Journal of Animal Science

Mehdi, Eslamizad,;Lisa-Maria, Tümmler,;Michael, Derno,;Matthias, Hoch,;Björn, Kuhla,

doi: 10.1093/jas/sky337pmid: 30256955

Abstract The pressure-based noseband sensor system (RWS: RumiWatch System; ITIN + HOCH GmbH Feeding Technology, Liestal, Switzerland) has recently been validated for the measurement of rumination time in mature cows. We aimed in this study at developing a similar pressure-based system for monitoring rumination in young dairy calves. To this end, a vegetable oil–filled silicon tube with a built-in pressure sensor (outer diameter 5.7 mm, length 38 cm) was attached to the noseband of a calf halter. In contrast to the RWS developed for mature cows, the accelerometer, the battery, the data logger, and the SD card of the RWS were integrated into 1 box to reduce the weight of the RWS to 0.35 kg. The box was attached to the halter so that it was located behind the right ear of the calf. Ten pre-weaned German Holstein calves (49–106 kg BW and 33–63 days of age) were equipped with the RWS. Calves were milk-fed thrice a day and offered hay and commercial starter for ad libitum intake. In parallel, animals were monitored by a video camera connected to a video recorder for 12 h. Two independent observers assessed the video records to obtain a reliable gold standard for the evaluation of the newly developed RWS. Data obtained by either RWS or visual video observation were processed as min rumination per h, yielding a total of 120 pairs of values (12 pairs per animal) for regression analysis. Assessment of 2 independent observers were highly correlated (r = 0.99). Results indicated relatively low random error between results obtained from the RWS (on y-axis) and video observations (on x-axis) (R2 = 0.82). However, the intercept of the regression line (y = 7.70 + 0.64 x) was significantly different from zero (P < 0.01) and the 95% confidence interval of the slope (0.79–0.94) did not include the value of 1. This translates to a significant systemic error resulting in overestimation of rumination time which is attributable to nutritive and nonnutritive oral activities that almost exclusively lasted for up to 10 min. Exclusion of false positive rumination signals lasting less than 10 or 5 consecutive min from the analysis reduced the random and systemic errors of the model (R2 = 0.86 and 0.93, respectively). We conclude that the newly developed RWS can be used to provide accurate measurement of rumination time in young calves. However, an extra programmed algorithm in the evaluation software is recommended to make the system more user-friendly for measurements on calves. INTRODUCTION Measuring rumination behavior in calves is of great importance for monitoring rumen development and roughage intake as well as calf health and well-being. Rumination is absent in newborn calves and its development is critical for stabilizing rumen fermentation and the development of normal rumen function (Swanson and Harris, 1958; Baldwin et al., 2004). Rumination time is positively correlated with dry feed intake (Swanson and Harris, 1958) and dry feed intake is directly linked with calf growth and future milk production (Gelsinger et al., 2016). In addition, a reduction in rumination time is considered as an indicator of anxiety in cattle (Borderas et al., 2008) and associated with increases in serum cortisol levels under stressful situations (Bristow and Holmes, 2007). Traditional methods of measuring rumination by direct observation are very laborious and time consuming (Beauchemin et al., 1989; Elischer et al., 2013). Therefore, developing and validating automated equipment for monitoring rumination in calves are warranted. Various systems have been developed and validated for automatically measuring feeding behavior of adults cows among which are pressure transducer (Ruuska et al., 2016; Rombach et al., 2018), electrical switches (Luginbuhl et al., 1987), and electrical deformation sensors (Beauchemin et al., 1989). However, reports on the application of rumination loggers for the measurement of rumination time in dairy calves are rare. Hill et al. (2017) evaluated an ear-attached movement sensor, optimized for mature dairy cows, to record rumination in calves and reported very high precision of the device in 6-wk-old calves (R2 = 0.91) but not in 4-wk-old calves (R2 < 0.30). They speculated that correct ear placement of the device, presence of face flies around the calf, improper weight of the sensor for the calves, differences in jaw movement pattern between calves and adult cows and suckling behavior might have contributed to poor precision of the device. Burfeind et al. (2011) have also evaluated an automated system based on sounds (Hi-Tag electronic rumination monitoring system) to measure rumination in calves. Although the system provides a reasonable measure of rumination time in dairy cows, they reported that the algorithm had high variability for calves under 9 mo of age. Therefore, it seems that further studies are needed to introduce and validate electronic devices for the measurement of rumination behavior in calves. The RumiWatch system (RWS) has recently been validated for measuring eating, rumination, and drinking behavior in stall-fed (Ruuska et al., 2016) and grazing dairy cows (Rombach et al., 2018). The principle of the measurement is based on an oil-filled silicone tube containing a pressure sensor fastened in a halter over the cow’s nose and an accelerometer which is placed at the right side of the muzzle which detects the x-y-z position of the head. The y-axis of the accelerometer is oriented perpendicular to the floor, whereas the x- and z-axes describe the parallel plain of the ground. The weight of the RWS amounts to 2.5 kg and is very robust on mature dairy cows and although it records systematic errors for eating and drinking, results obtained for rumination time have only little random and systemic errors (Ruuska et al., 2016). Thus, we assumed that, the RWS can also be used to measure rumination time of calves after reducing the weight of the system and fitting the halter size to the head of the calf. Therefore, the objectives of the present study were to (1) miniaturize the RWS to the head of a calf and (2) evaluate the accuracy and precision of the miniaturized RWS halter for measuring rumination time of pre-weaned dairy calves in reference to visual observations on video camera. MATERIALS AND METHODS The experiment was carried out at Tiertechnikum of the Leibniz Institute for Farm Animal Biology (FBN), Dummerstorf, Germany. All procedures were approved by the ethics committee of the State Government in Mecklenburg-West Pomerania, Germany (LALLF M-V/TSD/7221.3–1.1–074/12). A vegetable oil–filled silicon tube with a built-in pressure sensor (outer diameter 5.7 mm, length 38 cm) was attached to the noseband of a calf halter. In contrast to the RWS developed for mature cows (Ruuska et al., 2016), the accelerometer, the battery, the data logger, and the SD card of the RWS were integrated into 1 box to reduce the weight of the RWS to 0.35 kg. The box was attached to the halter so that it was located behind the right ear (Fig. 1). This location should protect the box from external percussions and not hinder the calf during drinking. The data logger registered the pressure at a frequency of 10 signals per second (10 Hz). The battery is able to last for a maximum of 100 d and the halter has to be removed at least once a week for data transfer from the SD card to the computer using a USB cable. Data recorded by the RWS were converted to a mean of 1 min and transferred to an Excel file using RWS converter software (version 0.7.3.2; ITIN + HOCH GmbH Feeding Technology, Liestal, Switzerland), which was originally developed for mature cows. The software is based on a generic algorithm converting jaw movements and the head position into rumination, eating, and drinking events with 1 min resolution. Rumination was classified as “1” in the excel file. Figure 1. View largeDownload slide The RumiWatch System (RWS; ITIN + HOCH GmbH Feeding Technology, Liestal, Switzerland) composed of a halter with a noseband sensor comprising of an oil-filled silicon tube reaching from the left to the right side of the nose and the right cheek to end up in the pressure sensor (1). Behind the right ear, a box comprising the accelerometer, the battery, the data logger, and a SD card was fixed to the halter (2). Data recorded by the device are converted to an Excel file using the corresponding evaluation software (version 0.7.3.2). Figure 1. View largeDownload slide The RumiWatch System (RWS; ITIN + HOCH GmbH Feeding Technology, Liestal, Switzerland) composed of a halter with a noseband sensor comprising of an oil-filled silicon tube reaching from the left to the right side of the nose and the right cheek to end up in the pressure sensor (1). Behind the right ear, a box comprising the accelerometer, the battery, the data logger, and a SD card was fixed to the halter (2). Data recorded by the device are converted to an Excel file using the corresponding evaluation software (version 0.7.3.2). The length of the halter’s neck and muzzle strap was adjustable to different sizes of the calf head and the noseband was fastened in a way that allowed a space of a freely moving finger (1.5 × 3 cm) between the calf nose and the noseband (Fig. 1). Ten halter-trained female German Holstein dairy calves between 33 and 63 d of age (body weight ranging from 49 to 106 kg) were equipped with the RWS and kept individually in a respiration chamber (Derno et al., 2009) to which the animals were adapted before. Calves were bedded on wood shavings, milk-fed thrice a day at 0700, 1300, and 1700 h and had free access to hay, commercial starter (Bergophor Futtermittelfabrik GmbH & Co, Kulmbach, Germany) and water. A video camera (Panasonic wv – BP 100, Lens wv LA210C3 Focal lenght 2.1 mm, Matsushita Electric Industrial Co, Ltd, Osaka, Japan) was placed at the top corner of the chamber (horizontal and vertical distance of 1.70 m relative to feeding bin) and connected to a video recorder to monitor the whole space of the chamber using the software GeoVision GV–650 B (Taipei, Taiwan). The camera registered video clips with 5 min of length each for later evaluation. After installing the RWS to the calf’s head, data transmitted in the first 16 h as well as the video records were not considered for analysis to take into account the adaptation of calves to the halter and the chamber. On the day after, from 0600 to 1800 h, data from RWS and video records were used for analysis. To test if human direct observation can provide a reliable gold standard for the evaluation of the RWS, 2 independent observers evaluated video records for 2 periods of 90 min each in the morning and in the afternoon (0900–1030 and 1500–1630 h) for each of 10 calves. The observers recorded rumination time per 90 min yielding 20 pairs of values for correlation analysis. Afterward, one of the observers evaluated the whole video records of the calves for 12 continuous hours (0600–1800 h) to provide data for the evaluation of the RWS. All video clips were evaluated by the use of windows media player 12.0.7601 (2009 Microsoft Corporation) at a rate of 3.5×. Because video clips were 5 min in length, we divided them visually to 5 equal parts on VMP and registered the start and stop time of rumination events accepting an error of about 1 min at each event. On video observation, eating was assigned when the animal was standing at the feeding bin and showing jaw movement with the head moving toward the feed bin and hay basket regularly to pick the feed up. There were jaw movements immediately after eating sessions for very short periods (1 or 2 min) when the calf was no longer beside the feed bin. These activities were considered as chewing the last bite of the feed. The calves were confirmed as playing when they showed jaw movements while pointing the muzzle toward objects such as bars in the chamber for suckling or tongue playing. There were times that the animal exhibited chewing activity while laying down that was interrupted by moving the head toward the ground to pick the wood shaving up which were considered as eating wood shavings. Finally, rumination was confirmed when the calves showed regular jaw movements interrupted by regurgitation and swallow cycles with the head remaining in a constant position. In cases that the calf was laying with the backside toward the camera, rumination was indicated by sudden contraction of the abdominal area followed by the regular movement of the ears. Because video clips were displayed at a faster rate than reality, such ear movements seemed like a vibration of the ears and head of the animal which were easily detectable. There are several reports in the literature that the pattern of the ear movement during rumination is totally different from that during eating or resting and these differences can be used to monitor feeding behavior of cattle (Bikker et al., 2014; Wolfger et al., 2015). Rumination times recorded by either video observation or the RWS were processed as min of rumination per h during 12 h of monitoring period yielding a total of 120 pairs of values (12 pairs for each of 10 calves) for statistical analysis using SAS software (version 9.4, SAS institute Inc, Cary, NC). To determine interobserver reliability, Pearson coefficient of correlation between data recorded by 2 independent observers were calculated and paired t-test was performed to compare the means between observers. To test the agreement of the rumination time recorded by the RWS and the video observation, regression analysis was performed with RWS results being on y and video observation on x axis. Because the deviation of each individual observation from the regression line was attributable to either animal or unexplained residual, the random coefficient regression model was used (Ruuska et al., 2016): Yij=B0+B1Xij+si+biXij+eij where B0 is the overall intercept (fixed effect), B1 is the overall regression coefficient of Y on X (fixed effect), si is the random effect of animal i (i = 1, …, n), bi is the random effect of animal i on the regression coefficient of Y on X in animal i, eij is the unexplained residual error, and j is the number of observations for each animal. Adjusted Y values for the random effect of animal were calculated by adding Y values on the overall regression line and residual between an individual observation ij and Y value on the regression line of animal i. Adjusted Y values were then fitted against X values using the REG procedure of SAS (Ruuska et al., 2016). Accordingly, coefficient of determination (R2) was calculated as an indicator of random error and intercept and slope of the regression line represented systemic error. Assuming perfect agreement between RWS and visual observations on the video, the hypothesis was that the slope of the regression line would be 1 and the intercept 0 (Daigle and Siegford, 2014). Deviation of the intercept from 0 was interpreted from the P value of the intercept, whereas the 95% confidence interval was used to interpret the deviation of the slope from 1. To further illustrate the probable systemic error of RWS, the average duration of rumination (RWS and video observation) as h/12 h was tested by paired t-test using procedure TTEST of SAS. RESULTS AND DISCUSSION Calves consumed on average 0.14 kg hay (on DM base) on the experimental day and spent on average 2.46 h ruminating during the 12-h monitoring period on the same day (Table 1). It has been indicated that dietary forage intake is the main determinant of rumination behavior in young calves (Swanson et al., 1958; Borderas et al., 2008; Laarman and Oba, 2011; Castells et al., 2012, 2013; Terré et al., 2013). Terré et al. (2013) demonstrated that provision of forage to pre-weaned calves significantly increased rumination time while NDF content of the starter had no effect on rumination behavior. Borderas et al. (2008) reported that 3-wk-old calves injected with mild doses of lipopolysaccharides reduced their rumination time that was accompanied by a reduction in time spent eating hay but not concentrate. Rumination time has been reported to be positively correlated with dry feed intake in pre-weaned calves (Swanson and Harris, 1958; r = 0.75). However, the correlation between hay intake adjusted for BW and rumination time was not significant in our study (0.34; P = 0.33). One possible explanation is that rumination time was measured for 12 h, whereas hay intake was recorded after 24 h as calves might have ruminated for different times when they were not observed. Another factor can be the differences in age of the calves studied. It has been reported that calf age affects rumination behavior with older calves spending less time ruminating per pound of dry feed ingested (Swanson and Harris, 1958). Furthermore, differences in BW and the level of milk consumption might have also weakened the correlation between hay intake and rumination time (Table 1). Table 1. Individual data on BW, hay, milk and concentrate intake, rumination time, and correlations between observed and predicted rumination times Intake, kg DM/d Animal Age (d) BW (kg) Hay1 Hay (%BW) Milk1 Concentrate1 RT2, h/12 h RF03 RF10 RF5 1 36 52 0.16 0.32 0.69 0.06 3.70 0.83 0.86 0.97 2 40 49 0.12 0.25 0.59 0.17 3.27 0.87 0.97 0.99 3 39 53 0.08 0.15 0.70 0.05 1.78 0.73 0.93 0.94 4 62 68 0.13 0.19 0.85 0.76 1.57 0.60 0.94 0.92 5 62 106 0.32 0.30 2.04 0.34 1.85 0.73 0.98 0.98 6 58 75 0.17 0.22 0.99 0.70 3.22 0.85 0.93 0.98 7 63 70 0.05 0.07 0.93 0.54 1.98 0.63 0.83 0.94 8 53 80 0.34 0.42 1.01 0.45 2.42 0.69 0.98 0.98 9 33 63 0 0 1.66 0.02 2.02 0.79 0.94 0.99 10 36 52 0.06 0.12 0.70 0.20 2.83 0.89 0.83 0.93 Average 48.2 66.8 0.14 0.20 1.02 0.33 2.46 0.76 0.92 0.96 Intake, kg DM/d Animal Age (d) BW (kg) Hay1 Hay (%BW) Milk1 Concentrate1 RT2, h/12 h RF03 RF10 RF5 1 36 52 0.16 0.32 0.69 0.06 3.70 0.83 0.86 0.97 2 40 49 0.12 0.25 0.59 0.17 3.27 0.87 0.97 0.99 3 39 53 0.08 0.15 0.70 0.05 1.78 0.73 0.93 0.94 4 62 68 0.13 0.19 0.85 0.76 1.57 0.60 0.94 0.92 5 62 106 0.32 0.30 2.04 0.34 1.85 0.73 0.98 0.98 6 58 75 0.17 0.22 0.99 0.70 3.22 0.85 0.93 0.98 7 63 70 0.05 0.07 0.93 0.54 1.98 0.63 0.83 0.94 8 53 80 0.34 0.42 1.01 0.45 2.42 0.69 0.98 0.98 9 33 63 0 0 1.66 0.02 2.02 0.79 0.94 0.99 10 36 52 0.06 0.12 0.70 0.20 2.83 0.89 0.83 0.93 Average 48.2 66.8 0.14 0.20 1.02 0.33 2.46 0.76 0.92 0.96 1Total amount of milk replacer, hay and concentrate consumed on the video observation day (0001–2400 h). 2RT = rumination time observed on video (0600–1800 h). 3Concordance correlation coefficient (R) between observed and predicted rumination time (F0 = including raw data of the RWS (no filtering); F10 = excludes rumination events lasting less than 10 min as recoded by the RWS; F5 = excludes rumination events lasting less than five min as recoded by the RWS). View Large Table 1. Individual data on BW, hay, milk and concentrate intake, rumination time, and correlations between observed and predicted rumination times Intake, kg DM/d Animal Age (d) BW (kg) Hay1 Hay (%BW) Milk1 Concentrate1 RT2, h/12 h RF03 RF10 RF5 1 36 52 0.16 0.32 0.69 0.06 3.70 0.83 0.86 0.97 2 40 49 0.12 0.25 0.59 0.17 3.27 0.87 0.97 0.99 3 39 53 0.08 0.15 0.70 0.05 1.78 0.73 0.93 0.94 4 62 68 0.13 0.19 0.85 0.76 1.57 0.60 0.94 0.92 5 62 106 0.32 0.30 2.04 0.34 1.85 0.73 0.98 0.98 6 58 75 0.17 0.22 0.99 0.70 3.22 0.85 0.93 0.98 7 63 70 0.05 0.07 0.93 0.54 1.98 0.63 0.83 0.94 8 53 80 0.34 0.42 1.01 0.45 2.42 0.69 0.98 0.98 9 33 63 0 0 1.66 0.02 2.02 0.79 0.94 0.99 10 36 52 0.06 0.12 0.70 0.20 2.83 0.89 0.83 0.93 Average 48.2 66.8 0.14 0.20 1.02 0.33 2.46 0.76 0.92 0.96 Intake, kg DM/d Animal Age (d) BW (kg) Hay1 Hay (%BW) Milk1 Concentrate1 RT2, h/12 h RF03 RF10 RF5 1 36 52 0.16 0.32 0.69 0.06 3.70 0.83 0.86 0.97 2 40 49 0.12 0.25 0.59 0.17 3.27 0.87 0.97 0.99 3 39 53 0.08 0.15 0.70 0.05 1.78 0.73 0.93 0.94 4 62 68 0.13 0.19 0.85 0.76 1.57 0.60 0.94 0.92 5 62 106 0.32 0.30 2.04 0.34 1.85 0.73 0.98 0.98 6 58 75 0.17 0.22 0.99 0.70 3.22 0.85 0.93 0.98 7 63 70 0.05 0.07 0.93 0.54 1.98 0.63 0.83 0.94 8 53 80 0.34 0.42 1.01 0.45 2.42 0.69 0.98 0.98 9 33 63 0 0 1.66 0.02 2.02 0.79 0.94 0.99 10 36 52 0.06 0.12 0.70 0.20 2.83 0.89 0.83 0.93 Average 48.2 66.8 0.14 0.20 1.02 0.33 2.46 0.76 0.92 0.96 1Total amount of milk replacer, hay and concentrate consumed on the video observation day (0001–2400 h). 2RT = rumination time observed on video (0600–1800 h). 3Concordance correlation coefficient (R) between observed and predicted rumination time (F0 = including raw data of the RWS (no filtering); F10 = excludes rumination events lasting less than 10 min as recoded by the RWS; F5 = excludes rumination events lasting less than five min as recoded by the RWS). View Large Assessments of the videos by 2 independent observers were highly correlated (r = 0.99, n = 20; P < 0.01) and the difference between observers was not significant (29.65 ± 2.90 vs. 29.65 ± 2.93 min/90 min; paired t-test; P = 1.00). Minor differences between observers (0, 1, or 2 min in each rumination event) mainly was due to the visual estimates of the start and the end of rumination events by the observers with 1 observer capturing a half minute of rumination and the other recording the following minute as the start of rumination. The same error applies in case of termination of rumination events. Therefore, human direct observation provided a reliable gold standard for the evaluation of the RWS in the present study. Results of the regression analysis of RWS against visual observation along with paired t-test comparison are presented in Table 2. The regression model (Column MF0 in Table 2) revealed relatively high correlation between RWS results and video documentation (R2 = 0.82). However, the 95% confidence interval of the slope (0.79–0.94) did not include value 1 and the intercept of the regression line (7.07 ± 0.64) differed significantly from 0. These results indicate a systemic error in RWS overestimating rumination time in calves. This was further confirmed by significantly greater rumination time recorded by RWS than video observations (3.69 vs. 2.46 h/12 h, paired t-test, P < 0.01). Contrary to our results, Ruuska et al. (2016) have reported high accuracy of the RWS measuring rumination time in mature dairy cows. This discrepancy can be attributed to differences in feeding and nonnutritive oral behaviors between calves and mature dairy cows which affect the signaling of the RWS differently. The algorithm used in RWS software to differentiate eating and rumination is based on both, the jaw movements and changes in the head position of mature cows. As far as the head remains in a constant position, jaw movements are classified as rumination by the software which is also applicable to young calves. However, this criteria resulted in some false positive rumination signals (no rumination but detected as such by RWS) in calves in the present work. Dairy calves separated from their dam express nonnutritive oral activities such as sham chewing, licking, suckling or nibbling on objects, or tongue-playing apart from suckling and ingestive chewing (Veissier et al., 2013). These behaviors are considered abnormal, are mostly performed by animals living in inappropriate captive environments (Mason, 1991), and increase by social deprivation (Veissier et al., 1998). Some of these behaviors such as nibbling and biting of substrates probably derive from the normal ontogeny of grazing in pre-ruminants (Veissier et al., 1998) and from an intrinsic need for exploring (Sato and Wood-Guch, 1988). Nonnutritive oral behaviors, therefore, indicate that without the opportunity to graze, in the absence of an appropriate amount of roughage, or in a poorly stimulating living environment such as limited freedom for movement and experience, calves redirect their grazing, ruminating, and exploring behaviors toward inappropriate objects (Leruste et al., 2014). Although we did not measure actual time spent performing nonnutritive oral behaviors in our study, these activities were a main contributor to false positive classifications (43% of total) when evaluated in 5 out of 10 calves over a 12-h period. In agreement with our study, Hill et al. (2017) also speculated that object-suckling behavior of calves has resulted in poor precision of an ear-tag system based on adult cow algorithm for the measurement of rumination in 4-wk-old calves. In addition, dairy calves still lack a well-developed solid feed intake behavior as adult cows show (Miller-Cushon et al., 2015) and therefore their jaw and head movements during eating might have not resembled those of mature cows, resulting in false positive classification by the RWS in this study. Differences in jaw movement pattern between calves and mature cows have also been reported to cause inaccuracies in determining rumination time with an ear-attached movement sensor, originally developed for adult cows, in other studies (Burfeind et al., 2011; Hill et al., 2017). Additionally, we speculate that the positioning of the hay basket has also affected the signaling of the RWS. Calves in the present study received hay in a basket which was attached to the wall of the chamber and thus ate with a horizontally stretched neck and a lifted head. Adult cows, in contrast, eat with bowed neck and ruminate with a straighter head. This speculation was supported with the fact that 48% of total false positive signals corresponded to eating in video records of 5 calves. Therefore, it appears that calf-specific nutritive and nonnutritive oral activities have mostly contributed to false positive rumination signals and overestimation of rumination time in our trial. Table 2. Comparison of rumination times of 10 dairy calves recorded by the RumiWatch system (RWS; ITIN + HOCH GmbH Feeding Technology, Liestal, Switzerland) and visual observations on a video Approach MF01 MF102 MF53 A) Rumination time, h/12 h  RWS 3.69 ± 0.20 2.03 ± 0.20 2.43 ± 0.20  Video 2.46 ± 0.23 2.46 ± 0.23 2.46 ± 0.23  P value4 <0.01 0.001 0.60 B) Linear regression model  Slope 0.87 0.88 0.93  95% CI range for slope 0.79–0.94 0.82–0.95 0.88–0.98  Intercept (SEM) 7.70 ± 0.64 -0.75 ± 0.57 0.73 ± 0.40   R2 0.82 0.86 0.93   P value5 <0.01 0.19 0.07 Approach MF01 MF102 MF53 A) Rumination time, h/12 h  RWS 3.69 ± 0.20 2.03 ± 0.20 2.43 ± 0.20  Video 2.46 ± 0.23 2.46 ± 0.23 2.46 ± 0.23  P value4 <0.01 0.001 0.60 B) Linear regression model  Slope 0.87 0.88 0.93  95% CI range for slope 0.79–0.94 0.82–0.95 0.88–0.98  Intercept (SEM) 7.70 ± 0.64 -0.75 ± 0.57 0.73 ± 0.40   R2 0.82 0.86 0.93   P value5 <0.01 0.19 0.07 Data were processed as rumination time (min/h) in 12 one-h periods per each calf yielding a total of 120 pairs of values. 1Model includes raw data of the RWS (no filtering). 2Model excludes rumination events lasting less than 10 min as recoded by the RWS. 3Model excludes rumination events lasting less than five min as recoded by the RWS. 4Comparison of rumination times determined by RWS vs. video using paired t-test. 5Deviation of the intercept from zero. View Large Table 2. Comparison of rumination times of 10 dairy calves recorded by the RumiWatch system (RWS; ITIN + HOCH GmbH Feeding Technology, Liestal, Switzerland) and visual observations on a video Approach MF01 MF102 MF53 A) Rumination time, h/12 h  RWS 3.69 ± 0.20 2.03 ± 0.20 2.43 ± 0.20  Video 2.46 ± 0.23 2.46 ± 0.23 2.46 ± 0.23  P value4 <0.01 0.001 0.60 B) Linear regression model  Slope 0.87 0.88 0.93  95% CI range for slope 0.79–0.94 0.82–0.95 0.88–0.98  Intercept (SEM) 7.70 ± 0.64 -0.75 ± 0.57 0.73 ± 0.40   R2 0.82 0.86 0.93   P value5 <0.01 0.19 0.07 Approach MF01 MF102 MF53 A) Rumination time, h/12 h  RWS 3.69 ± 0.20 2.03 ± 0.20 2.43 ± 0.20  Video 2.46 ± 0.23 2.46 ± 0.23 2.46 ± 0.23  P value4 <0.01 0.001 0.60 B) Linear regression model  Slope 0.87 0.88 0.93  95% CI range for slope 0.79–0.94 0.82–0.95 0.88–0.98  Intercept (SEM) 7.70 ± 0.64 -0.75 ± 0.57 0.73 ± 0.40   R2 0.82 0.86 0.93   P value5 <0.01 0.19 0.07 Data were processed as rumination time (min/h) in 12 one-h periods per each calf yielding a total of 120 pairs of values. 1Model includes raw data of the RWS (no filtering). 2Model excludes rumination events lasting less than 10 min as recoded by the RWS. 3Model excludes rumination events lasting less than five min as recoded by the RWS. 4Comparison of rumination times determined by RWS vs. video using paired t-test. 5Deviation of the intercept from zero. View Large Despite above limitations, our modification of the RWS data suggests that the system is still applicable for the measurement of rumination in calves. On the video, we observed that nutritive and nonnutritive oral activities causing false positive rumination signals at the RWS lasted in the majority of the cases less than 5 and in few other cases less than 10 consecutive min. Therefore, we first filtered out rumination events lasting less than 10 min and performed the regression using the MF10 model (Table 2). Although the precision and accuracy of the model improved significantly with the new criteria (MF10: y = 0.88x; 95% CI of the slope= 0.82–0.95; R2 = 0.86), we detected a risk of underestimation of rumination time with paired t-test analysis (Table 2; P = 0.001). However, when the filter threshold was decreased from 10 to 5 consecutive min, the precision and accuracy of the model was further increased (MF5: y = 0.93x; 95% CI of the slope = 0.88–0.98; R2 = 0.93, Table 2). High accuracy of new criteria in determining rumination time is also confirmed by the results of paired t-test comparison (P = 0.60; Table 2). These results are comparable to those reported by Ruuska et al. (2016) evaluating the RWS for measuring of rumination time in adult dairy cows (slope of 0.88; 95% CI = 0.73–1.02 and R2 = 0.93). Therefore, it seems that rumination signals recorded by the RWS lasting less than 5 consecutive min are caused by other activities than real rumination and should be ignored when using the system for calves. This would provide precise and accurate measurement of rumination time in pre-weaned dairy calves for research and practical purposes. We speculated that there might be anatomical or physiological constraints for the use of the RWS. Therefore, after adjusting to the random effect of animal, the concordance correlation coefficient was calculated between observed and predicted rumination times separately for each calf (Table 1). Results of the analysis revealed relatively high variation in precision of the device among calves before filtering out the data (F0). This variation can be attributed to variation in the number of false positive signals. As described above, eating and nonnutritive oral activities mostly contributed to false positive signals. Although we did not measure actual time spent eating, calves also showed large variation in dry feed intake (Table 1) and nutritional factors have been reported to influence nonnutritive oral behaviors as well (Leruste et al., 2014). After filtering out the data (F10 and F5), however, the variation among individual animals decreased markedly (Table 1). This indicates that variation in nonnutritive oral activities and feed intake behavior are major issues when the device data are not processed according to the criteria defined in our study. In conclusion, the RWS was relatively free from random error in predicting rumination time (R2 = 0.82). Due to significant systemic errors, however, the software developed for the detection of rumination in mature cows should be used with minor modifications for the measurement of rumination time in calves. Based on our observation in this study, ignoring rumination signals lasting less than 5 consecutive min (in almost all cases being false positive signals) would markedly improve the precision and accuracy of the RWS for the application in dairy calves. Defining extra algorithm in the evaluation software based on our findings is encouraged to make the system more user-friendly. Conflict of interest statement. None declare. Footnotes 1 We thank T. Lenke, K. Pilz, D. Oswald, R. Gaeth, and A. Schulz (all of Leibniz Institute for Farm Animal Biology, FBN) for animal feeding and care. The results reported herein have been obtained in frame of the JPI FACCE program “RumenStability” and were financially supported by the German Federal Ministry of Food and Agriculture (BMEL) through the Federal Office for Agriculture and Food (BEL), grant number 2814ERA04A. LITERATURE CITED Baldwin , R. L. VI. , K. R. McLeod , J. L. Klotz , and R. N. Heitmann . 2004 . Rumen development, intestinal growth and hepatic metabolism in the pre- and post-weaning ruminant . J. Dairy Sci . 87 ( E Suppl .): E55 – E65 . doi: 10.3168/jds.S0022-0302(04)70061-2 Google Scholar Crossref Search ADS Beauchemin , K. A. , S. Zelin , D. Genner , and J. G. Buchanan-Smith . 1989 . An automatic system for quantification of eating and ruminating activities of dairy cattle housed in stalls . J. Dairy Sci . 72 : 2746 – 2759 . doi: 10.3168/jds.S0022-0302(89)79418-2 Google Scholar Crossref Search ADS Bikker , J. P. , H. van Laar , P. Rump , J. Doorenbos , K. van Meurs , G. M. Griffioen , and J. Dijkstra . 2014 . Technical note: evaluation of an ear-attached movement sensor to record cow feeding behavior and activity . J. Dairy Sci . 97 : 2974 – 2979 . doi: 10.3168/jds.2013-7560 Google Scholar Crossref Search ADS PubMed Borderas , T. F. , A. M. de Passillé , and J. Rushen . 2008 . Behavior of dairy calves after a low dose of bacterial endotoxin . J. Anim. Sci . 86 : 2920 – 2927 . doi: 10.2527/jas.2008-0926 Google Scholar Crossref Search ADS PubMed Bristow , D. J. , and D. S. Holmes . 2007 . Cortisol levels and anxiety-related behaviors in cattle . Physiol. Behav . 90 : 626 – 628 . doi: 10.1016/j.physbeh.2006.11.015 Google Scholar Crossref Search ADS PubMed Burfeind , O. , K. Schirmann , M. A. von Keyserlingk , D. M. Veira , D. M. Weary , and W. Heuwieser . 2011 . Evaluation of a system for monitoring rumination in heifers and calves . J. Dairy Sci . 94 : 426 – 430 . doi: 10.3168/jds.2010-3239 Google Scholar Crossref Search ADS PubMed Castells , L. , A. Bach , G. Araujo , C. Montoro , and M. Terré . 2012 . Effect of different forage sources on performance and feeding behavior of holstein calves . J. Dairy Sci . 95 : 286 – 293 . doi: 10.3168/jds.2011-4405 Google Scholar Crossref Search ADS PubMed Castells , L. , A. Bach , A. Aris , and M. Terré . 2013 . Effects of forage provision to young calves on rumen fermentation and development of the gastrointestinal tract . J. Dairy Sci . 96 : 5226 – 5236 . doi: 10.3168/jds.2012-6419 Google Scholar Crossref Search ADS PubMed Daigle , C. L. , and J. M. Siegford . 2014 . When continuous observations just won’t do: developing accurate and efficient sampling strategies for the laying hen . Behav. Processes 103 : 58 – 66 . doi: 10.1016/j.beproc.2013.11.002 Google Scholar Crossref Search ADS PubMed Derno , M. , H. G. Elsner , E. A. Paetow , H. Scholze , and M. Schweigel . 2009 . Technical note: a new facility for continuous respiration measurements in lactating cows . J. Dairy Sci . 92 : 2804 – 2808 . doi: 10.3168/jds.2008-1839 Google Scholar Crossref Search ADS PubMed Elischer , M. F. , M. E. Arceo , E. L. Karcher , and J. M. Siegford . 2013 . Validating the accuracy of activity and rumination monitor data from dairy cows housed in a pasture-based automatic milking system . J. Dairy Sci . 96 : 6412 – 6422 . doi: 10.3168/jds.2013-6790 Google Scholar Crossref Search ADS PubMed Gelsinger , S. L. , A. J. Heinrichs , and C. M. Jones . 2016 . A meta-analysis of the effects of preweaned calf nutrition and growth on first-lactation performance . J. Dairy Sci . 99 : 6206 – 6214 . doi: 10.3168/jds.2015-10744 Google Scholar Crossref Search ADS PubMed Hill , T. M. , F. X. Suarez-Mena , W. Hu , T. S. Dennis , R. L. Schlotterbeck , L. L. Timms , and L. E. Hulbert . 2017 . Technical note: evaluation of an ear-attached movement sensor to record rumination, eating, and activity behaviors in 1-month-old calves . The Prof. Anim. Sci . 33 : 743 – 747 . doi: 10.15232/pas.2017-01623 Google Scholar Crossref Search ADS Laarman , A. H. , and M. Oba . 2011 . Short communication: effect of calf starter on rumen ph of holstein dairy calves at weaning . J. Dairy Sci . 94 : 5661 – 5664 . doi: 10.3168/jds.2011-4273 Google Scholar Crossref Search ADS PubMed Leruste , H. , M. Brscic , G. Cozzi , B. Kemp , M. Wolthuis-Fillerup , B. J. Lensink , E. A. Bokkers , and C. G. van Reenen . 2014 . Prevalence and potential influencing factors of non-nutritive oral behaviors of veal calves on commercial farms . J. Dairy Sci . 97 : 7021 – 7030 . doi: 10.3168/jds.2014-7917 Google Scholar Crossref Search ADS PubMed Luginbuhl , J. M. , K. R. Pond , J. C. Russ , and J. C. Burns . 1987 . A simple electronic device and computer interface system for monitoring chewing behavior of stall-fed ruminant animals . J. Dairy Sci . 70 : 1307 – 1312 . doi: 10.3168/jds.S0022-0302(87)80146-7 Google Scholar Crossref Search ADS PubMed Mason , G. J . 1991 . Stereotypies: a critical review . Anim. Behav . 41 : 1015 – 1037 . doi: 10.1016/S0003-3472(05)80640-2 Google Scholar Crossref Search ADS Miller-Cushon , E. K. , and T. J. DeVries . 2015 . Invited review: development and expression of dairy calf feeding behaviour . Can. J. Anim. Sci . 95 : 341 – 350 . doi: 10.4141/cjas-2014-163 Google Scholar Crossref Search ADS Rombach , M. , A. Münger , J. Niederhauser , K. H. Südekum , and F. Schori . 2018 . Evaluation and validation of an automatic jaw movement recorder (rumiwatch) for ingestive and rumination behaviors of dairy cows during grazing and supplementation . J. Dairy Sci . 101 : 2463 – 2475 . doi: 10.3168/jds.2016-12305 Google Scholar Crossref Search ADS PubMed Ruuska , S. , S. Kajava , M. Mughal , N. Zehner , and J. Mononen . 2016 . Validation of a pressure sensor-based system for measuring eating, rumination and drinking behavior of dairy cattle . Appl. Anim. Behav. Sci . 174 : 19 – 23 . doi: 10.1016/j.applanim.2015.11.005 Google Scholar Crossref Search ADS Sato , S. , and D. G. M. Wood-Gush . 1988 . The development of behavior in beef suckler calves . Biol. Behav . 13 : 126 – 142 . Swanson , E. W. , and J. D. Harris Jr . 1958 . Development of rumination in the young calf . J. Dairy Sci . 41 : 1768 – 1776 . doi: 10.3168/jds.S0022-0302(58)91161-5 Google Scholar Crossref Search ADS Terré , M. , E. Pedrals , A. Dalmau , and A. Bach . 2013 . What do preweaned and weaned calves need in the diet: a high fiber content or a forage source ? J. Dairy Sci . 96 : 5217 – 5225 . doi: 10.3168/jds.2012-6304 Google Scholar Crossref Search ADS PubMed Veissier , I. , S. Caré , and D. Pomiès . 2013 . Suckling, weaning, and the development of oral behaviours in dairy calves . Appl. Anim. Behav. Sci . 147 : 11 – 18 . doi: 10.1016/j.applanim.2013.05.002 Google Scholar Crossref Search ADS Veissier , I. , A. R. Ramirez de la Fe , and P. Pradel . 1998 . Nonnutritive oral activities and stress responses of veal calves in relation to feeding and housing conditions . Appl. Anim. Behav. Sci . 57 : 35 – 49 . doi: 10.1016/S0168-1591(97)00108-1 Google Scholar Crossref Search ADS Wolfger , B. , E. Timsit , E. A. Pajor , N. Cook , H. W. Barkema , and K. Orsel . 2015 . Technical note: accuracy of an ear tag-attached accelerometer to monitor rumination and feeding behavior in feedlot cattle . J. Anim. Sci . 93 : 3164 – 3168 . doi: 10.2527/jas.2014-8802 Google Scholar Crossref Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the American Society of Animal Science. All rights reserved. For permissions, please e-mail: [email protected]. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Yu, Haipeng; Spangler, Matthew L; Lewis, Ronald M; Morota, Gota

