genswein et al., 2003). However, ruminal pH data are limited in previous studies, so it is unclear whether increased day-to-day variation in DMI may attribute to metabolic disorders in cattle (Gibb and McAllister, 1999) or if it can predispose cattle to metabolic disorders (Galyean and Hubbert, 1995).

Previous research has shown that monensin supplementation favorably affects feed efficiency. Angus steers fed a high-concentrate diet with monensin exhibited decreased feed intake, slower eating rates, and fewer feeding events, while performance was unaffected compared with control animals (Baile et al., 1979). In transition dairy cows, Mullins et al. (2012) found that monensin reduced the interval between meal events during both prepartum and postpartum periods. These effects are similar to the differences observed between low- and high-RFI animals observed in the current study. Few studies have examined the effects of monensin on day-to-day variation in feeding behavior traits; however, several studies have reported that monensin appears to stabilize the day-to-day variation in feed intake (Burrin et al., 1988; Soto-Navarro et al., 2000; Erickson et al., 2003; Millen et al., 2015). The lower day-to-day variation in feed intake and feeding behavior may be indicative of a more favorable rumen environment, which results in improved performance and feed efficiency (Cooper et al., 1997). The similarity in feeding behavior patterns observed between monensin fed vs. control cattle and low- vs. high-RFI animals further indicates that differences in feeding behavior contribute to observed variation in feed efficiency.

Competition for feed bunk space has also been associated with increased day-to-day variation in feeding behavior. DeVries and von Keyserlingk, (2009) reported increased day-to-day variation in feeding behavior of heifers with limited access to the feed bunk in a study conducted with an Instetec system, even though feed was offered ad libitum. Additionally, Hosseinkhani et al. (2008) found that although increased competition in dairy cows increased eating rate and decreased BV duration, sorting remains the same, indicating that diet selection alone could not explain the difference in feed efficiency. However, this finding is contrary to that reported in beef steers, where social dominance was evaluated based upon an animal’s propensity to displace another animal from the feed bunk (Val-Laillet et al., 2008). Increased bunk displacement was indicative of reduced DMI, G:F, and increased RFI (Haskell et al., 2019). Animals with a higher displacement index also exhibited increased day-to-day variation in feeding behavior and reduced feed efficiency, indicating that the two may be related (Haskell et al., 2019).

Correlations between performance and feeding behavior

BV and HD duration were highly correlated (P < 0.05) with DMI (0.42 and 0.41, respectively; Table 5). Kayser and Hill (2013) reported that HD duration was positively correlated with DMI (0.37 and 0.52; P < 0.01) for Angus and Hereford bulls, respectively. Similar correlations were reported by Lancaster et al. (2009) in Angus bulls (0.38; P < 0.05). The phenotypic correlations between RFI and BV frequency or BV duration were 0.38 and 0.55 (P < 0.05), respectively. HD duration was highly correlated (P < 0.05) with RFI, which is similar to other studies using Angus and Hereford bulls (0.41 and 0.59, respectively; Kayser and Hill 2013). Additionally, the ratio traits BV duration per meal duration and HD duration per meal duration were highly correlated with RFI (0.47 and 0.42; P < 0.01, respectively). Latency to approach the bunk following feed delivery has been described as a dominance trait (Haskell et al., 2019) and negatively associated with feed intake (Haskell et al., 2019). This may explain why TTB is the only feeding behavior trait negatively correlated (−0.15; P < 0.05) with RFI, indicating that greater latency to feed following delivery resulted in decreased DMI (−0.10; P < 0.05), ADG (−0.21; P < 0.05), and G:F (−0.14; P < 0.05).

As feeding behavior traits were positively correlated with both DMI and RFI, a DMI covariate model was used to determine if differences in feeding behavior patterns due to RFI classification were associated with differences in DMI (Table 6). As expected, DMI was a significant covariate for the frequency and duration of BV and meal events (P = 0.07 for meal duration) and HD duration. However, the effects of RFI remained significant, with low-RFI steers having fewer and shorter BV and meal events, and shorter HD duration compared with steers with high RFI. Although DMI tended (P < 0.10) to be a significant covariate for TTB, RFI classification did not affect TTB when covariately adjusted for DMI. These results indicate that steers with divergent RFI have distinct feeding behavior patterns that are independent of differences in DMI.

Summary and Conclusions

The results from this study indicate that steers with low-RFI phenotypes exhibited distinctive feeding behavior patterns including fewer and shorter BV and meal events, and less HD duration than steers with high-RFI phenotypes. Additionally, HD duration per BV and meal duration and frequency of BV events per meal were less in low-RFI steers suggesting that high-RFI steers exhibited more intense feeding behavior patterns while consuming feed, which may be associated with altered appetite signaling mechanisms. Furthermore, the results from this study demonstrate that low-RFI steers exhibit less day-to-day variation of DMI and feeding behavior patterns, which may contribute to a more favorable rumen environment and greater efficiency of feed utilization.

Glossary

Abbreviations

ADG

average daily gain

AFD

assigned feed disappearance

BV

bunk visit

BW

body weight

DMI

dry matter intake

HD

head-down

LM

longissimus muscle

NFI

nonfeeding intervals

RFI

residual feed intake

RFID

radio frequency identification

RG

residual gain

RMSE

root mean square error

TTB

time to bunk

Conflict of interest statement

The authors declare no real or perceived conflicts of interest.