doi: 10.1093/jas/sky316pmid: 30165381

Abstract Genetic connectedness assesses the extent to which estimated breeding values can be fairly compared across management units. Ranking of individuals across units based on best linear unbiased prediction (BLUP) is reliable when there is a sufficient level of connectedness due to a better disentangling of genetic signal from noise. Connectedness arises from genetic relationships among individuals. Although a recent study showed that genomic relatedness strengthens the estimates of connectedness across management units compared with that of pedigree, the relationship between connectedness measures and prediction accuracies only has been explored to a limited extent. In this study, we examined whether increased measures of connectedness led to higher prediction accuracies evaluated by a cross-validation (CV) based on computer simulations. We applied prediction error variance of the difference, coefficient of determination (CD), and BLUP-type prediction models to data simulated under various scenarios. We found that a greater extent of connectedness enhanced accuracy of whole-genome prediction. The impact of genomics was more marked when large numbers of markers were used to infer connectedness and evaluate prediction accuracy. Connectedness across units increased with the proportion of connecting individuals and this increase was associated with improved accuracy of prediction. The use of genomic information resulted in increased estimates of connectedness and improved prediction accuracies compared with those of pedigree-based models when there were enough markers to capture variation due to QTL signals. INTRODUCTION Genetic connectedness quantifies the extent of risk associated with the comparisons of estimated breeding values (EBV) across management units (Foulley et al., 1990). Best linear unbiased prediction (BLUP) of EBV can be fairly compared across units in the presence of a sufficient level of connectedness. On the other hand, an insufficient level of connectedness increases the risk of uncertainty in EBV comparisons when selecting individuals across units due to imperfect uncoupling of genetic signal from noise. A number of studies have shown that increasing pedigree-based connectedness through exchange of common reference sires can result in more accurate comparisons of genetic values of individuals from different management units (Foulley et al., 1983; Hanocq et al., 1996; Kuehn et al., 2008). The magnitude of estimates of connectedness is a function of genetic relatedness or relationships among individuals. Despite the critical importance of connectedness towards enabling genetic evaluations, the impact of genomic information on the degree of connectedness relative to pedigree only has been explored to a limited extent. Use of genomics can affect genetic evaluations in 2 related but different contexts. One is related to determining whether EBV can be safely compared across management units and the other is related to enhancing the reliability of EBV. In the former context, Yu et al. (2017) employed 3 measures of connectedness to examine the extent to which genomic information increases the estimates of connectedness. They found that the use of genomic relatedness improved genetic connectedness measures across management units compared with the use of pedigree relationships. However, it remains an open question as to whether increased connectedness observed by genomic relatedness also leads to increased prediction accuracy of genetic values across management units. Although improving the quality of breeding value comparisons and improving the accuracy of genomic prediction have been discussed in different contexts historically, it is worth investigating how these 2 items are related to each other. The objectives of this study were to examine how choice of relationship matrices and connectedness statistics affect the estimates of connectedness under various simulated scenarios and to assess the relationship between connectedness level and genome-enabled prediction accuracy. In addition, a guideline with respect to a sufficient level of connectedness is discussed. MATERIALS AND METHODS Data Simulation Ten replicates of genotypes and phenotypes were simulated using the QMSim software (Sargolzaei and Schenkel, 2009) with details summarized in Figure 1. One single historical population with 1,100 generations was simulated with the forward-in-time approach to create the initial linkage disequilibrium (LD) and mutation-drift equilibrium. The mating system was based on the random union of gametes sampled from sires and dams and the only evolutionary forces simulated were mutation and drift. The first 1,000 historical generations had a constant size of 1,000 per generation and then linearly decreased from 1,000 to 320 in the last hundred historical generations to account for population bottlenecks. The numbers of individuals from each sex were equal across the historical generations except the last historical generation which included a random sample of 20 males and 300 females (generation 0). Figure 1. Open in new tabDownload slide Genomic data simulation parameters. SNPs, QTLs, and h2 represent total single nucleotide polymorphisms, quantitative trait loci, and trait heritability, respectively. Simulations were carried out across 2 different h2 (0.8 and 0.2), 2 different numbers of QTLs (1,015 and 290), and 2 different SNP densities (50,000 and 5,000). Using the 20 males and 300 females as founder animals, the population size was expanded by simulating 7 generations (genreations 1 to 7) with the total population size approximately equal to 2,210. Each dam had 1 or 2 progenies within each generation with the probability of 0.95 and 0.05, respectively. As with the historical population, the mating was at random without selection and proportion of male progeny was 50%. The replacement rates of sires and dams were 0.6 and 0.2, respectively. Phenotypes with heritability levels of 0.2 and 0.8 were simulated with phenotypic variance of 1.0, where the overall heritability was accounted for by the variance of QTL additive genetic effects assuming no extra polygenic effect. Allelic effects of QTLs were sampled from a gamma distribution with a shape parameter of 0.4 and a corresponding scale parameter to ensure that the sum of QTLs variances was equal to the predefined QTL variances. The residual effects were randomly sampled from a Gaussian distribution with a mean of 0 and variance equal to heritability. The overall phenotypic effects were the sum of QTL effects and residual effects. Pedigree information was recorded in the recent population from generations 0 to 7. Genotypic data were simulated for individuals (n = 2,210) in generations 1 to 7 coupled with 5,000 or 50,000 biallelic single nucleotide polymorphisms (SNPs) markers evenly distributed across 29 pairs of autosomes with each chromosome length of 100 cM. The number of autosomes and total chromosome length followed those of the bovine genome. Additionally, 290 or 1,015 randomly distributed QTLs were simulated: the former is equivalent to 10 QTLs per chromosome and the latter corresponds to 35 QTLs per chromosome. Markers and QTLs were simulated with a starting allele frequency of 0.5 and a recurrent mutation rate of 2.5 × 10−5 was used to create mutation-drift equilibrium in historical generations. In generation 1,100, markers and QTLs with minor allele frequency greater than 0.05 were randomly drawn from the segregating loci. Only SNPs but not QTLs were used to infer measures of connectedness and to assess accuracy of prediction. Management Units Simulation The management units were simulated in 2 steps following Yu et al. (2017): 1) individuals were classified into clusters and 2) clusters were assigned to management units (Figure 2). First, 10 individuals were chosen to represent medoids and then 10 distinctive groups were formed by assigning the remaining individuals to the closest medoid using the k-medoid algorithm (Kaufman and Rousseeuw, 1990). The size of 10 distinctive groups ranged from 91 to 590, varying slightly between replications. A dissimilarity matrix was created from the A (numerator relationship) matrix by calculating the distance between highest similarity and each similarity coefficient such that the largest similarity coefficient becomes zero. Clustering based on the k-medoid algorithm coupled with the dissimilarity matrix resulted in higher relationship coefficients within a cluster than between clusters. Figure 2. Open in new tabDownload slide Management unit (MU) simulation scenarios. (A) Scenario 1 (least connected design). Individuals within clusters 1 to 5 were assigned to MU1 and clusters 6 to 10 were assigned to MU2. (B) Scenarios 2 to 6 (partially connected to connected). The degree of connectedness was gradually increased by exchanging 10% (Scenario 2), 20% (Scenario 3), 30% (Scenario 4), 40% (Scenario 5), and 50% (Scenario 6) of randomly sampled individuals between MU1 and MU2. Scenario 6 corresponds to the connected design. Two management units were simulated with individuals within clusters assigned to a management unit in 6 ways. In Scenario 1, a least connected design was simulated by assigning individuals within clusters 1 to 5 into management unit 1 (MU1) and clusters 6 to 10 into management unit 2 (MU2). In Scenarios 2 to 6, the degree of genetic link was gradually increased by exchanging 10%, 20%, 30%, 40%, and 50% of randomly sampled individuals between MU1 and MU2. Prediction Error Variance Prediction error variance (PEV) can be derived from a linear mixed model, y=Xb+Zg+ε, where y ⁠, b ⁠, g ⁠, and ε refer to a vector of phenotypes, fixed effects, random additive genetic effects, and residuals, respectively. The incidence matrices X and Z connect fixed effects and random additive genetic effects with phenotypes. The joint distribution of random effects is as follows: (ygε)∼N[(Xb00),(ZKσg2Z'+Iσε2ZKσg2Iσε2Kσg2Z′Kσg20Iσε20Iσε2)], where σg2 is the additive genetic variance, σε2 is the residual variance, and K represents a relationship matrix, which will be defined in a later section. Following the mixed model equation of Henderson (1984), [X′XX′ZZ′XZ′Z+K−1λ][b^g^]=[X′yZ′y](1) where λ is a ratio of variance components which equals to σε2σg2 ⁠. BLUP of g is given by g^=(Z′MZ+K−1λ)−1Z′My, where M=I−X(X'X)−X' is the absorption matrix for fixed effects. Then, the PEV of g is given by (Henderson, 1984) PEV(g)=Var(g^−g) =Var(g|g^) =(Z′MZ+K−1λ)−1σε2 =C22σε2, where C22 denotes the lower right quadrant of the inverse of coefficient matrix in equation 1. Genetic Connectedness Two statistics applied in Yu et al. (2017) were used to measure connectedness in this study. The first one is the prediction error variance of the differences (PEVD) of EBV between individuals from different management units (Kennedy and Trus, 1993). A pair-wise comparison between ith and jth individuals is given by the variance of g^i−g^j PEVD(g^i−g^j)=[PEV(g^i)+PEV(g^j)−2PEC(g^i,g^j)] =(Cii22−Cij22−Cji22+Cjj22)σε2 =(Cii22+Cjj22−2Cij22)σε2, where ii and jj refer to the diagonal elements of the C22 matrix corresponding to ith and jth individuals, respectively, and ij denotes the off-diagonal elements of C22 matrix. The summary connectedness of PEVD across all pairs of comparisons in a contrast notation is defined as follows (Laloë, 1993): PEVD(x)=x'C22xσε2, where the sum of elements in a contrast vector x is zero. For instance, a pair-wise comparison between i′th and j′th management units with ni′ and nj′ individuals, the contrast vector x will be set as 1/ni′ −1/nj′ and 0 corresponding to individual belonging to i′th, j′th, and remaining units. The boundary of PEVD is not restricted, with a lower value indicating stronger connectedness. To express connectedness independent of unit of measurement, PEVD was scaled by additive genetic variance (Kuehn et al., 2008; Yu et al., 2017). The generalized CD measures the precision of EBV (Laloë, 1993). Different from PEVD, CD penalizes connectedness measurements if the genetic variability is too small across populations, CDij=var(g)−var(g|g^)var(g) =1−var(g|g^)var(g) =1−λCii22+Cjj22−2Cij22Kii+Kjj−2Kij, where CDij denotes a pair-wise comparison between ith and jth individuals. A summary CD of contrast between any management unit is defined as follows (Laloë et al., 1996): CD(x)=1−var(x'g|g^)var(x'g) =1−λx'C22xx'Kx, where x is the vector of contrast defined earlier. This statistic ranges from 0 to 1 and measures the accuracy of the design. A larger value suggests a stronger estimate of connectedness among management units. Relationship Matrix Any kind of (semi)-positive definite relationship matrices can be used to define K (Morota and Gianola, 2014). We used 3 types of K in this study constructed from different sources. The numerator relationship matrix (⁠ K = A ⁠) measures the expected additive genetic relationship coefficient between individuals on the basis of pedigree information. The diagonal elements are 1+F ⁠, where F represents inbreeding coefficient and off-diagonal elements are equal to twice the kinship coefficients. The construction of the A matrix was based on tracing all individuals extending over 8 generations to account for historical information and animals from generations 1 to 7 were used for analysis. This matrix expresses relationships as identical by descent (IBD) as it measures the probability of alleles inherited from the same ancestor by tracing pedigree (Wright, 1922). In contrast, a genomic relationship matrix (⁠ K = G ⁠) measures the molecular similarity among individuals. A typical G matrix is obtained as a function of the gene content matrix (⁠ S ⁠) including elements of 0, 1, and 2 corresponding to the number of reference alleles. The distribution of jth marker follows the binomial distribution of s.j∼B(2pj,2pj(1−pj)) ⁠, where pj is the allele frequency of jth marker. The G matrix of VanRaden (2008) is obtained as follows: G=WW′m, where w.j is the standardized gene content equal to s.j−2pj2pj(1−pj) and m is the total number of markers. One item that needs to be addressed when the A and G matrices are compared is that they are not on the same scale. For instance, the A matrix represents relationships among individuals and inbreeding level as deviations from the unrelated base population; conversely the G matrix expresses those relationships relative to the allele frequencies in the current generation. The following K = G* matrix rescales G to the same base population as in A by adjusting the inbreeding coefficient level in G similar to that of A ⁠, G*=(1−F¯)G+2F¯J, where F¯ and J refer to the average inbreeding coefficient of whole population in the A matrix and the n × n square matrix filled with 1, respectively (Powell et al., 2010). Whole-Genome Prediction Model The relationship between connectedness and prediction accuracy was investigated with a standard BLUP model, y=1m+g+ε,(2) where y ⁠, m ⁠, g ⁠, and ε refer to a vector of observed phenotypes, intercept, random additive genetic effects, and residuals, respectively. The model was treated under a Bayesian framework, where m was set as a flat prior, with the prior distributions for genetic and residual effects, (gε)∼N[(00),(Kσg200Iσε2)], where K is 1 of 3 (semi)-positive definite relationship matrices described earlier and I refers to the identity matrix. The variance components σg2 and σε2 represent variance of additive genetic effects and residual variance, respectively. The scaled inverse χ2 distribution was assigned to σg2 and σε2 by setting the degrees of freedom (⁠ df ⁠) equal to 5 and choosing the scale parameter S by equating the mode of scaled inverse χ2 distribution Sdf+2 to the quantity of R2Vyn−1∑i=1n∑j=1mxij2 ⁠, where R2 is the expected proportion of phenotypic variance (⁠ Vy ⁠) explained by the regression and n−1∑i=1n∑j=1mxij2 refers to the average sum squares of the genotypes (Pérez and de los Campos, 2014). Here R2 was set to 0.5 according to Pérez and de los Campos (2014). The prediction accuracy was evaluated by 2-fold CV, where the 2 management units were treated as the training and testing sets instead of randomly partitioning all individuals into 2 sets. The variance components were inferred from the data and the predictive ability of the model was calculated as the Pearson correlation between predicted genetic values and true genetic values in the testing set. Throughout this study, the BGLR R package was used to fit equation 2. A Gibbs sampler was run for 10,000 iterations, where the first 2,000 samples were discarded as burn-in. A total of 8,000 samples coupled with a thinning rate of 5 were used to infer posterior means. Criterion for Connectedness Measures The challenge with discussing connectedness is that there is no clear standard or benchmark for true connectedness. Although zero connectedness may be an indicator of possible bias, this issue has been discussed since Foulley et al. (1990). In this respect, Kuehn et al. (2008) proposed threshold values for moderate and strong levels of connectedness based on the relationship between prediction error correlation and model-based mean squared error. In this study, we provide a guideline for connectedness measures in terms of whole-genome prediction by performing CV. Note that prediction accuracy may simply increase as PEVD continues to decrease no matter how individuals across management units become genetically alike. On the other hand, measures of CD start to decrease as in Yu et al. (2017) when across management units include individuals that are too genetically similar. CD is suited for deriving a criterion because there is no point in enhancing prediction accuracy by simply reducing relatedness variability. Therefore, we explored the approximate threshold of CD that yields a reasonable prediction accuracy while maintaining genetic diversity in a population (Laloë, 1993; Laloë et al., 1996). RESULTS Figure 3 displays relationships between 2 management units with 5,000 markers used to compute 3 relationship matrices (⁠ A ⁠, G ⁠, and G* ⁠) according to 6 simulated management unit scenarios. For each scenario, average relationships were the highest for A and the smallest for G ⁠, and G* produced relationships somewhere between A and G ⁠. Relationships increased when more individuals were exchanged between the 2 units. This increasing relationship pattern was observed regardless of relationship matrices used. A similar tendency was shown when the number of markers was equal to 50,000 (result not shown). Figure 3. Open in new tabDownload slide Average relationship coefficients across management units with 5,000 markers over 2 heritability levels and 2 different numbers of quantitative trait loci. S1 to S6 denotes management unit simulation scenarios 1, 2, 3, 4, 5, and 6, respectively. The magnitude of connectedness level steadily increased from S1 to S6. We compared pedigree-based A ⁠, genome-based G ⁠, and rescaled genome-based G* relationship kernel matrices. Prediction Error Variance of the Difference The relationships between measures of connectedness and prediction accuracies obtained from the Bayesian BLUP model are shown in Figures 4 and 5. The prediction accuracies in Figures 4 and 5 are identical as they are based on the same simulations. Figure 4 depicts connectedness measured as PEVD of contrast with smaller values inferring increased connectedness. Generally, increased connectedness measures and prediction accuracies were observed as more individuals from the same clusters were shared between management units, regardless of h2 levels, type of kernel matrices, the number of QTLs, and marker density. Similarly, standard errors of estimates over 10 replicates ranged from 0.008 to 0.068 for prediction accuracy, and from 0.001 to 0.002 for PEVD, regardless of h2 levels, type of kernel matrices, the number of QTLs, and marker density. In Figure 4A with 290 QTLs and 5,000 markers, the G and G* matrices delivered similar or stronger connectedness measures and higher prediction accuracies than those of the A matrix. The results from G* strongly resembled those of G in terms of measures of connectedness and prediction accuracies. When marker density increased to 50,000, with the same number of QTLs, slightly improved prediction accuracies and increased estimates of connectedness were observed (Figure 4B). Stronger connectedness and higher prediction accuracy were shown with G and G* than A The pattern in Figure 4C with 1,015 QTLs and 5,000 markers resembled that of Figure 4A; however, we observed marginally decreased genomic prediction accuracies. Figure 4D with 1,015 QTLs and 50,000 markers presented the clearest pattern: the G and G* matrices consistently produced stronger estimates of connectedness and higher prediction accuracies than those of the A regardless of simulation scenarios and h2 levels. Figure 4. Open in new tabDownload slide Relationship between connectedness and prediction accuracy. PEVD and PA denote prediction error variance of the differences and prediction accuracy, respectively. PA was defined as the correlation between phenotypes and estimated breeding values cor(g,g^) Connectedness of pedigree-based A genome-based G and rescaled genome-based G* within 6 management units simulation scenarios across 2 heritabilities were compared with their prediction accuracies in each graph. (A) 290 QTLs and 5,000 markers. (B) 290 QTLs and 50,000 markers. (C) 1,015 QTLs and 5,000 markers. (D) 1,015 QTLs and 50,000 markers. Figure 5. Open in new tabDownload slide Relationship between connectedness and prediction accuracy. CD and PA denote coefficient of determination and prediction accuracy, respectively. PA was defined as the correlation between phenotypes and estimated breeding values cor(g,g^) Connectedness of pedigree-based A genome-based G and rescaled genome-based G* within 6 management units simulation scenarios across 2 heritabilities were compared with their prediction accuracies in each graph. (A) 290 QTLs and 5,000 markers. (B) 290 QTLs and 50,000 markers. (C) 1,015 QTLs and 5,000 markers. (D) 1,015 QTLs and 50,000 markers. Coefficient of Determination The change of prediction accuracies with the increasing proportion of linked individuals quantified with CD of contrast is shown in Figure 5, where larger CD values suggest stronger connectedness. The standard errors of estimates for CD through 10 replicates varied from 0.004 to 0.057, regardless of h2 levels, type of kernel matrices, the number of QTLs, and marker density. In general, the prediction accuracy improved when more individuals from the same clusters were assigned across units. Within each scenario, the estimates of CD increased up to Scenario 3 and decreased at Scenario 4 because CD penalized connectedness measures for reduced genetic variability. This corresponded to 20% exchange rate. In Figure 5A with 290 QTLs and 5,000 markers, similar or stronger connectedness and higher prediction accuracies were observed by the G matrix than those using A for all scenarios. An analogous tendency was identified in Figure 5C with 1,015 QTLs and 5,000 markers, except that marginal reduction of genomic prediction accuracies was observed. With 290 QTLs and an increased number of markers (50,000), both genomic prediction accuracies and estimates of connectedness increased slightly (Figure 5B). Overall, G and G* presented stronger estimates of connectedness and higher prediction accuracies than those of A ⁠. Clearer differences were observed when increasing the number of QTLs to 1,015 (Figure 5D). The G matrix clearly yielded higher estimates of connectedness and higher prediction accuracies when compared with A ⁠. The performances of G* were very similar to those of G in CD across all cases. DISCUSSION The concept of connectedness dates back to estimability in experimental design in the sense of all-or-none connectedness (Weeks and Williams, 1964; Eccleston and Hedayat, 1974). A dataset can be seen as connected if merging cells in a cross-table are possible such that all filled cells are connected (Searle, 1986). It was later extended to a random effect model or BLUP genetic evaluation known as reference sire progeny testing schemes by Foulley et al. (1983, 1990) and Miraei Ashtiani and James (1991). The central idea is when sires from 1 management unit are compared against sires in another unit, at least 1 sire should be tested in both units. Such common sires are known as link sires or reference sires. These authors investigated the efficient strategy of reference sire used to minimize PEVD between EBV by identifying the optimal number of progeny. Since then connectedness based on pedigree information has taken center stage in both theoretical development and real data applications (e.g., Laloë (1993), Hanocq and Boichard (1999), and Kuehn et al. (2008)). In addition, non-PEV-based genetic connectedness metrics have been developed (e.g., Foulley et al. (1992)). Connectedness is often used as an indicator of the robustness of genetic evaluation comparisons, where a higher level of connectedness suggests more reliable comparison of EBV across units. Past studies found that BLUP evaluations correctly yielded the likely ranking of individuals distributed across units when connectedness was present. Although research in pedigree-based connectedness is still critical, as shown in Yu et al. (2017) and in the current study, availability of genomic information now offers an opportunity to revisit a number of critical questions related to connectedness, such as how prediction accuracy is influenced given the level of connectedness between management units. The extent of connectedness level boils down to the ability of K to capture relationships among individuals. Connectedness increases with stronger across unit genetic relationship and it decreases with stronger within unit relationship (Kennedy and Trus, 1993). Advantages of genomic over pedigree relationships are as follows: 1) genomic measures relatedness arising from more distant ancestors than those included in a pedigree and 2) genomic captures the variation in realized kinship arising from the stochastic effects of Mendelian sampling and recombination. We tested 3 types of K to capture the relationship among individuals in this study. The 2 matrices A and G mainly differ in 1) the distinction between IBD and IBS and 2) the relationships are relative to the baseline population vs. current population. The G* relationship matrix helps us to put A and G on a similar scale. Although those factors contributed to the improved quality of genetic evaluation design with the increased proportion of connecting individuals as shown in Yu et al. (2017), the relationship between connectedness level and CV-derived prediction accuracy has been yet-to-be answered. The present study aimed to bridge this gap by applying PEVD and CD of contrasts to simulated phenotypes, pedigrees, genomics, and management units. Note that the magnitude of the differences in results may be observed when applied to real data compared with the simulation results shown in this study. Relationship Between Connectedness and Prediction Accuracy We used contrasts of PEVD and CD to investigate the relationship between connectedness and prediction accuracy. We found prediction accuracy improved with increased capturing of connectedness between units. This suggests that increase in the accuracy of the EBV comparison is positively associated with an increase in accuracy of CV-based prediction. In general, genomic prediction accuracy improved as more markers were used to infer a genomic relationship matrix and as more QTLs contributed to the genetic variation given plenty of markers. These can be attributed to the fact that 1) the greater the number of markers, the better capturing of QTL relationships among individuals (Ober et al., 2012) and 2) genomic best linear unbiased prediction (GBLUP) performs better when the number of QTLs is large, because of its infinitesimal model assumption (Daetwyler et al., 2010). This result may change when an alternative whole-genome prediction model is used instead of GBLUP. For instance, a BayesB type of model performs well when the number of QTLs is small (Daetwyler et al., 2010). Measures of connectedness increased as more markers were used to characterize connectedness. When more markers were used, genomic information captures more variation in relationships which results in increased measures of connectedness. Across 6 management unit scenarios, the extent of connectedness measured by PEVD and prediction accuracy from BLUP were higher as the proportion of individuals exchanged between the 2 units increased. The measurement of PEVD decreases when the number of markers increase regardless of QTL numbers and h2 levels. This was not always the case in CD because this statistic penalizes connectedness estimates when the amount of genetic variability across units was small. The G and G* matrices clearly outperformed that of A in prediction and also produced increased measures of connectedness (Figures 4 and 5). Interestingly, although the average relationship of individuals across management units computed from the G* matrix was more similar with that of A than G (Figures 3), the results of connectedness estimates and prediction accuracies obtained from the G* matrix were more similar with those of G (Figures 4 and 5). This is most likely because of the similar variation in relationships across management units captured by G and G* ⁠, which play an important role in measures of connectedness and prediction accuracies. The effect of scaling G to be more similar to A was minimal for PEVD and CD as G* produced increased measure of connectedness compared with that of A ⁠. This is in agreement with Yu et al. (2017) where they found that genome-based connectedness consistently increased estimates of connectedness in most cases regardless of rescaling G to the level of A ⁠. In addition, we observed marginally decreased genomic prediction accuracies when the number of QTLs was increased while the number of SNPs remained constant (Figures 4A vs. 4C and 5A vs. 5C). This is because the number of parameters we need to accurately predict increased and a sufficient number of markers is required to establish a sufficient level of LD to capture QTL signals. With more QTL, more markers are needed for them to contribute to or enhance prediction accuracy. This observation can also be supported theoretically from interactive deterministic genomic prediction accuracy simulators (Morota, 2017). What is the Sufficient Level of Connectedness? The extent to which a design is genetically connected or not has been the subject of discussion in the literature (e.g., Petersen (1978) and Fernando et al. (1983)). These authors proposed statistical approaches to determine the presence or absence of connectedness. A related question is to find a desired or sufficient level of connectedness based on connectedness metrics as in Kuehn et al. (2008). Here CD statistic offers an important insight because it accounts for the reduction of connectedness due to reduced genetic variability between individuals under comparison. This pattern was also observed by using both pedigree and genome-based CD connectedness in Yu et al. (2017). From the perspective of designing a breeding program, increasing connectedness simply by making individuals genetically similar to each other should be avoided (Laloë, 1993). Thus, the use of CD allows us to identify an upper limit of sufficient CD value that gives a reasonable prediction accuracy while maintaining the variability of relatedness. The CD began to fall around 20% exchange rate and the threshold CD value was in the range of 0.7 to 0.9 across simulation scenarios. When the measures of CD exceeded this threshold, prediction accuracy continued to improve in a mild degree or stayed the same, whereas connectedness estimates started to decrease. Although this cutoff value slightly varies among different scenarios (Yu et al., 2017), the CD metric can be used to optimize selective genotyping and phenotyping along the lines of Rincent et al. (2012) and Isidro et al. (2015). In contrast, when connectedness was determined with PEVD, prediction accuracy and connectedness both continued to increase when shifting more individuals across management units, thereby increasing genetic similarity. Such is clearly not a desired property in designing a breeding program. CONCLUSIONS In general, connectedness measures and prediction accuracies increased as more individuals from the same clusters were shared across management units. We found prediction accuracy improved with increased capturing of connectedness across units suggesting that increase in the accuracy of the EBV comparison is positively associated with increase in accuracy of CV-based prediction. This was entirely true for PEVD and partly so for CD. The impact of genomics was more marked compared with pedigree when a sufficient number of markers was present to capture QTLs. Although there is a need to establish increased levels of connectedness, simply increasing connectedness results in rapid decrease of relatedness variability which may not be desired in a breeding program. Use of CD allows us to find a connectedness level that gives a reasonable prediction accuracy while maintaining genetic diversity in a population. Footnotes 1 This work was supported in part by the University of Nebraska startup funds to G.M. LITERATURE CITED Daetwyler , H. D. , R. Pong-Wong, B. Villanueva, and J. A. Woolliams. 2010 . The impact of genetic architecture on genome-wide evaluation methods . Genetics 185 : 1021 – 1031 . doi: 10.1534/genetics.110.116855 . Google Scholar Crossref Search ADS PubMed WorldCat Eccleston , J. , and A. Hedayat. 1974 . On the theory of connected designs: characterization and optimality . Ann. Stat . 2 : 1238 – 1255 . Google Scholar Crossref Search ADS WorldCat Fernando , R. , D. Gianola, and M. Grossman. 1983 . Identifying all connected subsets in a two-way classification without interaction . J. Dairy Sci . 66 : 1399 – 1402 . Google Scholar Crossref Search ADS WorldCat Foulley , J. L. , J. Bouix, B. Goffinet, and M. J. Elsen. 1990 . Connectedness in genetic evaluation . In: D. Gianola, and K. Hammond, editors. Advances in statistical methods for genetic improvement of livestock . Springer Verlag , Heidelberg, Germany . p. 277 – 308 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Foulley , J. L. , E. Hanocq, and D. Boichard. 1992 . A criterion for measuring the degree of connectedness in linear models of genetic evaluation . Genet. Sel. Evol . 24 : 315 – 330 . Google Scholar Crossref Search ADS WorldCat Foulley , J. , L. Schaeffer, H. Song, and J. Wilton. 1983 . Progeny group size in an organized progeny test program of ai beef bulls using reference sires . Can. J. Anim. Sci . 63 : 17 – 26 . Google Scholar Crossref Search ADS WorldCat Hanocq , E. , and D. Boichard. 1999 . Connectedness in the french holstein cattle population . Genet. Sel. Evol . 31 : 163 . Google Scholar Crossref Search ADS WorldCat Hanocq , E. , D. Boichard, and J. L. Foulley. 1996 . A simulation study of the effect of connectedness on genetic trend . Genet. Sel. Evol . 28 : 67 . Google Scholar Crossref Search ADS WorldCat Henderson , C. R . 1984 . Applications of linear models in animal breeding . 3rd ed. L. R. Schaeffer, editor. Univ. of Guelph , Guelph . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Isidro , J. , J. L. Jannink, D. Akdemir, J. Poland, N. Heslot, and M. E. Sorrells. 2015 . Training set optimization under population structure in genomic selection . Theor. Appl. Genet . 128 : 145 – 158 . doi: 10.1007/s00122-014-2418-4 Google Scholar Crossref Search ADS PubMed WorldCat Kaufman , L. and P. Rousseeuw. 1990 . Finding groups in data: an introduction to cluster analysis . John Wiley and Sons , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Kennedy , B. W. , and D. Trus. 1993 . Considerations on genetic connectedness between management units under an animal model . J. Anim. Sci . 71 : 2341 – 2352 . Google Scholar Crossref Search ADS PubMed WorldCat Kuehn , L. A. , D. R. Notter, G. J. Nieuwhof, and R. M. Lewis. 2008 . Changes in connectedness over time in alternative sheep sire referencing schemes . J. Anim. Sci . 86 : 536 – 544 . doi: 10.2527/jas.2007-0256 Google Scholar Crossref Search ADS PubMed WorldCat Laloë , D . 1993 . Precision and information in linear models of genetic evaluation . Genet. Sel. Evol . 25 : 557 . Google Scholar Crossref Search ADS WorldCat Laloë , D. , F. Phocas, and F. Ménissier. 1996 . Considerations on measures of precision and connectedness in mixed linear models of genetic evaluation . Genet. Sel. Evol . 28 : 359 . Google Scholar Crossref Search ADS WorldCat Miraei Ashtiani , S. , and J. James. 1991 . Efficient use of link rams in merino sire reference schemes . In: Proc. 9th Conf. Aust. Assoc. Anim. Breed. Genet . University of Melbourne , Melbourne, Australia . p. 24 – 27 . Morota , G . 2017 . Shinygpas: interactive genomic prediction accuracy simulator based on deterministic formulas . Genet. Sel. Evol . 49 : 91 . doi: 10.1186/s12711-017-0368-4 Google Scholar Crossref Search ADS PubMed WorldCat Morota , G. , and D. Gianola. 2014 . Kernel-based whole-genome prediction of complex traits: a review . Front. Genet . 5 : 363 . doi: 10.3389/fgene.2014.00363 Google Scholar PubMed OpenURL Placeholder Text WorldCat Ober , U. , J. F. Ayroles, E. A. Stone, S. Richards, D. Zhu, R. A. Gibbs, C. Stricker, D. Gianola, M. Schlather, T. F. Mackay,et al. 2012 . Using whole-genome sequence data to predict quantitative trait phenotypes in Drosophila melanogaster . Plos Genet . 8 : e1002685 . doi: 10.1371/journal.pgen.1002685 Google Scholar Crossref Search ADS PubMed WorldCat Pérez , P. , and G. de los Campos. 2014 . Genome-wide regression and prediction with the BGLR statistical package . Genetics 198 : 483 – 495 . doi: 10.1534/genetics.114.164442 Google Scholar Crossref Search ADS PubMed WorldCat Petersen , P . 1978 . A test for connectedness fitted for the two-way blup-sire evaluation . Acta Agric. Scand . 28 : 360 – 362 . Google Scholar Crossref Search ADS WorldCat Powell , J. E. , P. M. Visscher, and M. E. Goddard. 2010 . Reconciling the analysis of IBD and IBS in complex trait studies . Nat. Rev. Genet . 11 : 800 – 805 . doi: 10.1038/nrg2865 Google Scholar Crossref Search ADS PubMed WorldCat Rincent , R. , D. Laloë, S. Nicolas, T. Altmann, D. Brunel, P. Revilla, V. M. Rodríguez, J. Moreno-Gonzalez, A. Melchinger, E. Bauer,et al. 2012 . Maximizing the reliability of genomic selection by optimizing the calibration set of reference individuals: comparison of methods in two diverse groups of maize inbreds (Zea mays L.) . Genetics 192 : 715 – 728 . doi: 10.1534/genetics.112.141473 Google Scholar Crossref Search ADS PubMed WorldCat Sargolzaei , M. , and F. S. Schenkel. 2009 . Qmsim: a large-scale genome simulator for livestock . Bioinformatics 25 : 680 – 681 . doi: 10.1093/bioinformatics/btp045 Google Scholar Crossref Search ADS PubMed WorldCat Searle , S . 1986 . Linear models . John Wiley & Sons , New York . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC VanRaden , P. M . 2008 . Efficient methods to compute genomic predictions . J. Dairy Sci . 91 : 4414 – 4423 . doi: 10.3168/jds.2007-0980 Google Scholar Crossref Search ADS PubMed WorldCat Weeks , D. , and D. Williams. 1964 . A note on the determination of connectedness in an n-way cross classification . Technometrics 6 : 319 – 324 . Google Scholar Crossref Search ADS WorldCat Wright , S . 1922 . Coefficients of inbreeding and relationship . Am. Nat . 56 : 330 – 338 . Google Scholar Crossref Search ADS WorldCat Yu , H. , M. L. Spangler, R. M. Lewis, and G. Morota. 2017 . Genomic relatedness strengthens genetic connectedness across management units . G3 (Bethesda) . 7 : 3543 – 3556 . doi: 10.1534/g3.117.300151 Google Scholar Crossref Search ADS PubMed WorldCat © The Author(s) 2018. Published by Oxford University Press on behalf of the American Society of Animal Science. This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact [email protected] © The Author(s) 2018. Published by Oxford University Press on behalf of the American Society of Animal Science.

Tiphaine, Macé,;Eliel, González-García,;Julien, Pradel,;Sara, Parisot,;Fabien, Carrière,;Sebastien, Douls,;Didier, Foulquié,;Dominique, Hazard,

doi: 10.1093/jas/sky318pmid: 30085118

Abstract Animal robustness may be defined as a complex trait characterizing the ability of an individual to be adapted, productive and healthy under contrasted and fluctuating environmental situations. Such a trait is now considered an essential criterion in order to meet the more ambitious goals of farming sustainability. In ruminants, one of the key mechanisms via which robustness is expressed is the capacity to mobilize or restore body reserves (BR) to cope with the challenges of negative energy balances. The objectives of this work were as follows: 1) to estimate the genetic parameters related to BR dynamics in ewes over successive production cycles and 2) to investigate BR management relationships between different physiological stages. For this, historical individual BW and BCS data from 2,920 phenotyped ewes were used for genetic analysis. The changes in BW (∆BW) and BCS (∆BCS) over time were analyzed. Eight physiological stages were considered to investigate these changes over time: mating, early pregnancy, mid-pregnancy, lambing, early suckling, mid-suckling, weaning, and postweaning. The estimated heritability were low for both ∆BW (h2 = 0.13 to 0.18) and ∆BCS (h2 = 0.04 to 0.16). Moderate to high positive genetic correlations (from 0.48 to 0.91) were obtained between BR mobilization phases and between BR accretion phases. Similarly, moderate to high negative genetic correlations (from −0.36 to −0.75) were estimated between the BR mobilization and accretion periods, suggesting that mechanisms driving BR mobilization and accretion processes were genetically correlated. This is the first study in ruminants that demonstrate that the extent and temporal changes in profiles of BR mobilization and accretion are heritable and genetically linked, indicating that such traits could be considered in genetic programs aimed at improving robustness. Nevertheless, further research is needed for a more comprehensive understanding of BR dynamics, notably by including other physiological parameters (i.e., metabolites and hormones) and additional information on the productive and reproductive life of the ewe. INTRODUCTION In the future, breeding strategies will include more robust animals in order to improve the sustainability of farming systems (De La Torre et al., 2015; Friggens et al., 2017). It is expected that ruminants will depend more and more on grazing land in harsh environments, which will expose them to more uncertain feed resources due to the unpredictability of the production environment (O’Mara, 2012). Therefore, improving the individual biological capacities of adaptation to such environments is a priority. In that regard, an optimal management of body reserves (BR; i.e., lipid mobilization and accretion processes), an indicator of individual metabolic plasticity (Friggens et al., 2004; Blanc et al., 2006), is considered a relevant trait for these purposes (Phocas et al., 2016). The combined measurement of the BCS and BW has been used in previous studies to analyze the genetic variability of body condition (Edmonson et al., 1989; Arango et al., 2002; Safari et al., 2005; Gizaw et al., 2007; Shackell et al., 2011). The genetic determinism of BCS, however, has only been investigated using single time-point measurements. In the present study, we propose to use variations of BW and BCS over time (i.e., gain or loss of BW and BCS at key physiological stages) to assess BR in a dynamic management perspective while studying its genetic determinism over successive productive cycles with a multitrait approach. Such variations of BW and BCS reflect the biological ability of the ewe to use BR, more or less efficiently, when facing alternating situations (challenges) leading to positive or negative energy balances. To develop this approach, which to the best of our knowledge has not been addressed in previous work, we used historical BW and BCS data (2002 to 2015) recorded longitudinally in Romane meat ewes reared under extensive rangeland conditions. Our objectives were as follows: 1) to analyze, by assessing BW (∆BW) and BCS (∆BCS) changes over time, the genetic components of BR dynamics in ewes over successive production cycles and 2) to investigate potential relationships between the BR dynamics and different physiological stages. We hypothesized the existence of the genetic determinism of BR dynamics in sheep as well as related effects between stages and productive cycles. MATERIALS AND METHODS Animals and Experimental Farming System The experiments described here fully comply with applicable legislation on research involving animal subjects in accordance with the European Union Council directive (2010/63/UE). The researchers carrying out the experiments were certified by the relevant French governmental authority. All experimental procedures were approved and performed under the guidelines for the care and use of experimental animals stated in the ethics policy of the French Ministry of Agriculture. The experimental animals were Romane sheep, a composite line obtained from Romanov and Berrichon du Cher breeds (Ricordeau et al., 1992). The ewes were reared exclusively outdoor on about 280 ha of rangelands at the INRA Experimental Farm La Fage in a flock comprising 250 reproductive females (Causse du Larzac 43°54′54.52″N; 3°05′38.11″E; approximately 800 m.a.s.l, Roquefort-sur-Soulzon, Aveyron, south of France). The rangeland is on a limestone plateau and is composed of about 25% shrubs and 75% grass such as Bromus erectus, Stipa pennata, Carex humilis, and Festuca duriscola. The rangeland of this farm is divided into “native” (nonfertilized) and fertilized paddocks used during the spring (from lambing to weaning; Molénat et al., 2005). From 2002 to 2015, the average annual temperature was 9.8 °C and the average annual precipitation was 910 mm. During the summer, which is hot and dry (average temperature of 17.8 °C and average rainfall for the season of 142 mm, from June to August), the grass stops growing and regreens in the beginning of autumn. In the winter (average temperature of 2.14 °C and average rainfall of 183 mm, from December to February), the flock is supplemented with hay or silage. Further details regarding the climatic conditions and overall management practices established in this system were reported by Molénat et al. (2005), González-García et al. (2014), and González-García and Hazard (2016). Before 2010, females were mated at 7 or 19 mo of age depending on their BW. After 2010, all first matings were performed at 19 mo old. Mating occurred in the autumn to obtain a peak of lambing at the beginning of spring (usually mid-April; González-García et al., 2014). Romane ewes produced on average 2.2 live lambs per lambing in the period comprised for this study. Lambs were weaned at approximately 2.5 mo (i.e., 75 d). In these experimental conditions, the culling rate considered an annual replacement of 30% of the ewes. Historical Data and Variables The BW and BCS measurements used in this study were collected regularly during each female’s productive cycle according to a physiological stage schedule. A maximum of 2,920 records per trait were registered over a 14-yr period (2002 to 2015) with 1,146 females in cycle 1 (i.e., first production cycle, from first to second mating), 1,068 in cycle 2, and 707 in cycle 3 and more. To assess BCS, the original grid described by Russel et al. (1969) was used and subdivided into a 1/10 scale, i.e., a scale from 1 to 5 with 0.1 increments. The BCS measurements were performed by the 2 same operators throughout the entire 14-yr period of this study, with regular training sessions for adjustments and calibration. All measurements were recorded in the Geedoc database (https://germinal.toulouse.inra.fr/~mcbatut/GEEDOC/) as well as other additional pedigree information and animal production performance data. To analyze individual performances related to BR mobilization or accretion, the differences in BW or BCS between 2 physiological stages were calculated and used for interpretation as follows: BWi−j= BWj− BWi[I]; BCSi−j= BCSj− BCSi[I] where i and j are 2 successive physiological stages. The intervals were chosen considering key physiological stages which represented periods of BR mobilization and accretion throughout a typical production cycle. Three intervals were defined in the BR mobilization period: 1) difference between mid-pregnancy (Pb; 80 ± 14 days after mating [DAM], in January) for BW or early pregnancy for BCS (Pa; 40 ± 15 DAM, in December) and lambing (L; 160 ± 16 DAM, in April) (BW-Pb:L; BCS-Pa:L); 2) difference between lambing and early suckling (Sa; 190 ± 14 DAM, in April) (BW-L:Sa; BCS-L:Sa); and 3) difference between pregnancy and weaning (W; 250 ± 16 DAM, in June) (BW-Pb:W; BCS-Pa:W). Three intervals were also defined for the BR accretion period: 1) difference between mating (M; 15 d before mating, in November) and pregnancy (BW-M:Pb; BCS-M:Pa); 2) difference between weaning and postweaning (Wp; 310 ± 12 DAM, in August) (BW-W:Wp; BCS-W:Wp); and 3) difference between W and M (BW-W:M; BCS-W:M) (Figure 1). These intervals were selected in order to illustrate short- and long-term variations of BW and BCS during the BR mobilization or accretion periods throughout each productive cycle. Figure 1. View largeDownload slide Average of body weight (A) and body condition score (B) over a productive cycle. Values are raw means for all ewes and parities involved in the study and for each time point. Each color represents a specific period defined to estimate variation of BW or BCS. Figure 1. View largeDownload slide Average of body weight (A) and body condition score (B) over a productive cycle. Values are raw means for all ewes and parities involved in the study and for each time point. Each color represents a specific period defined to estimate variation of BW or BCS. Descriptive Statistics Deviations from normality were inspected using the UNIVARIATE procedure of SAS (version 9.4; SAS Institute Inc., Cary, NC). None of the variables were transformed given a major deviation from normality was not observed. Analyses of variance taking into account the repeated measures (MIXED procedure of SAS) were used to test the relevant effects and interactions in order to determine factors of variation for BW and BCS. These analyses resulted in the identification of the following significant fixed effects: the Age at first lambing and the Parity of the ewe, the Litter size, and the Year of measurement. The Age effect took into account ewes lambing for the first time at 1 or 2 yr old (classes 1 and 2, respectively). The Parity effect took into account first, second, and third or more lambing (classes 1, 2, and 3, respectively). The Litter effect was classified according to the number of lambs born that remained with the dam during the S stage (i.e., class 1, singletons from L until W; class 2, ewes lambing twins and suckling one; class 3, ewes lambing and suckling twins; and class 4, ewes lambing and suckling more than 2 lambs). The litter effect considered only the litter of the corresponding cycle and not of the previous cycle. Finally, 14 yr were analyzed. The first-order interactions between Age × Litter and Parity × Litter were tested. An effect was considered significant if P < 0.05. Genetic Analyses The variance components for ∆BW and ∆BCS were estimated by restricted maximum-likelihood methodology applied to an animal model using ASREML 1.0 software (Gilmour et al., 2001). Age, Parity, Litter, and Year were considered as fixed effects. Random effects included additive genetic and permanent environmental effects of the ewe. Analyses assumed a repeatability model with measurements across productive cycles considered to be the same traits with constant variances. The following animal mixed model was fitted as follows: y= Xβ+ Zaa+ Wcc+ e [II] where y is the vector of observations corresponding to the trait(s) in the analysis; β is the vector of fixed effects; and a and c are the vectors of random ewe additive genetic and permanent environmental effects with incidence matrices X, Za, and Wc, respectively, and e is the vector of residual effects. The following (co)variance structure of random effects was assumed as follows: Var [ace ] =[ Ga⊗A000Pc⊗I000R⊗I ] where Ga is a (co)variance matrix for direct additive genetic effects; A is the numerator relationship matrix; Pc is a (co)variance matrix for the ewe permanent environmental effects; R is a (co)variance matrix for residual effects; I are identity matrices of appropriate size; and ⊗ is the direct matrix product. Univariate analyses were performed to estimate variances for each trait. Bivariate analyses were performed to estimate genetic and phenotypic correlations between traits. Variance estimates in the 2-trait analyses were very similar to those from single-trait analyses. From the variance components, 3 parameters were defined as follows: 1) heritability or proportion of total phenotypic variance attributed to the additive genetic effect, h2 = σ2a/(σ2a + σ2c + σ2e); 2) proportion of total phenotypic variance attributed to the permanent environmental effect, c2 =σ2c /(σ2a + σ2c + σ2e); and 3) proportion of total phenotypic variance attributed to the residual effect, e2 = σ2e/(σ2a + σ2c + σ2e). In addition, repeatability (r) was defined as the sum of h2 and c2. RESULTS Biological and Year of Measurement Effects The differences in BW and BCS, ∆BW and ∆BCS, respectively, calculated for the 1,146 ewes over several productive cycles were significantly affected by the Parity and the Age at first lambing of the ewe, the Litter and the Year (Table 1). These fixed factors had significant (P < 0.01) effects on ∆BW, except for Age over the BW-W:Wp period. Globally, a significant (P < 0.01) increase in the BW loss was observed for BW-Pb:W, BW-Pb:L and a decrease in the BW loss was observed for BW-L:Sa between parities 1 and 2. An increase was also observed in the BW gain for BW-M:Pb and BW-W:M with parity, whereas a decrease was observed for the BW gain during BW-W:Wp (Table 2). Similarly, the BCS loss observed for BCS-Pa:W and BCS-Pa:L decreased with parity, whereas an increase was observed in the BCS gain for BCS-M:Pa between parities 1 and 2. A decrease in the BCS gain was observed with parity for BCS-W:M and for BCS-W:Wp between parities 2 and 3, and parity had no effect for the period BCS-L:Sa. The Litter effect was significant (P < 0.001) for all the ∆BCSs evaluated, except for BCS:M-Pa (Table 1). Overall, greater BW gain for the physiological periods BW-M:Pb, BW-W:Wp, and BW-W:M and greater BW loss for BW-Pb:W and BW-Pb:L were observed for larger litter sizes (Table 2). On the contrary, the Litter effect on ∆BW was much lower for BW-L:Sa. The same tendencies were observed for ∆BCS. Regarding the Age effect, the weight gain over the BW-W:M and BW-M:Pb periods decreased with the increase of age at first lambing, whereas the weight loss during BW-Pb:W and BW-Pb:L increased. For BW-L:Sa period, BW gain was observed for ewes that lambed at 1 yr old and weight loss was observed for ewes that lambed at 2 yr old. Regarding ∆BCS, the BCS gain observed over the BCS-M:Pa, BCS-W:Wp, and BCS-W:M periods increased with the increase of age at first lambing. The BCS loss for BCS-Pa:W, BCS-Pa:L, and BCS-L:Sa periods was slightly higher in ewes lambing at 2 yr old than in ewes lambing at 1 yr old. Table 1. Effects of the ewes’ parity and age at first lambing, the litter size, and the year of measurement on the average BW and BCS changes over successive physiological stages Variable n Mean1 (SD) Parity Litter size Age at first lambing Year BW-M:Pb 2663 8.23 (4.68) *** *** *** *** BW-Pb:W 2627 −5.74 (6.66) *** *** *** *** BW-Pb:L 2598 −3.33 (5.45) *** *** *** *** BW-L:Sa 2717 0.02 (4.08) ** *** *** *** BW-W:Wp 2078 0.92 (4.05) *** *** NS *** BW-W:M 1388 4.13 (4.32) *** *** *** *** BCS-M:Pa 2706 0.12 (0.21) ** NS *** *** BCS-Pa:W 2698 −0.43 (0.27) *** *** *** *** BCS-Pa:L 2920 −0.32 (0.26) *** *** *** *** BCS-L:Sa 2811 −0.11 (0.21) NS *** *** *** BCS-W:Wp 1889 0.11 (0.21) ** *** *** *** BCS-W:M 1320 0.25 (0.23) *** *** *** *** Variable n Mean1 (SD) Parity Litter size Age at first lambing Year BW-M:Pb 2663 8.23 (4.68) *** *** *** *** BW-Pb:W 2627 −5.74 (6.66) *** *** *** *** BW-Pb:L 2598 −3.33 (5.45) *** *** *** *** BW-L:Sa 2717 0.02 (4.08) ** *** *** *** BW-W:Wp 2078 0.92 (4.05) *** *** NS *** BW-W:M 1388 4.13 (4.32) *** *** *** *** BCS-M:Pa 2706 0.12 (0.21) ** NS *** *** BCS-Pa:W 2698 −0.43 (0.27) *** *** *** *** BCS-Pa:L 2920 −0.32 (0.26) *** *** *** *** BCS-L:Sa 2811 −0.11 (0.21) NS *** *** *** BCS-W:Wp 1889 0.11 (0.21) ** *** *** *** BCS-W:M 1320 0.25 (0.23) *** *** *** *** 1BW variation (kg) and BCS variation (points). n = number of records; SD = standard deviation; NS = nonsignificant; BW = body weight; BCS = body condition score; M = mating; Pa = early pregnancy; Pb = midpregnancy; L = lambing; Sa = early suckling; Sb = midsuckling; W = weaning; Wp = postweaning. View Large Table 1. Effects of the ewes’ parity and age at first lambing, the litter size, and the year of measurement on the average BW and BCS changes over successive physiological stages Variable n Mean1 (SD) Parity Litter size Age at first lambing Year BW-M:Pb 2663 8.23 (4.68) *** *** *** *** BW-Pb:W 2627 −5.74 (6.66) *** *** *** *** BW-Pb:L 2598 −3.33 (5.45) *** *** *** *** BW-L:Sa 2717 0.02 (4.08) ** *** *** *** BW-W:Wp 2078 0.92 (4.05) *** *** NS *** BW-W:M 1388 4.13 (4.32) *** *** *** *** BCS-M:Pa 2706 0.12 (0.21) ** NS *** *** BCS-Pa:W 2698 −0.43 (0.27) *** *** *** *** BCS-Pa:L 2920 −0.32 (0.26) *** *** *** *** BCS-L:Sa 2811 −0.11 (0.21) NS *** *** *** BCS-W:Wp 1889 0.11 (0.21) ** *** *** *** BCS-W:M 1320 0.25 (0.23) *** *** *** *** Variable n Mean1 (SD) Parity Litter size Age at first lambing Year BW-M:Pb 2663 8.23 (4.68) *** *** *** *** BW-Pb:W 2627 −5.74 (6.66) *** *** *** *** BW-Pb:L 2598 −3.33 (5.45) *** *** *** *** BW-L:Sa 2717 0.02 (4.08) ** *** *** *** BW-W:Wp 2078 0.92 (4.05) *** *** NS *** BW-W:M 1388 4.13 (4.32) *** *** *** *** BCS-M:Pa 2706 0.12 (0.21) ** NS *** *** BCS-Pa:W 2698 −0.43 (0.27) *** *** *** *** BCS-Pa:L 2920 −0.32 (0.26) *** *** *** *** BCS-L:Sa 2811 −0.11 (0.21) NS *** *** *** BCS-W:Wp 1889 0.11 (0.21) ** *** *** *** BCS-W:M 1320 0.25 (0.23) *** *** *** *** 1BW variation (kg) and BCS variation (points). n = number of records; SD = standard deviation; NS = nonsignificant; BW = body weight; BCS = body condition score; M = mating; Pa = early pregnancy; Pb = midpregnancy; L = lambing; Sa = early suckling; Sb = midsuckling; W = weaning; Wp = postweaning. View Large Table 2. Changes in body weight (∆BW, kg) and body condition score (∆BCS, points) as a function of the physiological stage, parity, litter size, and age at first lambing of the ewes Variable Parity Litter size Age at first lambing 1 2 3 and + 1 2 3 4 and + 1 2 n (%) 39 36 25 20 44 28 8 46 54 BW-M:Pb 7.32 7.83 8.81 7.00 8.38 7.80 8.85 8.33 7.62 BW-Pb:W −3.7 −6.20 −8.49 −2.83 −5.31 −7.19 −9.25 −4.64 −7.65 BW-Pb:L −0.85 −2.70 −4.33 0.01 −3.2 −2.12 −5.31 −1.36 −3.89 BW-L:Sa −0.39 0.14 0.00 0.49 0.50 −0.32 −0.02 0.46 −0.13 BW-W:Wp 2.05 1.17 1.03 1.10 1.44 1.91 1.95 1.46 1.75 BW-W:M 4.60 4.91 . 3.74 4.11 5.37 5.79 5.28 4.23 BCS-M:Pa 0.09 0.17 0.17 0.14 0.12 0.15 0.15 0.12 0.16 BCS-Pa:W −0.52 −0.40 −0.40 −0.34 −0.41 −0.49 −0.54 0 −0.42 −0.48 BCS-Pa:L −0.39 −0.27 −0.26 −0.18 −0.33 −0.29 −0.40 −0.29 −0.33 BCS-L:Sa −0.09 −0.10 −0.12 −0.08 −0.06 −0.16 −0.12 −0.09 −0.12 BCS-W:Wp 0.09 0.12 0.10 0.07 0.09 0.12 0.11 0.07 0.13 BCS-W:M 0.22 0.18 . 0.14 0.17 0.24 0.25 0.16 0.24 Variable Parity Litter size Age at first lambing 1 2 3 and + 1 2 3 4 and + 1 2 n (%) 39 36 25 20 44 28 8 46 54 BW-M:Pb 7.32 7.83 8.81 7.00 8.38 7.80 8.85 8.33 7.62 BW-Pb:W −3.7 −6.20 −8.49 −2.83 −5.31 −7.19 −9.25 −4.64 −7.65 BW-Pb:L −0.85 −2.70 −4.33 0.01 −3.2 −2.12 −5.31 −1.36 −3.89 BW-L:Sa −0.39 0.14 0.00 0.49 0.50 −0.32 −0.02 0.46 −0.13 BW-W:Wp 2.05 1.17 1.03 1.10 1.44 1.91 1.95 1.46 1.75 BW-W:M 4.60 4.91 . 3.74 4.11 5.37 5.79 5.28 4.23 BCS-M:Pa 0.09 0.17 0.17 0.14 0.12 0.15 0.15 0.12 0.16 BCS-Pa:W −0.52 −0.40 −0.40 −0.34 −0.41 −0.49 −0.54 0 −0.42 −0.48 BCS-Pa:L −0.39 −0.27 −0.26 −0.18 −0.33 −0.29 −0.40 −0.29 −0.33 BCS-L:Sa −0.09 −0.10 −0.12 −0.08 −0.06 −0.16 −0.12 −0.09 −0.12 BCS-W:Wp 0.09 0.12 0.10 0.07 0.09 0.12 0.11 0.07 0.13 BCS-W:M 0.22 0.18 . 0.14 0.17 0.24 0.25 0.16 0.24 Data (Least squares means) are expressed in kg of BW changes and points of BCS changes (based on the scale of Russel et al. (1969)). n = number of records in percentage; . = no data; BW = body weight; BCS = body condition score; M = mating; Pa = early pregnancy; Pb = midpregnancy; L = lambing; Sa = early suckling; Sb = midsuckling; W = weaning; Wp = postweaning. View Large Table 2. Changes in body weight (∆BW, kg) and body condition score (∆BCS, points) as a function of the physiological stage, parity, litter size, and age at first lambing of the ewes Variable Parity Litter size Age at first lambing 1 2 3 and + 1 2 3 4 and + 1 2 n (%) 39 36 25 20 44 28 8 46 54 BW-M:Pb 7.32 7.83 8.81 7.00 8.38 7.80 8.85 8.33 7.62 BW-Pb:W −3.7 −6.20 −8.49 −2.83 −5.31 −7.19 −9.25 −4.64 −7.65 BW-Pb:L −0.85 −2.70 −4.33 0.01 −3.2 −2.12 −5.31 −1.36 −3.89 BW-L:Sa −0.39 0.14 0.00 0.49 0.50 −0.32 −0.02 0.46 −0.13 BW-W:Wp 2.05 1.17 1.03 1.10 1.44 1.91 1.95 1.46 1.75 BW-W:M 4.60 4.91 . 3.74 4.11 5.37 5.79 5.28 4.23 BCS-M:Pa 0.09 0.17 0.17 0.14 0.12 0.15 0.15 0.12 0.16 BCS-Pa:W −0.52 −0.40 −0.40 −0.34 −0.41 −0.49 −0.54 0 −0.42 −0.48 BCS-Pa:L −0.39 −0.27 −0.26 −0.18 −0.33 −0.29 −0.40 −0.29 −0.33 BCS-L:Sa −0.09 −0.10 −0.12 −0.08 −0.06 −0.16 −0.12 −0.09 −0.12 BCS-W:Wp 0.09 0.12 0.10 0.07 0.09 0.12 0.11 0.07 0.13 BCS-W:M 0.22 0.18 . 0.14 0.17 0.24 0.25 0.16 0.24 Variable Parity Litter size Age at first lambing 1 2 3 and + 1 2 3 4 and + 1 2 n (%) 39 36 25 20 44 28 8 46 54 BW-M:Pb 7.32 7.83 8.81 7.00 8.38 7.80 8.85 8.33 7.62 BW-Pb:W −3.7 −6.20 −8.49 −2.83 −5.31 −7.19 −9.25 −4.64 −7.65 BW-Pb:L −0.85 −2.70 −4.33 0.01 −3.2 −2.12 −5.31 −1.36 −3.89 BW-L:Sa −0.39 0.14 0.00 0.49 0.50 −0.32 −0.02 0.46 −0.13 BW-W:Wp 2.05 1.17 1.03 1.10 1.44 1.91 1.95 1.46 1.75 BW-W:M 4.60 4.91 . 3.74 4.11 5.37 5.79 5.28 4.23 BCS-M:Pa 0.09 0.17 0.17 0.14 0.12 0.15 0.15 0.12 0.16 BCS-Pa:W −0.52 −0.40 −0.40 −0.34 −0.41 −0.49 −0.54 0 −0.42 −0.48 BCS-Pa:L −0.39 −0.27 −0.26 −0.18 −0.33 −0.29 −0.40 −0.29 −0.33 BCS-L:Sa −0.09 −0.10 −0.12 −0.08 −0.06 −0.16 −0.12 −0.09 −0.12 BCS-W:Wp 0.09 0.12 0.10 0.07 0.09 0.12 0.11 0.07 0.13 BCS-W:M 0.22 0.18 . 0.14 0.17 0.24 0.25 0.16 0.24 Data (Least squares means) are expressed in kg of BW changes and points of BCS changes (based on the scale of Russel et al. (1969)). n = number of records in percentage; . = no data; BW = body weight; BCS = body condition score; M = mating; Pa = early pregnancy; Pb = midpregnancy; L = lambing; Sa = early suckling; Sb = midsuckling; W = weaning; Wp = postweaning. View Large The Year effect was always highly significant (P < 0.001), whatever the physiological stage. The first-order interactions between the fixed effects considered here were not significant. Heritabilities and Genetic Correlations Estimates of variance components for ∆BW and ∆BCS are presented in Table 3. Direct heritabilities for ∆BW and ∆BCS ranged between 0.04 ± 0.02 and 0.18 ± 0.04. The highest heritabilities were found for BW-W:Wp (0.18 ± 0.04), BW-Pb:L (0.17 ± 0.03), and BW-W:M (0.17 ± 0.03), whereas the lowest heritability values were obtained for BCS-L:Sa (0.04 ± 0.02) and BCS-M:Pa (0.07 ± 0.02). Low heritabilities were found for BCS-Pa:W (0.16 ± 0.03) and BW-M:Pb (0.16 ± 0.03). Estimates for the fraction of permanent environmental variance were null except for BW-M:Pb, BW-Pb:W, and BCS-L:Sa. Repeatability estimates ranged from 0.08 ± 0.02 to 0.17 ± 0.02 for BCS-L:Sa and BCS-Pa:W, respectively. For ∆BW, repeatability estimates varied from 0.17 ± 0.03 to 0.21 ± 0.02 for BW-Pb:L and BW-M:Pb, respectively. Phenotypic variance was high for BW-Pb:W, BW-Pb:L, BW-W:M, BCS-Pa:W, and BCS-Pa:L, and lower for BW-M:Pb, BW-W:Wp, BW-W:M, BCS-M:Pa, BCS-W:Wp, and BCS-W:M periods (Table 3). Table 3. Estimates (± standard errors) of variance components, heritability, and repeatability for BW and BCS changes Variable h2 c2 e2 R σ2p BW-M:Pb 0.16 (0.03) 0.05 (0.03) 0.79 (0.02) 0.21 (0.02) 9.13 (0.27) BW-Pb:W 0.13 (0.03) 0.06 (0.03) 0.81 (0.03) 0.19 (0.03) 15.57 (0.46) BW-Pb:L 0.17 (0.03) 0.00 (0.03) 0.83 (0.02) 0.17 (0.03) 17.80 (0.54) BW-L:Sa 0.13 (0.03) 0.01 (0.02) 0.86 (0.02) 0.14 (0.02) 15.26 (0.44) BW-W:Wp 0.18 (0.04) 0.03 (0.03) 0.79 (0.03) 0.21 (0.03) 6.95 (0.23) BW-W:M 0.17 (0.03) 0.00 (0.00) 0.83 (0.04) 0.17 (0.04) 10.28 (0.40) BCS-M:Pa 0.07 (0.02) 0.00 (0.02) 0.92 (0.02) 0.08 (0.02) 0.041 (0.001) BCS-Pa:W 0.16 (0.03) 0.02 (0.03) 0.83 (0.02) 0.17 (0.02) 0.056 (0.002) BCS-Pa:L 0.10 (0.02) 0.02 (0.02) 0.88 (0.02) 0.12 (0.02) 0.055 (0.001) BCS-L:Sa 0.04 (0.02) 0.04 (0.02) 0.92 (0.02) 0.08 (0.02) 0.030 (0.001) BCS-W:Wp 0.10 (0.03) 0.01 (0.03) 0.89 (0.03) 0.11 (0.03) 0.032 (0.001) BCS-W:M 0.10 (0.04) 0.01 (0.05) 0.89 (0.05) 0.11 (0.05) 0.039 (0.002) Variable h2 c2 e2 R σ2p BW-M:Pb 0.16 (0.03) 0.05 (0.03) 0.79 (0.02) 0.21 (0.02) 9.13 (0.27) BW-Pb:W 0.13 (0.03) 0.06 (0.03) 0.81 (0.03) 0.19 (0.03) 15.57 (0.46) BW-Pb:L 0.17 (0.03) 0.00 (0.03) 0.83 (0.02) 0.17 (0.03) 17.80 (0.54) BW-L:Sa 0.13 (0.03) 0.01 (0.02) 0.86 (0.02) 0.14 (0.02) 15.26 (0.44) BW-W:Wp 0.18 (0.04) 0.03 (0.03) 0.79 (0.03) 0.21 (0.03) 6.95 (0.23) BW-W:M 0.17 (0.03) 0.00 (0.00) 0.83 (0.04) 0.17 (0.04) 10.28 (0.40) BCS-M:Pa 0.07 (0.02) 0.00 (0.02) 0.92 (0.02) 0.08 (0.02) 0.041 (0.001) BCS-Pa:W 0.16 (0.03) 0.02 (0.03) 0.83 (0.02) 0.17 (0.02) 0.056 (0.002) BCS-Pa:L 0.10 (0.02) 0.02 (0.02) 0.88 (0.02) 0.12 (0.02) 0.055 (0.001) BCS-L:Sa 0.04 (0.02) 0.04 (0.02) 0.92 (0.02) 0.08 (0.02) 0.030 (0.001) BCS-W:Wp 0.10 (0.03) 0.01 (0.03) 0.89 (0.03) 0.11 (0.03) 0.032 (0.001) BCS-W:M 0.10 (0.04) 0.01 (0.05) 0.89 (0.05) 0.11 (0.05) 0.039 (0.002) h2 = heritability; c2 = proportion of total phenotypic variance due to ewe permanent environmental effect; e2 = proportion of total phenotypic variance due to temporary environmental effects; r = repeatability; σ2p = total phenotypic variance; BW = body weight; BCS = body condition score; M = mating; Pa = early pregnancy; Pb = midpregnancy; L = lambing; Sa = early suckling; Sb = midsuckling; W = weaning; Wp = postweaning. View Large Table 3. Estimates (± standard errors) of variance components, heritability, and repeatability for BW and BCS changes Variable h2 c2 e2 R σ2p BW-M:Pb 0.16 (0.03) 0.05 (0.03) 0.79 (0.02) 0.21 (0.02) 9.13 (0.27) BW-Pb:W 0.13 (0.03) 0.06 (0.03) 0.81 (0.03) 0.19 (0.03) 15.57 (0.46) BW-Pb:L 0.17 (0.03) 0.00 (0.03) 0.83 (0.02) 0.17 (0.03) 17.80 (0.54) BW-L:Sa 0.13 (0.03) 0.01 (0.02) 0.86 (0.02) 0.14 (0.02) 15.26 (0.44) BW-W:Wp 0.18 (0.04) 0.03 (0.03) 0.79 (0.03) 0.21 (0.03) 6.95 (0.23) BW-W:M 0.17 (0.03) 0.00 (0.00) 0.83 (0.04) 0.17 (0.04) 10.28 (0.40) BCS-M:Pa 0.07 (0.02) 0.00 (0.02) 0.92 (0.02) 0.08 (0.02) 0.041 (0.001) BCS-Pa:W 0.16 (0.03) 0.02 (0.03) 0.83 (0.02) 0.17 (0.02) 0.056 (0.002) BCS-Pa:L 0.10 (0.02) 0.02 (0.02) 0.88 (0.02) 0.12 (0.02) 0.055 (0.001) BCS-L:Sa 0.04 (0.02) 0.04 (0.02) 0.92 (0.02) 0.08 (0.02) 0.030 (0.001) BCS-W:Wp 0.10 (0.03) 0.01 (0.03) 0.89 (0.03) 0.11 (0.03) 0.032 (0.001) BCS-W:M 0.10 (0.04) 0.01 (0.05) 0.89 (0.05) 0.11 (0.05) 0.039 (0.002) Variable h2 c2 e2 R σ2p BW-M:Pb 0.16 (0.03) 0.05 (0.03) 0.79 (0.02) 0.21 (0.02) 9.13 (0.27) BW-Pb:W 0.13 (0.03) 0.06 (0.03) 0.81 (0.03) 0.19 (0.03) 15.57 (0.46) BW-Pb:L 0.17 (0.03) 0.00 (0.03) 0.83 (0.02) 0.17 (0.03) 17.80 (0.54) BW-L:Sa 0.13 (0.03) 0.01 (0.02) 0.86 (0.02) 0.14 (0.02) 15.26 (0.44) BW-W:Wp 0.18 (0.04) 0.03 (0.03) 0.79 (0.03) 0.21 (0.03) 6.95 (0.23) BW-W:M 0.17 (0.03) 0.00 (0.00) 0.83 (0.04) 0.17 (0.04) 10.28 (0.40) BCS-M:Pa 0.07 (0.02) 0.00 (0.02) 0.92 (0.02) 0.08 (0.02) 0.041 (0.001) BCS-Pa:W 0.16 (0.03) 0.02 (0.03) 0.83 (0.02) 0.17 (0.02) 0.056 (0.002) BCS-Pa:L 0.10 (0.02) 0.02 (0.02) 0.88 (0.02) 0.12 (0.02) 0.055 (0.001) BCS-L:Sa 0.04 (0.02) 0.04 (0.02) 0.92 (0.02) 0.08 (0.02) 0.030 (0.001) BCS-W:Wp 0.10 (0.03) 0.01 (0.03) 0.89 (0.03) 0.11 (0.03) 0.032 (0.001) BCS-W:M 0.10 (0.04) 0.01 (0.05) 0.89 (0.05) 0.11 (0.05) 0.039 (0.002) h2 = heritability; c2 = proportion of total phenotypic variance due to ewe permanent environmental effect; e2 = proportion of total phenotypic variance due to temporary environmental effects; r = repeatability; σ2p = total phenotypic variance; BW = body weight; BCS = body condition score; M = mating; Pa = early pregnancy; Pb = midpregnancy; L = lambing; Sa = early suckling; Sb = midsuckling; W = weaning; Wp = postweaning. View Large Regarding ∆BW, high negative genetic correlations were found between BW-Pb:W and BW-W:Wp, and BW-Pb:L and BW-L:Sa (Tables 4 and 5). Moderate negative genetic correlations were found between BW-M:Pb and BW-W:Wp, BW-M:Pb and BW-W:M, BW-Pb:W and BW-W:M, and BW-L:Sa and BW-W:M, respectively. A moderate positive genetic correlation was found between BW-Pb:W and BW-BW-L:Sa, and a high positive genetic correlation was found between BW-W:Wp and BW-W:M. Phenotypic correlations followed the same tendencies as genetic correlations but with lower values. The highest positive phenotypic correlation was found between BW-W:Wp and BW-W:M, whereas the highest negative phenotypic correlation was observed between BW-M:Pb and BW-Pb:W. Table 4. Genetic and phenotypic correlations (±standard errors) for BW changes Variable BW-M:Pb BW-Pb:W BW-Pb:L BW-L:Sa BW-W:Wp BW-W:M BW-M:Pb – −0.17 (0.15) −0.16 (0.14) 0.18 (0.15) −0.45 (0.14) −0.37 (0.16) BW-Pb:W −0.47 (0.02) – 0.03 (0.16) 0.48 (0.16) −0.68 (0.11) −0.50 (0.14) BW-Pb:L −0.37 (0.02) 0.51 (0.02) – −0.63 (0.10) 0.03 (0.15) −0.03 (0.16) BW-L:Sa −0.03 (0.02) 0.16 (0.02) −0.37 (0.02) – −0.30 (0.16) −0.36 (0.17) BW-W:Wp −0.04 (0.02) −0.45 (0.02) −0.06 (0.03) −0.10 (0.02) – 0.91 (0.06) BW-W:M −0.03 (0.03) −0.43 (0.02) −0.05 (0.03) −0.10 (0.03) 0.68 (0.01) – Variable BW-M:Pb BW-Pb:W BW-Pb:L BW-L:Sa BW-W:Wp BW-W:M BW-M:Pb – −0.17 (0.15) −0.16 (0.14) 0.18 (0.15) −0.45 (0.14) −0.37 (0.16) BW-Pb:W −0.47 (0.02) – 0.03 (0.16) 0.48 (0.16) −0.68 (0.11) −0.50 (0.14) BW-Pb:L −0.37 (0.02) 0.51 (0.02) – −0.63 (0.10) 0.03 (0.15) −0.03 (0.16) BW-L:Sa −0.03 (0.02) 0.16 (0.02) −0.37 (0.02) – −0.30 (0.16) −0.36 (0.17) BW-W:Wp −0.04 (0.02) −0.45 (0.02) −0.06 (0.03) −0.10 (0.02) – 0.91 (0.06) BW-W:M −0.03 (0.03) −0.43 (0.02) −0.05 (0.03) −0.10 (0.03) 0.68 (0.01) – View Large Table 4. Genetic and phenotypic correlations (±standard errors) for BW changes Variable BW-M:Pb BW-Pb:W BW-Pb:L BW-L:Sa BW-W:Wp BW-W:M BW-M:Pb – −0.17 (0.15) −0.16 (0.14) 0.18 (0.15) −0.45 (0.14) −0.37 (0.16) BW-Pb:W −0.47 (0.02) – 0.03 (0.16) 0.48 (0.16) −0.68 (0.11) −0.50 (0.14) BW-Pb:L −0.37 (0.02) 0.51 (0.02) – −0.63 (0.10) 0.03 (0.15) −0.03 (0.16) BW-L:Sa −0.03 (0.02) 0.16 (0.02) −0.37 (0.02) – −0.30 (0.16) −0.36 (0.17) BW-W:Wp −0.04 (0.02) −0.45 (0.02) −0.06 (0.03) −0.10 (0.02) – 0.91 (0.06) BW-W:M −0.03 (0.03) −0.43 (0.02) −0.05 (0.03) −0.10 (0.03) 0.68 (0.01) – Variable BW-M:Pb BW-Pb:W BW-Pb:L BW-L:Sa BW-W:Wp BW-W:M BW-M:Pb – −0.17 (0.15) −0.16 (0.14) 0.18 (0.15) −0.45 (0.14) −0.37 (0.16) BW-Pb:W −0.47 (0.02) – 0.03 (0.16) 0.48 (0.16) −0.68 (0.11) −0.50 (0.14) BW-Pb:L −0.37 (0.02) 0.51 (0.02) – −0.63 (0.10) 0.03 (0.15) −0.03 (0.16) BW-L:Sa −0.03 (0.02) 0.16 (0.02) −0.37 (0.02) – −0.30 (0.16) −0.36 (0.17) BW-W:Wp −0.04 (0.02) −0.45 (0.02) −0.06 (0.03) −0.10 (0.02) – 0.91 (0.06) BW-W:M −0.03 (0.03) −0.43 (0.02) −0.05 (0.03) −0.10 (0.03) 0.68 (0.01) – View Large Table 5. Genetic and phenotypic correlations (±standard errors) for BCS changes Variable BCS-M:Pa BCS-Pa:W BCS-Pa:L BCS-L:Sa BCS-W:Wp BCS-W:M BCS-M:Pa – −0.52 (0.13) −0.44 (0.15) −0.27 (0.26) −0.02 (0.23) −0.32 (0.22) BCS-Pa:W −0.47 (0.02) – 0.70 (0.09) 0.52 (0.19) −0.62 (0.14) −0.75 (0.14) BCS-Pa:L −0.45 (0.02) 0.61 (0.01) – 0.01 (0.23) −0.39 (0.19) −0.56 (0.22) BCS-L:Sa −0.00 (0.02) 0.14 (0.02) −0.34 (0.02) – −0.47 (0.25) −0.58 (0.21) BCS-W:Wp −0.00 (0.02) −0.42 (0.02) −0.04 (0.02) −0.03 (0.02) – 0.57 (0.17) BCS-W:M 0.01 (0.03) −0.43 (0.02) −0.10 (0.03) −0.03 (0.03) 0.55 (0.02) – Variable BCS-M:Pa BCS-Pa:W BCS-Pa:L BCS-L:Sa BCS-W:Wp BCS-W:M BCS-M:Pa – −0.52 (0.13) −0.44 (0.15) −0.27 (0.26) −0.02 (0.23) −0.32 (0.22) BCS-Pa:W −0.47 (0.02) – 0.70 (0.09) 0.52 (0.19) −0.62 (0.14) −0.75 (0.14) BCS-Pa:L −0.45 (0.02) 0.61 (0.01) – 0.01 (0.23) −0.39 (0.19) −0.56 (0.22) BCS-L:Sa −0.00 (0.02) 0.14 (0.02) −0.34 (0.02) – −0.47 (0.25) −0.58 (0.21) BCS-W:Wp −0.00 (0.02) −0.42 (0.02) −0.04 (0.02) −0.03 (0.02) – 0.57 (0.17) BCS-W:M 0.01 (0.03) −0.43 (0.02) −0.10 (0.03) −0.03 (0.03) 0.55 (0.02) – Phenotypic correlations below the diagonal; genetic correlations above the diagonal; BW: Body Weight; BCS: Body Condition Score; M: Mating; Pa: Early pregnancy; Pb: Mid-pregnancy; L: Lambing; Sa: Early suckling; Sb: Mid-suckling; W: Weaning; Wp: Post-weaning. View Large Table 5. Genetic and phenotypic correlations (±standard errors) for BCS changes Variable BCS-M:Pa BCS-Pa:W BCS-Pa:L BCS-L:Sa BCS-W:Wp BCS-W:M BCS-M:Pa – −0.52 (0.13) −0.44 (0.15) −0.27 (0.26) −0.02 (0.23) −0.32 (0.22) BCS-Pa:W −0.47 (0.02) – 0.70 (0.09) 0.52 (0.19) −0.62 (0.14) −0.75 (0.14) BCS-Pa:L −0.45 (0.02) 0.61 (0.01) – 0.01 (0.23) −0.39 (0.19) −0.56 (0.22) BCS-L:Sa −0.00 (0.02) 0.14 (0.02) −0.34 (0.02) – −0.47 (0.25) −0.58 (0.21) BCS-W:Wp −0.00 (0.02) −0.42 (0.02) −0.04 (0.02) −0.03 (0.02) – 0.57 (0.17) BCS-W:M 0.01 (0.03) −0.43 (0.02) −0.10 (0.03) −0.03 (0.03) 0.55 (0.02) – Variable BCS-M:Pa BCS-Pa:W BCS-Pa:L BCS-L:Sa BCS-W:Wp BCS-W:M BCS-M:Pa – −0.52 (0.13) −0.44 (0.15) −0.27 (0.26) −0.02 (0.23) −0.32 (0.22) BCS-Pa:W −0.47 (0.02) – 0.70 (0.09) 0.52 (0.19) −0.62 (0.14) −0.75 (0.14) BCS-Pa:L −0.45 (0.02) 0.61 (0.01) – 0.01 (0.23) −0.39 (0.19) −0.56 (0.22) BCS-L:Sa −0.00 (0.02) 0.14 (0.02) −0.34 (0.02) – −0.47 (0.25) −0.58 (0.21) BCS-W:Wp −0.00 (0.02) −0.42 (0.02) −0.04 (0.02) −0.03 (0.02) – 0.57 (0.17) BCS-W:M 0.01 (0.03) −0.43 (0.02) −0.10 (0.03) −0.03 (0.03) 0.55 (0.02) – Phenotypic correlations below the diagonal; genetic correlations above the diagonal; BW: Body Weight; BCS: Body Condition Score; M: Mating; Pa: Early pregnancy; Pb: Mid-pregnancy; L: Lambing; Sa: Early suckling; Sb: Mid-suckling; W: Weaning; Wp: Post-weaning. View Large Regarding ∆BCS, high negative genetic correlations were found between BCS-Pa:W and BCS-W:Wp, BCS-Pa:W and BCS-W:M, BCS-Pa:L and BCS-W:M, and BCS-L:Sa and BCS-W:M, respectively (Tables 4 and 5). Moderate negative genetic correlations were found between BCS-M:Pa and BCS-Pa:W, BCS-M:Pa and BCS-Pa:L, BCS-Pa:L and BCS-W:Wp, and BCS-L:Sa and BCS-W:Wp, respectively. A moderate positive genetic correlation was found between BCS-Pa:W and BCS-L:Sa. High positive genetic correlations were found between BCS-Pa:W and BCS-Pa:L, and BCS-W:Wp and BCS-W:M. Phenotypic correlations followed the same tendencies as genetic correlations but with lower values. The highest positive phenotypic correlation was found between BCS-Pa:W and BCS-Pa:L, whereas the highest negative phenotypic correlation was found between BCS-Pa:W and BCS-W:M. When considering the correlations between ∆BW and ∆BCS, a high negative genetic correlation was observed between BCS-Pa:W and BW-W:Wp (Tables 6 and 7). Moderate negative genetic correlations were observed between BCS-Pa:W and BW-W:M, BCS-Pa:L and BW-W:Wp, BCS-L:Sa and BW-W:Wp, BCS-W:Wp and BW-L:Sa, BCS-W:M and BW-Pb:W, and BCS-W:M and BW-Pb:L, respectively. Moderate positive genetic correlations were observed between BCS-M:Pa and BW-M:Pb, BCS-Pa:W and BW-Pb:W, BCS-Pa:W and BW-L:Sa, BCS-Pa:L and BW-Pb:L, and BCS-W:Wp and BW-W:Wp, respectively. High positive genetic correlations were observed between BCS-W:Wp and BW-W:M, BCS-W:M and BW-W:Wp, and BCS-W:M and BW-W:M. Phenotypic correlations followed the same tendencies as genetic correlations but with lower values. The highest positive phenotypic correlation was observed between BCS-Pa:W and BW-Pb:W, whereas the highest negative phenotypic correlation was found between BCS-Pa:W and BW-W:M. Table 6. Genetic correlations (± standard errors) between BW and BCS changes Variables BW-M:Pb BW-Pb:W BW-Pb:L BW-L:Sa BW-W:Wp BW-W:M BCS-M:Pa 0.46 (0.16) −0.25 (0.19) 0.02 (0.18) −0.01 (0.19) 0.30 (0.18) 0.03 (0.20) BCS-Pa:W 0.01 (0.14) 0.54 (0.12) 0.19 (0.14) 0.29 (0.15) −0.73 (0.11) −0.46 (0.15) BCS-Pa:L 0.19 (0.16) 0.20 (0.16) 0.31 (0.14) 0.12 (0.16) −0.48 (0.14) −0.32 (0.17) BCS-L:Sa −0.14 (0.23) 0.09 (0.24) −0.09 (0.23) 0.18 (0.23) −0.40 (0.21) −0.11 (0.25) BCS-W:Wp −0.18 (0.18) −0.44 (0.19) −0.26 (0.17) −0.40 (0.19) 0.37 (0.18) 0.58 (0.17) BCS-W:M −0.38 (0.24) −0.52 (0.19) −0.37 (0.18) −0.32 (0.22) 0.60 (0.18) 0.68 (0.16) Variables BW-M:Pb BW-Pb:W BW-Pb:L BW-L:Sa BW-W:Wp BW-W:M BCS-M:Pa 0.46 (0.16) −0.25 (0.19) 0.02 (0.18) −0.01 (0.19) 0.30 (0.18) 0.03 (0.20) BCS-Pa:W 0.01 (0.14) 0.54 (0.12) 0.19 (0.14) 0.29 (0.15) −0.73 (0.11) −0.46 (0.15) BCS-Pa:L 0.19 (0.16) 0.20 (0.16) 0.31 (0.14) 0.12 (0.16) −0.48 (0.14) −0.32 (0.17) BCS-L:Sa −0.14 (0.23) 0.09 (0.24) −0.09 (0.23) 0.18 (0.23) −0.40 (0.21) −0.11 (0.25) BCS-W:Wp −0.18 (0.18) −0.44 (0.19) −0.26 (0.17) −0.40 (0.19) 0.37 (0.18) 0.58 (0.17) BCS-W:M −0.38 (0.24) −0.52 (0.19) −0.37 (0.18) −0.32 (0.22) 0.60 (0.18) 0.68 (0.16) View Large Table 6. Genetic correlations (± standard errors) between BW and BCS changes Variables BW-M:Pb BW-Pb:W BW-Pb:L BW-L:Sa BW-W:Wp BW-W:M BCS-M:Pa 0.46 (0.16) −0.25 (0.19) 0.02 (0.18) −0.01 (0.19) 0.30 (0.18) 0.03 (0.20) BCS-Pa:W 0.01 (0.14) 0.54 (0.12) 0.19 (0.14) 0.29 (0.15) −0.73 (0.11) −0.46 (0.15) BCS-Pa:L 0.19 (0.16) 0.20 (0.16) 0.31 (0.14) 0.12 (0.16) −0.48 (0.14) −0.32 (0.17) BCS-L:Sa −0.14 (0.23) 0.09 (0.24) −0.09 (0.23) 0.18 (0.23) −0.40 (0.21) −0.11 (0.25) BCS-W:Wp −0.18 (0.18) −0.44 (0.19) −0.26 (0.17) −0.40 (0.19) 0.37 (0.18) 0.58 (0.17) BCS-W:M −0.38 (0.24) −0.52 (0.19) −0.37 (0.18) −0.32 (0.22) 0.60 (0.18) 0.68 (0.16) Variables BW-M:Pb BW-Pb:W BW-Pb:L BW-L:Sa BW-W:Wp BW-W:M BCS-M:Pa 0.46 (0.16) −0.25 (0.19) 0.02 (0.18) −0.01 (0.19) 0.30 (0.18) 0.03 (0.20) BCS-Pa:W 0.01 (0.14) 0.54 (0.12) 0.19 (0.14) 0.29 (0.15) −0.73 (0.11) −0.46 (0.15) BCS-Pa:L 0.19 (0.16) 0.20 (0.16) 0.31 (0.14) 0.12 (0.16) −0.48 (0.14) −0.32 (0.17) BCS-L:Sa −0.14 (0.23) 0.09 (0.24) −0.09 (0.23) 0.18 (0.23) −0.40 (0.21) −0.11 (0.25) BCS-W:Wp −0.18 (0.18) −0.44 (0.19) −0.26 (0.17) −0.40 (0.19) 0.37 (0.18) 0.58 (0.17) BCS-W:M −0.38 (0.24) −0.52 (0.19) −0.37 (0.18) −0.32 (0.22) 0.60 (0.18) 0.68 (0.16) View Large Table 7. Phenotypic correlations (±standard errors) between BW and BCS changes Variables BW-M:Pb BW-Pb:W BW-Pb:L BW-L:Sa BW-W:Wp BW-W:M BCS-M:Pa 0.12 (0.02) −0.04 (0.02) 0.02 (0.02) 0.00 (0.02) 0.08 (0.02) 0.05 (0.03) BCS-Pa:W −0.00 (0.02) 0.36 (0.02) 0.17 (0.02) 0.08 (0.02) -0.20 (0.02) -0.21 (0.03) BCS-Pa:L 0.02 (0.02) 0.22 (0.02) 0.31 (0.02) 0.03 (0.02) -0.12 (0.02) -0.12 (0.03) BCS-L:Sa −0.05 (0.02) 0.10 (0.02) −0.01 (0.02) 0.14 (0.02) -0.02 (0.02) -0.03 (0.03) BCS-W:Wp −0.03 (0.02) −0.07 (0.03) −0.06 (0.03) −0.05 (0.02) 0.24 (0.02) 0.14 (0.03) BCS-W:M −0.04 (0.03) −0.14 (0.03) −0.08 (0.03) −0.04 (0.03) 0.22 (0.03) 0.26 (0.03) Variables BW-M:Pb BW-Pb:W BW-Pb:L BW-L:Sa BW-W:Wp BW-W:M BCS-M:Pa 0.12 (0.02) −0.04 (0.02) 0.02 (0.02) 0.00 (0.02) 0.08 (0.02) 0.05 (0.03) BCS-Pa:W −0.00 (0.02) 0.36 (0.02) 0.17 (0.02) 0.08 (0.02) -0.20 (0.02) -0.21 (0.03) BCS-Pa:L 0.02 (0.02) 0.22 (0.02) 0.31 (0.02) 0.03 (0.02) -0.12 (0.02) -0.12 (0.03) BCS-L:Sa −0.05 (0.02) 0.10 (0.02) −0.01 (0.02) 0.14 (0.02) -0.02 (0.02) -0.03 (0.03) BCS-W:Wp −0.03 (0.02) −0.07 (0.03) −0.06 (0.03) −0.05 (0.02) 0.24 (0.02) 0.14 (0.03) BCS-W:M −0.04 (0.03) −0.14 (0.03) −0.08 (0.03) −0.04 (0.03) 0.22 (0.03) 0.26 (0.03) BW = body weight; BCS = body condition score; M = mating; Pa = early pregnancy; Pb = midpregnancy; L = lambing; Sa = early suckling; Sb = midsuckling; W = weaning; Wp = postweaning. View Large Table 7. Phenotypic correlations (±standard errors) between BW and BCS changes Variables BW-M:Pb BW-Pb:W BW-Pb:L BW-L:Sa BW-W:Wp BW-W:M BCS-M:Pa 0.12 (0.02) −0.04 (0.02) 0.02 (0.02) 0.00 (0.02) 0.08 (0.02) 0.05 (0.03) BCS-Pa:W −0.00 (0.02) 0.36 (0.02) 0.17 (0.02) 0.08 (0.02) -0.20 (0.02) -0.21 (0.03) BCS-Pa:L 0.02 (0.02) 0.22 (0.02) 0.31 (0.02) 0.03 (0.02) -0.12 (0.02) -0.12 (0.03) BCS-L:Sa −0.05 (0.02) 0.10 (0.02) −0.01 (0.02) 0.14 (0.02) -0.02 (0.02) -0.03 (0.03) BCS-W:Wp −0.03 (0.02) −0.07 (0.03) −0.06 (0.03) −0.05 (0.02) 0.24 (0.02) 0.14 (0.03) BCS-W:M −0.04 (0.03) −0.14 (0.03) −0.08 (0.03) −0.04 (0.03) 0.22 (0.03) 0.26 (0.03) Variables BW-M:Pb BW-Pb:W BW-Pb:L BW-L:Sa BW-W:Wp BW-W:M BCS-M:Pa 0.12 (0.02) −0.04 (0.02) 0.02 (0.02) 0.00 (0.02) 0.08 (0.02) 0.05 (0.03) BCS-Pa:W −0.00 (0.02) 0.36 (0.02) 0.17 (0.02) 0.08 (0.02) -0.20 (0.02) -0.21 (0.03) BCS-Pa:L 0.02 (0.02) 0.22 (0.02) 0.31 (0.02) 0.03 (0.02) -0.12 (0.02) -0.12 (0.03) BCS-L:Sa −0.05 (0.02) 0.10 (0.02) −0.01 (0.02) 0.14 (0.02) -0.02 (0.02) -0.03 (0.03) BCS-W:Wp −0.03 (0.02) −0.07 (0.03) −0.06 (0.03) −0.05 (0.02) 0.24 (0.02) 0.14 (0.03) BCS-W:M −0.04 (0.03) −0.14 (0.03) −0.08 (0.03) −0.04 (0.03) 0.22 (0.03) 0.26 (0.03) BW = body weight; BCS = body condition score; M = mating; Pa = early pregnancy; Pb = midpregnancy; L = lambing; Sa = early suckling; Sb = midsuckling; W = weaning; Wp = postweaning. View Large DISCUSSION The BR management over time, named BR dynamics (i.e., biological ability to mobilize and restore BR over time) in the present study, is presently considered as an interesting biological component to include in current and future animal breeding programs to increase robustness in ruminants, in order to improve the sustainability of livestock systems (De La Torre et al., 2015; Phocas et al., 2016; Friggens et al., 2017). This can be applied for systems with a wide range of farming environments and is particularly beneficial to those having to cope for instance with important fluctuations in feed resources. In this study, we confirmed our global hypothesis regarding the existence of genetic determinism for ∆BW and ∆BCS in sheep. Overall BR Dynamics Combined BCS and BW measurements are considered to be typical and relatively easily monitored parameters to describe the level of BR in sheep (Thorup et al., 2012; Brown et al., 2015; Morel et al., 2016; Puillet and Martin, 2017). Very few studies, however, are available in the literature regarding BR dynamics over complete and successive production cycles. To investigate the genetic determinism of BR dynamics using a multitrait approach, we defined several time periods based on key physiological stages over a representative production cycle. These periods were also defined in order to assess ewes’ ability to mobilize and to restore BR over time and to take into account the maximal amplitude and/or slopes of BR variations. Overall, in our experimental conditions, BR accretion was observed from weaning to early pregnancy. The increase in ∆BW and ∆BCS during the postweaning period is probably due to a decrease in the ewes’ energy requirements. The increase in ∆BW during early pregnancy is most probably related to the anabolism of pregnancy combined with the onset of fetus growth (Bauman and Currie, 1980), but also to the favorable herbage availability and quality on the rangeland at this season of the year (González-García et al., 2014). The BR mobilization was observed from midpregnancy to weaning in our conditions even if a slight, brief increase in BR was observed after early suckling, followed by a decrease in BR at midsuckling. The decrease in BR during pregnancy and suckling is likely due to the known negative energy balance of this physiological stage. The progressive increase in the nutrient requirements of the dam, to support fetus growth and development during pregnancy and then milk production after lambing, contrasts with the linear decrease in the physical intake capacity as pregnancy progresses and the physiological incapacity of the female to meet the peak in nutrient and energy demand that occurs during the first weeks after parturition (Nielsen et al., 2003; Smith et al., 2017). The biological factors that have an effect on BR changes are similar to those affecting body condition measured at single time point (Marı́a and Ascaso, 1999; González-García et al., 2014, 2015). Concerning the litter effect, an overall increase in BR mobilization could be observed via ∆BW and ∆BCS measurements as the litter size increased. This was expected because of the related higher energy requirements for ewes suckling multiple lambs. The ∆BW also increased somewhat as parity increased, especially between mating and midpregnancy and during the postweaning period, which suggests that ewes learn how to better manage BR with age (Clawson et al., 1991). Concerning the effect of age at first lambing, ewes had higher gain and lower loss of BW when they lambed at 1 yr of age. This might be explained by the breeding system used in our experimental farm where heavier ewes were early mated (i.e., at 7 mo of age), whereas lighter ewes were mated later (i.e., at 19 mo of age). However, the ∆BCS were slightly influenced by the age at first lambing. The year also affected BR variations in our conditions, probably due to variations in biomass availability in the rangeland induced by differences in climatic conditions, thus leading to different nutritional and energy balances among years. Genetic Parameters The heritabilities estimated in the present study for ∆BW and ∆BCS measured over time, and based on repeated measurements (i.e., from 0.13 ± 0.03 to 0.18 ± 0.04 and from 0.04 ± 0.02 to 0.16 ± 0.03, respectively), suggest that the biological capacities determining the nature of BR accretion and mobilization processes are heritable and could be transmitted across generations. To the best of our knowledge, this is the first work reporting heritability estimations considering a dynamic approach of ∆BW and ∆BCS throughout a complete productive cycle (e.g., over several production cycles) in ruminants. Our estimates for ∆BW provide consistent values with those reported previously in sheep for BW gain or loss (Rose et al., 2013). Another study estimated the heritability of a ∆BCS measurement in Holstein cows between the first and tenth weeks after calving and estimated a heritability of 0.09 ± 0.04 (Pryce et al., 2001), which is in agreement with our findings. The heritability estimations calculated for ∆BCS and ∆BW over time were lower than those that we obtained for BCS and BW at a given point, which range from 0.26 to 0.37 for BCS and from 0.44 to 0.58 for BW (Supplementary Table 1). Our single time-point BW and BCS heritability estimations were consistent with previous findings in cows and sheep (Koenen et al., 2001; Mao et al., 2004; Safari et al., 2005; Gizaw et al., 2007; Shackell et al., 2011). Thus, amplitude of variations in BR, i.e., BR mobilization and accretion ability, would be under genetic control but in a lower extent than level of BR. However, although more heritable traits would be better suited to faster genetic progress, BW and BCS levels did not represent the ability of ewes to mobilize or to restore BR as indicated by absence or few moderate genetic correlations between amplitude in BCS or BW variations and levels. The low proportion of total phenotypic variance for ∆BW and ∆BCS due to the ewe permanent environmental effect suggested that no additional ewe effects apart from the genetic effect affected BR dynamics. Therefore, the repeatability for BR dynamics was close to the heritability. Interestingly, a larger permanent ewe environmental effect was observed, mainly for BW, for measurements at single time points (Supplementary Table 2). This suggests that no additional ewe effects apart from the genetic effect influenced the BW, and BCS in a lower extent, of a ewe when BW was considered at a given single time point. High positive genetic correlations for ∆BW and ∆BCS between the early postweaning period and the entire postweaning period confirm that BR accretion (i.e., with the inherent anabolic processes) were predominant during this period. Surprisingly, the increase in ∆BW during the first half of pregnancy was negatively correlated with the increase in ∆BW during the postweaning period and the increase in ∆BCS during early pregnancy was not genetically correlated with the increase in ∆BCS during postweaning. Such negative correlations could be due to different biological mechanisms involved in BW recovery during the postweaning period and the increase in BW related to the growth of the fetus during pregnancy. These results contrasted with the high positive genetic correlations found between single time-point measurements of BW and BCS during the BR accretion period (i.e., during the beginning of postweaning and the entire postweaning period; Supplementary Table 3). The differences in the high or moderate correlations for single time-point measurements and the nonzero correlations among ∆BCS and ∆BW might be explained by the higher heritability of BW and BCS levels than ∆BW and ∆BCS. The long BR mobilization period from the second month of pregnancy to weaning was correlated with shorter periods (i.e., from midpregnancy to lambing; from lambing to early suckling) within the same mobilization period, both for ∆BW and ∆BCS. The high positive genetic relationships observed between ∆BCS over the 3 periods (i.e., from midpregnancy to lambing; from lambing to early suckling; and from midpregnancy to weaning) defined to investigate the decrease in BR confirmed that these periods were representative of BR mobilization. In addition, the negative genetic relationships observed between ∆BCS over these 3 periods and the 3 periods characterizing BR increase (i.e., during the beginning of postweaning; during the entire postweaning period; and from mating to midpregnancy) suggest that the series of mechanisms involved either in BR mobilization and/or accretion processes are probably driven by different physiological factors, but are also genetically related. This result is consistent with common biochemical mechanisms involved in the regulation of lipogenesis and lipolysis in adipose tissue (Chilliard et al., 1998). The ∆BCS values were generally independent of the level of BCS (or BW) measured at single time points throughout the productive cycle (e.g., fat ewes did not show higher BR mobilization and vice versa). However, some ∆BCS (and ∆BW) values were genetically correlated with the level of BCS (or BW) (Supplementary Tables 4–6). The increase in ∆BCS during early pregnancy was positively related to several single time-point BCS measurements (e.g., BCS at early pregnancy, at midpregnancy, at lambing, and at midsuckling). The loss in BCS during pregnancy was negatively correlated with the level of BCS at early pregnancy. These results may be related to the ewes’ fitness; indeed, it appeared that the ewes that could not mobilize enough BR during pregnancy tended to activate self-preservation mechanisms. Surprisingly, a similar kind of relationship was not observed during the suckling period. The present study also highlights high positive genetic correlations between ∆BW and ∆BCS over the same physiological periods. This suggests that variations in both parameters are physiologically concomitant but also genetically linked. Surprisingly, no correlations were found between ∆BW and ∆BCS at early suckling. This could be explained by other factors that greatly affect the manifestation and extent of the expression of these parameters at this stage (e.g., litter size effects on energy balance and year effects). Apart from the correlations at the same physiological stage, almost all the correlations between the BR accretion and mobilization phases were negative. This confirms that ∆BW and ∆BCS are 2 genetically related and somehow mutually exclusive traits, i.e., with well-known successive and complementary physiological cycles. In conclusion, the present study shows that BR dynamics is a heritable trait and could be a good trait to be considered for improving individual robustness in sheep. Some intervals of BR changes appeared to be more relevant for genetic selection purposes due to higher heritabilities and/or higher genetic correlations between periods. However, to define the best time interval for genetic selection of BR dynamics, further studies are needed to address hypotheses helping to explain BR changes over time. It would be also interesting to investigate additional direct or indirect relationships between BR dynamics and productive and reproductive traits before considering such traits in selection. Further research efforts are necessary, i.e., including complementary traits (e.g., plasma metabolites and hormone profiles), to determine direct relationships between BR dynamics and the energy metabolism of the animal. This would lead to a better understanding of BR accretion and mobilization mechanisms and help identify individuals or periods with excessive BR mobilization that could have a significant impact on health or survival. Footnotes 1 The authors would like to thank the INRA Divisions of Animal Genetics and Animal Physiology and Livestock Systems for their financial support over the years in providing salaries, experimental costs, and computational facilities. The authors are indebted to all the staff of La Fage experimental farm for animal management and care as well as for their key role in the experimental data collection. LITERATURE CITED Arango , J. A. , L. V. Cundiff , and L. D. Van Vleck . 2002 . Genetic parameters for weight, weight adjusted for body condition score, height, and body condition score in beef cows . J. Anim. Sci . 80 : 3112 – 3122 . Google Scholar Crossref Search ADS PubMed Bauman , D. E. , and W. B. Currie . 1980 . Partitioning of nutrients during pregnancy and lactation: a review of mechanisms involving homeostasis and homeorhesis . J. Dairy Sci . 63 : 1514 – 1529 . Google Scholar Crossref Search ADS PubMed Blanc , F. , F. Bocquier , J. Agabriel , P. D’hour , and Y. Chilliard . 2006 . Adaptive abilities of the females and sustainability of ruminant livestock systems. A review . Anim. Res . 55 : 489 – 510 . doi:10.1051/animres:2006040 Google Scholar Crossref Search ADS Brown , D. J. , D. B. Savage , G. N. Hinch , and S. Hatcher . 2015 . Monitoring liveweight in sheep is a valuable management strategy: a review of available technologies . Anim. Prod. Sci . 55 : 427 – 436 . doi:10.1071/AN13274 Google Scholar Crossref Search ADS Chilliard , Y. , F. Bocquier , and M. Doreau . 1998 . Digestive and metabolic adaptations of ruminants to undernutrition, and consequences on reproduction . Reprod. Nutr. Dev . 38 : 131 – 152 . Google Scholar Crossref Search ADS PubMed Clawson , A. J. , J. D. Garlich , M. T. Coffey , and W. G. Pond . 1991 . Nutritional, physiological, genetic, sex, and age effects on fat-free dry matter composition of the body in avian, fish, and mammalian species: a review . J. Anim. Sci . 69 : 3617 – 3644 . Google Scholar Crossref Search ADS PubMed De La Torre , A. , E. Recoules , F. Blanc , I. Ortigues-Marty , P. D’Hour , and J. Agabriel . 2015 . Changes in calculated residual energy in variable nutritional environments: an indirect approach to apprehend suckling beef cows’ robustness . Livest. Sci . 176 : 75 – 84 . doi:10.1016/j.livsci.2015.03.008 Google Scholar Crossref Search ADS Edmonson , A. J. , I. J. Lean , L. D. Weaver , T. Farver , and G. Webster . 1989 . A body condition scoring chart for holstein dairy cows . J. Dairy Sci . 72 : 68 – 78 . doi:10.3168/jds.S0022-0302(89)79081-0 Google Scholar Crossref Search ADS Friggens , N. C. , F. Blanc , D. P. Berry , and L. Puillet . 2017 . Review: deciphering animal robustness. A synthesis to facilitate its use in livestock breeding and management . Animal 11 : 2237 – 2251 . doi:10.1017/S175173111700088X Google Scholar Crossref Search ADS PubMed Friggens , N. C. , K. L. Ingvartsen , and G. C. Emmans . 2004 . Prediction of body lipid change in pregnancy and lactation . J. Dairy Sci . 87 : 988 – 1000 . doi:10.3168/jds.S0022-0302(04)73244-0 Google Scholar Crossref Search ADS PubMed Gilmour , A. R. , B. R. Cullis , S. J. Welham , and R. Thompson . 2001 . ASREML Reference Manual . 239. Gizaw , S. , S. Lemma , H. Komen , and J. A. M. Van Arendonk . 2007 . Estimates of genetic parameters and genetic trends for live weight and fleece traits in Menz sheep . Small Rumin. Res . 70 : 145 – 153 . doi:10.1016/j.smallrumres.2006.02.007 Google Scholar Crossref Search ADS González-García , E. , V. Gozzo de Figuereido , D. Foulquie , E. Jousserand , P. Autran , S. Camous , A. Tesniere , F. Bocquier , and M. Jouven . 2014 . Circannual body reserve dynamics and metabolic profile changes in romane ewes grazing on rangelands . Domest. Anim. Endocrinol . 46 : 37 – 48 . doi:10.1016/j.domaniend.2013.10.002 Google Scholar Crossref Search ADS PubMed González-García , E. , and D. Hazard . 2016 . Growth rates of Romane ewe lambs and correlated effects of being mated as hoggets or two-tooth ewes on first offspring performance . Livest. Sci . 189 : 63 – 69 . doi:10.1016/j.livsci.2016.04.022 Google Scholar Crossref Search ADS González-García , E. , A. Tesniere , S. Camous , F. Bocquier , F. Barillet , and P. Hassoun . 2015 . The effects of parity, litter size, physiological state, and milking frequency on the metabolic profile of lacaune dairy ewes . Domest. Anim. Endocrinol . 50 : 32 – 44 . doi:10.1016/j.domaniend.2014.07.001 Google Scholar Crossref Search ADS PubMed Koenen , E. P. , R. F. Veerkamp , P. Dobbelaar , and G. De Jong . 2001 . Genetic analysis of body condition score of lactating dutch holstein and red-and-white heifers . J. Dairy Sci . 84 : 1265 – 1270 . doi:10.3168/jds.S0022-0302(01)74588-2 Google Scholar Crossref Search ADS PubMed Mao , I. L. , K. Sloniewski , P. Madsen , and J. Jensen . 2004 . Changes in body condition score and in its genetic variation during lactation . Livest. Prod. Sci . 89 : 55 – 65 . doi:10.1016/j.livprodsci.2003.12.005 Google Scholar Crossref Search ADS Marı́a , G. A. , and M. S. Ascaso . 1999 . Litter size, lambing interval and lamb mortality of Salz, Rasa Aragonesa, Romanov and F1 ewes on accelerated lambing management . Small Rumin. Res . 32 : 167 – 172 . doi:10.1016/S0921-4488(98)00176-X Google Scholar Crossref Search ADS Molénat , G. , D. Foulquié , P. Autran , J. Bouix , D. Hubert , M. Jacquin , and B. Bibe . 2005 . Pour un élevage ovin allaitant performant et durable sur parcours: un système expérimental sur le Causse du Larzac . INRA Prod. Anim . 18 : 323 – 338 . Morel , P. C. H. , N. M. Schreurs , R. A. Corner-Thomas , A. W. Greer , C. M. C. Jenkinson , A. L. Ridler , and P. R. Kenyon . 2016 . Live weight and body composition associated with an increase in body condition score of mature ewes and the relationship to dietary energy requirements . Small Rumin. Res . 143 : 8 – 14 . doi:10.1016/j.smallrumres.2016.08.014 Google Scholar Crossref Search ADS Nielsen , H. M. , N. C. Friggens , P. Løvendahl , J. Jensen , and K. L. Ingvartsen . 2003 . Influence of breed, parity, and stage of lactation on lactational performance and relationship between body fatness and live weight . Livest. Prod. Sci . 79 : 119 – 133 . doi:10.1016/S0301-6226(02)00146-X Google Scholar Crossref Search ADS O’Mara , F. P . 2012 . The role of grasslands in food security and climate change . Ann. Bot . 110 : 1263 – 1270 . doi:10.1093/aob/mcs209 Google Scholar Crossref Search ADS PubMed Phocas , F. , C. Belloc , J. Bidanel , L. Delaby , J. Y. Dourmad , B. Dumont , P. Ezanno , L. Fortun-Lamothe , G. Foucras , B. Frappat , et al. 2016 . Review: towards the agroecological management of ruminants, pigs and poultry through the development of sustainable breeding programmes: I-selection goals and criteria . Animal 10 : 1749 – 1759 . doi:10.1017/S1751731116000926 Google Scholar Crossref Search ADS PubMed Pryce , J. E. , M. P. Coffey , and G. Simm . 2001 . The relationship between body condition score and reproductive performance . J. Dairy Sci . 84 : 1508 – 1515 . doi:10.3168/jds.S0022-0302(01)70184-1 Google Scholar Crossref Search ADS PubMed Puillet , L. , and O. Martin . 2017 . A dynamic model as a tool to describe the variability of lifetime body weight trajectories in livestock females . J. Anim. Sci . 95 : 4846 – 4856 . doi:10.2527/jas2017.1803 Google Scholar Crossref Search ADS PubMed Ricordeau , G. , L. Tchamitchian , J. C. Brunel , T. C. Nguyen , and D. François . 1992 . La race ovine INRA 401: un exemple de souche synthétique . INRA Prod. Anim . (hs): 255 – 262 . Rose , G. , A. Kause , H. A. Mulder , J. H. van der Werf , A. N. Thompson , M. B. Ferguson , and J. A. van Arendonk . 2013 . Merino ewes can be bred for body weight change to be more tolerant to uncertain feed supply . J. Anim. Sci . 91 : 2555 – 2565 . doi:10.2527/jas.2012-5539 Google Scholar Crossref Search ADS PubMed Russel , A. J. F. , J. M. Doney , and R. G. Gunn . 1969 . Subjective assessment of body fat in live sheep . J. Agric. Sci . 72 : 451 – 454 . doi:10.1017/S0021859600024874 Google Scholar Crossref Search ADS Safari , E. , N. M. Fogarty , and A. R. Gilmour . 2005 . A review of genetic parameter estimates for wool, growth, meat and reproduction traits in sheep . Livest. Prod. Sci . 92 : 271 – 289 . doi:10.1016/j.livprodsci.2004.09.003 Google Scholar Crossref Search ADS Shackell , G. H. , N. G. Cullen , and G. J. Greer . 2011 . Genetic parameters associated with adult ewe liveweight and body condition . In: Proc. 19th Assoc. Adv. Anim. Breed. Genet . p. 103 – 106 . Smith , G. L. , N. C. Friggens , C. J. Ashworth , and M. G. G. Chagunda . 2017 . Association between body energy content in the dry period and post-calving production disease status in dairy cattle . Animal 11 : 1590 – 1598 . doi:10.1017/S1751731117000040 Google Scholar Crossref Search ADS PubMed Thorup , V. M. , D. Edwards , and N. C. Friggens . 2012 . On-farm estimation of energy balance in dairy cows using only frequent body weight measurements and body condition score . J. Dairy Sci . 95 : 1784 – 1793 . doi:10.3168/jds.2011-4631 Google Scholar Crossref Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the American Society of Animal Science. All rights reserved. For permissions, please e-mail: [email protected]. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

McRae, Kathryn M; Rowe, Suzanne J; Baird, Hayley J; Bixley, Matthew J; Clarke, Shannon M

doi: 10.1093/jas/sky323pmid: 30099550

Abstract Pneumonia is an important issue for sheep production, leading to reduced growth rate and a predisposition to pleurisy. The objective of this study was to identify loci associated with pneumonic lesions and pleurisy in New Zealand progeny test lambs. The lungs from 3,572 progeny-test lambs were scored for presence and severity of pneumonic lesions and pleurisy at slaughter. Animals were genotyped using the Illumina Ovine Infinium HD SNP BeadChip (606,006 markers). The heritability of lung lesion score and pleurisy were calculated using the genomic relationship matrix, and genome-wide association analyses were conducted using EMMAX and haplotype trend regression. At slaughter, 35% of lambs had pneumonic lesions, with 9% showing lesions on more than half of any individual lobe. The number of lambs recorded as having pleurisy by the processing plants was 9%. Heritability estimates for pneumonic lesions and pleurisy scores adjusted for heteroscedasticity (CPSa and PLEURa) were 0.16 (± 0.03) and 0.05 (± 0.02), respectively. Five single-nucleotide polymorphisms (SNPs) were significantly associated with pneumonic lesions at the genome-wide level, and additional 37 SNPs were suggestively significant. Four SNPs were significantly associated with pleurisy, with an additional 11 SNPs reaching the suggestive level of significance. There were no regions that overlapped between the 2 traits. Multiple SNPs were in regions that contained genes involved in either the DNA damage response or the innate immune response, including several that had previously been reported to have associations with respiratory disease. Both EMMAX and HTR analyses of pleurisy data showed a significant peak on chromosome 2, located downstream from the transcription factor SP3. SP3 activates or suppresses the expression of numerous genes, including several genes with known functions in the immune system. This study identified several SNPs associated with genes involved in both the innate immune response and the response to DNA damage that are associated with pneumonic lesions and pleurisy in lambs at slaughter. Additionally, the identification in sheep of several SNPs within genes that have previously been associated with the respiratory system in cattle, pigs, rats, and mice indicates that there may be common pathways that underlie the response to invasion by respiratory pathogens in multiple species. INTRODUCTION Chronic nonprogressive pneumonia is the most common form of ovine pneumonia in New Zealand, and is an important issue for sheep production, leading to reduced growth rate (Kirton et al., 1976; Alley, 1987; Goodwin et al., 2004; McRae et al., 2016) and a predisposition to pleurisy (Alley, 2002). There is well-documented evidence for between-animal variation in the ability of livestock to resist multiple diseases of economic importance, including respiratory disease (Bishop and Morris, 2007; Davies et al., 2009). Previous work has established that the heritability of pneumonic lesions at slaughter in New Zealand mixed breed progeny tested lambs is 0.07 ± 0.02 (McRae et al., 2016), which is comparable to estimates of respiratory disease in cattle (Snowder, 2009). With growing pressure to reduce the use of antibiotics and drugs in agriculture, these heritable differences mean that improvement of animal health through genetic selection for enhanced resistance can be used as a complementary approach to current methods for disease control (Goddard, 2012). More fundamentally, genomics, through tools such as genome-wide association studies (GWAS), can also be used to further increase our understanding of the genetic mechanisms underlying the host response to disease, and compare these mechanisms between breeds or species. Discovering regions of the genome associated with resistance or susceptibility may also lead to the development of new diagnostic tools and alternative treatments. The aim of this study was to utilize genotype data to identify regions of the ovine genome associated with pneumonic lesions and pleurisy in New Zealand lambs. MATERIALS AND METHODS Animals were managed in accordance with the provisions of the New Zealand Animal Welfare Act 1999, and the New Zealand Codes of Welfare developed under sections 68–79 of the Act. Animals The lungs from a total of 3,572 ewe and ram lambs from 4 flocks were scored for the presence and severity of pneumonic lesions. Lambs were from 3 South Island (Flocks A, B, and C) and one North Island (Flock D) progeny test flocks. All flocks were fixed-date slaughters, which took place when lambs were between 4 and 8 mo of age. Dams were composites of the main dual-purpose sheep breeds used in New Zealand, including Romney, Coopworth, Perendale, and Texel. Sires were a mixture of dual-purpose and terminal sire composites. Phenotypic Measurements The methodology for scoring pneumonic lesions has been previously described (Baird et al., 2012; McRae et al., 2016). Briefly, lungs were scored at chain speed postslaughter at the processing plant. The “consolidated pneumonia score” (CPS) system has a range from 0 to 2, where 0 = no lesions present; 1 = any individual lobe with up to 50% of the lobe affected and 2 = any individual lobe with greater than 50% of the lobe affected. Pneumonic lesions were defined as compacted, dark purple-red areas of the lung that were firm to touch. Information on lamb carcasses that were identified as having pleurisy and detained for trimming was obtained from the processing plants. Data cleaning consisted of removal of records with 1) missing values, and 2) contemporary groups (CG) containing less than 5 observations. CG was defined as flock, birth year, sex, weaning mob, and slaughter date; animals needed to have all of these in common to be considered in the same CG. Weaning mob was obtained from Sheep Improvement Limited (SIL), the New Zealand sheep genetic evaluation database. To adjust for heteroscedasticity, CPS (initially scored as 0, 1, or 2) was scaled using the formula CPSa = CPS/SQRT[CPSm*(2-CPSm)], where m is the mean incidence rate within the CG where phenotypic score is being adjusted. Pleurisy (initially coded as 0 or 1) values were also transformed using the formula PLEURa = PLEUR/SQRT[PLEURm*(1-PLEURm)]. Genotypes and Quality Control Genomic DNA was extracted from ear tissue samples collected from lambs at tailing, using a high-throughput DNA extraction method (Clarke et al., 2014). Animals were genotyped with the Illumina Ovine Infinium HD SNP BeadChip (606,006 markers) according to the manufacturer’s protocol. Genome coordinates of each single-nucleotide polymorphism (SNP) were based on the OARv3.1 ovine genome assembly (Jiang et al., 2014). Quality control checks excluded markers that appeared nonautosomal (including pseudoautosomal), had a call rate below 90%, and/or had a minor allele frequency (MAF) ≤0.01. Individuals were excluded from the analysis if there was more than 5% genotyping failure. After quality control measures, 3,546 phenotyped animals were available, with 537,117 of the initial 606,006 SNPs utilized for analysis. Heritability Variance components were estimated using restricted maximum likelihood (REML) procedures fitting an animal model in ASReml (Gilmour et al., 2015), with the genomic relationship matrix (GRM) estimated in GenABEL (Aulchenko et al., 2007) using HD genotypes. Heritabilities were obtained by running a univariate analysis on the respective traits. Data analysis models for both CPSa and PLEURa, included CG as a fixed effect (McRae et al., 2016). Genome-wide Association Analyses Pneumonic lesion and pleurisy data were analyzed using values adjusted for heteroscedasticity (CPSa and PLEURa). Pneumonic lesion data was also analyzed by only including animals with no lesions and those with severe lesions [CPSa (0&2)]. Genome-wide association analyses were performed using SNP & Variation Suite v8.4.0 (Golden Helix, Inc., Bozeman, MT, www.goldenhelix.com) using 2 of the following approaches: 1) Efficient Mixed-Model Association eXpedited (EMMAX) using identity-by-state (IBS), and 2) haplotype trend regression (HTR) with a 3-SNP sliding window. Analyses were performed on adjusted values, with CG fitted as a covariate. Genome browse software was used to visualize results with an added track of Ovis aries genes from Ensembl 84. After Bonferonni correction, thresholds were 9.31 × 10−8 and 1.86 × 10−6 for genome-wide significance (P < 0.05) and suggestive significance (P < 0.1), respectively. RESULTS Incidence of Pneumonia and Pleurisy In total, 3,572 lungs were scored for pneumonic lesions from lambs born between 2013 and 2015 (Table 1). Of these, 1,234 (35%) had lesions, with 329 (9%) showing lesions on more than 50% of any individual lobe (CPS of 2). The number of lambs recorded as having pleurisy by the processing plants was 310 (9%). The incidence of pneumonia was significantly higher in 2014-born lambs than those born in 2013 or 2015 (P < 0.001). Of the 310 animals recorded as having pleurisy, 118 (38%) had a CPS of 0, 71 (23%) had a score of 1, 60 (19%) had a score of 1, and 61 (20%) were unable to be scored due to the lungs being retained in the carcass. Table 1. Incidence of pneumonic lesions and pleurisy by flock and year of birth Flock . Year born . Lungs scored . CPSa > 0 . CPSa = 2 . Pleurisy . A 2013 292 52 (18%) 20 (7%) 25 (9%) 2014 483 194 (40%) 57 (12%) 70 (14%) 2015 467 98 (21%) 29 (6%) 77 (16%) B 2014 766 334 (44%) 80 (10%) 59 (8%) 2015 292 46 (16%) 11 (4%) 6 (2%) C 2015 56 14 (25%) 3 (5%) 13 (23%) D 2014 1,216 496 (41%) 129 (11%) 60 (5%) Total 3,572 1,234 (35%) 329 (9%) 310 (9%) Flock . Year born . Lungs scored . CPSa > 0 . CPSa = 2 . Pleurisy . A 2013 292 52 (18%) 20 (7%) 25 (9%) 2014 483 194 (40%) 57 (12%) 70 (14%) 2015 467 98 (21%) 29 (6%) 77 (16%) B 2014 766 334 (44%) 80 (10%) 59 (8%) 2015 292 46 (16%) 11 (4%) 6 (2%) C 2015 56 14 (25%) 3 (5%) 13 (23%) D 2014 1,216 496 (41%) 129 (11%) 60 (5%) Total 3,572 1,234 (35%) 329 (9%) 310 (9%) aCPS = Consolidated Pneumonia Score, where 0 = no lesions present; 1 = individual lobes with up to 50% of the lobe affected and 2 = individual lobes with greater than 50% of the lobe affected. Open in new tab Table 1. Incidence of pneumonic lesions and pleurisy by flock and year of birth Flock . Year born . Lungs scored . CPSa > 0 . CPSa = 2 . Pleurisy . A 2013 292 52 (18%) 20 (7%) 25 (9%) 2014 483 194 (40%) 57 (12%) 70 (14%) 2015 467 98 (21%) 29 (6%) 77 (16%) B 2014 766 334 (44%) 80 (10%) 59 (8%) 2015 292 46 (16%) 11 (4%) 6 (2%) C 2015 56 14 (25%) 3 (5%) 13 (23%) D 2014 1,216 496 (41%) 129 (11%) 60 (5%) Total 3,572 1,234 (35%) 329 (9%) 310 (9%) Flock . Year born . Lungs scored . CPSa > 0 . CPSa = 2 . Pleurisy . A 2013 292 52 (18%) 20 (7%) 25 (9%) 2014 483 194 (40%) 57 (12%) 70 (14%) 2015 467 98 (21%) 29 (6%) 77 (16%) B 2014 766 334 (44%) 80 (10%) 59 (8%) 2015 292 46 (16%) 11 (4%) 6 (2%) C 2015 56 14 (25%) 3 (5%) 13 (23%) D 2014 1,216 496 (41%) 129 (11%) 60 (5%) Total 3,572 1,234 (35%) 329 (9%) 310 (9%) aCPS = Consolidated Pneumonia Score, where 0 = no lesions present; 1 = individual lobes with up to 50% of the lobe affected and 2 = individual lobes with greater than 50% of the lobe affected. Open in new tab Heritability The heritability estimated for CPSa was 0.16 (± 0.03) and for PLEURa was 0.05 (± 0.02). This is slightly higher than the previously published estimates of 0.07 ± 0.02 and 0.02 ± 0.01, respectively (McRae et al., 2016). This is likely to be due to the use of a GRM rather than recorded pedigree information in estimating the heritability in the current analysis; the 1,216 lambs from Flock D were included in both studies, however only sire information is recorded for these animals, therefore using a GRM rather than pedigree is a more accurate estimation of relatedness. The genetic correlation between the 2 traits was 0.58 (± 0.16), and the phenotypic correlation was (0.15 ± 0.02), which was in line with previous estimates. Genome-wide Association Analyses When adjusted pneumonic lesion (CPSa) information from all animals was included, there were no SNPs that passed the threshold for suggestive significance in the EMMAX analysis (Fig. 1A). In the HTR analysis, however, 4 regions, on chromosomes 3, 6, 8, and 13, were significant (Fig. 1B), with a further 31 SNPs passing the level for suggestive significance (Table 2). When only the extreme animals were included (i.e., animals with no lesions compared to those with severe lesions; CPS of 0 vs. CPS of 2), several SNPs in each analysis were of suggestive significance, although none reached the genome-wide level of significance (Fig. 1C and D). The top 3 SNPs in the EMMAX analysis were all intronic variants, in the LSAMP, PPIL6, and KCNMA1 genes (Table 2). The top SNPs in the HTR analysis included the same SNP in EYA4 that reached the suggestive significance level in the CPSa HTR analysis, along with 2 missense variants in exon 2 of ATAD5. Both EMMAX and HTR analyses of pleurisy data showed a significant peak on chromosome 2 (Fig. 2). Additionally, there were multiple suggestively significant intergenic SNPs on chromosomes 8 and 11 (Table 2). Table 2. SNPs suggestively and significantly associated with consolidated pneumonia score (CPS) and pleurisy in New Zealand lambs Testa . Analysisb . P-valuec . Chr . Position . RSID . Gened . Gene named . Variant consequence (impact)d . CPSa HTR 1.72E-06 1 198656040 rs421794454 ENSOARG00000020512 RFC4 Intron variant HTR 1.02E-06 1 198674451 rs416081302 Multiple small nucleolar RNAs Upstream gene variant HTR 8.00E-07 2 158991503 rs407273673 HTR 1.15E-06 2 158998132 rs417102378 HTR 8.05E-07 3 197683811 rs426850802 HTR 8.74E-07 3 197705247 rs405096150 HTR 1.72E-07 3 197720146 rs430716198 HTR 1.34E-07 3 197720936 rs412869687 HTR 3.80E-07 3 197824787 rs407726225 HTR 4.75E-08* 3 197825391 rs424070250 HTR 1.06E-06 4 100364303 rs403394816 HTR 6.45E-08* 6 77843695 rs399606595 HTR 1.63E-09* 8 7733798 rs429357466 HTR 7.65E-09* 8 7743164 rs400905064 HTR 4.29E-07 8 12061122 rs425423371 HTR 1.01E-06 8 12061431 rs402511423 HTR 1.85E-06 8 12149272 rs423436094 HTR 6.85E-07 8 46753989 rs422310670 HTR 4.64E-07 8 58610312 rs418966278 ENSOARG00000014564 EYA4 Intron variant HTR 1.10E-06 8 65817474 rs419752214 HTR 2.43E-07 8 65827376 rs399425501 HTR 3.76E-07 8 65834390 rs429368446 HTR 6.06E-07 9 56587375 rs424700173 HTR 8.75E-07 12 12552747 rs408273790 HTR 1.01E-06 13 8840940 rs400804234 HTR 3.36E-08* 13 8848881 rs417728121 HTR 2.36E-07 13 8850719 rs423025524 HTR 4.67E-07 13 8855334 rs416260513 HTR 1.49E-06 13 8860196 rs420406541 HTR 3.06E-07 13 8864106 rs428620400 HTR 4.21E-07 13 8871693 rs430812458 HTR 9.09E-07 13 8872416 rs404379883 HTR 8.25E-07 18 51969325 rs419274927 ENSOARG00000026456 Novel lincRNA Noncoding transcript variant HTR 1.26E-06 20 18894279 rs401389671 HTR 1.37E-06 20 23420251 rs417356235 HTR 1.10E-06 20 23423229 rs399485900 CPSa (0&2) EMMAX 7.91E-07 1 178857472 rs410655004 ENSOARG00000019641 LSAMP Intron variant EMMAX 1.42E-06 8 27846793 rs430024463 ENSOARG00000010461 PPIL6 Intron variant EMMAX 1.29E-06 25 32872254 rs422854508 ENSOARG00000009163 KCNMA1 Intron variant HTR 9.49E-07 4 110104884 rs425466808 HTR 1.39E-06 8 7743164 rs400905064 HTR 2.23E-07 8 46753989 rs422310670 HTR 1.43E-06 8 58610312 rs418966278 ENSOARG00000014564 EYA4 Intron variant HTR 8.91E-07 11 17720321 rs419581914 ENSOARG00000012322 ATAD5 Missense variant (moderate) HTR 1.79E-06 11 17720415 rs400520703 ENSOARG00000012322 ATAD5 Missense variant (moderate) PLEURa EMMAX 3.10E-09* 2 134984962 rs398681238 EMMAX 2.32E-08* 2 134985148 rs424471052 EMMAX 1.57E-06 2 135163372 rs415671617 EMMAX 1.48E-06 2 242723012 rs421193149 EMMAX 1.58E-06 11 15261261 rs420254502 EMMAX 3.29E-07 11 15262540 rs409974296 EMMAX 1.80E-07 11 15265356 rs417033802 HTR 1.38E-07 2 134976058 rs404285802 ENSOARG00000000469 SP3 Downstream gene variant HTR 7.66E-08* 2 134979525 rs428634189 HTR 1.53E-08* 2 134984962 rs398681238 HTR 6.59E-07 2 134985148 rs424471052 HTR 1.88E-07 2 134998369 rs414115266 HTR 1.82E-06 2 135006264 rs412779979 HTR 1.72E-06 8 13863996 rs414046873 HTR 9.86E-07 8 88651287 rs412134993 HTR 4.03E-07 8 88659717 rs398705894 Testa . Analysisb . P-valuec . Chr . Position . RSID . Gened . Gene named . Variant consequence (impact)d . CPSa HTR 1.72E-06 1 198656040 rs421794454 ENSOARG00000020512 RFC4 Intron variant HTR 1.02E-06 1 198674451 rs416081302 Multiple small nucleolar RNAs Upstream gene variant HTR 8.00E-07 2 158991503 rs407273673 HTR 1.15E-06 2 158998132 rs417102378 HTR 8.05E-07 3 197683811 rs426850802 HTR 8.74E-07 3 197705247 rs405096150 HTR 1.72E-07 3 197720146 rs430716198 HTR 1.34E-07 3 197720936 rs412869687 HTR 3.80E-07 3 197824787 rs407726225 HTR 4.75E-08* 3 197825391 rs424070250 HTR 1.06E-06 4 100364303 rs403394816 HTR 6.45E-08* 6 77843695 rs399606595 HTR 1.63E-09* 8 7733798 rs429357466 HTR 7.65E-09* 8 7743164 rs400905064 HTR 4.29E-07 8 12061122 rs425423371 HTR 1.01E-06 8 12061431 rs402511423 HTR 1.85E-06 8 12149272 rs423436094 HTR 6.85E-07 8 46753989 rs422310670 HTR 4.64E-07 8 58610312 rs418966278 ENSOARG00000014564 EYA4 Intron variant HTR 1.10E-06 8 65817474 rs419752214 HTR 2.43E-07 8 65827376 rs399425501 HTR 3.76E-07 8 65834390 rs429368446 HTR 6.06E-07 9 56587375 rs424700173 HTR 8.75E-07 12 12552747 rs408273790 HTR 1.01E-06 13 8840940 rs400804234 HTR 3.36E-08* 13 8848881 rs417728121 HTR 2.36E-07 13 8850719 rs423025524 HTR 4.67E-07 13 8855334 rs416260513 HTR 1.49E-06 13 8860196 rs420406541 HTR 3.06E-07 13 8864106 rs428620400 HTR 4.21E-07 13 8871693 rs430812458 HTR 9.09E-07 13 8872416 rs404379883 HTR 8.25E-07 18 51969325 rs419274927 ENSOARG00000026456 Novel lincRNA Noncoding transcript variant HTR 1.26E-06 20 18894279 rs401389671 HTR 1.37E-06 20 23420251 rs417356235 HTR 1.10E-06 20 23423229 rs399485900 CPSa (0&2) EMMAX 7.91E-07 1 178857472 rs410655004 ENSOARG00000019641 LSAMP Intron variant EMMAX 1.42E-06 8 27846793 rs430024463 ENSOARG00000010461 PPIL6 Intron variant EMMAX 1.29E-06 25 32872254 rs422854508 ENSOARG00000009163 KCNMA1 Intron variant HTR 9.49E-07 4 110104884 rs425466808 HTR 1.39E-06 8 7743164 rs400905064 HTR 2.23E-07 8 46753989 rs422310670 HTR 1.43E-06 8 58610312 rs418966278 ENSOARG00000014564 EYA4 Intron variant HTR 8.91E-07 11 17720321 rs419581914 ENSOARG00000012322 ATAD5 Missense variant (moderate) HTR 1.79E-06 11 17720415 rs400520703 ENSOARG00000012322 ATAD5 Missense variant (moderate) PLEURa EMMAX 3.10E-09* 2 134984962 rs398681238 EMMAX 2.32E-08* 2 134985148 rs424471052 EMMAX 1.57E-06 2 135163372 rs415671617 EMMAX 1.48E-06 2 242723012 rs421193149 EMMAX 1.58E-06 11 15261261 rs420254502 EMMAX 3.29E-07 11 15262540 rs409974296 EMMAX 1.80E-07 11 15265356 rs417033802 HTR 1.38E-07 2 134976058 rs404285802 ENSOARG00000000469 SP3 Downstream gene variant HTR 7.66E-08* 2 134979525 rs428634189 HTR 1.53E-08* 2 134984962 rs398681238 HTR 6.59E-07 2 134985148 rs424471052 HTR 1.88E-07 2 134998369 rs414115266 HTR 1.82E-06 2 135006264 rs412779979 HTR 1.72E-06 8 13863996 rs414046873 HTR 9.86E-07 8 88651287 rs412134993 HTR 4.03E-07 8 88659717 rs398705894 aAnalyses were performed on consolidated pneumonia and pleurisy scores after adjustment for heteroscedasticity (CPSa and PLEURa, respectively). For CPSa data, analyses were performed using all animals, or only including animals with scores of 0 or 2 [CPSa (0&2)]. bGenome-wide association analyses were conducted using 2 approaches: 1) Efficient Mixed-Model Association eXpedited (EMMAX) using identity-by-state (IBS), and 2) haplotype trend regression (HTR) with a 3-SNP sliding window. Contemporary group (sex, birth year, flock, weaning mob, and kill date) was fitted as a covariate in all analyses. cAfter Bonferonni correction, thresholds were 9.31 × 10−8 and 1.86 × 10−6 for genome-wide significance (P < 0.05*) and suggestive significance (P < 0.1), respectively. dGene names and variant consequences are based on Ensembl Release 84. Open in new tab Table 2. SNPs suggestively and significantly associated with consolidated pneumonia score (CPS) and pleurisy in New Zealand lambs Testa . Analysisb . P-valuec . Chr . Position . RSID . Gened . Gene named . Variant consequence (impact)d . CPSa HTR 1.72E-06 1 198656040 rs421794454 ENSOARG00000020512 RFC4 Intron variant HTR 1.02E-06 1 198674451 rs416081302 Multiple small nucleolar RNAs Upstream gene variant HTR 8.00E-07 2 158991503 rs407273673 HTR 1.15E-06 2 158998132 rs417102378 HTR 8.05E-07 3 197683811 rs426850802 HTR 8.74E-07 3 197705247 rs405096150 HTR 1.72E-07 3 197720146 rs430716198 HTR 1.34E-07 3 197720936 rs412869687 HTR 3.80E-07 3 197824787 rs407726225 HTR 4.75E-08* 3 197825391 rs424070250 HTR 1.06E-06 4 100364303 rs403394816 HTR 6.45E-08* 6 77843695 rs399606595 HTR 1.63E-09* 8 7733798 rs429357466 HTR 7.65E-09* 8 7743164 rs400905064 HTR 4.29E-07 8 12061122 rs425423371 HTR 1.01E-06 8 12061431 rs402511423 HTR 1.85E-06 8 12149272 rs423436094 HTR 6.85E-07 8 46753989 rs422310670 HTR 4.64E-07 8 58610312 rs418966278 ENSOARG00000014564 EYA4 Intron variant HTR 1.10E-06 8 65817474 rs419752214 HTR 2.43E-07 8 65827376 rs399425501 HTR 3.76E-07 8 65834390 rs429368446 HTR 6.06E-07 9 56587375 rs424700173 HTR 8.75E-07 12 12552747 rs408273790 HTR 1.01E-06 13 8840940 rs400804234 HTR 3.36E-08* 13 8848881 rs417728121 HTR 2.36E-07 13 8850719 rs423025524 HTR 4.67E-07 13 8855334 rs416260513 HTR 1.49E-06 13 8860196 rs420406541 HTR 3.06E-07 13 8864106 rs428620400 HTR 4.21E-07 13 8871693 rs430812458 HTR 9.09E-07 13 8872416 rs404379883 HTR 8.25E-07 18 51969325 rs419274927 ENSOARG00000026456 Novel lincRNA Noncoding transcript variant HTR 1.26E-06 20 18894279 rs401389671 HTR 1.37E-06 20 23420251 rs417356235 HTR 1.10E-06 20 23423229 rs399485900 CPSa (0&2) EMMAX 7.91E-07 1 178857472 rs410655004 ENSOARG00000019641 LSAMP Intron variant EMMAX 1.42E-06 8 27846793 rs430024463 ENSOARG00000010461 PPIL6 Intron variant EMMAX 1.29E-06 25 32872254 rs422854508 ENSOARG00000009163 KCNMA1 Intron variant HTR 9.49E-07 4 110104884 rs425466808 HTR 1.39E-06 8 7743164 rs400905064 HTR 2.23E-07 8 46753989 rs422310670 HTR 1.43E-06 8 58610312 rs418966278 ENSOARG00000014564 EYA4 Intron variant HTR 8.91E-07 11 17720321 rs419581914 ENSOARG00000012322 ATAD5 Missense variant (moderate) HTR 1.79E-06 11 17720415 rs400520703 ENSOARG00000012322 ATAD5 Missense variant (moderate) PLEURa EMMAX 3.10E-09* 2 134984962 rs398681238 EMMAX 2.32E-08* 2 134985148 rs424471052 EMMAX 1.57E-06 2 135163372 rs415671617 EMMAX 1.48E-06 2 242723012 rs421193149 EMMAX 1.58E-06 11 15261261 rs420254502 EMMAX 3.29E-07 11 15262540 rs409974296 EMMAX 1.80E-07 11 15265356 rs417033802 HTR 1.38E-07 2 134976058 rs404285802 ENSOARG00000000469 SP3 Downstream gene variant HTR 7.66E-08* 2 134979525 rs428634189 HTR 1.53E-08* 2 134984962 rs398681238 HTR 6.59E-07 2 134985148 rs424471052 HTR 1.88E-07 2 134998369 rs414115266 HTR 1.82E-06 2 135006264 rs412779979 HTR 1.72E-06 8 13863996 rs414046873 HTR 9.86E-07 8 88651287 rs412134993 HTR 4.03E-07 8 88659717 rs398705894 Testa . Analysisb . P-valuec . Chr . Position . RSID . Gened . Gene named . Variant consequence (impact)d . CPSa HTR 1.72E-06 1 198656040 rs421794454 ENSOARG00000020512 RFC4 Intron variant HTR 1.02E-06 1 198674451 rs416081302 Multiple small nucleolar RNAs Upstream gene variant HTR 8.00E-07 2 158991503 rs407273673 HTR 1.15E-06 2 158998132 rs417102378 HTR 8.05E-07 3 197683811 rs426850802 HTR 8.74E-07 3 197705247 rs405096150 HTR 1.72E-07 3 197720146 rs430716198 HTR 1.34E-07 3 197720936 rs412869687 HTR 3.80E-07 3 197824787 rs407726225 HTR 4.75E-08* 3 197825391 rs424070250 HTR 1.06E-06 4 100364303 rs403394816 HTR 6.45E-08* 6 77843695 rs399606595 HTR 1.63E-09* 8 7733798 rs429357466 HTR 7.65E-09* 8 7743164 rs400905064 HTR 4.29E-07 8 12061122 rs425423371 HTR 1.01E-06 8 12061431 rs402511423 HTR 1.85E-06 8 12149272 rs423436094 HTR 6.85E-07 8 46753989 rs422310670 HTR 4.64E-07 8 58610312 rs418966278 ENSOARG00000014564 EYA4 Intron variant HTR 1.10E-06 8 65817474 rs419752214 HTR 2.43E-07 8 65827376 rs399425501 HTR 3.76E-07 8 65834390 rs429368446 HTR 6.06E-07 9 56587375 rs424700173 HTR 8.75E-07 12 12552747 rs408273790 HTR 1.01E-06 13 8840940 rs400804234 HTR 3.36E-08* 13 8848881 rs417728121 HTR 2.36E-07 13 8850719 rs423025524 HTR 4.67E-07 13 8855334 rs416260513 HTR 1.49E-06 13 8860196 rs420406541 HTR 3.06E-07 13 8864106 rs428620400 HTR 4.21E-07 13 8871693 rs430812458 HTR 9.09E-07 13 8872416 rs404379883 HTR 8.25E-07 18 51969325 rs419274927 ENSOARG00000026456 Novel lincRNA Noncoding transcript variant HTR 1.26E-06 20 18894279 rs401389671 HTR 1.37E-06 20 23420251 rs417356235 HTR 1.10E-06 20 23423229 rs399485900 CPSa (0&2) EMMAX 7.91E-07 1 178857472 rs410655004 ENSOARG00000019641 LSAMP Intron variant EMMAX 1.42E-06 8 27846793 rs430024463 ENSOARG00000010461 PPIL6 Intron variant EMMAX 1.29E-06 25 32872254 rs422854508 ENSOARG00000009163 KCNMA1 Intron variant HTR 9.49E-07 4 110104884 rs425466808 HTR 1.39E-06 8 7743164 rs400905064 HTR 2.23E-07 8 46753989 rs422310670 HTR 1.43E-06 8 58610312 rs418966278 ENSOARG00000014564 EYA4 Intron variant HTR 8.91E-07 11 17720321 rs419581914 ENSOARG00000012322 ATAD5 Missense variant (moderate) HTR 1.79E-06 11 17720415 rs400520703 ENSOARG00000012322 ATAD5 Missense variant (moderate) PLEURa EMMAX 3.10E-09* 2 134984962 rs398681238 EMMAX 2.32E-08* 2 134985148 rs424471052 EMMAX 1.57E-06 2 135163372 rs415671617 EMMAX 1.48E-06 2 242723012 rs421193149 EMMAX 1.58E-06 11 15261261 rs420254502 EMMAX 3.29E-07 11 15262540 rs409974296 EMMAX 1.80E-07 11 15265356 rs417033802 HTR 1.38E-07 2 134976058 rs404285802 ENSOARG00000000469 SP3 Downstream gene variant HTR 7.66E-08* 2 134979525 rs428634189 HTR 1.53E-08* 2 134984962 rs398681238 HTR 6.59E-07 2 134985148 rs424471052 HTR 1.88E-07 2 134998369 rs414115266 HTR 1.82E-06 2 135006264 rs412779979 HTR 1.72E-06 8 13863996 rs414046873 HTR 9.86E-07 8 88651287 rs412134993 HTR 4.03E-07 8 88659717 rs398705894 aAnalyses were performed on consolidated pneumonia and pleurisy scores after adjustment for heteroscedasticity (CPSa and PLEURa, respectively). For CPSa data, analyses were performed using all animals, or only including animals with scores of 0 or 2 [CPSa (0&2)]. bGenome-wide association analyses were conducted using 2 approaches: 1) Efficient Mixed-Model Association eXpedited (EMMAX) using identity-by-state (IBS), and 2) haplotype trend regression (HTR) with a 3-SNP sliding window. Contemporary group (sex, birth year, flock, weaning mob, and kill date) was fitted as a covariate in all analyses. cAfter Bonferonni correction, thresholds were 9.31 × 10−8 and 1.86 × 10−6 for genome-wide significance (P < 0.05*) and suggestive significance (P < 0.1), respectively. dGene names and variant consequences are based on Ensembl Release 84. Open in new tab Figure 1. Open in new tabDownload slide Manhattan plot of genome-wide association analysis for consolidated pneumonia score (CPS) in New Zealand lambs. Analyses were performed on all animals (A and B), or only including animals with scores of 0 or 2 (C and D). Genome-wide association analyses were conducted using 2 approaches: 1) Efficient Mixed-Model Association eXpedited (EMMAX) using identity-by-state (IBS) (A and C), and 2) haplotype trend regression (HTR) with a 3-SNP sliding window (B and D). Analyses were performed on pneumonic lesion scores after adjustment for heteroscedasticity, with contemporary group fitted as a covariate. Figure 2. Open in new tabDownload slide Manhattan plot of genome-wide association analysis for pleurisy in New Zealand lambs. Genome-wide association analyses were conducted using 2 approaches: 1) Efficient Mixed-Model Association eXpedited (EMMAX) using identity-by-state (IBS) (A), and 2) haplotype trend regression (HTR) with a 3-SNP sliding window (B). Analyses were performed on pleurisy scores after adjustment for heteroscedasticity, with contemporary group fitted as a covariate. DISCUSSION In sheep, as with other ruminants, respiratory disease such as pneumonia is etiologically complex, resulting from a complex interaction between multiple infectious agents and the host, which is often compromised by physical and physiological stress. GWAS help provide an understanding of the genes and pathways involved in the response to disease. GWAS in both dairy (Neibergs et al., 2014) and beef (Keele et al., 2015) cattle have identified multiple loci associated with bovine respiratory disease complex (BRDC). Neibergs et al. (2014) discovered candidate loci involved in viral susceptibility, viral entry into cells, and modulation of inflammation in a case–control analysis of preweaned Holstein calves. A GWAS of lung lesions in beef cattle identified SNPs near candidate genes involved in functions such as tissue repair and regeneration, cell proliferation, apoptosis, and immunity (Keele et al., 2015). The majority of SNPs associated with pneumonic lesions in this study were in intergenic regions of the sheep genome. Intergenic variants within RFC4, EYA4, and a novel lincRNA were suggestively associated with pneumonic lesions when including all the data and variants within LSAMP, PPIL6, and KCNMA1 reached suggestive significance when only including the extreme animals. Additionally, 2 missense variants in exon 2 of ATAD5 also reached the suggestively significant level in the analysis of the extreme animals. EYA4, ATAD5, and RFC4 all have roles in the response to DNA damage. Eyes Absent (EYA) proteins are implicated in a diverse range of processes, including DNA damage repair and innate immunity (Tadjuidje and Hegde, 2013). EYA4 has been shown to enhance the innate immune response to viruses through stimulating the interferon regulatory factor 3 (IRF3)-mediated transcription of IFN-β and CXCL10 in response to undigested DNA (Okabe et al., 2009). EYA4 has been associated with familial lung cancer risk (Wilson et al., 2014), and a SNP located within 15 kb of EYA4 has been significantly associated with lung lesions in commercial beef cattle (Keele et al., 2015). The replication factor C (RFC) complex, composed of subunits RFC1-5, also plays an essential role in DNA replication and repair in eukaryotes (Kim and MacNeill, 2003). Additionally, several RFC-like complexes (RLC), made up of RFC2-5 and an alternative subunit that replaces RFC1, have been reported, including ATAD5-RLC (Ben-Aroya et al., 2003). Atad5+/− mice show high levels of genomic instability (Bell et al., 2011), and delayed DNA replication and cell division, leading to an altered adaptive immune response though reduced immunoglobulin class switching (Zanotti et al., 2015). The identification of 2 suggestively significant SNPs within genes that form ATAD5-RLC highlights the potential importance of this complex in the host response to respiratory challenge. Although not their primary role, both PPIL6 and KCNMA1 have previously been associated with the respiratory system. A QTL containing the cyclophilin-like PPIL6 was associated with the variability of immune response in a crossbred swine population postinfluenza vaccination (Zanella et al., 2015). The potassium channel gene KCNMA1 was expressed at significantly higher levels in the lungs of asthmatic rats compared to those of control rats (Yin et al., 2008), and is differentially methylated during normal development in the mouse and human lung (Cuna et al., 2015). The significant peak on chromosome 2 associated with pleurisy was detected using 2 independent methods. This peak is located downstream from the transcription factor SP3, which is involved in the activation or suppression the expression of numerous genes, including the interferon regulatory factor IRF3 and IL-10, an anti-inflammatory cytokine (Tone et al., 2000). Of interest is that Sp3 knockout mice die at birth of respiratory failure, although only minor structural abnormalities are observed in the lungs (Bouwman et al., 2000). As mentioned above, IRF3 is involved in the innate response to viral infection (Xu et al., 2012), and several bovine viral pathogens including bovine herpesvirus 1 (BHV-1) and bovine diarrhoea virus (BVDV) target IRF3 activity, halting the interferon response (Srikumaran et al., 2007). As with other farmed ruminants, in sheep, pneumonia is etiologically complex. While Mannheimia haemolytica is considered to be the predominant agent responsible for lung damage, multiple viruses (Davies et al., 1977; Davies et al., 1982; Davies and Jones, 1985) have also been shown to play a role through compromising the respiratory system, allowing secondary invasion by bacteria (Brogden et al., 1998). An enhanced immune response to viruses could therefore result in a reduced chance of developing lung damage. As previously discussed, pneumonia can arise through a combination of a variety of environmental and pathogenic factors. Despite the complex nature of this disease, previous research in both sheep and cattle has shown that there is an underlying genetic component in the variation observed between animals in their susceptibility to pneumonia (Snowder, 2009; McRae et al., 2016). This indicates that it is possible to select for animals with the ability to withstand and/or recover from infection that can be the result of multiple causative factors. This study identified several SNPs associated with genes involved in both the innate immune response and the response to DNA damage that are associated with pneumonic lesions and pleurisy in lambs at slaughter. Additionally, the identification in sheep of several SNPs within genes that had previously been reported to be involved in the respiratory system in cattle, pigs, rats, and mice indicates that there may be common genetic pathways underlying the response to respiratory disease in multiple mammalian species. Footnotes 1 This work was supported by FarmIQ, AgResearch Core, and Beef + Lamb New Zealand Genetics funding. The flocks involved in this study were from FarmIQ and Pastoral Greenhouse Gas Research Consortium (PGgRc) funded progeny test programs. The authors would like to acknowledge the AgResearch Animal Genomics field staff and Silver Fern Farms staff from the Takapau and Finegand processing plants for their help in data collection. The Illumina Ovine Infinium® HD SNP BeadChip was used with the kind permission of the International Sheep Genomics Consortium (www.sheephapmap.org). LITERATURE CITED Alley , M. R . 1987 . The effect of chronic non-progressive pneumonia on weight gain of pasture fed lambs . N Z Vet J . 35 : 163 – 166 . doi: 10.1080/00480169.1987.35429 Google Scholar Crossref Search ADS PubMed WorldCat Alley , M. R . 2002 . Pneumonia in sheep in New Zealand: an overview . N Z Vet J . 50 ( 3 Suppl ): 99 – 101 . doi: 10.1080/ 00480169.2002.36281 Google Scholar Crossref Search ADS PubMed WorldCat Aulchenko , Y. S. , S. Ripke, A. Isaacs, and C. M. van Duijn. 2007 . Genabel: an R library for genome-wide association analysis . Bioinformatics . 23 : 1294 – 1296 . doi: 10.1093/bioinformatics/btm108 Google Scholar Crossref Search ADS PubMed WorldCat Baird , H. J. , S. M. Clarke, and P. L. Johnson. 2012 . Brief communication: development of a visual scoring system for ovine pneumonia at the processing plant . Proc N Z Soc Anim . 72 : 169 – 171 . Google Scholar OpenURL Placeholder Text WorldCat Bell , D. W. , N. Sikdar, K. Y. Lee, J. C. Price, R. Chatterjee, H. D. Park, J. Fox, M. Ishiai, M. L. Rudd, L. M. Pollock, et al. ; NISC Comparative Sequencing Program . 2011 . Predisposition to cancer caused by genetic and functional defects of mammalian atad5 . Plos Genet . 7 : e1002245 . doi: 10.1371/journal.pgen.1002245 Google Scholar Crossref Search ADS PubMed WorldCat Ben-Aroya , S. , A. Koren, B. Liefshitz, R. Steinlauf, and M. Kupiec. 2003 . ELG1, a yeast gene required for genome stability, forms a complex related to replication factor C . Proc. Natl. Acad. Sci. USA . 100 : 9906 – 9911 . doi: 10.1073/pnas.1633757100 Google Scholar Crossref Search ADS WorldCat Bishop , S. C. , and C. A. Morris. 2007 . Genetics of disease resistance in sheep and goats . Small Ruminant Res . 70 : 48 – 59 . doi: 10.1016/j.smallrumres.2007.01.006 Google Scholar Crossref Search ADS WorldCat Bouwman , P. , H. Göllner, H. P. Elsässer, G. Eckhoff, A. Karis, F. Grosveld, S. Philipsen, and G. Suske. 2000 . Transcription factor sp3 is essential for post-natal survival and late tooth development . Embo J . 19 : 655 – 661 . doi: 10.1093/emboj/19.4.655 Google Scholar Crossref Search ADS PubMed WorldCat Brogden , K. A. , H. D. Lehmkuhl, and R. C. Cutlip. 1998 . Pasteurella haemolytica complicated respiratory infections in sheep and goats . Vet. Res . 29 : 233 – 254 . Google Scholar PubMed OpenURL Placeholder Text WorldCat Clarke , S. M. , H. M. Henry, K. G. Dodds, T. W. Jowett, T. R. Manley, R. M. Anderson, and J. C. McEwan. 2014 . A high throughput single nucleotide polymorphism multiplex assay for parentage assignment in New Zealand sheep . Plos One . 9 : e93392 . doi: 10.1371/journal.pone.0093392 Google Scholar Crossref Search ADS PubMed WorldCat Cuna , A. , B. Halloran, O. Faye-Petersen, D. Kelly, D. K. Crossman, X. Cui, K. Pandit, N. Kaminski, S. Bhattacharya, A. Ahmad, et al. 2015 . Alterations in gene expression and DNA methylation during murine and human lung alveolar septation . Am. J. Respir. Cell Mol. Biol . 53 : 60 – 73 . doi: 10.1165/rcmb.2014-0160OC Google Scholar Crossref Search ADS PubMed WorldCat Davies , D. H. , D. L. Dungworth, S. Humphreys, and A. J. Johnson. 1977 . Concurrent infection of lambs with parainfluenza virus type 3 and Pasteurella haemolytica . N. Z. Vet. J . 25 : 263 – 265 . doi: 10.1080/00480169.1977.34425 Google Scholar Crossref Search ADS PubMed WorldCat Davies , G. , S. Genini, S. C. Bishop, and E. Giuffra. 2009 . An assessment of opportunities to dissect host genetic variation in resistance to infectious diseases in livestock . Animal . 3 : 415 – 436 . doi: 10.1017/S1751731108003522 Google Scholar Crossref Search ADS PubMed WorldCat Davies , D. H. , M. Herceg, and D. C. Thurley. 1982 . Experimental infection of lambs with an adenovirus followed by Pasteurella haemolytica . Vet. Microbiol . 7 : 369 – 381 . doi: 10.1016/0378-1135(82)90017-7 Google Scholar Crossref Search ADS PubMed WorldCat Davies , D. H. , and S. Jones. 1985 . Serological evidence of respiratory syncytial virus infection in lambs . N. Z. Vet. J . 33 : 155 – 156 . doi: 10.1080/00480169.1985.35211 Google Scholar Crossref Search ADS PubMed WorldCat Gilmour , A. R. , B. J. Gogel, B. R. Cullis, S. J. Welham, and R. Thompson. 2015 . ASReml User Guide Release 4.1 Structural Specification . VSN International, Hemel Hempstead , HP1 1ES, UK . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Goddard , M. E . 2012 . Uses of genomics in livestock agriculture . Anim Prod Sci . 52 : 73 – 77 (Review). doi: 10.1071/AN11180 Google Scholar Crossref Search ADS WorldCat Goodwin , K. A. , R. Jackson, C. Brown, P. R. Davies, R. S. Morris, and N. R. Perkins. 2004 . Pneumonic lesions in lambs in New Zealand: patterns of prevalence and effects on production . N Z Vet J . 52 : 175 – 179 . doi: 10.1080/00480169.2004.36425 Google Scholar Crossref Search ADS PubMed WorldCat Jiang , Y. , M. Xie, W. Chen, R. T. Talbot, J. F. Maddox, T. Faraut, C. Wu, D. M. Muzny, Y. Li, W. Zhang,et al. 2014 . The sheep genome illuminates biology of the rumen and lipid metabolism . Science . 344 : 1168 – 1173 . doi: 10.1126/science.1252806 Google Scholar Crossref Search ADS PubMed WorldCat Keele , J. W. , L. A. Kuehn, T. G. McDaneld, R. G. Tait, S. A. Jones, T. P. Smith, S. D. Shackelford, D. A. King, T. L. Wheeler, A. K. Lindholm-Perry, et al. 2015 . Genomewide association study of lung lesions in cattle using sample pooling . J. Anim. Sci . 93 : 956 – 964 . doi: 10.2527/jas.2014-8492 Google Scholar Crossref Search ADS PubMed WorldCat Kim , J. , and S. A. MacNeill. 2003 . Genome stability: a new member of the RFC family . Curr. Biol . 13 : R873 – R875 . doi:10.1016/j.cub.2003.10.048 Google Scholar Crossref Search ADS PubMed WorldCat Kirton , A. H. , P. J. O’Hara, E. H. Shortridge, and D. O. Cordes. 1976 . Seasonal incidence of enzootic pneumonia and its effect on the growth of lambs . N Z Vet J . 24 : 59 – 64 . doi: 10.1080/00480169.1976.34284 Google Scholar Crossref Search ADS PubMed WorldCat McRae , K. M. , H. J. Baird, K. G. Dodds, M. J. Bixley, and S. M. Clarke. 2016 . Incidence and heritability of ovine pneumonia, and the relationship with production traits in New Zealand sheep . Small Ruminant Res . 145 : 136 – 141 . doi: 10.1016/j.smallrumres.2016.11.003 Google Scholar Crossref Search ADS WorldCat Neibergs , H. L. , C. M. Seabury, A. J. Wojtowicz, Z. Wang, E. Scraggs, J. N. Kiser, M. Neupane, J. E. Womack, A. Van Eenennaam, G. R. Hagevoort, et al. ; Bovine Respiratory Disease Complex Coordinated Agricultural Project Research Team . 2014 . Susceptibility loci revealed for bovine respiratory disease complex in pre-weaned Holstein calves . BMC Genomics . 15 : 1164 . doi: 10.1186/1471-2164-15-1164 Google Scholar Crossref Search ADS PubMed WorldCat Okabe , Y. , T. Sano, and S. Nagata. 2009 . Regulation of the innate immune response by threonine-phosphatase of eyes absent . Nature . 460 : 520 – 524 . doi: 10.1038/nature08138 Google Scholar Crossref Search ADS PubMed WorldCat Snowder , G . 2009 . Genetics, environment and bovine respiratory disease . Anim. Health Res. Rev . 10 : 117 – 119 . doi: 10.1017/S1466252309990144 Google Scholar Crossref Search ADS PubMed WorldCat Srikumaran , S. , C. L. Kelling, and A. Ambagala. 2007 . Immune evasion by pathogens of bovine respiratory disease complex . Anim. Health Res. Rev . 8 : 215 – 229 . doi: 10.1017/S1466252307001326 Google Scholar Crossref Search ADS PubMed WorldCat Tadjuidje , E. , and R. S. Hegde. 2013 . The eyes absent proteins in development and disease . Cell. Mol. Life Sci . 70 : 1897 – 1913 . doi: 10.1007/s00018-012-1144-9 Google Scholar Crossref Search ADS PubMed WorldCat Tone , M. , M. J. Powell, Y. Tone, S. A. Thompson, and H. Waldmann. 2000 . IL-10 gene expression is controlled by the transcription factors sp1 and sp3 . J. Immunol . 165 : 286 – 291 . Google Scholar Crossref Search ADS PubMed WorldCat Wilson , I. M. , E. A. Vucic, K. S. Enfield, K. L. Thu, Y. A. Zhang, R. Chari, W. W. Lockwood, N. Radulovich, D. T. Starczynowski, J. P. Banáth, et al. 2014 . EYA4 is inactivated biallelically at a high frequency in sporadic lung cancer and is associated with familial lung cancer risk . Oncogene . 33 : 4464 – 4473 . doi: 10.1038/onc.2013.396 Google Scholar Crossref Search ADS PubMed WorldCat Xu , H. G. , R. Jin, W. Ren, L. Zou, Y. Wang, and G. P. Zhou. 2012 . Transcription factors sp1 and sp3 regulate basal transcription of the human IRF-3 gene . Biochimie . 94 : 1390 – 1397 . doi: 10.1016/j.biochi.2012.03.011 Google Scholar Crossref Search ADS PubMed WorldCat Yin , L. M. , G. H. Jiang, Y. Wang, Y. Wang, Y. Y. Liu, W. R. Jin, Z. Zhang, Y. D. Xu, and Y. Q. Yang. 2008 . Serial analysis of gene expression in a rat lung model of asthma . Respirology . 13 : 972 – 982 . doi: 10.1111/j.1440-1843.2008.01398.x Google Scholar PubMed OpenURL Placeholder Text WorldCat Zanella , R. , D. Gava, J. d. e. O. Peixoto, R. Schaefer, J. R. Ciacci-Zanella, N. Biondo, M. V. da Silva, M. E. Cantão, and M. C. Ledur. 2015 . Unravelling the genetic components involved in the immune response of pigs vaccinated against influenza virus . Virus Res . 210 : 327 – 336 . doi: 10.1016/j.virusres.2015.09.003 Google Scholar Crossref Search ADS PubMed WorldCat Zanotti , K. J. , R. W. Maul, D. P. Castiblanco, W. Yang, Y. J. Choi, J. T. Fox, K. Myung, H. Saribasak, and P. J. Gearhart. 2015 . ATAD5 deficiency decreases B cell division and Igh recombination . J. Immunol . 194 : 35 – 42 . doi: 10.4049/jimmunol.1401158 Google Scholar Crossref Search ADS PubMed WorldCat © The Author(s) 2018. Published by Oxford University Press on behalf of the American Society of Animal Science. This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact [email protected] © The Author(s) 2018. Published by Oxford University Press on behalf of the American Society of Animal Science.

Antonio, Torres-Vázquez, José;van der Werf, Julius H J, ;A, Clark, Samuel

doi: 10.1093/jas/sky325pmid: 30124864

Abstract Genetic and phenotypic parameters for feed efficiency, growth, and carcass traits for Australian Angus beef cattle were estimated. Growth traits included birth weight (BWT), 200-d weight (200dWT), 400-d weight (400dWT), and 600-d weight (600dWT). Traits associated with feed efficiency were average daily weight gain (ADG), metabolic midweight, average of daily feed intake (FI), feed conversion ratio (FCR), residual feed intake (RFI), and residual gain (RG). Carcass traits involved were carcass eye muscle area (CEMA), carcass intramuscular fat (IMF), subcutaneous fat depths at the 12th/13th rib (CRIB), rump P8 fat depth (P8FAT), and carcass weight (CWT). For growth traits, heritability estimates ranged from 0.14 ± 0.03 for 200dWT to 0.48 ± 0.06 for 600dWT. For feed efficiency traits, direct heritability estimates for FI, FCR, RFI, and RG were 0.55 ± 0.08, 0.20 ± 0.06, 0.40 ± 0.07, and 0.19 ± 0.06, respectively. High heritability estimates were observed for CEMA, IMF, P8FAT, and CWT of 0.52 ± 0.09, 0.61 ± 0.09, 0.55 ± 0.09, and 0.66 ± 0.09, respectively. Strong positive genetic correlations were found for FI with 200dWT, 400dWT, and 600dWT of 0.68 ± 0.09, 0.42 ± 0.11, and 0.61 ± 0.07, respectively. Weak genetic correlations were observed between RFI and growth traits. For carcass traits, genetic correlations between RFI and CEMA, IMF, CRIB, P8FAT, CWT were −0.19 ± 0.14, 0.31 ± 0.14, 0.18 ± 0.16, 0.24 ± 0.13, and 0.40 ± 0.12, respectively. There was a tendency for low to moderate unfavorable genetic associations between feed efficiency traits, evaluated as RFI and RG, with growth and carcass traits. This implies that selection for RFI would have slight negative impacts on growth and reduce carcass quality. To avoid this, it would be necessary to build selection indices to select feed efficient animals without compromising growth and meat quality. INTRODUCTION The cost of feed is a major expense in beef cattle production systems (Archer et al., 1999). Due to this cost, the efficiency of converting feed into useable animal products, commonly referred to as feed efficiency, has become a common objective in many beef cattle breeding programs (Arthur and Herd, 2008; Berry and Crowley, 2013). Feed efficiency in beef cattle breeding programs has been commonly targeted using residual feed intake (RFI), which is defined as the difference between actual and predicted intake based on its live weight and growth rate over a given period (Koch et al., 1963). In Australia, the interest in feed efficiency has received much attention over the last 2 decades, with many studies designed to better understand the genetic relationships between feed efficiency and other traits in the breeding objective. Previous studies have documented genetic variation between RFI and its component traits, as well as genetic associations between RFI with growth and carcass traits in growing cattle destined for markets that utilize short-grain feeding periods (Archer et al., 1999; Berry and Crowley, 2013). These studies were often constrained by the number of animals tested for all traits due to the high cost of recording such data. Despite the attention that RFI has received, there is not a general consensus relative to the genetic relationship among growth, carcass, and feed efficiency traits (Archer et al., 1999; Hill, 2012; Berry and Crowley, 2013). These genetic correlations strongly affect the ability to select for improved feed efficiency alongside improvement of growth and carcass quality. Consequently, the objective of this study was to estimate genetic and phenotypic parameters for feed efficiency traits with growth and carcass traits under a long-grain finishing feedlot regime for an Australian Angus cattle population. MATERIALS AND METHODS Animal Care Animal records for feedlot based traits were collected with animal ethics approval AEC12-082. Data for growth and carcass traits for the animals used in this study were provided by the Angus Society of Australia (ASA) which were collected as part of routine commercial animal mamangement and, therefore, were not subject to animal care and animal ethics committee approval. Data All phenotypic data were collected on a group of Angus steers and heifers from the Angus Sire Benchmarking Program (ASBP, also known as the Angus Beef Information Nucleus). This structured dataset represented a progeny test of registered Angus sires from herds located in New South Wales and Victoria, Australia (Banks, 2011). Growth data included records for 6,371 animals, born between 2011 and 2016, for birth weight (BWT), 200-d weight (200dWT), 400-d weight (400dWT), and 600-d weight (600dWT). Feed efficiency data were collected on 2,220 Angus steers of the ASBP, from 2013 to 2017 at Tullimba research feedlot (30°20′S, 151°10′E, altitude 560 m) near Kingstown, NSW, Australia. Initially, animals’ age ranged from 500 to 600 d and animals weighed an average of 578 kg. Upon entry to the feedlot, animals were fed with a conditioned diet for 21-d period and they were subsequently measured for body weight (BW, kg; fortnightly) and daily feed intake (FI) over an additional 70-d (approximately) test period. During the test period, animals had ad libitum access to a full mixed ration composed of 74.8% tempered barley, 4.6% cotton hulls, 6% cottonseed, 5% mill run, 4.6% chopped hay, and 5% liquid mineral supplement. Daily individual FI (kg/d) was measured using the GrowSafe automatic feeding system installed at Tullimba, Kingstown, NSW, Australia (GrowSafe Systems Ltd., Airdrie, Alberta, Canada). The automatic feeding system was described and validated by Basarab et al. (2002, 2003). During the test period, average daily weight gain (ADG; kg/d) was calculated as the regression of weight on time (d), whereas metabolic weight at the midpoint of the test period (MMWT) was obtained as the midpoint BW raised to the 0.73 power (BW0.73) (Arthur et al., 2001b; Berry and Crowley, 2013). Feed conversion ratio (FCR) was obtained as the average daily FI divided by the ADG. RFI (kg/d) was estimated as the residual of regressing FI on ADG and MMWT with contemporary group (CG) included in the model, where the CG effect was defined as the concatenation of herd, year of birth, birth type (single or twin), breeder-defined management group, observation date, and age. Similarly, residual gain (RG; kg/d) was estimated as the residual of regressing ADG on FI and MMWT in the model that included the CG effect as defined previously (Arthur and Herd, 2008; Berry and Crowley, 2012; Berry and Crowley, 2013). On completion of FI testing, the steers were feed for a further ~180 d to give a total grain finishing period of ~270 d as described by Duff et al. (2018). At the completion of the long-grain finishing period, carcass traits were recorded. Measured traits included carcass eye muscle area (CEMA), intramuscular fat (IMF), subcutaneous fat depth at the 12th/13th rib (CRIB), rump P8 fat depth (P8FAT), and carcass weight (CWT), with an average age at slaughter of 793 d for all traits. CGs for each trait were formed according to the BREEDPLAN format, which in this case included herd, year of birth, sex, birth type (single or twin), breeder-defined management group, trial, day of measurement, and age, with animals subdivided by age in slices of 45 d (Graser et al., 2005). Duplicated records and CG of less than 5 animals or with incomplete information were eliminated. The final number of CG for each trait was 138 for BWT and 200dWT; 85 for 400dWT; 108 for 600dWT; 80 for ADG, MMWT, FI, FCR, RFI, and RG; and 51 for CEMA, IMF, CRIB, P8FAT, and CWT. Descriptive statistics of the data after editing are in Table 1. Ranges of age for 200dWT, 400dWT, and 600dWT were 97 to 290, 291 to 492, and 493 to 820 d, respectively, with averages of 188, 404, and 552 d, respectively. The final pedigree file included ancestors over 14 generations with a total of 14,662 animals involving 1,454 sires and 7,835 dams; with 232 sires and 4,341 dams having progeny with phenotypic records. Table 1. Descriptive statistics for feed efficiency, growth, and carcass traits Trait n Mean Min1 Max2 SD CV, % Birth weight, kg 5,920 38.04 18.00 61.00 5.23 13.75 200-d weight, kg 5,764 229.37 88.00 394.00 44.49 19.40 400-d weight, kg 3,204 371.27 206.00 548.00 59.90 16.13 600-d weight, kg 3,513 511.81 289.00 882.00 106.83 20.87 Average daily gain, kg d 1,998 1.59 0.44 3.07 0.35 22.03 Metabolic midweight, kg 1,998 103.78 87.45 121.73 6.18 5.96 Feed intake, kg d 1,998 14.90 9.24 20.76 1.88 12.59 Feed conversion ratio 1,998 9.78 4.82 33.78 2.24 22.87 Residual feed intake, kg d 1,998 0.00 −5.62 4.08 1.11 – Residual gain, kg d 1,998 0.00 −0.97 1.11 0.24 – Carcass eye muscle area, mm2 1,634 90.23 66.00 124.00 9.31 10.32 Carcass intramuscular fat, % 1,382 10.02 3.20 25.10 3.25 32.38 Fat depths at the 12th/13th rib, mm 1,612 18.89 7.00 40.00 5.44 28.78 Rump P8 fat depth, mm 1,636 22.81 9.00 43.00 6.14 26.92 Carcass weight, kg 1,640 458.08 338.10 568.60 36.53 7.97 Trait n Mean Min1 Max2 SD CV, % Birth weight, kg 5,920 38.04 18.00 61.00 5.23 13.75 200-d weight, kg 5,764 229.37 88.00 394.00 44.49 19.40 400-d weight, kg 3,204 371.27 206.00 548.00 59.90 16.13 600-d weight, kg 3,513 511.81 289.00 882.00 106.83 20.87 Average daily gain, kg d 1,998 1.59 0.44 3.07 0.35 22.03 Metabolic midweight, kg 1,998 103.78 87.45 121.73 6.18 5.96 Feed intake, kg d 1,998 14.90 9.24 20.76 1.88 12.59 Feed conversion ratio 1,998 9.78 4.82 33.78 2.24 22.87 Residual feed intake, kg d 1,998 0.00 −5.62 4.08 1.11 – Residual gain, kg d 1,998 0.00 −0.97 1.11 0.24 – Carcass eye muscle area, mm2 1,634 90.23 66.00 124.00 9.31 10.32 Carcass intramuscular fat, % 1,382 10.02 3.20 25.10 3.25 32.38 Fat depths at the 12th/13th rib, mm 1,612 18.89 7.00 40.00 5.44 28.78 Rump P8 fat depth, mm 1,636 22.81 9.00 43.00 6.14 26.92 Carcass weight, kg 1,640 458.08 338.10 568.60 36.53 7.97 1Min is the minimum value. 2Max is the maximum value. View Large Table 1. Descriptive statistics for feed efficiency, growth, and carcass traits Trait n Mean Min1 Max2 SD CV, % Birth weight, kg 5,920 38.04 18.00 61.00 5.23 13.75 200-d weight, kg 5,764 229.37 88.00 394.00 44.49 19.40 400-d weight, kg 3,204 371.27 206.00 548.00 59.90 16.13 600-d weight, kg 3,513 511.81 289.00 882.00 106.83 20.87 Average daily gain, kg d 1,998 1.59 0.44 3.07 0.35 22.03 Metabolic midweight, kg 1,998 103.78 87.45 121.73 6.18 5.96 Feed intake, kg d 1,998 14.90 9.24 20.76 1.88 12.59 Feed conversion ratio 1,998 9.78 4.82 33.78 2.24 22.87 Residual feed intake, kg d 1,998 0.00 −5.62 4.08 1.11 – Residual gain, kg d 1,998 0.00 −0.97 1.11 0.24 – Carcass eye muscle area, mm2 1,634 90.23 66.00 124.00 9.31 10.32 Carcass intramuscular fat, % 1,382 10.02 3.20 25.10 3.25 32.38 Fat depths at the 12th/13th rib, mm 1,612 18.89 7.00 40.00 5.44 28.78 Rump P8 fat depth, mm 1,636 22.81 9.00 43.00 6.14 26.92 Carcass weight, kg 1,640 458.08 338.10 568.60 36.53 7.97 Trait n Mean Min1 Max2 SD CV, % Birth weight, kg 5,920 38.04 18.00 61.00 5.23 13.75 200-d weight, kg 5,764 229.37 88.00 394.00 44.49 19.40 400-d weight, kg 3,204 371.27 206.00 548.00 59.90 16.13 600-d weight, kg 3,513 511.81 289.00 882.00 106.83 20.87 Average daily gain, kg d 1,998 1.59 0.44 3.07 0.35 22.03 Metabolic midweight, kg 1,998 103.78 87.45 121.73 6.18 5.96 Feed intake, kg d 1,998 14.90 9.24 20.76 1.88 12.59 Feed conversion ratio 1,998 9.78 4.82 33.78 2.24 22.87 Residual feed intake, kg d 1,998 0.00 −5.62 4.08 1.11 – Residual gain, kg d 1,998 0.00 −0.97 1.11 0.24 – Carcass eye muscle area, mm2 1,634 90.23 66.00 124.00 9.31 10.32 Carcass intramuscular fat, % 1,382 10.02 3.20 25.10 3.25 32.38 Fat depths at the 12th/13th rib, mm 1,612 18.89 7.00 40.00 5.44 28.78 Rump P8 fat depth, mm 1,636 22.81 9.00 43.00 6.14 26.92 Carcass weight, kg 1,640 458.08 338.10 568.60 36.53 7.97 1Min is the minimum value. 2Max is the maximum value. View Large Statistical Analyses For each trait, an optimal model was derived by testing the significance of fixed effects. The fixed effects tested were CG for each trait as well as linear and quadratic covariates of age (except BWT) and dam age; for carcass traits (other than CWT), linear and quadratic covariates of CWT were also tested. Analyses of the fixed effects were done with JMP version 14 software package (SAS Institute Inc., Cary, NC). Variance components and heritabilities were estimated with univariate animal models. Models for the analysis included the significant fixed effects shown in Table 2. For all traits, a direct additive genetic effect was included as a random effect; and for BWT and 200dWT, a maternal genetic effect was also included as a random effect. Genetic and phenotypic correlations were estimated using bivariate models with similar fixed (Table 2) and random effects as the univariate models. Univariate and bivariate models were analyzed using ASReml software (Gilmour et al., 2009). Table 2. Significant fixed effects for growth, feed efficiency, and carcass quality traits Fix effect1 Trait CG Age Age2 Dam Dam2 CWT CWT2 Birth weight, kg <0.001*** <0.001*** <0.001*** 200-d weight, kg <0.001*** <0.001*** <0.001*** <0.001*** <0.001*** 400-d weight, kg <0.001*** <0.001*** 0.022* 0.012* 0.011* 600-d weight, kg <0.001*** <0.001*** <0.001*** <0.001*** Average daily gain, kg d <0.001*** Metabolic midweight, kg <0.001*** Feed intake, kg d <0.001*** Feed conversion ratio <0.001*** Residual feed intake, kg d <0.001*** Residual gain, kg d <0.001*** Carcass eye muscle area, mm2 <0.001*** 0. 018* 0.007** Carcass intramuscular fat, % <0.001*** 0.044* Fat depths at the 12th/13th rib, mm <0.001*** 0.004** 0.002** 0.006** Rump P8 fat depth, mm <0.001*** 0.003** Carcass weight, kg <0.001*** Fix effect1 Trait CG Age Age2 Dam Dam2 CWT CWT2 Birth weight, kg <0.001*** <0.001*** <0.001*** 200-d weight, kg <0.001*** <0.001*** <0.001*** <0.001*** <0.001*** 400-d weight, kg <0.001*** <0.001*** 0.022* 0.012* 0.011* 600-d weight, kg <0.001*** <0.001*** <0.001*** <0.001*** Average daily gain, kg d <0.001*** Metabolic midweight, kg <0.001*** Feed intake, kg d <0.001*** Feed conversion ratio <0.001*** Residual feed intake, kg d <0.001*** Residual gain, kg d <0.001*** Carcass eye muscle area, mm2 <0.001*** 0. 018* 0.007** Carcass intramuscular fat, % <0.001*** 0.044* Fat depths at the 12th/13th rib, mm <0.001*** 0.004** 0.002** 0.006** Rump P8 fat depth, mm <0.001*** 0.003** Carcass weight, kg <0.001*** 1CG = contemporary group effect; Age = age in days when the trait was measured; Age2 = age of the measurement squared; Dam = age of the dam in days; Dam2 = age of the dam squared; CWT = carcass weight, kg; CWT2 = carcass weight squared, kg2. *P < 0.05; **P < 0.01; ***P < 0.001. View Large Table 2. Significant fixed effects for growth, feed efficiency, and carcass quality traits Fix effect1 Trait CG Age Age2 Dam Dam2 CWT CWT2 Birth weight, kg <0.001*** <0.001*** <0.001*** 200-d weight, kg <0.001*** <0.001*** <0.001*** <0.001*** <0.001*** 400-d weight, kg <0.001*** <0.001*** 0.022* 0.012* 0.011* 600-d weight, kg <0.001*** <0.001*** <0.001*** <0.001*** Average daily gain, kg d <0.001*** Metabolic midweight, kg <0.001*** Feed intake, kg d <0.001*** Feed conversion ratio <0.001*** Residual feed intake, kg d <0.001*** Residual gain, kg d <0.001*** Carcass eye muscle area, mm2 <0.001*** 0. 018* 0.007** Carcass intramuscular fat, % <0.001*** 0.044* Fat depths at the 12th/13th rib, mm <0.001*** 0.004** 0.002** 0.006** Rump P8 fat depth, mm <0.001*** 0.003** Carcass weight, kg <0.001*** Fix effect1 Trait CG Age Age2 Dam Dam2 CWT CWT2 Birth weight, kg <0.001*** <0.001*** <0.001*** 200-d weight, kg <0.001*** <0.001*** <0.001*** <0.001*** <0.001*** 400-d weight, kg <0.001*** <0.001*** 0.022* 0.012* 0.011* 600-d weight, kg <0.001*** <0.001*** <0.001*** <0.001*** Average daily gain, kg d <0.001*** Metabolic midweight, kg <0.001*** Feed intake, kg d <0.001*** Feed conversion ratio <0.001*** Residual feed intake, kg d <0.001*** Residual gain, kg d <0.001*** Carcass eye muscle area, mm2 <0.001*** 0. 018* 0.007** Carcass intramuscular fat, % <0.001*** 0.044* Fat depths at the 12th/13th rib, mm <0.001*** 0.004** 0.002** 0.006** Rump P8 fat depth, mm <0.001*** 0.003** Carcass weight, kg <0.001*** 1CG = contemporary group effect; Age = age in days when the trait was measured; Age2 = age of the measurement squared; Dam = age of the dam in days; Dam2 = age of the dam squared; CWT = carcass weight, kg; CWT2 = carcass weight squared, kg2. *P < 0.05; **P < 0.01; ***P < 0.001. View Large In matrix notation, univariate animal models for 400dWT, 600dWT, feed efficiency, and carcass traits can be represented as follows: y = Xb + Z1u + e (1) For BWT and 200dWT, the univariate maternal effect models can be represented as follows: y = Xb + Z1u + Z2m+ e (2) where y is the vector of the phenotypes for the traits; b is the vector of fixed effects for the analyzed traits; u is the vector which contains animal random effects; m is the vector of random maternal genetic effects of the dams (model 2); X and Z1 are the incidence matrices relating observations to fixed and animal effects, respectively; Z2 is the incidence matrix relating observations to maternal effects (model 2); and e is the vector of residual effects for the analyzed traits. Model 2 assumed that direct and maternal genetic effects were uncorrelated, i.e., σAM = 0. For model 1, the expectations and variance matrices for random vectors are described as follows: E[ yue ]= [ Xb00 ];  V[ ue ] The expectations and variance matrices for random vectors in model 2 are described as follows: E[ yume ]= [ Xb000 ];  V[ ume ] The bivariate animal models involving feed efficiency and carcass traits can be represented as follows: Y = Xb + Z1u + e (3) Bivariate models involving BWT and 200dWT can be represented as follows: Y = Xb + Z1u + Z2m+ e (4) where Y is the vector of records for the traits; b is the vector of fixed effects for the analyzed traits; u is the vector which contains animal random effects; m is the vector of random maternal genetic effects of the dams (model 4); X and Z1 are the incidence matrices relating observations to fixed and animal effects, respectively; Z2 is the incidence matrix relating observations to maternal effects (model 4); and e is the vector of residual effects for the analyzed traits. Model 4 assumed that direct and maternal genetic effects were uncorrelated, i.e., σAM = 0. For model 3, the expectations and variance matrices for random vectors are described as follows: E[ Yue ]= [ Xb00 ]; V[ ue ]= [ GR ]= [ A⊗Gu0 0 I0⊗R] For model 4, the expectations and variance matrices for random vectors are described as follows: E [ Yume ]= [ Xb000 ]; V[ ume ] = [ Gu00 0Gm0 00R ]= [ A⊗Gu00 A⊗0Gm0 00 I0⊗R ] where Gu, Gm, and R denote 2 × 2 matrices containing additive genetic, maternal genetic (model 4), and residual (co) variance components, respectively; A is the numerator relationship matrix; I0 is an identity matrix for the total number of observations; and ⊗ is the Kronecker product. RESULTS AND DISCUSSION Descriptive statistics for the studied traits are summarized in Table 1. Variance components and heritability estimates are presented in Table 3. In this study, high estimates of heritability were observed for carcass traits (from 0.34 to 0.66) compared with growth (from 0.14 to 0.48) and feed efficiency traits (from 0.19 to 0.55). Table 3. Variance component and heritability estimates (SE) using univariate models for growth, feed efficiency, and carcass traits in Angus cattle Parameter1 Trait2 σ2a σ2m σ2e σ2p h2a h2m BWT 6.25 ± 0.90 2.37 ± 0.47 10.38 ± 0.63 19.00 ± 0.40 0.33 ± 0.04 0.12 ± 0.02 200dWT 63.48 ± 12.81 111.95 ± 11.37 276.18 ± 12.57 451.61 ± 9.01 0.14 ± 0.03 0.25 ± 0.02 400dWT 219.70 ± 42.25 – 622.74 ± 38.58 842.44 ± 22.36 0.26 ± 0.05 – 600dWT 603.63 ± 81.59 – 655.99 ± 65.58 1,259.60 ± 34.78 0.48 ± 0.06 – ADG 0.03 ± 0.01 – 0.05 ± 0.01 0.08 ± 0.00 0.33 ± 0.07 – MMWT 10.28 ± 1.81 – 11.83 ± 1.53 22.11 ± 0.77 0.46 ± 0.07 – FI 1.14 ± 0.18 – 0.94 ± 0.15 2.07 ± 0.07 0.55 ± 0.08 – FCR 0.71 ± 0.22 – 2.83 ± 0.22 3.54 ± 0.12 0.20 ± 0.06 – RFI 0.52 ± 0.10 – 0.78 ± 0.09 1.30 ± 0.04 0.40 ± 0.07 – RG 0.01 ± 0.00 – 0.05 ± 0.00 0.06 ± 0.00 0.19 ± 0.06 – CEMA 31.34 ± 5.87 – 28.63 ± 4.85 59.97 ± 2.36 0.52 ± 0.09 – IMF 5.60 ± 0.99 – 3.56 ± 0.80 9.16 ± 0.40 0.61 ± 0.09 – CRIB 8.83 ± 2.22 – 16.87 ± 1.97 25.69 ± 0.98 0.34 ± 0.08 – P8FAT 16.62 ± 3.01 – 13.63 ± 2.46 30.25 ± 1.20 0.55 ± 0.09 – CWT 740.85 ± 122 – 384.22 ± 96.93 1,125.10 ± 46.15 0.66 ± 0.09 – Parameter1 Trait2 σ2a σ2m σ2e σ2p h2a h2m BWT 6.25 ± 0.90 2.37 ± 0.47 10.38 ± 0.63 19.00 ± 0.40 0.33 ± 0.04 0.12 ± 0.02 200dWT 63.48 ± 12.81 111.95 ± 11.37 276.18 ± 12.57 451.61 ± 9.01 0.14 ± 0.03 0.25 ± 0.02 400dWT 219.70 ± 42.25 – 622.74 ± 38.58 842.44 ± 22.36 0.26 ± 0.05 – 600dWT 603.63 ± 81.59 – 655.99 ± 65.58 1,259.60 ± 34.78 0.48 ± 0.06 – ADG 0.03 ± 0.01 – 0.05 ± 0.01 0.08 ± 0.00 0.33 ± 0.07 – MMWT 10.28 ± 1.81 – 11.83 ± 1.53 22.11 ± 0.77 0.46 ± 0.07 – FI 1.14 ± 0.18 – 0.94 ± 0.15 2.07 ± 0.07 0.55 ± 0.08 – FCR 0.71 ± 0.22 – 2.83 ± 0.22 3.54 ± 0.12 0.20 ± 0.06 – RFI 0.52 ± 0.10 – 0.78 ± 0.09 1.30 ± 0.04 0.40 ± 0.07 – RG 0.01 ± 0.00 – 0.05 ± 0.00 0.06 ± 0.00 0.19 ± 0.06 – CEMA 31.34 ± 5.87 – 28.63 ± 4.85 59.97 ± 2.36 0.52 ± 0.09 – IMF 5.60 ± 0.99 – 3.56 ± 0.80 9.16 ± 0.40 0.61 ± 0.09 – CRIB 8.83 ± 2.22 – 16.87 ± 1.97 25.69 ± 0.98 0.34 ± 0.08 – P8FAT 16.62 ± 3.01 – 13.63 ± 2.46 30.25 ± 1.20 0.55 ± 0.09 – CWT 740.85 ± 122 – 384.22 ± 96.93 1,125.10 ± 46.15 0.66 ± 0.09 – 1σ2a = additive genetic variance; σ2m = maternal genetic variance; σ2e = residual variance; σ2p = phenotypic variance; h2a = additive heritability; h2m = maternal heritability. 2BWT = birth weight; 200dWT = 200-d weight; 400dWT = 400-d weight; 600dWT = 600-d weight; ADG = average daily gain; MMWT = metabolic midweight; FI = feed intake; FCR = feed conversion ratio; RFI = residual feed intake; RG = residual gain; CEMA = carcass eye muscle area; IMF = carcass intramuscular fat; CRIB = subcutaneous fat depths at the 12th/13th rib; P8FAT = rump P8 fat depth; CWT = carcass weight. View Large Table 3. Variance component and heritability estimates (SE) using univariate models for growth, feed efficiency, and carcass traits in Angus cattle Parameter1 Trait2 σ2a σ2m σ2e σ2p h2a h2m BWT 6.25 ± 0.90 2.37 ± 0.47 10.38 ± 0.63 19.00 ± 0.40 0.33 ± 0.04 0.12 ± 0.02 200dWT 63.48 ± 12.81 111.95 ± 11.37 276.18 ± 12.57 451.61 ± 9.01 0.14 ± 0.03 0.25 ± 0.02 400dWT 219.70 ± 42.25 – 622.74 ± 38.58 842.44 ± 22.36 0.26 ± 0.05 – 600dWT 603.63 ± 81.59 – 655.99 ± 65.58 1,259.60 ± 34.78 0.48 ± 0.06 – ADG 0.03 ± 0.01 – 0.05 ± 0.01 0.08 ± 0.00 0.33 ± 0.07 – MMWT 10.28 ± 1.81 – 11.83 ± 1.53 22.11 ± 0.77 0.46 ± 0.07 – FI 1.14 ± 0.18 – 0.94 ± 0.15 2.07 ± 0.07 0.55 ± 0.08 – FCR 0.71 ± 0.22 – 2.83 ± 0.22 3.54 ± 0.12 0.20 ± 0.06 – RFI 0.52 ± 0.10 – 0.78 ± 0.09 1.30 ± 0.04 0.40 ± 0.07 – RG 0.01 ± 0.00 – 0.05 ± 0.00 0.06 ± 0.00 0.19 ± 0.06 – CEMA 31.34 ± 5.87 – 28.63 ± 4.85 59.97 ± 2.36 0.52 ± 0.09 – IMF 5.60 ± 0.99 – 3.56 ± 0.80 9.16 ± 0.40 0.61 ± 0.09 – CRIB 8.83 ± 2.22 – 16.87 ± 1.97 25.69 ± 0.98 0.34 ± 0.08 – P8FAT 16.62 ± 3.01 – 13.63 ± 2.46 30.25 ± 1.20 0.55 ± 0.09 – CWT 740.85 ± 122 – 384.22 ± 96.93 1,125.10 ± 46.15 0.66 ± 0.09 – Parameter1 Trait2 σ2a σ2m σ2e σ2p h2a h2m BWT 6.25 ± 0.90 2.37 ± 0.47 10.38 ± 0.63 19.00 ± 0.40 0.33 ± 0.04 0.12 ± 0.02 200dWT 63.48 ± 12.81 111.95 ± 11.37 276.18 ± 12.57 451.61 ± 9.01 0.14 ± 0.03 0.25 ± 0.02 400dWT 219.70 ± 42.25 – 622.74 ± 38.58 842.44 ± 22.36 0.26 ± 0.05 – 600dWT 603.63 ± 81.59 – 655.99 ± 65.58 1,259.60 ± 34.78 0.48 ± 0.06 – ADG 0.03 ± 0.01 – 0.05 ± 0.01 0.08 ± 0.00 0.33 ± 0.07 – MMWT 10.28 ± 1.81 – 11.83 ± 1.53 22.11 ± 0.77 0.46 ± 0.07 – FI 1.14 ± 0.18 – 0.94 ± 0.15 2.07 ± 0.07 0.55 ± 0.08 – FCR 0.71 ± 0.22 – 2.83 ± 0.22 3.54 ± 0.12 0.20 ± 0.06 – RFI 0.52 ± 0.10 – 0.78 ± 0.09 1.30 ± 0.04 0.40 ± 0.07 – RG 0.01 ± 0.00 – 0.05 ± 0.00 0.06 ± 0.00 0.19 ± 0.06 – CEMA 31.34 ± 5.87 – 28.63 ± 4.85 59.97 ± 2.36 0.52 ± 0.09 – IMF 5.60 ± 0.99 – 3.56 ± 0.80 9.16 ± 0.40 0.61 ± 0.09 – CRIB 8.83 ± 2.22 – 16.87 ± 1.97 25.69 ± 0.98 0.34 ± 0.08 – P8FAT 16.62 ± 3.01 – 13.63 ± 2.46 30.25 ± 1.20 0.55 ± 0.09 – CWT 740.85 ± 122 – 384.22 ± 96.93 1,125.10 ± 46.15 0.66 ± 0.09 – 1σ2a = additive genetic variance; σ2m = maternal genetic variance; σ2e = residual variance; σ2p = phenotypic variance; h2a = additive heritability; h2m = maternal heritability. 2BWT = birth weight; 200dWT = 200-d weight; 400dWT = 400-d weight; 600dWT = 600-d weight; ADG = average daily gain; MMWT = metabolic midweight; FI = feed intake; FCR = feed conversion ratio; RFI = residual feed intake; RG = residual gain; CEMA = carcass eye muscle area; IMF = carcass intramuscular fat; CRIB = subcutaneous fat depths at the 12th/13th rib; P8FAT = rump P8 fat depth; CWT = carcass weight. View Large Heritability estimates for growth traits ranged from 0.14 for 200dWT to 0.48 for 600dWT. For BWT, both direct (0.33 ± 0.04) and maternal genetic (0.12 ± 0.02) were similar to the ranges of 0.34 to 0.52, and 0.07 to 0.13, respectively, published by Meyer (1992) in Australian Angus cattle. However, the direct heritability estimate for 200dWT (0.14 ± 0.03) was lower than the range (of 0.19 to 0.44) documented by Meyer (1992). The maternal heritability estimate for 200dWT in this study (0.25 ± 0.02) was higher than recent estimates in beef cattle (Torres-Vázquez and Spangler, 2016). A possible reason for the high maternal heritability estimate could be the amount of information available for dams in this dataset. There was a lack of pedigree information on the dams and dams rarely had more than one progeny which made it difficult to separate maternal genetic and environmental components. Despite the problems with the structure of the data, some other authors have published similar maternal heritabilities in beef cattle consistent with the finding of this study (Trus and Wilton, 1988; Hetzel et al., 1990). The estimate of direct heritability for 400dWT (0.26 ± 0.05) was in the range of 0.21 to 0.31 published for several authors and summarized by Meyer (1992) in different beef cattle populations; however, our estimate for 600dWT (0.48 ± 0.06) was high when compared with the range summarized (0.26 to 0.43) for this trait. The higher heritability estimate for 600dWT observed in this study was more in line with papers by Meyer (2005) and Jeyaruban et al. (2009) (of 0.44 and 0.40, respectively), which illustrate that estimates of variance components can substantially change over time and therefore need to be reestimated as the population changes. The heritability estimated for ADG (0.33 ± 0.07), FCR (0.20 ± 0.06), and RFI (0.40 ± 0.07) (Table 3) was similar to the pooled heritability estimates published by Berry and Crowley (2013), using a meta-analysis of 39 scientific publications on feed efficiency traits in growing animals. Furthermore, the heritability estimate for MMWT (of 0.46 ± 0.07) in this study was similar to the estimate of 0.40 ± 0.02 provided by Arthur et al. (2001b) for Australian Angus cattle. Our estimate for FI (0.55 ± 0.08) was higher than previously estimated by Arthur et al. (2001a) of 0.39 in Australian Angus cattle and more recently by Mao et al. (2013) of 0.39 in a Canadian Angus population. In our study, FI was more heritable than RFI which is in contrast to previous studies (Arthur et al., 2001b; Mao et al., 2013). Interestingly, the heritability estimates for RG (0.19 ± 0.06) was lower than the mean heritability estimate of 0.28 ± 0.03 published by Berry and Crowley (2013) in growing cattle. A possible explanation for the differences observed between the current study and those previously mentioned is animals in this study that were substantially older (~520 d) than those utilized by Arthur et al. (2001b) (~268 d) and Mao et al. (2013) (~330 d) and those summarized by Berry and Crowley (2013). The differences in heritability estimates confirm that feed efficiency traits can change over an animal’s lifetime (i.e., feed efficiency traits measured at postweaning are different from late stage feedlot and cow feed efficiency) as suggested by Arthur et al. (2004). Furthermore, maternal and permanent environmental variance components for feed efficiency traits were not evident for this study because animals were measured at a later age than most previous studies. In the current study, estimates of both maternal genetic and maternal permanent environmental components for feed efficiency traits were zero, which is in agreement with the results from several other authors suggesting that the maternal components for feed efficiency traits are low (Hoque et al., 2007; Crowley et al., 2010). Carcass traits in the present work were moderately to highly heritable, which is in agreement with previous work in beef cattle (Meyer et al., 2004; Rios-Utrera and Van Vleck, 2004). In general, heritability estimates for carcass traits were higher than those estimated for growth and feed efficiency traits. Our heritability estimates for CEMA (0.52 ± 0.09) and CRIB (0.34 ± 0.08) were close to the estimate of 0.49 ± 0.14 and 0.35 ± 0.12, respectively, published by Mao et al. (2013) in Angus and Charolais steers. Furthermore, the estimates for CEMA were similar to the estimates reported by Meyer et al. (2004) in Australian Hereford cattle (0.59 to 0.67). The heritability estimate for CRIB (0.34 ± 0.08) was close to the estimate published by Reverter et al. (2003), of 0.41, in tropically adapted beef breeds in Australia, and Meyer et al. (2004), from 0.25 to 0.31, in Australian Hereford bulls. This illustrates that for traits such as CEMA- and CRIB-estimated heritabilities seem to be relatively similar across breed types, the sex of the animal, and age of measurement. The heritability estimate for P8FAT (0.55 ± 0.09) was above the range of 0.20 to 0.30 Meyer et al. (2004), and Reverter et al. (2003) in temperate (0.36) and tropically adapted breeds (0.30), respectively. The estimate for IMF (0.61 ± 0.09) was higher than what was recently estimated in Angus animals by Mao et al. (2013) (0.37 ± 0.11) and also higher that those estimated by several authors in different beef cattle populations (Reverter et al., 2003; Meyer et al., 2004; Mateescu et al., 2015). Beef cattle in Australia are produced in 3 major finishing systems: 1) grass fed beef production with no grain feeding; 2) short period of grain finishing (<150 d on feed); and 3) long periods of grain finishing (>200 d on feed). This last period of grain finishing is generally associated with high-value markets that require a premium meat quality and large carcasses. Carcass traits in this study were measured after a long feeding period (250 to 270 d) with heavy animals at slaughter (458 kg) and perhaps animals could express more genetic variation compared with those fed for a short period (for 140 d) or pasture-based trials as observed in previous studies. Genetic and phenotypic correlations estimated between growth traits are shown in Table 4. In the present work, the genetic correlations between BWT with the other growth traits were moderated with a range from 0.50 to 0.53, and between the other growth traits these correlations were higher (from 0.92 to 0.96). Genetic correlations, between growth traits, abound in the scientific literature and tend to be from moderate to high for different beef cattle populations, suggesting that the expression of these growth traits would be determined for the same group of genes (Meyer, 1992; Davis, 1993; Koots et al., 1994). Table 4. Estimates of genetic (above diagonal) and phenotypic (below diagonal) correlations with their standard errors between growth traits Trait1 BWT 200dWT 400dWT 600dWT BWT 0.50 ± 0.10 0.53 ± 0.08 0.53 ± 0.07 200dWT 0.37 ± 0.01 0.96 ± 0.03 0.92 ± 0.03 400dWT 0.36 ± 0.02 0.70 ± 0.01 0.92 ± 0.03 600dWT 0.39 ± 0.02 0.64 ± 0.01 0.77 ± 0.01 Trait1 BWT 200dWT 400dWT 600dWT BWT 0.50 ± 0.10 0.53 ± 0.08 0.53 ± 0.07 200dWT 0.37 ± 0.01 0.96 ± 0.03 0.92 ± 0.03 400dWT 0.36 ± 0.02 0.70 ± 0.01 0.92 ± 0.03 600dWT 0.39 ± 0.02 0.64 ± 0.01 0.77 ± 0.01 1BWT = birth weight; 200dWT = 200-d weight; 400dWT = 400-d weight; 600dWT = 600-d weight. View Large Table 4. Estimates of genetic (above diagonal) and phenotypic (below diagonal) correlations with their standard errors between growth traits Trait1 BWT 200dWT 400dWT 600dWT BWT 0.50 ± 0.10 0.53 ± 0.08 0.53 ± 0.07 200dWT 0.37 ± 0.01 0.96 ± 0.03 0.92 ± 0.03 400dWT 0.36 ± 0.02 0.70 ± 0.01 0.92 ± 0.03 600dWT 0.39 ± 0.02 0.64 ± 0.01 0.77 ± 0.01 Trait1 BWT 200dWT 400dWT 600dWT BWT 0.50 ± 0.10 0.53 ± 0.08 0.53 ± 0.07 200dWT 0.37 ± 0.01 0.96 ± 0.03 0.92 ± 0.03 400dWT 0.36 ± 0.02 0.70 ± 0.01 0.92 ± 0.03 600dWT 0.39 ± 0.02 0.64 ± 0.01 0.77 ± 0.01 1BWT = birth weight; 200dWT = 200-d weight; 400dWT = 400-d weight; 600dWT = 600-d weight. View Large Genetic and phenotypic correlations among measures of feed efficiency and growth traits are summarized in Tables 5 and 6, respectively. As expected, RFI and RG were phenotypically independent of the components production traits of ADG and MMWT, and FI and MMWT, respectively (Table 5). For RFI, there were positive, unfavorable genetic associations with ADG (0.34) and MMWT (0.18). These findings are contrary to those found by Arthur et al. (2001b), who found weak negative, favorable associations between RFI and its component traits. Furthermore, the genetic correlation between RFI and FI (0.83) was stronger than the estimate of 0.51 published by Mao et al. (2013). The correlations estimated in this study suggest that although RFI is phenotypically independent of ADG and MMWT, it is not genetically independent of older animals as observed in this study. It also implies that selection for more feed efficient animals would reduce FI, but would also decrease ADG. Although contrary to Arthur et al. (2001b), similar findings have been observed in other cattle breeds (Berry and Crowley, 2013; Ceacero et al., 2016). Following the same trend as the correlation between RFI and FI, the genetic correlation between RG and ADG was strongly positive which has been observed previously by Crowley et al. (2010). Similarly, the genetic correlation estimated between FCR and ADG (−0.69 ± 0.09) was negative and favorable, in agreement with other studies in beef cattle (Arthur et al., 2001a; Arthur et al., 2001b; Schenkel et al., 2004). The genetic correlation between FCR and RFI was moderate (0.20 ± 0.16) and lower than those reported by Arthur et al. (2001b) and Mao et al. (2013) in younger growing cattle. Interestingly, Berry and Crowley (2013) in a meta-analysis documented that this correlation can range widely from −0.62 to 0.76 with an average of 0.39. A possible explanation for the large differences is that RFI and FCR are unique traits across studies where the animals used range in age and growth stages from young growing cattle to more mature animals (as used in this study). Table 5. Estimates of genetic (above diagonal) and phenotypic (below diagonal) correlations with their standard errors between feed efficiency traits Trait1 ADG MMWT FI FCR RFI RG ADG 0.63 ± 0.10 0.72 ± 0.08 −0.69 ± 0.09 0.34 ± 0.14 0.81 ± 0.06 MMWT 0.33 ± 0.02 0.67 ± 0.07 −0.21 ± 0.17 0.18 ± 0.13 0.21 ± 0.17 FI 0.45 ± 0.02 0.54 ± 0.02 −0.12 ± 0.16 0.83 ± 0.04 0.19 ± 0.16 FCR −0.78 ± 0.01 −0.02 ± 0.02 0.05 ± 0.02 0.20 ± 0.16 −0.92 ± 0.04 RFI 0.01 ± 0.02 0.00 ± 0.02 0.80 ± 0.01 0.37 ± 0.02 −0.13 ± 0.17 RG 0.89 ± 0.01 0.00 ± 0.02 0.00 ± 0.02 −0.90 ± 0.00 −0.34 ± 0.02 Trait1 ADG MMWT FI FCR RFI RG ADG 0.63 ± 0.10 0.72 ± 0.08 −0.69 ± 0.09 0.34 ± 0.14 0.81 ± 0.06 MMWT 0.33 ± 0.02 0.67 ± 0.07 −0.21 ± 0.17 0.18 ± 0.13 0.21 ± 0.17 FI 0.45 ± 0.02 0.54 ± 0.02 −0.12 ± 0.16 0.83 ± 0.04 0.19 ± 0.16 FCR −0.78 ± 0.01 −0.02 ± 0.02 0.05 ± 0.02 0.20 ± 0.16 −0.92 ± 0.04 RFI 0.01 ± 0.02 0.00 ± 0.02 0.80 ± 0.01 0.37 ± 0.02 −0.13 ± 0.17 RG 0.89 ± 0.01 0.00 ± 0.02 0.00 ± 0.02 −0.90 ± 0.00 −0.34 ± 0.02 1ADG = average daily gain; MMWT = metabolic midweight; FI = feed intake; FCR = feed conversion ratio; RFI = residual feed intake; RG = residual gain. View Large Table 5. Estimates of genetic (above diagonal) and phenotypic (below diagonal) correlations with their standard errors between feed efficiency traits Trait1 ADG MMWT FI FCR RFI RG ADG 0.63 ± 0.10 0.72 ± 0.08 −0.69 ± 0.09 0.34 ± 0.14 0.81 ± 0.06 MMWT 0.33 ± 0.02 0.67 ± 0.07 −0.21 ± 0.17 0.18 ± 0.13 0.21 ± 0.17 FI 0.45 ± 0.02 0.54 ± 0.02 −0.12 ± 0.16 0.83 ± 0.04 0.19 ± 0.16 FCR −0.78 ± 0.01 −0.02 ± 0.02 0.05 ± 0.02 0.20 ± 0.16 −0.92 ± 0.04 RFI 0.01 ± 0.02 0.00 ± 0.02 0.80 ± 0.01 0.37 ± 0.02 −0.13 ± 0.17 RG 0.89 ± 0.01 0.00 ± 0.02 0.00 ± 0.02 −0.90 ± 0.00 −0.34 ± 0.02 Trait1 ADG MMWT FI FCR RFI RG ADG 0.63 ± 0.10 0.72 ± 0.08 −0.69 ± 0.09 0.34 ± 0.14 0.81 ± 0.06 MMWT 0.33 ± 0.02 0.67 ± 0.07 −0.21 ± 0.17 0.18 ± 0.13 0.21 ± 0.17 FI 0.45 ± 0.02 0.54 ± 0.02 −0.12 ± 0.16 0.83 ± 0.04 0.19 ± 0.16 FCR −0.78 ± 0.01 −0.02 ± 0.02 0.05 ± 0.02 0.20 ± 0.16 −0.92 ± 0.04 RFI 0.01 ± 0.02 0.00 ± 0.02 0.80 ± 0.01 0.37 ± 0.02 −0.13 ± 0.17 RG 0.89 ± 0.01 0.00 ± 0.02 0.00 ± 0.02 −0.90 ± 0.00 −0.34 ± 0.02 1ADG = average daily gain; MMWT = metabolic midweight; FI = feed intake; FCR = feed conversion ratio; RFI = residual feed intake; RG = residual gain. View Large Table 6. Estimates of genetic (above diagonal) and phenotypic (below diagonal) correlations with their standard errors between carcass traits Trait1 CEMA IMF CRIB P8FAT CWT CEMA 0.06 ± 0.13 −0.08 ± 0.15 −0.22 ± 0.13 0.03 ± 0.13 IMF 0.03 ± 0.03 −0.11 ± 0.15 −0.07 ± 0.13 0.21 ± 0.12 CRIB −0.14 ± 0.03 0.04 ± 0.03 0.50 ± 0.12 0.14 ± 0.15 P8FAT −0.16 ± 0.03 0.01 ± 0.03 0.33 ± 0.02 0.27 ± 0.11 CWT 0.01 ± 0.04 0.09 ± 0.03 0.02 ± 0.04 0.24 ± 0.03 Trait1 CEMA IMF CRIB P8FAT CWT CEMA 0.06 ± 0.13 −0.08 ± 0.15 −0.22 ± 0.13 0.03 ± 0.13 IMF 0.03 ± 0.03 −0.11 ± 0.15 −0.07 ± 0.13 0.21 ± 0.12 CRIB −0.14 ± 0.03 0.04 ± 0.03 0.50 ± 0.12 0.14 ± 0.15 P8FAT −0.16 ± 0.03 0.01 ± 0.03 0.33 ± 0.02 0.27 ± 0.11 CWT 0.01 ± 0.04 0.09 ± 0.03 0.02 ± 0.04 0.24 ± 0.03 1CEMA = carcass eye muscle area; IMF = carcass intramuscular fat; CRIB = subcutaneous fat depths at the 12th/13th rib; P8FAT = rump P8 fat depth; CWT = carcass weight. View Large Table 6. Estimates of genetic (above diagonal) and phenotypic (below diagonal) correlations with their standard errors between carcass traits Trait1 CEMA IMF CRIB P8FAT CWT CEMA 0.06 ± 0.13 −0.08 ± 0.15 −0.22 ± 0.13 0.03 ± 0.13 IMF 0.03 ± 0.03 −0.11 ± 0.15 −0.07 ± 0.13 0.21 ± 0.12 CRIB −0.14 ± 0.03 0.04 ± 0.03 0.50 ± 0.12 0.14 ± 0.15 P8FAT −0.16 ± 0.03 0.01 ± 0.03 0.33 ± 0.02 0.27 ± 0.11 CWT 0.01 ± 0.04 0.09 ± 0.03 0.02 ± 0.04 0.24 ± 0.03 Trait1 CEMA IMF CRIB P8FAT CWT CEMA 0.06 ± 0.13 −0.08 ± 0.15 −0.22 ± 0.13 0.03 ± 0.13 IMF 0.03 ± 0.03 −0.11 ± 0.15 −0.07 ± 0.13 0.21 ± 0.12 CRIB −0.14 ± 0.03 0.04 ± 0.03 0.50 ± 0.12 0.14 ± 0.15 P8FAT −0.16 ± 0.03 0.01 ± 0.03 0.33 ± 0.02 0.27 ± 0.11 CWT 0.01 ± 0.04 0.09 ± 0.03 0.02 ± 0.04 0.24 ± 0.03 1CEMA = carcass eye muscle area; IMF = carcass intramuscular fat; CRIB = subcutaneous fat depths at the 12th/13th rib; P8FAT = rump P8 fat depth; CWT = carcass weight. View Large Among carcass traits, the genetic correlation between CRIB and P8FAT was the highest and positive (0.50 ± 0.12), followed by the association between P8FAT and CWT (0.27 ± 0.11), and IMF and CWT (0.21 ± 0.12). Our estimated genetic correlation between CRIB with P8FAT was similar to the estimate from Robinson and Oddy (2004) in feedlot-finished beef cattle. Genetic correlations between CWT with the other carcass traits were weak to moderate, suggesting that selection for heavier CWTs would result in higher values for IMF, CRIB, and P8FAT (Table 6). These positive genetic correlations agreed with the estimates published by Hoque et al. (2006) in Japanese Black cattle. Of specific interest in the present work were the genetic correlations between feed efficiency traits with growth and carcass traits. In general, RFI presented stronger genetic associations with growth traits compared with RG and FCR (Table 7). The genetic correlation between RFI with 200dWT and 600dWT was slightly positive. This result contrasts those found by Arthur et al. (2001b) and Jeyaruban et al. (2009) who have published negative but favorable genetic associations between RFI with 200dWT and 600dWT. The genetic correlation between RFI and 400dWT was close to zero. In Angus cattle, negative and favorable genetic correlations have been published by Arthur et al. (2001b) and Jeyaruban et al. (2009). The differences observed between the current study and those by Arthur et al. (2001b) and Jeyaruban et al. (2009) may be attributed to the difference in performance between bulls and steers. The previous mentioned studies included many bull records from industry seedstock herds, whereas the current study is limited to commercial steers and heifers. It also suggests that the environments where steers were recorded (from essentially commercial environments) may be different from those where bulls were raised. In this study, the genetic correlations for FCR and RG with growth and carcass traits were very similar but of the opposite sign, which reflects the direction of selection (increases in RG and decreases in FCR are desired), which further confirms the strong relationship between FCR and RG observed in this study. Table 7. Estimates of genetic and phenotypic correlations with their standard errors for growth and carcass traits with feed efficiency traits Trait1 ADG MMWT FI FCR RFI RG Genetic correlations BWT 0.27 ± 0.12 0.30 ± 0.10 0.20 ± 0.10 −0.09 ± 0.15 0.00 ± 0.12 0.19 ± 0.15 200dWT 0.45 ± 0.14 0.92 ± 0.04 0.68 ± 0.09 0.05 ± 0.18 0.25 ± 0.14 −0.06 ± 0.18 400dWT 0.27 ± 0.14 0.90 ± 0.04 0.42 ± 0.11 0.10 ± 0.17 −0.02 ± 0.14 −0.12 ± 0.17 600dWT 0.53 ± 0.10 0.98 ± 0.01 0.61 ± 0.07 −0.12 ± 0.15 0.15 ± 0.11 0.09 ± 0.15 CEMA 0.06 ± 0.15 0.12 ± 0.14 −0.05 ± 0.13 −0.05 ± 0.18 −0.19 ± 0.14 0.08 ± 0.18 IMF 0.11 ± 0.15 0.20 ± 0.13 0.30 ± 0.12 0.06 ± 0.18 0.31 ± 0.14 −0.12 ± 0.18 CRIB 0.23 ± 0.17 −0.23 ± 0.15 0.11 ± 0.15 −0.23 ± 0.20 0.18 ± 0.16 0.28 ± 0.20 P8FAT 0.27 ± 0.14 −0.03 ± 0.13 0.20 ± 0.12 −0.12 ± 0.17 0.24 ± 0.13 0.24 ± 0.17 CWT 0.71 ± 0.09 0.78 ± 0.05 0.73 ± 0.07 −0.26 ± 0.16 0.40 ± 0.12 0.33 ± 0.16 Phenotypic correlations BWT 0.13 ± 0.02 0.33 ± 0.02 0.20 ± 0.02 0.01 ± 0.02 0.01 ± 0.03 0.01 ± 0.02 200dWT 0.14 ± 0.02 0.64 ± 0.01 0.33 ± 0.02 0.07 ± 0.02 0.01 ± 0.02 −0.08 ± 0.02 400dWT 0.10 ± 0.03 0.82 ± 0.01 0.34 ± 0.03 0.11 ± 0.03 −0.08 ± 0.03 −0.15 ± 0.03 600dWT 0.34 ± 0.02 0.90 ± 0.00 0.48 ± 0.02 −0.07 ± 0.02 −0.01 ± 0.02 0.05 ± 0.02 CEMA −0.03 ± 0.03 −0.11 ± 0.03 −0.09 ± 0.03 0.00 ± 0.03 −0.05 ± 0.03 0.01 ± 0.03 IMF 0.03 ± 0.03 0.05 ± 0.03 0.08 ± 0.03 0.01 ± 0.03 0.07 ± 0.03 −0.01 ± 0.03 CRIB 0.02 ± 0.03 −0.11 ± 0.03 0.04 ± 0.03 −0.02 ± 0.03 0.06 ± 0.03 0.02 ± 0.03 P8FAT 0.09 ± 0.03 0.11 ± 0.03 0.17 ± 0.03 −0.01 ± 0.03 0.12 ± 0.03 0.02 ± 0.03 CWT 0.43 ± 0.02 0.72 ± 0.01 0.52 ± 0.02 −0.14 ± 0.03 0.11 ± 0.03 0.17 ± 0.03 Trait1 ADG MMWT FI FCR RFI RG Genetic correlations BWT 0.27 ± 0.12 0.30 ± 0.10 0.20 ± 0.10 −0.09 ± 0.15 0.00 ± 0.12 0.19 ± 0.15 200dWT 0.45 ± 0.14 0.92 ± 0.04 0.68 ± 0.09 0.05 ± 0.18 0.25 ± 0.14 −0.06 ± 0.18 400dWT 0.27 ± 0.14 0.90 ± 0.04 0.42 ± 0.11 0.10 ± 0.17 −0.02 ± 0.14 −0.12 ± 0.17 600dWT 0.53 ± 0.10 0.98 ± 0.01 0.61 ± 0.07 −0.12 ± 0.15 0.15 ± 0.11 0.09 ± 0.15 CEMA 0.06 ± 0.15 0.12 ± 0.14 −0.05 ± 0.13 −0.05 ± 0.18 −0.19 ± 0.14 0.08 ± 0.18 IMF 0.11 ± 0.15 0.20 ± 0.13 0.30 ± 0.12 0.06 ± 0.18 0.31 ± 0.14 −0.12 ± 0.18 CRIB 0.23 ± 0.17 −0.23 ± 0.15 0.11 ± 0.15 −0.23 ± 0.20 0.18 ± 0.16 0.28 ± 0.20 P8FAT 0.27 ± 0.14 −0.03 ± 0.13 0.20 ± 0.12 −0.12 ± 0.17 0.24 ± 0.13 0.24 ± 0.17 CWT 0.71 ± 0.09 0.78 ± 0.05 0.73 ± 0.07 −0.26 ± 0.16 0.40 ± 0.12 0.33 ± 0.16 Phenotypic correlations BWT 0.13 ± 0.02 0.33 ± 0.02 0.20 ± 0.02 0.01 ± 0.02 0.01 ± 0.03 0.01 ± 0.02 200dWT 0.14 ± 0.02 0.64 ± 0.01 0.33 ± 0.02 0.07 ± 0.02 0.01 ± 0.02 −0.08 ± 0.02 400dWT 0.10 ± 0.03 0.82 ± 0.01 0.34 ± 0.03 0.11 ± 0.03 −0.08 ± 0.03 −0.15 ± 0.03 600dWT 0.34 ± 0.02 0.90 ± 0.00 0.48 ± 0.02 −0.07 ± 0.02 −0.01 ± 0.02 0.05 ± 0.02 CEMA −0.03 ± 0.03 −0.11 ± 0.03 −0.09 ± 0.03 0.00 ± 0.03 −0.05 ± 0.03 0.01 ± 0.03 IMF 0.03 ± 0.03 0.05 ± 0.03 0.08 ± 0.03 0.01 ± 0.03 0.07 ± 0.03 −0.01 ± 0.03 CRIB 0.02 ± 0.03 −0.11 ± 0.03 0.04 ± 0.03 −0.02 ± 0.03 0.06 ± 0.03 0.02 ± 0.03 P8FAT 0.09 ± 0.03 0.11 ± 0.03 0.17 ± 0.03 −0.01 ± 0.03 0.12 ± 0.03 0.02 ± 0.03 CWT 0.43 ± 0.02 0.72 ± 0.01 0.52 ± 0.02 −0.14 ± 0.03 0.11 ± 0.03 0.17 ± 0.03 1BWT = birth weight; 200dWT = 200-d weight; 400dWT = 400-d weight; 600dWT = 600-d weight; ADG = average daily gain; MMWT = metabolic midweight; FI = feed intake; FCR = feed conversion ratio; RFI = residual feed intake; RG = residual gain; CEMA = carcass eye muscle area; IMF = carcass intramuscular fat; CRIB = subcutaneous fat depths at the 12th/13th rib; P8FAT = rump P8 fat depth; CWT = carcass weight. View Large Table 7. Estimates of genetic and phenotypic correlations with their standard errors for growth and carcass traits with feed efficiency traits Trait1 ADG MMWT FI FCR RFI RG Genetic correlations BWT 0.27 ± 0.12 0.30 ± 0.10 0.20 ± 0.10 −0.09 ± 0.15 0.00 ± 0.12 0.19 ± 0.15 200dWT 0.45 ± 0.14 0.92 ± 0.04 0.68 ± 0.09 0.05 ± 0.18 0.25 ± 0.14 −0.06 ± 0.18 400dWT 0.27 ± 0.14 0.90 ± 0.04 0.42 ± 0.11 0.10 ± 0.17 −0.02 ± 0.14 −0.12 ± 0.17 600dWT 0.53 ± 0.10 0.98 ± 0.01 0.61 ± 0.07 −0.12 ± 0.15 0.15 ± 0.11 0.09 ± 0.15 CEMA 0.06 ± 0.15 0.12 ± 0.14 −0.05 ± 0.13 −0.05 ± 0.18 −0.19 ± 0.14 0.08 ± 0.18 IMF 0.11 ± 0.15 0.20 ± 0.13 0.30 ± 0.12 0.06 ± 0.18 0.31 ± 0.14 −0.12 ± 0.18 CRIB 0.23 ± 0.17 −0.23 ± 0.15 0.11 ± 0.15 −0.23 ± 0.20 0.18 ± 0.16 0.28 ± 0.20 P8FAT 0.27 ± 0.14 −0.03 ± 0.13 0.20 ± 0.12 −0.12 ± 0.17 0.24 ± 0.13 0.24 ± 0.17 CWT 0.71 ± 0.09 0.78 ± 0.05 0.73 ± 0.07 −0.26 ± 0.16 0.40 ± 0.12 0.33 ± 0.16 Phenotypic correlations BWT 0.13 ± 0.02 0.33 ± 0.02 0.20 ± 0.02 0.01 ± 0.02 0.01 ± 0.03 0.01 ± 0.02 200dWT 0.14 ± 0.02 0.64 ± 0.01 0.33 ± 0.02 0.07 ± 0.02 0.01 ± 0.02 −0.08 ± 0.02 400dWT 0.10 ± 0.03 0.82 ± 0.01 0.34 ± 0.03 0.11 ± 0.03 −0.08 ± 0.03 −0.15 ± 0.03 600dWT 0.34 ± 0.02 0.90 ± 0.00 0.48 ± 0.02 −0.07 ± 0.02 −0.01 ± 0.02 0.05 ± 0.02 CEMA −0.03 ± 0.03 −0.11 ± 0.03 −0.09 ± 0.03 0.00 ± 0.03 −0.05 ± 0.03 0.01 ± 0.03 IMF 0.03 ± 0.03 0.05 ± 0.03 0.08 ± 0.03 0.01 ± 0.03 0.07 ± 0.03 −0.01 ± 0.03 CRIB 0.02 ± 0.03 −0.11 ± 0.03 0.04 ± 0.03 −0.02 ± 0.03 0.06 ± 0.03 0.02 ± 0.03 P8FAT 0.09 ± 0.03 0.11 ± 0.03 0.17 ± 0.03 −0.01 ± 0.03 0.12 ± 0.03 0.02 ± 0.03 CWT 0.43 ± 0.02 0.72 ± 0.01 0.52 ± 0.02 −0.14 ± 0.03 0.11 ± 0.03 0.17 ± 0.03 Trait1 ADG MMWT FI FCR RFI RG Genetic correlations BWT 0.27 ± 0.12 0.30 ± 0.10 0.20 ± 0.10 −0.09 ± 0.15 0.00 ± 0.12 0.19 ± 0.15 200dWT 0.45 ± 0.14 0.92 ± 0.04 0.68 ± 0.09 0.05 ± 0.18 0.25 ± 0.14 −0.06 ± 0.18 400dWT 0.27 ± 0.14 0.90 ± 0.04 0.42 ± 0.11 0.10 ± 0.17 −0.02 ± 0.14 −0.12 ± 0.17 600dWT 0.53 ± 0.10 0.98 ± 0.01 0.61 ± 0.07 −0.12 ± 0.15 0.15 ± 0.11 0.09 ± 0.15 CEMA 0.06 ± 0.15 0.12 ± 0.14 −0.05 ± 0.13 −0.05 ± 0.18 −0.19 ± 0.14 0.08 ± 0.18 IMF 0.11 ± 0.15 0.20 ± 0.13 0.30 ± 0.12 0.06 ± 0.18 0.31 ± 0.14 −0.12 ± 0.18 CRIB 0.23 ± 0.17 −0.23 ± 0.15 0.11 ± 0.15 −0.23 ± 0.20 0.18 ± 0.16 0.28 ± 0.20 P8FAT 0.27 ± 0.14 −0.03 ± 0.13 0.20 ± 0.12 −0.12 ± 0.17 0.24 ± 0.13 0.24 ± 0.17 CWT 0.71 ± 0.09 0.78 ± 0.05 0.73 ± 0.07 −0.26 ± 0.16 0.40 ± 0.12 0.33 ± 0.16 Phenotypic correlations BWT 0.13 ± 0.02 0.33 ± 0.02 0.20 ± 0.02 0.01 ± 0.02 0.01 ± 0.03 0.01 ± 0.02 200dWT 0.14 ± 0.02 0.64 ± 0.01 0.33 ± 0.02 0.07 ± 0.02 0.01 ± 0.02 −0.08 ± 0.02 400dWT 0.10 ± 0.03 0.82 ± 0.01 0.34 ± 0.03 0.11 ± 0.03 −0.08 ± 0.03 −0.15 ± 0.03 600dWT 0.34 ± 0.02 0.90 ± 0.00 0.48 ± 0.02 −0.07 ± 0.02 −0.01 ± 0.02 0.05 ± 0.02 CEMA −0.03 ± 0.03 −0.11 ± 0.03 −0.09 ± 0.03 0.00 ± 0.03 −0.05 ± 0.03 0.01 ± 0.03 IMF 0.03 ± 0.03 0.05 ± 0.03 0.08 ± 0.03 0.01 ± 0.03 0.07 ± 0.03 −0.01 ± 0.03 CRIB 0.02 ± 0.03 −0.11 ± 0.03 0.04 ± 0.03 −0.02 ± 0.03 0.06 ± 0.03 0.02 ± 0.03 P8FAT 0.09 ± 0.03 0.11 ± 0.03 0.17 ± 0.03 −0.01 ± 0.03 0.12 ± 0.03 0.02 ± 0.03 CWT 0.43 ± 0.02 0.72 ± 0.01 0.52 ± 0.02 −0.14 ± 0.03 0.11 ± 0.03 0.17 ± 0.03 1BWT = birth weight; 200dWT = 200-d weight; 400dWT = 400-d weight; 600dWT = 600-d weight; ADG = average daily gain; MMWT = metabolic midweight; FI = feed intake; FCR = feed conversion ratio; RFI = residual feed intake; RG = residual gain; CEMA = carcass eye muscle area; IMF = carcass intramuscular fat; CRIB = subcutaneous fat depths at the 12th/13th rib; P8FAT = rump P8 fat depth; CWT = carcass weight. View Large Regardless of the high standard errors for RFI, positive but undesirable genetic associations of these traits with carcass traits were estimated, meaning that selecting more efficient animals would decrease values for CRIB and P8FAT (Table 7). McDonagh et al. (2001) documented, after 1 generation of divergent selection, that high-efficient steers fed for between 112 and 180 d had less subcutaneous fat over their rib and rump, but similar cross-sectional area of the eye muscle than low efficient steers. This phenomenon has also been observed by several others (Arthur et al., 2001b; Robinson and Oddy, 2004; Ceacero et al., 2016), and our results further confirm that fat animals tended to be less efficient when efficiency is defined using RFI. In contrast, there were positive and unfavorable genetic correlations for RFI with IMF (0.31 ± 0.14). Similar findings have been reported in other beef cattle populations where more efficient cattle tend to reduce carcass quality traits (Archer et al., 1999; Nkrumah et al., 2007; Berry and Crowley, 2013). Positive, but unfavorable, genetic correlations were estimated between RFI and CWT (0.40 ± 0.12), which is different from the correlations between RFI and CWT (0.12 ± 0.20) observed by Mao et al. (2013). One potential reason for the differences observed between this study and Mao et al. (2013) is that animals in the current study were older, heavier, and fed grain for a longer period. This also suggests that as animals are fed for longer periods the adjustment for growth (ADG) and MMWT no longer ensure low correlations between RFI and other growth-related production traits. Several authors have documented negative properties for ratio traits such as FCR because they do not guarantee the selection for the most efficient animals (Gunsett, 1984; Bishop et al., 1991; Arthur et al., 2001b; Berry and Crowley, 2013). The genetic correlations estimated in this study were close to zero or slightly negative (favorable) between FCR and carcass traits. This has been observed previously by Arthur et al. (2001b) and Mao et al. (2013). The strong genetic correlations between FCR and RG (−0.92 ± 0.04), and between RG with ADG (0.81 ± 0.06) were consistent with those published by Crowley et al. (2010). This suggests that RG targeted similar outcomes to FCR whilst avoiding the negative properties of ratio traits. Genetic and phenotypic correlations between FI and growth and carcass traits were all higher than those estimated for RFI. This is not surprising given that RFI is essentially FI with the phenotypic variation due to growth (ADG) and maintenance (MMWT) removed. Such results agree with previous studies by Arthur et al. (2001b) and Mao et al. (2013) and those summarized by Berry and Crowley (2013). Medium and high positive genetic correlations between growth traits and CWT were found in this study (Table 8), and have also been published by other authors in beef cattle (Crews et al., 2004; Bouquet et al., 2010). Genetic correlations between P8FAT and CRIB with growth traits were negative, from moderate to weak, indicating that selection for faster growth rates would slightly decrease carcass rib and rump fat for this population. There were low, near zero phenotypic correlations between IMF and growth traits (and CWT), which is in agreement with previous studies (Reverter et al., 2003; Meyer et al., 2004). In contrast, previous studies that estimated the genetic correlation between growth traits and IMF suggested that there was a negative correlation between such traits (Meyer et al., 2004). In this study, CWT was positively correlated with growth traits. In addition, IMF was positively associated with 200dWT, 400dWT, and 600dWT (Table 8). In Australia, both growth and marbling (IMF) have been a large part of the breeding objective for Angus cattle, and the trend of selection for high growth and high marbling may be observed in these genetic correlations. Table 8. Estimates of genetic and phenotypic correlations with their standard errors for carcass traits with growth traits Trait1 CEMA IMF CRIB P8FAT CWT Genetic correlations BWT 0.05 ± 0.13 −0.09 ± 0.12 −0.27 ± 0.14 −0.03 ± 0.12 0.26 ± 0.11 200dWT 0.07 ± 0.15 0.18 ± 0.15 −0.47 ± 0.16 −0.06 ± 0.15 0.64 ± 0.09 400dWT 0.05 ± 0.15 0.21 ± 0.14 −0.39 ± 0.16 −0.23 ± 0.14 0.56 ± 0.10 600dWT 0.10 ± 0.12 0.19 ± 0.12 −0.26 ± 0.14 −0.08 ± 0.12 0.75 ± 0.05 Phenotypic correlations BWT −0.08 ± 0.03 −0.04 ± 0.03 −0.17 ± 0.03 −0.06 ± 0.03 0.23 ± 0.03 200dWT −0.07 ± 0.03 −0.05 ± 0.03 −0.09 ± 0.03 0.11 ± 0.03 0.46 ± 0.02 400dWT −0.12 ± 0.04 0.04 ± 0.04 −0.09 ± 0.04 0.10 ± 0.03 0.54 ± 0.02 600dWT −0.13 ± 0.03 0.04 ± 0.03 −0.14 ± 0.03 0.08 ± 0.03 0.66 ± 0.01 Trait1 CEMA IMF CRIB P8FAT CWT Genetic correlations BWT 0.05 ± 0.13 −0.09 ± 0.12 −0.27 ± 0.14 −0.03 ± 0.12 0.26 ± 0.11 200dWT 0.07 ± 0.15 0.18 ± 0.15 −0.47 ± 0.16 −0.06 ± 0.15 0.64 ± 0.09 400dWT 0.05 ± 0.15 0.21 ± 0.14 −0.39 ± 0.16 −0.23 ± 0.14 0.56 ± 0.10 600dWT 0.10 ± 0.12 0.19 ± 0.12 −0.26 ± 0.14 −0.08 ± 0.12 0.75 ± 0.05 Phenotypic correlations BWT −0.08 ± 0.03 −0.04 ± 0.03 −0.17 ± 0.03 −0.06 ± 0.03 0.23 ± 0.03 200dWT −0.07 ± 0.03 −0.05 ± 0.03 −0.09 ± 0.03 0.11 ± 0.03 0.46 ± 0.02 400dWT −0.12 ± 0.04 0.04 ± 0.04 −0.09 ± 0.04 0.10 ± 0.03 0.54 ± 0.02 600dWT −0.13 ± 0.03 0.04 ± 0.03 −0.14 ± 0.03 0.08 ± 0.03 0.66 ± 0.01 1BWT = birth weight; 200dWT = 200-d weight; 400dWT = 400-d weight; 600dWT = 600-d weight; CEMA = carcass eye muscle area; IMF = carcass intramuscular fat; CRIB = subcutaneous fat depths at the 12th/13th rib; P8FAT = rump P8 fat depth; CWT = carcass weight. View Large Table 8. Estimates of genetic and phenotypic correlations with their standard errors for carcass traits with growth traits Trait1 CEMA IMF CRIB P8FAT CWT Genetic correlations BWT 0.05 ± 0.13 −0.09 ± 0.12 −0.27 ± 0.14 −0.03 ± 0.12 0.26 ± 0.11 200dWT 0.07 ± 0.15 0.18 ± 0.15 −0.47 ± 0.16 −0.06 ± 0.15 0.64 ± 0.09 400dWT 0.05 ± 0.15 0.21 ± 0.14 −0.39 ± 0.16 −0.23 ± 0.14 0.56 ± 0.10 600dWT 0.10 ± 0.12 0.19 ± 0.12 −0.26 ± 0.14 −0.08 ± 0.12 0.75 ± 0.05 Phenotypic correlations BWT −0.08 ± 0.03 −0.04 ± 0.03 −0.17 ± 0.03 −0.06 ± 0.03 0.23 ± 0.03 200dWT −0.07 ± 0.03 −0.05 ± 0.03 −0.09 ± 0.03 0.11 ± 0.03 0.46 ± 0.02 400dWT −0.12 ± 0.04 0.04 ± 0.04 −0.09 ± 0.04 0.10 ± 0.03 0.54 ± 0.02 600dWT −0.13 ± 0.03 0.04 ± 0.03 −0.14 ± 0.03 0.08 ± 0.03 0.66 ± 0.01 Trait1 CEMA IMF CRIB P8FAT CWT Genetic correlations BWT 0.05 ± 0.13 −0.09 ± 0.12 −0.27 ± 0.14 −0.03 ± 0.12 0.26 ± 0.11 200dWT 0.07 ± 0.15 0.18 ± 0.15 −0.47 ± 0.16 −0.06 ± 0.15 0.64 ± 0.09 400dWT 0.05 ± 0.15 0.21 ± 0.14 −0.39 ± 0.16 −0.23 ± 0.14 0.56 ± 0.10 600dWT 0.10 ± 0.12 0.19 ± 0.12 −0.26 ± 0.14 −0.08 ± 0.12 0.75 ± 0.05 Phenotypic correlations BWT −0.08 ± 0.03 −0.04 ± 0.03 −0.17 ± 0.03 −0.06 ± 0.03 0.23 ± 0.03 200dWT −0.07 ± 0.03 −0.05 ± 0.03 −0.09 ± 0.03 0.11 ± 0.03 0.46 ± 0.02 400dWT −0.12 ± 0.04 0.04 ± 0.04 −0.09 ± 0.04 0.10 ± 0.03 0.54 ± 0.02 600dWT −0.13 ± 0.03 0.04 ± 0.03 −0.14 ± 0.03 0.08 ± 0.03 0.66 ± 0.01 1BWT = birth weight; 200dWT = 200-d weight; 400dWT = 400-d weight; 600dWT = 600-d weight; CEMA = carcass eye muscle area; IMF = carcass intramuscular fat; CRIB = subcutaneous fat depths at the 12th/13th rib; P8FAT = rump P8 fat depth; CWT = carcass weight. View Large Genetic parameters estimated in the current study may be useful for calculating the prediction of genetic values, direct and correlated selection response, and for developing economic selection indices. They are also key to understanding the current makeup of the Australia Angus population. It is important to acknowledge that information on genetic parameters for feed efficiency traits is still limited when compared with that available for growth traits. Fewer studies are available because feed efficiency traits and carcass quality traits are expensive and difficult to measure in beef cattle. This has meant that there are limited studies that have attempted to quantify the genetic relationships between feed efficiency traits and carcass traits. These studies are limited by the number of observations and therefore large standard errors have been reported. Such large standard errors have also made it difficult to generalize conclusions across relatively underpowered studies. Large standard errors could be, in part, due to small number of animal being measured but also due to inaccuracies in the measurement of FI and growth (Hill, 2012). RFI has been the preferred feed efficiency trait for genetic improvement of feed efficiency in beef cattle. RFI and RG as linear indexes increase the response to selection compared with some disadvantages of ratio traits (Gunsett, 1984). A major difference between most studies is related to the differences in age of measurements for feed efficiency and carcass traits. It is often difficult to generalize and compare the impact of selecting for RFI across populations because for most studies RFI may in fact be a different unique trait, with all studies being influenced by differences in management conditions, diets (e.g., ad libitum feeding, restricted feeding, and composition or diet type), finishing systems, and breeds, making it difficult to generalize conclusions across studies. Results of this study suggest the existence of low and positive (unfavorable) genetic associations between feed efficiency measured as RFI with meat quality traits measured under a long feeding (~270 d) production system. This implies that long-term selection for RFI could negatively affect meat quality carcass traits which are highly valuable for the markets targeted in this production system. It also illustrates the need for a balanced selection index, considering all other economically important traits and that single trait selection for RFI may have undesirable outcomes for many production traits. Further studies that expand the number of records and test different production systems (pasture based and short grain feeding) are essential to elucidate the direction of genetic correlations to design optimal breeding programs. In conclusion, all traits were from moderately to highly heritable, indicating that all traits would respond favorably to selection. Response to selection could be higher for RFI compared with FCR and RG. However, selection for feed efficient animals based on RFI would result in cattle with lighter weights and lower meat quality. To avoid these problems, it would be necessary to build selection indices to select efficient animals with favorable weights and beef quality. Footnotes 1 We thank the Angus Society of Australia for their support of this research and access to the data for this study through the Angus Sire Benchmarking Project (ASBP). Meat and Livestock Australia have provided funding support for feed intake data collection (B.SBP.0089) and cofunding support for the ASBP (PSH.0528). Generous support has also been provided to the ASBP by numerous Angus bull breeders, cooperator herds, supply chain partners, technicians, and research organizations. LITERATURE CITED Archer , J. A. , E. C. Richardson , R. M. Herd , and P. F. Arthur . 1999 . Potential for selection to improve efficiency of feed use in beef cattle: a review . Aust. J. Agric. Res . 50 : 147 – 161 . doi: https://doi.org/10.1071/A98075 Google Scholar Crossref Search ADS Arthur , P. F. , J. A. Archer , and R. M. Herd . 2004 . Feed intake and efficiency in beef cattle: overview of recent Australian research and challenges for the future . Aust. J. Exp. Agric . 44 : 361 – 369 . doi: https://doi.org/10.1071/EA02162 Google Scholar Crossref Search ADS Arthur , P. F. , J. A. Archer , D. J. Johnston , R. M. Herd , E. C. Richardson , and P. F. Parnell . 2001b . Genetic and phenotypic variance and covariance components for feed intake, feed efficiency, and other postweaning traits in angus cattle . J. Anim. Sci . 79 : 2805 – 2811 . doi: https://doi.org/10.2527/2001.79112805x Google Scholar Crossref Search ADS Arthur , J. P. F. , and R. M. Herd . 2008 . Residual feed intake in beef cattle . R. Bras. Zootec . 37 ( Suppl .): 269 – 279 . doi: https://doi.org/10.1590/S1516-35982008001300031 Google Scholar Crossref Search ADS Arthur , P. F. , G. Renand , and D. Krauss . 2001a . Genetic and phenotypic relationships among different measures of growth and feed efficiency in young charolais bulls . Livest. Prod. Sci . 68 : 131 – 139 . doi: https://doi.org/10.1016/S0301-6226(00)00243-8 Google Scholar Crossref Search ADS Banks , R. G . 2011 . Progress in implementation of beef information nucleus portfolio in the Australian beef industry . Proc. Assoc. Adv. Anim. Breed. Genet . 19 : 399 – 402 . http://www.aaabg.org/livestocklibrary/2011/banks399.pdf (accessed June 5 2018). Basarab , J. A. , M. A. Price , J. L. Aalhus , E. K. Okine , W. M. Snelling , and K. L. Lyle . 2003 . Residual feed intake and body composition in young growing cattle . Can. J. Anim. Sci . 83 : 189 – 204 . doi: https://doi.org/10.4141/A02-065 Google Scholar Crossref Search ADS Basarab , J. A. , K. Toogood , K. L. Lyle , and V. Ohama . 2002 . Validation of the GrowSafe® System for determining individual animal feed intake. Final report. In Net feed intake in young growing cattle. Canadian Alberta Beef Industry Development Fund, Project Number: 98AB249. Alberta Agriculture, Food and Rural Development, Lacombe Research Centre, Lacombe, Alberta, AB. Final Report. p. 13–32 . Berry , D. P. , and J. J. Crowley . 2012 . Residual intake and body weight gain: a new measure of efficiency in growing cattle . J. Anim. Sci . 90 : 109 – 115 . doi: https://doi.org/10.2527/jas.2011-4245 Google Scholar Crossref Search ADS PubMed Berry , D. P. , and J. J. Crowley . 2013 . Cell biology symposium: genetics of feed efficiency in dairy and beef cattle . J. Anim. Sci . 91 : 1594 – 1613 . doi: https://doi.org/10.2527/jas.2012-5862 Google Scholar Crossref Search ADS PubMed Bishop , M. D. , M. E. Davis , W. R. Harvey , G. R. Wilson , and B. D. VanStavern . 1991 . Divergent selection for postweaning feed conversion in Angus beef cattle: II. Genetic and phenotypic correlations and realized heritability estimate . J. Anim. Sci . 69 : 4360 – 4367 . doi: https://doi.org/10.2527/1991.69114360x Google Scholar Crossref Search ADS PubMed Bouquet , A. , M.-N. Fouilloux , G. Renand , and F. Phocas . 2010 . Genetic parameters for growth, muscularity, feed efficiency and carcass traits of young beef bulls . Livest. Sci . 129 : 38 – 48 . doi: https://doi.org/10.1016/j.livsci.2009.12.010 Google Scholar Crossref Search ADS Ceacero , T. M. , M. E. Mercadante , J. N. Cyrillo , R. C. Canesin , S. F. Bonilha , and L. G. de Albuquerque . 2016 . Phenotypic and genetic correlations of feed efficiency traits with growth and carcass traits in Nellore cattle selected for postweaning weight . PLoS ONE 11 : e0161366 . doi: https://doi.org/10.1371/journal.pone.0161366 Google Scholar Crossref Search ADS PubMed Crews , D. H. , Jr. , M. Lowerison , N. Caron , and R. A. Kemp . 2004 . Genetic parameters among growth and carcass traits of Canadian Charolais cattle . Can. J. Anim. Sci . 84 : 589 – 597 . doi: https://doi.org/10.4141/A04-019 Google Scholar Crossref Search ADS Crowley , J. J. , M. McGee , D. A. Kenny , D. H. Crews , Jr , R. D. Evans , and D. P. Berry . 2010 . Phenotypic and genetic parameters for different measures of feed efficiency in different breeds of Irish performance-tested beef bulls . J. Anim. Sci . 88 : 885 – 894 . doi: https://doi.org/10.2527/jas.2009-1852 Google Scholar Crossref Search ADS PubMed Davis , G. P . 1993 . Genetic parameters for tropical beef cattle in northern Australia: a review . Aust. J. Agric. Res . 44 : 179 – 198 . doi: https://doi.org/10.1071/AR9930179 Google Scholar Crossref Search ADS Duff , C. , J. H. J. van der Werf , and S. A. Clark . 2018 . Comparison of two live-animal ultrasound systems to predict carcass intramuscular fat and marbling in Australian Angus cattle . In: Proc. 11th World Congr. Genet. Appl. Livest. Prod. No. Electronic Poster Session . Auckland, NZ . p. 262 . http://www.wcgalp.org/system/files/proceedings/2018/comparison-two-live-animal-ultrasound-systems-predict-carcase-intramuscular-fat-and-marbling.pdf (accessed June 5 2018). Gilmour , A. R. , B. J. Gogel , B. R. Cullis , and R. Thompson . 2009 . ASReml user guide release 3.0 . VSN Int. Ltd ., Hemel Hempstead, UK . Graser , H.-U. , B. Tier , D. J. Johnston , and S. A. Barwick . 2005 . Genetic evaluation for the beef industry in Australia . Aust. J. Agric. Res . 45 : 913 – 921 . doi: https://doi.org/10.1071/EA05075 Google Scholar Crossref Search ADS Gunsett , F. C ., 1984 . Linear index selection to improve traits defined as ratios . J. Anim. Sci . 59 : 1185 – 1193 . doi: https://doi.org/10.2527/jas1984.5951185x Google Scholar Crossref Search ADS Hetzel , D. J. S. , R. L. Quaas , G. W. Seifert , K. G. Bean , H. M. Burrow , and W. J. Aspden . 1990 . Genetic parameters for growth of tropical beef cattle . Proc. Aust. Assoc. Anim. Breed. Genet . 8 : 517 – 520 . http://www.aaabg.org/livestocklibrary/1990/ab90107.pdf (accessed June 4 2018). Hill , R. A . 2012 . Feed efficiency in the beef industry . John Wiley & Sons , Ames, IA . Hoque , M. A. , P. F. Arthur , K. Hiramoto , A. R. Gilmour , and T. Oikawa . 2007 . Variance components due to direct genetic, maternal genetic and permanent environmental effect for growth and feed-efficiency traits in young male Japanese black cattle . J. Anim. Breed. Genet . 124 : 102 – 107 . doi: https://doi.org/10.1111/j.1439-0388.2007.00648.x Google Scholar Crossref Search ADS PubMed Hoque , M. A. , P. F. Arthur , K. Hiramoto , and T. Oikawa . 2006 . Genetic parameters for carcass traits of field progeny and their relationships with feed efficiency traits of their sire population for Japanese Black cattle . Livest. Sci . 100 : 251 – 260 . doi: https://doi.org/10.1016/j.livsci.2005.09.006 Google Scholar Crossref Search ADS Jeyaruban , M. G. , D. J. Johnston , and H. U. Graser . 2009 . Genetic association of net feed intake measured at two stages with insulin-like growth factor-I, growth and ultrasound scanned traits in Angus cattle . Proc. Assoc. Adv. Anim. Breed. Genet . 18 : 584 – 587 . http://www.aaabg.org/proceedings18/files/jeyaruban584.pdf (accessed June 1, 2018). Koch , R. M. , L. A. Swiger , D. Chambers , and K. E. Gregory . 1963 . Efficiency of feed use in beef cattle . J. Anim. Sci . 22 : 486 – 494 . doi: https://doi.org/10.2527/jas1963.222486x Google Scholar Crossref Search ADS Koots , K. R. , J. P. Gibson , and J. W. Wilton . 1994 . Analyses of published genetic parameter estimates for beefproduction traits. 2. Phenotypic and genetic correlations . Anim. Breed. Abstr . 62 : 826 – 853 . Mao , F. , L. Chen , M. Vinsky , E. Okine , Z. Wang , J. Basarab , D. H. Crews , Jr , and C. Li . 2013 . Phenotypic and genetic relationships of feed efficiency with growth performance, ultrasound, and carcass merit traits in Angus and Charolais steers . J. Anim. Sci . 91 : 2067 – 2076 . doi: https://doi.org/10.2527/jas.2012-5470 Google Scholar Crossref Search ADS PubMed Mateescu , R. G. , D. J. Garrick , A. J. Garmyn , D. L. VanOverbeke , G. G. Mafi , and J. M. Reecy . 2015 . Genetic parameters for sensory traits in longissimus muscle and their associations with tenderness, marbling score, and intramuscular fat in Angus cattle . J. Anim. Sci . 93 : 21 – 27 . doi: https://doi.org/10.2527/jas.2014-8405 Google Scholar Crossref Search ADS PubMed McDonagh , M. D. , R. M. Herd , E. C. Richardson , V. H. Oddy , J. A. Archer , and P. F. Arthur . 2001 . Meat quality and the Calpain system of feedlot steers following a single generation of divergent selection for residual feed intake . Aust. J. Exp. Agric . 41 : 1013 – 1021 . doi: https://doi.org/10.1071/EA00024 Google Scholar Crossref Search ADS Meyer , K . 1992 . Variance components due to direct and maternal effects for growth traits of Australian beef cattle . Livest. Prod. Sci . 31 : 179 – 204 . doi: https://doi.org/10.1016/0301-6226(92)90017-X Google Scholar Crossref Search ADS Meyer , K . 2005 . Estimates of covariance functions for growth of Angus cattle from random regression analyses fitting B-spline functions . Proc. Aust. Assoc. Anim. Breed. Genet . 16 : 52 – 55 . http://www.aaabg.org/livestocklibrary/ 2005/052meyer.pdf (accessed June 5 2018). Meyer , K. , D. J. Johnston , and H.-U. Graser . 2004 . Estimates of the complete genetic covariance matrix for traits in multi-trait genetic evaluation of Australian Hereford cattle . Aust. J. Agric. Res . 55 : 195 – 210 . doi: https://doi.org/10.1071/AR03164 Google Scholar Crossref Search ADS Nkrumah , J. D. , D. H. Keisler , D. H. Crews Jr. , J. A. Basarab , Z. Wang , C. Li , M. A. Price , E. K. Okine , and S. S. Moore . 2007 . Genetic and phenotypic relationships of serum leptin concentration with performance, efficiency of gain, and carcass merit of feedlot cattle . J. Anim. Sci . 85 : 2147 – 2155 . doi: https://doi.org/10.2527/jas.2006–767 Google Scholar Crossref Search ADS PubMed Reverter , A. , D. J. Johnston , D. Perry , M. E. Goddard , and H. M. Burrow . 2003 . Genetic and phenotypic characterisation of animal, carcass, and meat quality traits from temperate and tropically adapted beef breeds. 2. Abattoir carcass traits . Aust. J. Agric. Res . 54 : 119 – 134 . doi: https://doi.org/10.1071/AR02086 Google Scholar Crossref Search ADS Rios-Utrera , A. , and L. D. Van Vleck . 2004 . Heritability estimates for carcass traits of cattle: a review . Genet. Mol. Res . 3 : 380 – 394 . https://digitalcommons.unl.edu/animalscifacpub/242/ (accessed March 1, 2018). Google Scholar PubMed Robinson , D. L. , and V. H. Oddy . 2004 . Genetic parameters for feed efficiency, fatness, muscle area and feeding behaviour of feedlot finished beef cattle . Livest. Prod. Sci . 90 : 255 – 270 . doi: https://doi.org/10.1016/j.livprodsci.2004.06.011 Google Scholar Crossref Search ADS Schenkel , F. S. , S. P. Miller , and J. W. Wilton . 2004 . Genetic parameters and breed differences for feed efficiency, growth and body composition traits of young beef bulls . Can. J. Anim. Sci . 84 : 177 – 185 . doi: https://doi.org/10.4141/A03-085 Google Scholar Crossref Search ADS Torres-Vázquez , J. A. , and M. L. Spangler . 2016 . Genetic parameters for docility, weaning weight, yearling weight, and intramuscular fat percentage in hereford cattle . J. Anim. Sci . 94 : 21 – 27 . doi: https://doi.org/10.2527/jas.2015-9566 Google Scholar Crossref Search ADS PubMed Trus , D. , and J. W. Wilton . 1988 . Genetic parameters for maternal traits in beef cattle . Can. J. Anim. Sci . 68 : 119 – 128 . doi: https://doi.org/10.4141/cjas88-011 Google Scholar Crossref Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of the American Society of Animal Science. All rights reserved. For permissions, please e-mail: [email protected]. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

T, Howard, Jeremy;A, Rathje, Tom;E, Bruns, Caitlyn;F, Wilson-Wells, Danielle;D, Kachman, Stephen;L, Spangler, Matthew

doi: 10.1093/jas/sky330pmid: 30107560

Abstract Across the majority livestock species, routinely collected genomic and pedigree information has been incorporated into evaluations using single-step methods. As a result, strategies that reduce genotyping costs without reducing the response to selection are important as they could have substantial economic impacts on breeding programs. Therefore, the objective of the current study was to investigate the impact of selectively genotyping selection candidates on the selection response using simulation. Populations were simulated to mimic the genome and population structure of a swine and cattle population undergoing selection on an index comprised of the estimated breeding values (EBV) for 2 genetically correlated quantitative traits. Ten generations were generated and genotyping began generation 7. Two phenotyping scenarios were simulated that assumed the first trait was recorded early in life on all individuals and the second trait was recorded on all versus a random subset of the individuals. The EBV were generated from a bivariate animal model. Multiple genotyping scenarios were generated that ranged from not genotyping any selection candidates, a proportion of the selection candidates based on either their index value or chosen at random, and genotyping all selection candidates. An interim index value was utilized to decide who to genotype for the selective genotype strategy. The interim value assumed only the first trait was observed and the only genotypic information available was on animals in previous generations. Within each genotyping scenario 25 replicates were generated. Within each genotyping scenario the mean response per generation and the degree to which EBV were inflated/deflated was calculated. Across both species and phenotyping strategies, the plateau of diminishing returns was observed when 60% of the selection candidates with the largest index values were genotyped. When randomly genotyping selection candidates, either 80 or 100% of the selection candidates needed to be genotyped for there not to be a reduction in the index response. Across both populations, no differences in the degree that EBV were inflated/deflated for either trait 1 or 2 were observed between nongenotyped and genotyped animals. The current study has shown that animals can be selectively genotyped in order to optimize the response to selection as a function of the cost to conduct a breeding program using single-step genomic best linear unbiased prediction. INTRODUCTION Across the majority of livestock species, it has become a routine practice to genotype a proportion of the selection candidates in order to obtain a more accurate prediction of an animal’s genetic merit (Berry et al., 2016; Knol et al., 2016). Furthermore, the incorporation of genomic information into routine genetic evaluations using multistep methods has, in general, been replaced with single-step methods. One of the issues with multiple-step methods is that they are more sensitive to biases when selective genotyping and phenotyping exists compared to single-step methods (Patry and Ducrocq, 2011; Masuda et al., 2017). One such single-step method, referred to as single-step genomic BLUP (ssGBLUP), utilizes a relationship matrix that blends full pedigree and genomic information to simultaneously evaluate genotyped and nongenotyped animals (Aguilar et al., 2010; Christensen and Lund, 2010). The ssGBLUP method does not rely on deregressed breeding values (Garrick et al., 2009), properly weights information from genotyped individuals and accounts for preselection bias of genomically selected parents without phenotypes (Legarra et al., 2014; Masuda et al., 2017). Due to single-step methods being less sensitive to scenarios where selection candidates are selectively genotyped, strategies that minimize the cost of genotyping, while not reducing the response to selection, can be investigated. In general, across multiple livestock species, prior to having decided which animals to genotype some, albeit limited, phenotypes of economic importance are collected. For example, in swine, birth weight and average daily gain in the nursery can be collected along with birth weight and weaning weight in beef cattle prior to making selection decisions. As a result, information on early life traits can be utilized when deciding which animals to genotype in order to reduce the need to genotype animals with a low probability of being selected. The impact of selectively genotyping selection candidates over multiple generations on the long-term response to selection when estimating breeding values using ssGBLUP is currently unknown. MATERIALS AND METHODS No animal care approval was required because all data were simulated. Simulated Data To determine the impact of different genotyping strategies on the response to selection, a simulation and the generation of estimated breeding values (EBV) was conducted using the Geno-Diver software (Howard et al., 2017, V3). In order to understand if differences existed across species that have multiple offspring versus a single offspring, genomes and population structures that mimicked swine and cattle populations were generated. Swine Genome and Population Structure For the swine population, a genome with 5 chromosomes, each with a length of 136 Mb, was simulated. A length of 136 Mb was chosen based on the mean length of the swine autosomal chromosome. Within Geno-Diver, MaCS (Chen et al., 2009), a coalescence-based simulation program, was called to generate sequence data for 1,300 haplotypes within each chromosome. To generate levels of linkage disequilibrium (LD) in the sequence data that are similar to a swine population, the “Ne100_Scen2” option within Geno-Diver was utilized. The LD decay in the founder population is outlined in Supplementary Figure S1. After generating sequence information, 1,000 quantitative trait loci (QTL) and a marker panel consisting of 15,000 neutral markers were generated. The QTL and markers were spread equally across all 5 chromosomes resulting in 200 and 3,000 QTL and markers, respectively, within each chromosome. The number of markers per chromosome was chosen to resemble a medium density marker panel (e.g., Illumina PorcineSNP60K BeadChip; Illumina Inc.). In order for a QTL or marker to be chosen from the full set of base haplotypes, the minor allele frequency (MAF) had to be greater than 0.01 and 0.05, respectively. The founder population consisted of 50 males and 400 females that were generated by randomly allocating base haplotypes, without replacement, to founder individuals across all chromosomes. Following the creation of the founder population, a forward-in-time simulation approach was utilized for a total of 10 generations. The population size for the forward-in-time portion was the same as the founder population and constant across generations. An animal was allowed to remain in the breeding population for a maximum of 8 generations. Male and female parents were replaced by selected offspring at a rate of 0.60 each generation. All parents were mated at random and each mating resulted in a total of 6 offspring. An offspring had an equal chance of being a male or female. Within a generation, a maximum of 2 selection candidates could be selected within each full-sib family. Cattle Genome and Population Structure For the cattle population, a genome with 5 chromosomes, each with a length of 87 Mb, was simulated. Similar to the swine population, a length of 87 Mb was chosen based on the mean length of the cattle autosomal chromosome. To generate levels of LD in the sequence data (n = 2,500 haplotypes) that is similar to a cattle population, the “Ne100_Scen1” within Geno-Diver was utilized when calling MaCS (Chen et al., 2009). Similar to the swine population, the LD decay in the founder population is outlined in Supplementary Figure S1. The “Ne100_Scen1” option generates lower levels of short-range LD compared to the “Ne100_Scen2” option that was utilized in the swine population. After generating sequence information, 1,000 QTL and a marker panel consisting of 8,750 neutral markers were generated and distributed equally across all 5 chromosomes. The number of markers per chromosome was chosen to resemble a medium density marker panel (e.g., Illumina BovineSNP50K BeadChip; Illumina Inc.). The founder population consisted of 50 males and 1,000 females that were generated from the base haplotypes across all chromosomes. Similar to the simulated swine population, a forward-in-time simulation approach was utilized for 10 generations and the population size was constant across generations. The male and female parents were replaced by selected offspring at a rate of 0.40 and 0.20, respectively, each generation. An animal was allowed to remain in the breeding population for a maximum of 10 generations. All parents were mated at random and each mating resulted in 1 offspring that had an equal chance of being a male or female. Genetic Architecture Across both species, 2 genetically correlated quantitative traits were simulated. Two traits were generated in order to simulate an early life trait that was recorded prior to deciding whether to genotype a selection candidate and a second trait that was not recorded until after selection. Within each trait, additive effects were sampled from a gamma distribution and a correlation of 0.25 between the additive effects for trait 1 and 2 was generated following the method described in Hayashi and Iwata (2013). A range of correlations were initially investigated and no major differences were observed in terms of the proportion of genotyped animals that resulted in a diminishing rate of returns relative to genetic gain (data not shown). As a result, only the scenario with a correlation of 0.25 between the additive effects is described herein. For simplicity, it was assumed that all QTL had an impact on both traits. The marginal distribution to generate additive effects across both traits was assigned a scale and shape parameter of 0.4 and 1.66, respectively. A correlation between the additive effects was generated from 3 independent gamma distributions, x1, x2, x3, which were a Gamma (0.10,1.66), Gamma (0.30,1.66), and Gamma (0.30,1.66), respectively. Samples from x1, x2, and x3 had an equal chance of being positive or negative. The additive QTL effects for trait 1 and 2 were generated as x1 + x2 and x1 + x3, respectively. The phenotype for individual i and trait j (yij) was generated as: yij= μj+∑q=1nQTLγiqajq+eij, where μj is the general mean for trait j, nQTL is the number of QTL, γiq is the genotype (i.e., 0 for the homozygote; 1 for the heterozygote; 2 for the alternative homozygote) for individual i at QTL q, ajq is the additive effect for trait j at QTL q, and eij is a random residual (e ∼ N(0, σe2 )) for individual i and trait j. The residuals were generated from independent normal distributions resulting in a residual covariance across traits being null. Across both traits, the additive effects were scaled to generate a trait with a heritability of 0.35. The phenotypic variance was set at 1.0; therefore, the residual variance was 0.65 across both traits. A range of heritability combinations were initially investigated and no major differences were observed in terms of the proportion of genotyped animals at the point of diminishing genetic gain (data not shown). Selection and Phenotype Information In order to build up the pedigree across both species, 2 generations of random selection and culling were utilized. For the remaining generations, animals were selected and culled based on an index comprised of the EBV for both traits. The index for individual i was constructed as outlined below: indexi=EBVTrait1σEBVTrait1∗0.20+ EBVTrait2σEBVTrait2∗0.80, where EBVTrait1 is the EBV for individual i for trait 1, EBVTrait2 is the EBV for individual i for trait 2, σEBVTrait1 is the standard deviation of EBV for trait 1 on animals born in generation 2 and σEBVTrait2 is the standard deviation of EBV for trait 2 on animals born in generation 2. The standard deviation across both EBV was calculated in generation 2 because it was the generation when selection began. The EBV that were used to generate the index were estimated based on a bivariate animal model as outlined below: y=Xb+Zu+e, where y is a vector of phenotypic observations, b is a vector of fixed effects, u is a vector of random additive genetic effects, e is a vector of random residuals, and X and Z are incidence matrices relating observations to the fixed and random additive genetic effects, respectively. The only fixed effect was the intercept. It was assumed that the var(u) = K⊗ G , var(e) = I⊗ R , and the cov(a,e) = 0, where G and R are 2 × 2 matrices of variance and covariance components for random animal and residual effects and K is a relationship kernel. Starting at generation 7 and continuing through all remaining generations, an animal had the potential to be genotyped. As a result, EBV from generations 3 to 6 were estimated using a relationship kernel based on pedigree information (A; u ~ N(0, σu2A )). For the remaining generations, EBV were estimated using a relationship kernel that is a blend of pedigree and genomic information (Aguilar et al., 2010; Christensen and Lund, 2010) referred to as H (u ~ N(0, σu2H )). When EBV were estimated using the A or H matrix, the method will be referred to as pBLUP and ssGBLUP, respectively. When constructing the inverse of H, an initial genomic relationship matrix (Graw) was constructed as Graw=MM′2∑pj(1−pj), where M is a genotype incidence matrix that has been centered based on allele frequencies (VanRaden, 2008) and p is the allele frequency of the second allele at the jth SNP. The allele frequencies were estimated from all genotyped animals that were utilized when estimating breeding values. As outlined in Vitezica et al. (2011), A22 and Graw need to be compatible. The A22 matrix refers to the pedigree-based relationship for genotyped animals and was constructed as outlined in Colleau (2002). Therefore, Graw was adjusted to make the mean diagonal and mean of all elements equal the mean diagonals and mean of all elements of A22 as outlined in Christensen et al. (2012). A weighted genomic relationship (Gw; 0.95Graw + 0.05A22) was utilized when blending genomic and pedigree information. Lastly, when constructing the inverse of H (H−1), the τ and Ω values for scaling the inverse of Gw and A22 were both set at 1.0. Across both species, 2 types of phenotyping scenarios were investigated in order to understand the impact of different genotyping strategies on a dense (i.e., growth rate) versus sparsely recorded trait (i.e., feed intake). Within both scenarios, the first trait was observed on all selection candidates and resembled an early life trait, but the second trait was either observed on all selection candidates (dense_dense) or only a random proportion of the selection candidates (dense_sparse). For trait 2 in the dense_sparse scenario, phenotypes were allocated randomly across all selection candidates prior to selection. The second trait was observed after an animal was selected and therefore selection candidates lacked phenotypic information for the second trait at the time of selection in the dense_sparse scenario. Given breeding values were estimated from a bivariate animal model, information on the second trait was generated based on the genetic correlation between trait 1 and 2. As a result, the EBV for the first and second trait was not the average EBV of the 2 parents. Within each sex, 20% and 40% of the selection candidates for the swine and cattle scenario, respectively, had phenotypes recorded for the second trait in the dense_sparse scenario. Genotyping Scenarios Starting at generation 7, ten different genotyping scenarios were generated that ranged from not genotyping any selection candidates, a proportion of the selection candidates based on either their interim index breeding value or chosen at random, and genotyping all selection candidates. These scenarios are outlined in Table 1. Within each phenotype scenario all genotyping scenarios were investigated. For the genotyping scenario where a proportion of the animals with the highest index breeding value were genotyped, an interim index value was calculated prior to a genotyping decision being made. An interim value was generated that assumed the first trait was observed while the second trait was not observed and the only genotypic information available was on animals in previous generations. It should be noted the interim value was only utilized to decide who to genotype and an updated index value that included genotypic information, if it was available, on the selection candidates was calculated prior to selection. Table 1. Summary of genotyping scenarios and total number of animal genotypes across all generations by species Genotyping scenario Summary Mean number genotyped1 Swine Beef pblup No parents and selection candidates are genotyped. 0 0 random20 All selected parents and 20% of the selection candidates genotyped at random. 3,605 2,735 index20 All selected parents and 20% of the selection candidates with the highest index breeding value. 3,293 2,498 random40 All selected parents and 40% of the selection candidates genotyped at random. 5,737 3,550 index40 All selected parents and 40% of the selection candidates with the highest index breeding value. 5,257 3,216 random60 All selected parents and 60% of the selection candidates genotyped at random. 7,928 4,371 index60 All selected parents and 60% of the selection candidates with the highest index breeding value. 7,650 4,050 random80 All selected parents and 80% of the selection candidates genotyped at random. 10,165 5,206 index80 All selected parents and 80% of the selection candidates with the highest index breeding value. 10,050 5,050 all All the parents when genotyping was started and all selection candidates for the remaining generations. 12,450 6,050 Genotyping scenario Summary Mean number genotyped1 Swine Beef pblup No parents and selection candidates are genotyped. 0 0 random20 All selected parents and 20% of the selection candidates genotyped at random. 3,605 2,735 index20 All selected parents and 20% of the selection candidates with the highest index breeding value. 3,293 2,498 random40 All selected parents and 40% of the selection candidates genotyped at random. 5,737 3,550 index40 All selected parents and 40% of the selection candidates with the highest index breeding value. 5,257 3,216 random60 All selected parents and 60% of the selection candidates genotyped at random. 7,928 4,371 index60 All selected parents and 60% of the selection candidates with the highest index breeding value. 7,650 4,050 random80 All selected parents and 80% of the selection candidates genotyped at random. 10,165 5,206 index80 All selected parents and 80% of the selection candidates with the highest index breeding value. 10,050 5,050 all All the parents when genotyping was started and all selection candidates for the remaining generations. 12,450 6,050 1Within a genotype scenario, the mean number of genotyped animals across all generations was averaged across the 2 phenotyping scenarios. View Large Table 1. Summary of genotyping scenarios and total number of animal genotypes across all generations by species Genotyping scenario Summary Mean number genotyped1 Swine Beef pblup No parents and selection candidates are genotyped. 0 0 random20 All selected parents and 20% of the selection candidates genotyped at random. 3,605 2,735 index20 All selected parents and 20% of the selection candidates with the highest index breeding value. 3,293 2,498 random40 All selected parents and 40% of the selection candidates genotyped at random. 5,737 3,550 index40 All selected parents and 40% of the selection candidates with the highest index breeding value. 5,257 3,216 random60 All selected parents and 60% of the selection candidates genotyped at random. 7,928 4,371 index60 All selected parents and 60% of the selection candidates with the highest index breeding value. 7,650 4,050 random80 All selected parents and 80% of the selection candidates genotyped at random. 10,165 5,206 index80 All selected parents and 80% of the selection candidates with the highest index breeding value. 10,050 5,050 all All the parents when genotyping was started and all selection candidates for the remaining generations. 12,450 6,050 Genotyping scenario Summary Mean number genotyped1 Swine Beef pblup No parents and selection candidates are genotyped. 0 0 random20 All selected parents and 20% of the selection candidates genotyped at random. 3,605 2,735 index20 All selected parents and 20% of the selection candidates with the highest index breeding value. 3,293 2,498 random40 All selected parents and 40% of the selection candidates genotyped at random. 5,737 3,550 index40 All selected parents and 40% of the selection candidates with the highest index breeding value. 5,257 3,216 random60 All selected parents and 60% of the selection candidates genotyped at random. 7,928 4,371 index60 All selected parents and 60% of the selection candidates with the highest index breeding value. 7,650 4,050 random80 All selected parents and 80% of the selection candidates genotyped at random. 10,165 5,206 index80 All selected parents and 80% of the selection candidates with the highest index breeding value. 10,050 5,050 all All the parents when genotyping was started and all selection candidates for the remaining generations. 12,450 6,050 1Within a genotype scenario, the mean number of genotyped animals across all generations was averaged across the 2 phenotyping scenarios. View Large Evaluation of Scenarios Within each genotyping scenario, a total of 25 replicates were generated. Within each replicate, the mean true breeding values (TBV) for trait 1 and 2 along with the mean true index value within each generation were utilized to calculate the mean response per generation. The mean response was calculated as the difference in the associated value for all animals born in generation 10 and all animals born in generation 2. Furthermore, the correlation between TBV and EBV for trait 1 and 2 on the selection candidates within each generation, referred to as accuracy, was calculated. Lastly, the degree to which EBV were inflated/deflated across different genotype scenarios for the selection candidates was quantified by the coefficient of regression of TBV on EBV. The expected coefficient of regression is a value of 1.0, which implies the EBV are not inflated/deflated. For each metric, the 95% confidence interval was calculated across all replicates based on a randomized complete block design with replicates (i.e., block) and genotype scenario considered fixed. RESULTS The mean index selection response per generation for the swine and cattle population across different genotyping and phenotyping scenarios is outlined in Fig. 1. Each of the scenarios displayed a plateauing pattern in the index response as the proportion of genotyped animals approached 100%. For the scenario where only a portion of animals were genotyped at random, increases in the index response were slower as the proportion of genotyped animals increased. Alternatively, when selection candidates were genotyped based on having higher index values, selection response was quicker suggesting that genotyping more individuals provided minimal improvement in the selection response. For example, across both species and phenotyping strategies genotyping the top 60% of the selection candidates based on their index value within each sex did not result in a statistical significant (P-value > 0.05) change in the selection response compared to genotyping all selection candidates. Alternatively, when genotyping individuals at random, 80% to 100% of the selection candidates needed to be genotyped to avoid a statistically significant reduction in the index response compared to genotyping all selection candidates. The mean response to selection for trait 1, trait 2, and the index across different genotyping and phenotyping scenarios for the swine and cattle population is outlined in Supplementary Tables S1 and S2, respectively. Across both species and phenotyping scenarios, the results for trait 2 in terms of the proportion genotyped without a significant reduction in the response to selection were similar to the index response results. Lastly, no major differences in the mean selection response for trait 1 existed across genotyping or phenotyping strategies for both species. This result is not unexpected given the phenotypes for trait 1 were observed and utilized when predicting the interim value and therefore contained more information prior to genotyping compared to trait 2 that was not observed on selection candidates. However, in general, across species and phenotyping strategies, a decrease in the response for trait 1 and an increase in the response for trait 2 were observed when EBV were estimated utilizing genomic and pedigree information (i.e., ssGBLUP across genotyping scenarios) compared to pedigree information only (i.e., pBLUP). Figure 1. View largeDownload slide Mean index true breeding value response per generation across different genotyping1 and phenotyping2 scenarios for simulated swine and cattle populations. 1The genotype strategy refers to, when applicable (i.e., 20% to 80% genotyped), the criteria used to determine who to genotype. For the EBV strategy, individuals in the top index true breeding value percentile were genotyped, whereas for the random strategy individuals were genotyped at random. 2The dense_dense phenotype strategy refers to all individuals obtaining a phenotype for both of the traits that are in the index. The dense_sparse phenotype strategy refers to all individuals obtaining a phenotype for the first trait while only a fraction (20% in swine and 40% in cattle) of the individuals obtained a phenotype for the second trait. 3The proportion genotyped refers to the proportion of selection candidates genotyped within each generation. The 0% refers to no animals genotyped (i.e., traditional pedigree-based selection), 20% to 80% refers to the proportion genotyped based on the genotyping scenario and 100 refers to all selection candidates being genotyped. Figure 1. View largeDownload slide Mean index true breeding value response per generation across different genotyping1 and phenotyping2 scenarios for simulated swine and cattle populations. 1The genotype strategy refers to, when applicable (i.e., 20% to 80% genotyped), the criteria used to determine who to genotype. For the EBV strategy, individuals in the top index true breeding value percentile were genotyped, whereas for the random strategy individuals were genotyped at random. 2The dense_dense phenotype strategy refers to all individuals obtaining a phenotype for both of the traits that are in the index. The dense_sparse phenotype strategy refers to all individuals obtaining a phenotype for the first trait while only a fraction (20% in swine and 40% in cattle) of the individuals obtained a phenotype for the second trait. 3The proportion genotyped refers to the proportion of selection candidates genotyped within each generation. The 0% refers to no animals genotyped (i.e., traditional pedigree-based selection), 20% to 80% refers to the proportion genotyped based on the genotyping scenario and 100 refers to all selection candidates being genotyped. Outlined in Table 2 is the mean accuracy of the EBV for nongenotyped and genotyped animals across different genotyping scenarios for the swine and cattle populations. In general, across both species and phenotype scenarios, the gain in accuracy for genotyped animals was negligible for trait 1 given phenotypic information was available at the time of selection. On the other hand, for trait 2 the EBV accuracy increased for genotyped animals as the proportion of animals increased and the degree at which the accuracy changed depended on the genotyping strategy. When animals were chosen to be genotyped based on their index value, accuracy increased to a greater extent as compared to genotyping the same proportion of individuals at random. The increase in accuracy was even larger for the phenotyping scenario where trait 2 was sparsely recorded (i.e., dense_sparse) compared to the scenario where both traits were densely recorded (i.e., dense_dense). However, across both phenotyping scenarios and both species, the accuracy when selectively genotyping was numerically lower compared to randomly genotyping selection candidates at the same proportion. Although, across both species and phenotyping strategies, genotyping 60% of the selection candidates based on their index value resulted in a negligible (P-value > 0.05) reduction in the index response compared to the scenario when all animals were genotyped. Lastly, minor changes in the accuracy for nongenotyped animals were observed for trait 1 and under random genotyping of individuals for trait 2. When selectively genotyping individuals, the accuracy for nongenotyped animals for trait 2 decreased as a greater proportion of the selection candidates with a high index value were genotyped. The selection candidates that were in the nongenotyped group had genotyped parents that were older and therefore this likely resulted in compatibility issues between G and A22. Although the error surrounding the accuracy for nongenotyped animals for trait 2 was larger compared to the accuracy for nongenotyped animals for trait 1. Table 2. Mean accuracy for selection candidates born after generation 6 across different genotyping1 and phenotyping scenarios2 for nongenotyped (NG) and genotyped (G) animals in the swine and cattle population Population Genotyping scenario Phenotype scenario 1 Phenotype scenario 2 Trait 1 Trait 2 Trait 1 Trait 2 NG G NG G NG G NG G Swine pblup 0.71 – 0.45 – 0.70 – 0.35 – random20 0.73 0.86 0.51 0.74 0.72 0.86 0.43 0.58 index20 0.71 0.85 0.44 0.65 0.71 0.84 0.37 0.47 random40 0.73 0.88 0.52 0.76 0.72 0.87 0.44 0.60 index40 0.71 0.87 0.39 0.68 0.70 0.86 0.33 0.50 random60 0.73 0.88 0.52 0.78 0.72 0.88 0.43 0.61 index60 0.71 0.88 0.36 0.71 0.70 0.88 0.31 0.53 random80 0.73 0.89 0.52 0.78 0.72 0.89 0.44 0.62 index80 0.70 0.89 0.33 0.74 0.70 0.88 0.26 0.57 all – 0.90 – 0.79 – 0.90 – 0.63 Mean 95% confidence interval range 0.05 0.04 0.09 0.15 0.05 0.05 0.11 0.16 Cattle pblup 0.70 – 0.47 – 0.70 – 0.40 – random20 0.73 0.86 0.52 0.77 0.72 0.86 0.46 0.66 index20 0.72 0.84 0.42 0.68 0.71 0.84 0.37 0.56 random40 0.73 0.87 0.52 0.77 0.73 0.86 0.48 0.68 index40 0.71 0.85 0.37 0.71 0.71 0.85 0.33 0.59 random60 0.73 0.88 0.53 0.78 0.73 0.87 0.48 0.68 index60 0.71 0.87 0.32 0.73 0.71 0.86 0.30 0.61 random80 0.74 0.88 0.53 0.79 0.73 0.88 0.47 0.69 index80 0.70 0.87 0.27 0.76 0.72 0.87 0.25 0.65 all – 0.88 – 0.79 – 0.88 – 0.69 Mean 95% confidence interval range 0.05 0.04 0.10 0.10 0.05 0.04 0.11 0.12 Population Genotyping scenario Phenotype scenario 1 Phenotype scenario 2 Trait 1 Trait 2 Trait 1 Trait 2 NG G NG G NG G NG G Swine pblup 0.71 – 0.45 – 0.70 – 0.35 – random20 0.73 0.86 0.51 0.74 0.72 0.86 0.43 0.58 index20 0.71 0.85 0.44 0.65 0.71 0.84 0.37 0.47 random40 0.73 0.88 0.52 0.76 0.72 0.87 0.44 0.60 index40 0.71 0.87 0.39 0.68 0.70 0.86 0.33 0.50 random60 0.73 0.88 0.52 0.78 0.72 0.88 0.43 0.61 index60 0.71 0.88 0.36 0.71 0.70 0.88 0.31 0.53 random80 0.73 0.89 0.52 0.78 0.72 0.89 0.44 0.62 index80 0.70 0.89 0.33 0.74 0.70 0.88 0.26 0.57 all – 0.90 – 0.79 – 0.90 – 0.63 Mean 95% confidence interval range 0.05 0.04 0.09 0.15 0.05 0.05 0.11 0.16 Cattle pblup 0.70 – 0.47 – 0.70 – 0.40 – random20 0.73 0.86 0.52 0.77 0.72 0.86 0.46 0.66 index20 0.72 0.84 0.42 0.68 0.71 0.84 0.37 0.56 random40 0.73 0.87 0.52 0.77 0.73 0.86 0.48 0.68 index40 0.71 0.85 0.37 0.71 0.71 0.85 0.33 0.59 random60 0.73 0.88 0.53 0.78 0.73 0.87 0.48 0.68 index60 0.71 0.87 0.32 0.73 0.71 0.86 0.30 0.61 random80 0.74 0.88 0.53 0.79 0.73 0.88 0.47 0.69 index80 0.70 0.87 0.27 0.76 0.72 0.87 0.25 0.65 all – 0.88 – 0.79 – 0.88 – 0.69 Mean 95% confidence interval range 0.05 0.04 0.10 0.10 0.05 0.04 0.11 0.12 1See Table 1 for a description of the genotyping scenarios. 2The dense_dense phenotype strategy refers to all individuals obtaining a phenotype for both of the traits that are in the index. The dense_sparse phenotype strategy refers to all individuals obtaining a phenotype for the first trait while only a fraction (20% in swine and 40% in cattle) of the individuals obtained a phenotype for the second trait. View Large Table 2. Mean accuracy for selection candidates born after generation 6 across different genotyping1 and phenotyping scenarios2 for nongenotyped (NG) and genotyped (G) animals in the swine and cattle population Population Genotyping scenario Phenotype scenario 1 Phenotype scenario 2 Trait 1 Trait 2 Trait 1 Trait 2 NG G NG G NG G NG G Swine pblup 0.71 – 0.45 – 0.70 – 0.35 – random20 0.73 0.86 0.51 0.74 0.72 0.86 0.43 0.58 index20 0.71 0.85 0.44 0.65 0.71 0.84 0.37 0.47 random40 0.73 0.88 0.52 0.76 0.72 0.87 0.44 0.60 index40 0.71 0.87 0.39 0.68 0.70 0.86 0.33 0.50 random60 0.73 0.88 0.52 0.78 0.72 0.88 0.43 0.61 index60 0.71 0.88 0.36 0.71 0.70 0.88 0.31 0.53 random80 0.73 0.89 0.52 0.78 0.72 0.89 0.44 0.62 index80 0.70 0.89 0.33 0.74 0.70 0.88 0.26 0.57 all – 0.90 – 0.79 – 0.90 – 0.63 Mean 95% confidence interval range 0.05 0.04 0.09 0.15 0.05 0.05 0.11 0.16 Cattle pblup 0.70 – 0.47 – 0.70 – 0.40 – random20 0.73 0.86 0.52 0.77 0.72 0.86 0.46 0.66 index20 0.72 0.84 0.42 0.68 0.71 0.84 0.37 0.56 random40 0.73 0.87 0.52 0.77 0.73 0.86 0.48 0.68 index40 0.71 0.85 0.37 0.71 0.71 0.85 0.33 0.59 random60 0.73 0.88 0.53 0.78 0.73 0.87 0.48 0.68 index60 0.71 0.87 0.32 0.73 0.71 0.86 0.30 0.61 random80 0.74 0.88 0.53 0.79 0.73 0.88 0.47 0.69 index80 0.70 0.87 0.27 0.76 0.72 0.87 0.25 0.65 all – 0.88 – 0.79 – 0.88 – 0.69 Mean 95% confidence interval range 0.05 0.04 0.10 0.10 0.05 0.04 0.11 0.12 Population Genotyping scenario Phenotype scenario 1 Phenotype scenario 2 Trait 1 Trait 2 Trait 1 Trait 2 NG G NG G NG G NG G Swine pblup 0.71 – 0.45 – 0.70 – 0.35 – random20 0.73 0.86 0.51 0.74 0.72 0.86 0.43 0.58 index20 0.71 0.85 0.44 0.65 0.71 0.84 0.37 0.47 random40 0.73 0.88 0.52 0.76 0.72 0.87 0.44 0.60 index40 0.71 0.87 0.39 0.68 0.70 0.86 0.33 0.50 random60 0.73 0.88 0.52 0.78 0.72 0.88 0.43 0.61 index60 0.71 0.88 0.36 0.71 0.70 0.88 0.31 0.53 random80 0.73 0.89 0.52 0.78 0.72 0.89 0.44 0.62 index80 0.70 0.89 0.33 0.74 0.70 0.88 0.26 0.57 all – 0.90 – 0.79 – 0.90 – 0.63 Mean 95% confidence interval range 0.05 0.04 0.09 0.15 0.05 0.05 0.11 0.16 Cattle pblup 0.70 – 0.47 – 0.70 – 0.40 – random20 0.73 0.86 0.52 0.77 0.72 0.86 0.46 0.66 index20 0.72 0.84 0.42 0.68 0.71 0.84 0.37 0.56 random40 0.73 0.87 0.52 0.77 0.73 0.86 0.48 0.68 index40 0.71 0.85 0.37 0.71 0.71 0.85 0.33 0.59 random60 0.73 0.88 0.53 0.78 0.73 0.87 0.48 0.68 index60 0.71 0.87 0.32 0.73 0.71 0.86 0.30 0.61 random80 0.74 0.88 0.53 0.79 0.73 0.88 0.47 0.69 index80 0.70 0.87 0.27 0.76 0.72 0.87 0.25 0.65 all – 0.88 – 0.79 – 0.88 – 0.69 Mean 95% confidence interval range 0.05 0.04 0.10 0.10 0.05 0.04 0.11 0.12 1See Table 1 for a description of the genotyping scenarios. 2The dense_dense phenotype strategy refers to all individuals obtaining a phenotype for both of the traits that are in the index. The dense_sparse phenotype strategy refers to all individuals obtaining a phenotype for the first trait while only a fraction (20% in swine and 40% in cattle) of the individuals obtained a phenotype for the second trait. View Large Outlined in Table 3 is the mean regression of TBV on the EBV for nongenotyped and genotyped animals across different genotyping scenarios for the swine and cattle population. Across both cattle and swine populations, no differences in the regression coefficient between nongenotyped and genotyped animals for either trait 1 or 2 were observed within a given genotyping and phenotyping scenario. Therefore, the EBV for genotyped animals for a given trait are not inflated/deflated to a greater degree compared to nongenotyped animals for the same trait within the same scenario. For the swine population, the degree of inflation/deflation in the EBV was minimal and the 95% CI contained the value of 1.0 across all genotyping and phenotyping strategies. Alternatively, for multiple genotyping scenarios and across both phenotyping scenarios the EBV in the cattle population were slightly deflated based on the 95% CI not containing 1.0. Deflated EBV occurred more often for trait 2 that was not observed on selection candidates. It should be noted that even though some genotyping scenarios resulted in deflated EBV, the regression coefficient between nongenotyped and genotyped animals was not statistically different (P-value > 0.05). Table 3. Mean inflation1 of breeding values in the selection candidates born after generation 6 across different genotyping2 and phenotyping scenarios3 for nongenotyped (NG) and genotyped (G) animals in the swine and cattle population Population Genotyping scenario Phenotype scenario 1 Phenotype scenario 2 Trait 1 Trait 2 Trait 1 Trait 2 NG G NG G NG G NG G Swine pblup 1.00 – 0.95 0.99 – 0.95 – random20 1.00 1.00 1.00 1.02 0.99 0.99 1.00 1.01 index20 0.99 1.00 1.03 1.05 0.99 0.99 1.07 1.04 random40 1.00 1.00 1.00 1.02 0.99 1.00 1.01 1.02 index40 1.00 1.01 1.03 1.05 0.99 1.01 1.07 1.03 random60 1.00 1.01 1.00 1.03 1.00 1.01 1.00 1.01 index60 1.00 1.02 1.01 1.04 0.99 1.01 1.06 1.03 random80 1.00 1.01 1.00 1.02 0.99 1.00 1.00 1.01 index80 1.00 1.02 1.04 1.03 1.00 1.02 1.00 1.03 all – 1.01 – 1.03 – 1.01 – 1.02 Mean 95% confidence interval range 0.07 0.05 0.21 0.13 0.08 0.05 0.32 0.19 Cattle pblup 1.00 – 0.99 – 1.00 – 1.00 – random20 1.01 1.02 1.08 1.13* 1.01 1.05* 1.08 1.12* index20 1.01 1.01 1.09 1.14* 1.01 1.03 1.12 1.12* random40 1.01 1.02 1.08 1.13* 1.01 1.04* 1.10 1.13* index40 1.01 1.01 1.08 1.13* 1.02 1.03 1.12 1.14* random60 1.01 1.03 1.08 1.12* 1.01 1.04* 1.10 1.13* index60 1.00 1.02 1.04 1.13* 1.02 1.04* 1.11 1.13* random80 1.02 1.03 1.08 1.12* 1.00 1.04* 1.07 1.12* index80 1.01 1.03 1.00 1.12* 1.03 1.04* 1.04 1.13* all – 1.03 – 1.11* – 1.03* – 1.12* Mean 95% confidence interval range 0.10 0.07 0.26 0.15 0.10 0.07 0.34 0.18 Population Genotyping scenario Phenotype scenario 1 Phenotype scenario 2 Trait 1 Trait 2 Trait 1 Trait 2 NG G NG G NG G NG G Swine pblup 1.00 – 0.95 0.99 – 0.95 – random20 1.00 1.00 1.00 1.02 0.99 0.99 1.00 1.01 index20 0.99 1.00 1.03 1.05 0.99 0.99 1.07 1.04 random40 1.00 1.00 1.00 1.02 0.99 1.00 1.01 1.02 index40 1.00 1.01 1.03 1.05 0.99 1.01 1.07 1.03 random60 1.00 1.01 1.00 1.03 1.00 1.01 1.00 1.01 index60 1.00 1.02 1.01 1.04 0.99 1.01 1.06 1.03 random80 1.00 1.01 1.00 1.02 0.99 1.00 1.00 1.01 index80 1.00 1.02 1.04 1.03 1.00 1.02 1.00 1.03 all – 1.01 – 1.03 – 1.01 – 1.02 Mean 95% confidence interval range 0.07 0.05 0.21 0.13 0.08 0.05 0.32 0.19 Cattle pblup 1.00 – 0.99 – 1.00 – 1.00 – random20 1.01 1.02 1.08 1.13* 1.01 1.05* 1.08 1.12* index20 1.01 1.01 1.09 1.14* 1.01 1.03 1.12 1.12* random40 1.01 1.02 1.08 1.13* 1.01 1.04* 1.10 1.13* index40 1.01 1.01 1.08 1.13* 1.02 1.03 1.12 1.14* random60 1.01 1.03 1.08 1.12* 1.01 1.04* 1.10 1.13* index60 1.00 1.02 1.04 1.13* 1.02 1.04* 1.11 1.13* random80 1.02 1.03 1.08 1.12* 1.00 1.04* 1.07 1.12* index80 1.01 1.03 1.00 1.12* 1.03 1.04* 1.04 1.13* all – 1.03 – 1.11* – 1.03* – 1.12* Mean 95% confidence interval range 0.10 0.07 0.26 0.15 0.10 0.07 0.34 0.18 1Inflation is coefficient of regression of TBV on EBV and values with an * have a 95% confidence interval that does not contain 1.0. 2See Table 1 for a description of the genotyping scenarios. 3The dense_dense phenotype strategy refers to all individuals obtaining a phenotype for both of the traits that are in the index. The dense_sparse phenotype strategy refers to all individuals obtaining a phenotype for the first trait while only a fraction (20% in swine and 40% in cattle) of the individuals obtained a phenotype for the second trait. View Large Table 3. Mean inflation1 of breeding values in the selection candidates born after generation 6 across different genotyping2 and phenotyping scenarios3 for nongenotyped (NG) and genotyped (G) animals in the swine and cattle population Population Genotyping scenario Phenotype scenario 1 Phenotype scenario 2 Trait 1 Trait 2 Trait 1 Trait 2 NG G NG G NG G NG G Swine pblup 1.00 – 0.95 0.99 – 0.95 – random20 1.00 1.00 1.00 1.02 0.99 0.99 1.00 1.01 index20 0.99 1.00 1.03 1.05 0.99 0.99 1.07 1.04 random40 1.00 1.00 1.00 1.02 0.99 1.00 1.01 1.02 index40 1.00 1.01 1.03 1.05 0.99 1.01 1.07 1.03 random60 1.00 1.01 1.00 1.03 1.00 1.01 1.00 1.01 index60 1.00 1.02 1.01 1.04 0.99 1.01 1.06 1.03 random80 1.00 1.01 1.00 1.02 0.99 1.00 1.00 1.01 index80 1.00 1.02 1.04 1.03 1.00 1.02 1.00 1.03 all – 1.01 – 1.03 – 1.01 – 1.02 Mean 95% confidence interval range 0.07 0.05 0.21 0.13 0.08 0.05 0.32 0.19 Cattle pblup 1.00 – 0.99 – 1.00 – 1.00 – random20 1.01 1.02 1.08 1.13* 1.01 1.05* 1.08 1.12* index20 1.01 1.01 1.09 1.14* 1.01 1.03 1.12 1.12* random40 1.01 1.02 1.08 1.13* 1.01 1.04* 1.10 1.13* index40 1.01 1.01 1.08 1.13* 1.02 1.03 1.12 1.14* random60 1.01 1.03 1.08 1.12* 1.01 1.04* 1.10 1.13* index60 1.00 1.02 1.04 1.13* 1.02 1.04* 1.11 1.13* random80 1.02 1.03 1.08 1.12* 1.00 1.04* 1.07 1.12* index80 1.01 1.03 1.00 1.12* 1.03 1.04* 1.04 1.13* all – 1.03 – 1.11* – 1.03* – 1.12* Mean 95% confidence interval range 0.10 0.07 0.26 0.15 0.10 0.07 0.34 0.18 Population Genotyping scenario Phenotype scenario 1 Phenotype scenario 2 Trait 1 Trait 2 Trait 1 Trait 2 NG G NG G NG G NG G Swine pblup 1.00 – 0.95 0.99 – 0.95 – random20 1.00 1.00 1.00 1.02 0.99 0.99 1.00 1.01 index20 0.99 1.00 1.03 1.05 0.99 0.99 1.07 1.04 random40 1.00 1.00 1.00 1.02 0.99 1.00 1.01 1.02 index40 1.00 1.01 1.03 1.05 0.99 1.01 1.07 1.03 random60 1.00 1.01 1.00 1.03 1.00 1.01 1.00 1.01 index60 1.00 1.02 1.01 1.04 0.99 1.01 1.06 1.03 random80 1.00 1.01 1.00 1.02 0.99 1.00 1.00 1.01 index80 1.00 1.02 1.04 1.03 1.00 1.02 1.00 1.03 all – 1.01 – 1.03 – 1.01 – 1.02 Mean 95% confidence interval range 0.07 0.05 0.21 0.13 0.08 0.05 0.32 0.19 Cattle pblup 1.00 – 0.99 – 1.00 – 1.00 – random20 1.01 1.02 1.08 1.13* 1.01 1.05* 1.08 1.12* index20 1.01 1.01 1.09 1.14* 1.01 1.03 1.12 1.12* random40 1.01 1.02 1.08 1.13* 1.01 1.04* 1.10 1.13* index40 1.01 1.01 1.08 1.13* 1.02 1.03 1.12 1.14* random60 1.01 1.03 1.08 1.12* 1.01 1.04* 1.10 1.13* index60 1.00 1.02 1.04 1.13* 1.02 1.04* 1.11 1.13* random80 1.02 1.03 1.08 1.12* 1.00 1.04* 1.07 1.12* index80 1.01 1.03 1.00 1.12* 1.03 1.04* 1.04 1.13* all – 1.03 – 1.11* – 1.03* – 1.12* Mean 95% confidence interval range 0.10 0.07 0.26 0.15 0.10 0.07 0.34 0.18 1Inflation is coefficient of regression of TBV on EBV and values with an * have a 95% confidence interval that does not contain 1.0. 2See Table 1 for a description of the genotyping scenarios. 3The dense_dense phenotype strategy refers to all individuals obtaining a phenotype for both of the traits that are in the index. The dense_sparse phenotype strategy refers to all individuals obtaining a phenotype for the first trait while only a fraction (20% in swine and 40% in cattle) of the individuals obtained a phenotype for the second trait. View Large DISCUSSION Using simulation, this study has provided evidence that animals can be selectively genotyped as a means to reduce the cost of genotyping without any reduction in the long-term genetic gain when breeding values are estimated using ssGBLUP. The use of genomic selection across the majority of livestock species has resulted in a large number of animals that are routinely genotyped. Therefore, methods that strategically select animals within a breeding program to genotype that reduce routine genotyping costs, without any reduction in the response to selection, are important to optimize the response to selection as a function of the cost to conduct a breeding program. Previous research has been conducted on the impact of different genotyping strategies (Lillehammer et al., 2011; Buch et al., 2012; Tribout et al., 2012; Lillehammer et al., 2013), although the impact of different genotyping strategies within the context of ssGBLUP has not been investigated. The use of ssGBLUP in routine evaluations is attractive because it is less sensitive to scenarios where animals are selectively genotyped and/or genomic preselection exists compared to multistep methods (Patry and Ducrocq, 2011; Masuda et al., 2017). As illustrated by Masuda et al. (2017), when incorporating genomic information into traditional pedigree-based EBV using multistep methods in dairy cattle, genomic preselection for genotyped sires and cows resulted in biased genetic trends across time. Furthermore, the authors found that the bias was reduced when EBV were estimated using ssGBLUP (Masuda et al., 2017). A plateau in the index response to selection as a greater proportion of the selection candidates were genotyped was observed when choosing animals to genotype with the highest index value compared to a nearly linear increase in the selection response as more animals were genotyped at random. A similar trend was observed for the response to selection for trait 2 when selectively versus randomly genotyping selection candidates. Across both species and phenotyping strategies, the plateau of diminishing returns was observed when only 60% of the selection candidates with the largest index values were genotyped. A similar result was observed by Tribout et al. (2012), such that genotyping a limited number of preselected candidates significantly reduced financial costs, while preserving most of the benefits in terms of genetic trends. As a result, the cost of genotyping can be reduced by not genotyping selection candidates that have a low probability of being selected. Phenotypic information from the first trait along with parent average information on the second trait was included when generating the interim index value, which was utilized to determine whether an animal was genotyped. Therefore, to some degree, information on the Mendelian sampling term for the second trait is generated through the genetic correlation with the first (observed) trait, although genotype information provides a more precise estimate of the Mendelian sampling term. When genotyping a proportion of the selection candidates at random, information on the parent average and Mendelian sampling values are not utilized when deciding who to genotype, both of which provide information on the probability of an animal being selected to serve as a parent. As a result, a greater proportion of animals needed to be genotyped to ensure all animals that have a high probably of being selected to serve as parents are genotyped, which is what was observed. For example, 80% to 100% of the selection candidates needed to be genotyped when genotyping was done at random in order for there not to be a reduction in the index response. The genotype proportion with diminishing returns is likely to be population specific and depends on the proportion of the selection candidates that are selected within a given generation and the mating design. For example, assortative mating plans result in a subset of the families with a high probability of generating selection candidates compared to random mating which was utilized in the simulation. As a result, the genotype proportion with diminishing returns needs to be taken in the context of a population breeding design. Lastly, when EBV were estimated using ssGBLUP instead of pBLUP, the selection response for trait 1 was reduced and increased for trait 2. When EBV were estimated with pBLUP, EBV for trait 2 had a lower accuracy and the resulting EBV were regressed more toward zero resulting in a lower EBV standard deviation compared to trait 1. As a result, under pBLUP the EBV for the second trait contributed less to the overall index compared to the EBV estimated using ssGBLUP. Genotyping an animal resulted in a large increase in the accuracy and an even larger increase was observed when a selection candidate did not have phenotypic information on the trait. For strategies that genotyped a certain proportion of the selection candidates, the increase in accuracy as a greater number of animals were genotyped was negligible for trait 1 as a result of phenotypic information being available at the time of selection. For strategies that genotyped a certain percentage of the selection candidates, the accuracy of genotyped animals for trait 2 increased as more selection candidates were genotyped and the increase in accuracy was dependent on the genotyping strategy. For the selective genotyping strategy, the accuracy increased to a greater extent as more selection candidates were genotyped compared to the random genotyping strategy. Across both species and both phenotyping strategies, when genotyping the same proportion of animals, the accuracy was numerically larger under the random scenario compared to the selective genotype strategy. The accuracy when selectively genotyping at a given percentage is, in part, lower than randomly genotyping at the same percentage due to only having a portion of the full-sib and/or half-sib families genotyped. As a result, additive genetic variation explained by the markers is not being fully captured, which is verified by a smaller numerical difference in the accuracy of selective versus random genotyping as a greater proportion of the animals are genotyped. It should be noted that the accuracy in this context is population-wide and does not reflect the standard error associated with an individual animal’s EBV. As a result, selective genotyping allows for one to obtain a more precise EBV prediction (i.e., individual animal accuracy) for animals which have a high probability of being parents without any significant reduction (P-value > 0.05) in the population-wide accuracy. For random selection, a EBV prediction was more accurate, but an animal with a low and high probability of being selected has an equal chance of getting genotyped. For example, when genotyping the same proportion of animals, the numerically largest difference in accuracy for selective genotyping versus random was observed at 20%, although the selection response was larger for the selective genotyping scenario versus the random genotyping scenario. This highlights the importance of genotyping selection candidates in order to obtain an estimate of the Mendelian sampling term. For the nongenotyped animals, minor changes in the accuracy were observed for trait 1 and when randomly genotyping selection candidates for trait 2. Lastly, the accuracy for the nongenotyped animals for trait 2 decreased as a greater proportion of the high index value selection candidates were genotyped, although the error surrounding the accuracy estimate was much larger for trait 2 compared to trait 1. In the nongenotyped group for trait 2, as more individuals were genotyped the nongenotyped group was comprised of selections candidates whose parents were older compared to the genotyped group. As a result, selection candidates with older genotyped parents along with changes in allele frequencies and the additive genetic variance across time likely resulted in compatibility issues between G and A22. In a real population, these issues are not likely to arise due to multiple traits being selected for simultaneously and as a result less change is expected for each trait. In order to verify that the decrease was partially explained by older genotyped parents with nongenotyped offspring, a simulation similar to the swine scenario, but with discrete generations (i.e., parents are only allowed to serve as parents for 1 generation) was generated (results not shown). With discrete generations, the accuracy for nongenotyped individuals on trait 2 no longer decreased as a greater proportion of the selection candidates were selectively genotyped. Across both species and phenotyping strategies and within each genotyping scenario for trait 1 and 2, the degree of inflation/deflation in EBV was similar across nongenotyped versus genotyped selection candidates. This is of primary importance in order to alleviate issues when comparing the EBV for animals that are not genotyped versus have genotyped information. Furthermore, across all genotype scenarios in the swine population, the 95% CI contained 1.0, although for some genotype scenarios in the cattle population the 95% CI did not contain 1.0. As outlined in Koivula et al. (2015) and more recently in Martini et al. (2017), different scaling values for G and A22 when setting up ssGBLUP will impact the degree that EBV are inflated/deflated. As a result, the choice of the blending factors can be optimized, although outside the scope of the current manuscript, in order to minimize the amount EBV are inflated/deflated. Across both phenotyping strategies the same plateau was observed in terms of the genotype proportion, but the response was lower in the dense_sparse scenario compared to the dense_dense across both populations. Therefore, optimizing the number of phenotypes and genotypes simultaneously needs to be investigated in order to further optimize the response to selection as a function of the cost to run a breeding program. Furthermore, under the dense_sparse scenario, it was assumed that within each sex, 20% and 40% of the selection candidates in the swine and cattle scenario obtained phenotypes for trait 2. A simplistic scenario was generated herein. Admittedly, in cases where traits are sex-limited, when the density of phenotypic information varies across sexes for other reasons, or when having phenotypic information on certain traits necessitates genotyping, the proportion of selection candidates that need to be genotyped could be impacted. CONCLUSIONS When simulating 2 phenotyping scenarios, the current study has shown that animals can be selectively genotyped in order to reduce the cost of genotyping animals, with minimal reduction in the response to selection. Using a simulated swine and cattle population, the plateau of diminishing returns was observed when only 60% of the selection candidates with the largest index values were genotyped. Therefore, selective genotype can be utilized to optimize the response to selection as a function of the cost to conduct a breeding program. Further research investigating the optimization of genotyping and phenotyping strategies is needed. ACKNOWLEDGMENTS This project is based on research that was partially supported by the Nebraska Agricultural Experiment Station with funding from the Hatch Act (accession number 1011203) through the USDA National Institute of Food and Agriculture. LITERATURE CITED Aguilar , I. , I. Misztal , D. L. Johnson , A. Legarra , S. Tsuruta , and T. J. Lawlor . 2010 . Hot topic: a unified approach to utilize phenotypic, full pedigree, and genomic information for genetic evaluation of Holstein final score . J. Dairy Sci . 93 : 743 – 752 . doi: https://doi.org/10.3168/jds.2009-2730 Google Scholar Crossref Search ADS PubMed Berry , D. P. , J. F. Garcia , and D. J. Garrick , 2016 . Development and implementation of genomic predictions in beef cattle . Anim. Front . 6 : 32 – 38 . doi: https://doi.org/10.2527/af.2016-0005 Google Scholar Crossref Search ADS Buch , L. H. , M. K. Sørensen , P. Berg , L. D. Pedersen , and A. C. Sørensen . 2012 . Genomic selection strategies in dairy cattle: strong positive interaction between use of genotypic information and intensive use of young bulls on genetic gain . J. Anim. Breed. Genet . 129 : 138 – 151 . doi: https://doi.org/10.1111/j.1439-0388.2011.00947.x Google Scholar Crossref Search ADS PubMed Chen , G. K. , P. Marjoram , and J. D. Wall . 2009 . Fast and flexible simulation of DNA sequence data . Genome Res . 19 : 136 – 142 . doi: https://doi.org/10.1101/gr.083634.108 Google Scholar Crossref Search ADS PubMed Christensen , O. F. , and M. S. Lund . 2010 . Genomic prediction when some animals are not genotyped . Genet. Sel. Evol . 42 : 2 . doi: https://doi.org/10.1186/1297-9686-42-2 Google Scholar Crossref Search ADS PubMed Christensen O. F. , P. Madsen , B. Nielsen , T. Ostersen , and G. Su . 2012 . Single-step methods for genomic evaluation in pigs. Animal 6:1565–1571. doi:10.1017/S1751731112000742 Colleau , J. J . 2002 . An indirect approach to the extensive calculation of relationship coefficients . Genet. Sel. Evol . 34 : 409 – 421 . doi: https://doi.org/10.1051/gse:2002015 Google Scholar Crossref Search ADS PubMed Garrick , D. J. , J. F. Taylor , and R. L. Fernando . 2009 . Deregressing estimated breeding values and weighting information for genomic regression analyses . Genet. Sel. Evol . 41 : 55 . doi: https://doi.org/10.1186/1297-9686-41-55 Google Scholar Crossref Search ADS PubMed Hayashi , T. , and H. Iwata . 2013 . A Bayesian method and its variational approximation for prediction of genomic breeding values in multiple traits . BMC Bioinformatics 14 : 34 . doi: https://doi.org/10.1186/1471-2105-14-34 Google Scholar Crossref Search ADS PubMed Howard , J. T. , F. Tiezzi , J. E. Pryce , and C. Maltecca . 2017 . Geno-Diver: a combined coalescence and forward-in-time simulator for populations undergoing selection for complex traits . J. Anim. Breed. Genet . 134 : 553 – 563 . doi: https://doi.org/10.1111/jbg.12277 Google Scholar Crossref Search ADS PubMed Knol , E. F. , B. Nielsen and P. W. Knap . 2016 . Genomic selection in commercial pig breeding . Anim. Front . 6 : 15 – 22 . doi: https://doi.org/10.2527/af.2016-0003 Google Scholar Crossref Search ADS Koivula , M. , I. Strandén , J. Pösö , G. P. Aamand , and E. A. Mäntysaari . 2015 . Single-step genomic evaluation using multitrait random regression model and test-day data . J. Dairy Sci . 98 : 2775 – 2784 . doi: https://doi.org/10.3168/jds.2014-8975 Google Scholar Crossref Search ADS PubMed Legarra , A. , O. F. Chistensen , I. Aguilar , and I. Misztal . 2014 . Single step, a general approach for genomic selection . Livest. Prod. Sci . 166 : 54 – 65 . doi: https://doi.org/10.1016/j.livsci.2014.04.029 Google Scholar Crossref Search ADS Lillehammer , M. , T. H. Meuwissen , and A. K. Sonesson . 2011 . Genomic selection for maternal traits in pigs . J. Anim. Sci . 89 : 3908 – 3916 . doi: https://doi.org/10.2527/jas.2011-4044 Google Scholar Crossref Search ADS PubMed Lillehammer , M. , T. H. Meuwissen , and A. K. Sonesson . 2013 . Genomic selection for two traits in a maternal pig breeding scheme . J. Anim. Sci . 91 : 3079 – 3087 . doi: https://doi.org/10.2527/jas.2012-5113 Google Scholar Crossref Search ADS PubMed Martini , J. W. R. , M. F. Schrauf , C. A. Garcia-Baccino , E. C. G. Pimentel , S. Munilla , A. Rogberg-Muñoz , R. J. C. Cantet , C. Reimer , N. Gao , V. Wimmer , et al. 2017 . The effect of the H-1 scaling factors τ and ω on the structure of H in the single-step procedure . Genet. Sel. Evol . 50 : 16 . doi: https://doi.org/10.1186/s12711-018-0386-x Google Scholar Crossref Search ADS Masuda , Y. , P. M. VanRaden , I. Misztal , and T. J. Lawlor . 2017 . Differing genetic trend estimates from traditional and genomic evaluations of genotyped animals as evidence of preselection bias in US Holsteins . J. Dairy Sci . 101 : 5194 – 5206 . doi: https://doi.org/10.3168/jds.2009–2730 Google Scholar Crossref Search ADS Patry , C. , and V. Ducrocq . 2011 . Accounting for genomic pre-selection in national BLUP evaluations in dairy cattle . Genet. Sel. Evol . 43 : 30 . doi: https://doi.org/10.1186/1297-9686-43-30 Google Scholar Crossref Search ADS PubMed Tribout , T. , C. Larzul , and F. Phocas . 2012 . Efficiency of genomic selection in a purebred pig male line . J. Anim. Sci . 90 : 4164 – 4176 . doi: https://doi.org/10.2527/jas.2012-5107 Google Scholar Crossref Search ADS PubMed VanRaden , P. M . 2008 . Efficient methods to compute genomic predictions . J. Dairy Sci . 91 : 4414 – 4423 . doi: https://doi.org/10.3168/jds.2007-0980 Google Scholar Crossref Search ADS PubMed Vitezica , Z. G. , I. Aguilar , I. Misztal , and A. Legarra . 2011 . Bias in genomic predictions for populations under selection . Genet. Res. (Camb) . 93 : 357 – 366 . doi: https://doi.org/10.1017/S001667231100022X Google Scholar Crossref Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the American Society of Animal Science. All rights reserved. For permissions, please e-mail: [email protected]. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Petrosus, Elizabeth; Silva, Ediane B; Lay, Don; Eicher, Susan D

doi: 10.1093/jas/sky312pmid: 30060210

Abstract Stress and anxiety have been associated with changes in the microbiota of the gut and ultimately diminished resistance to pathogens. The objective of this study was to observe intestinal microbiota and susceptibility to Salmonella associated with stress hormones, cortisol (CORT), and norepinephrine (NE), in piglets. At weaning, 90 piglets (15 for a Salmonella challenge) were trained to take the carrier (apple juice) orally. At 2 wk after weaning, pens of piglets were assigned randomly to 1 of 3 treatments: control (CNT), NE, or CORT. Blood samples were collected prior to treatment, then piglets were dosed orally with treatments twice on day 0; at 0800 and 1600 h. Control piglets were administered 6.1 mL of the carrier only, NE pigs were administered 40 mg/mL of NE-bitartrate salt dissolved in the carrier, and CORT pigs were administered 12 mg/mL of hydrocortisone acetate dissolved in the carrier. Jugular blood samples were collected prior to necropsies (n = 5/treatment) at 0800 and 1600 h on day 1, and at 0800 h on days 2, 7, and 14 after treatments were started. A subset of pigs were subjected to a 24-h Salmonella challenge. Jejunal and ileal tissues and jejunal, ileal, cecal, and rectal contents were collected and colonies were counted. Microbial data and blood samples were analyzed using mixed models with fixed effects of treatment and day. Cortisol-treated piglets exhibited a spike in plasma CORT concentrations at 0800 h day 1 (P = 0.001) accompanied by greater concentrations of cecal Escherichia coli (P < 0.05) and a shift in intestinal environment to favor coliforms on day 2 (P < 0.05). Salmonella concentrations from rectal contents tended (P = 0.07) to be suppressed by CORT. Lactic acid–producing bacteria rectal concentrations were greater (P = 0.03) in CORT pigs on 0800 h on day 1 then NE pigs and tended to be greater than CNT (P = 0.09) and were greater on day 14 for both CNT (P = 0.003) and NE (P = 0.02). Norepinephrine spiked in NE piglets at 0800 h on day 1 (P = 0.001), 1600 h day 1 (P = 0.004), through day 2 (P = 0.04). Intestinal environment of NE pigs shifted to favor ileal anaerobes (P ≤ 0.05) and facultative anaerobes (E. coli; P = 0.01) compared to CNT. However, Salmonella concentrations in rectal contents were suppressed by NE compared to CNT (P = 0.05). Oral administration of NE and CORT had the desired effect of increasing concentrations of stress hormones and resulted in microbiome shifts throughout the intestines. INTRODUCTION The gut and the brain communicate via the gut–brain axis. The gut is innervated with contacts to the nervous system, through which bidirectional communication between the gut and the brain occur (Lyte, 2014). Anxiety and other negative affective states therefore influence the microorganisms living in the gut. An increase in stress and anxiety and impairments in learning and memory have been associated with infection and changes from the normal gut microbiome (Goehler et al., 2007; Li et al., 2009; Gareau et al., 2011). Therefore, stress with greater cortisol (CORT) and catecholamines as observed in piglets during weaning, may be changing the gut microbiome and susceptibility to disease. One of the earliest sources of stress for production piglets is weaning. Both the HPA axis and SAM axis and their hormones have been studied as physiological causes of stress in weaned pigs (Stanton and Mueller, 1976). Neurotransmitters associated with that response have recently been implicated in gastrointestinal physiology and disease (Mittal et al., 2017); because the relationship of the gut system and enteric nervous system is bidirectional the relationship becomes complex. Neurotransmitters play a major role in gut homeostasis including the gut microbiome and overall gut motility and health. Additionally, catecholamines activate different receptors at different concentrations, further muddling our knowledge of this intricate model (Mittal et al., 2017). Catecholamines may come directly from the microbial population or transported by them to the site (Sudo, 2014). Additionally, exogenous manipulation of catecholamines has ameliorated symptoms of disease and disease (Liang et al., 2015) and catecholamines have been shown to effect immunity and pathogen growth and virulence (Freestone, 2013). Reduced intestine Lactobacillus counts have been associated with environmental stress; such as no bedding, food, or water (Tannock and Savage (1974), maternal separation (Bailey and Cow, 1999), and even high-stress periods in humans (Knowles et al., 2008). In contrast, epinephrine and norepinephrine (NE) are known to increase virulence of some foodborne pathogens (Lustri et al., 2017). Still few studies have examined the impact on microbiota due to acute psychological stressors such as weaning in pigs. Based on evidence from other species, we hypothesize that we can induce intestine microbial shifts with oral administration of NE or CORT. The objectives of this study were to administer CORT or NE to nursery pigs (after weaning stress has subsided) to determine changes in gut microbial communities and susceptibility to Salmonella. MATERIALS AND METHODS Animals and Treatments Animal procedures were approved by the Purdue Animal Care and Use Committee using the 3rd edition of Guide for the care and use of agricultural animals in research and teaching (2010). At weaning (19 ± 2 d of age, mean ± SD), Yorkshire by Landrace cross barrows and gilts (n = 75 trial piglets, 5 per treatment per necropsy; and 15 for a Salmonella challenge, 5 per treatment) were blocked by weight into 15 pens and balanced by sex. The study was conducted in 2 replications for logistical reasons. Additional companion piglets were added to bring the total piglets per pen to 8. Pigs were fed a nursery diet (Table 1) ad libitum. Mean piglet BW ± SD was 12.8 kg ± 1.81 at 2 wk after weaning when the experiment began. Because pigs are known to like apple juice and it would have minimal effect on the diet, the piglets were trained to take the carrier (apple juice) orally by syringe (Ardes France Mod Depose, Wabash Valley Feeds, West Lafayette, IN). This route of administration has 2 distinct advantages to other methods. The pigs would be maintained in their social housing, so there was not the isolation stress of single housing. The second method of i.v. collection requires the stress of that process. Therefore, the oral delivery was chosen to circumvent unnecessary social and handling stress. Two wk post-weaning (experimental day 0), piglets were randomly assigned to control, NE, or CORT treatment. On day 0, piglets were administered treatments twice, at 0800 and 1600 h. Control piglets received 6.1 mL of carrier only, NE pigs received 244 mg of NE-bitartrate salt (Sigma-Aldrich; St. Louis, MO) dissolved in 6.1 mL of carrier, and CORT piglets received 73.2 mg of hydrocortisone acetate (Sigma-Aldrich) dissolved in 6.1 mL of carrier. These concentrations were based on Pullinger et al., (2010) for NE concentrations and on Kranendonk et al., (2005) for oral CORT delivery. Table 1. Phase 4 nursery basal diet Ingredient . %, as fed . PU Corn 2006 NRC 62.540 SBM 31.340 Soybean oil 2.000 Limestone 1.180 MonoCal 1.020 Vitamin 0.250 TM 0.125 Se 600 0.050 Phytase 0.100 Salt 0.350 Lysine-HCL 0.360 DL-Methionine 0.160 l-Threonine 0.155 l-Tryptophan 0.020 Banmith, 48 0.100 Corn treatment premix (diet 1–6) 0.250 Total 100.000 Ingredient . %, as fed . PU Corn 2006 NRC 62.540 SBM 31.340 Soybean oil 2.000 Limestone 1.180 MonoCal 1.020 Vitamin 0.250 TM 0.125 Se 600 0.050 Phytase 0.100 Salt 0.350 Lysine-HCL 0.360 DL-Methionine 0.160 l-Threonine 0.155 l-Tryptophan 0.020 Banmith, 48 0.100 Corn treatment premix (diet 1–6) 0.250 Total 100.000 Open in new tab Table 1. Phase 4 nursery basal diet Ingredient . %, as fed . PU Corn 2006 NRC 62.540 SBM 31.340 Soybean oil 2.000 Limestone 1.180 MonoCal 1.020 Vitamin 0.250 TM 0.125 Se 600 0.050 Phytase 0.100 Salt 0.350 Lysine-HCL 0.360 DL-Methionine 0.160 l-Threonine 0.155 l-Tryptophan 0.020 Banmith, 48 0.100 Corn treatment premix (diet 1–6) 0.250 Total 100.000 Ingredient . %, as fed . PU Corn 2006 NRC 62.540 SBM 31.340 Soybean oil 2.000 Limestone 1.180 MonoCal 1.020 Vitamin 0.250 TM 0.125 Se 600 0.050 Phytase 0.100 Salt 0.350 Lysine-HCL 0.360 DL-Methionine 0.160 l-Threonine 0.155 l-Tryptophan 0.020 Banmith, 48 0.100 Corn treatment premix (diet 1–6) 0.250 Total 100.000 Open in new tab Sample Collection and Processing Prior to receiving treatments, at 0700 h and before euthanasia for each necropsy, jugular blood samples were collected into sodium heparin vacuum tubes (BD; Franklin Lakes, NJ) and the serum was harvested and frozen at −80°C until processed for CORT and NE concentrations. Trained individuals euthanized piglets with CO2 gas followed by exsanguination. Necropsies were performed at 0800 and 1600 h on day 1, and at 0800 h on days 2, 7, and 14. Jejunal tissue, ileal tissue, mesenteric lymph nodes (MLN), jejunal content, ileal content, cecal content, and rectal content were collected. Jejunal tissue (3 cm) was collected from the proximal jejunum. Ileal tissue (3 cm) was collected from the ileocecal junction. The tissues were flash frozen using liquid nitrogen, and stored on liquid nitrogen cooled ice for transportation to the lab where they were ground for collection of 1 g amounts. Content samples (up to 5 mL) were stored on ice for transportation to the lab where jejunal, ileal, cecal, and rectal contents were aliquoted into 1 g amounts. All samples were stored at −80°C until processing. Salmonella Challenge Because Salmonella is a common pathogens in neonatal pigs and has enhanced growth and virulence in response to catecholamines and alters genetic gene expression (Freestone, 2013; Bearson, 2016), we tested the ability of the stress hormones to reduce or exacerbate Salmonella attachment and translocation to MLN after weaning. At 0700 h on day 2 in replication 2, 5 pigs per treatment were moved from the nursery to the USDA BSL2 laboratory, where they were housed individually. Piglets were tested for fecal presumptive Salmonella (using plating on Rambach agar) on each of 2 separate days prior to delivery of Salmonella enterica serovar Typhimurium (ATCC strain χ4232, Walsh et al., 2012; Eicher et al., 2017). One milliliter containing 4 × 108 CFU/mL was delivered to each pig intranasally at immediately upon arrival. At 24-h post-infection (PI), necropsies were performed and ileal tissue, MLN, ileal content, cecal content, and rectal content were collected. Ileal tissue (1 g) was collected from the ileocecal junction. Ileal, cecal, and rectal contents were collected in 1 g amounts. Samples were transported and stored as above. Microbial Community Profiles Content samples (1 g) were serial diluted with buffered peptone water (BPW) (Fluka Analytical; St. Louis, MO). Tissue samples were diluted with 9 mL of BPW and homogenized using a mallet and stomacher (Seward; Bohemia, NY) machine. After homogenization, 3 mL of each sample were used for serial dilutions. Jejunal, ileal, and MLN tissue samples were diluted to 10−3. Jejunal and ileal content were diluted to 10−3 and cecum and rectal content were diluted to 10−6. Tissue samples were tested for total coliforms and Escherichia coli. Content samples were tested for total coliforms, aerobes, anaerobes, E. coli, lactic acid bacteria (LAB), and Enterococcus. To determine total coliforms and aerobes, 1 mL of samples were plated on 3M Petrifilm plates. To enumerate anaerobes, 10 µL of samples were plated on brain heart infusion agar in 12 well plates; E. coli, 20 µL of samples were plated on EMB agar in 6 well plates; LAB, 10 µL of samples were plated on MRS agar in 12 well plates; Enterococcus, 20 µL of samples were plated on m-enterococcus agar in 6 well plates. Plates testing for anaerobes and LAB were incubated for 18 to 24 h anaerobically at 37°C. Plates testing for total coliforms, aerobes, E. coli, and enterococcus were incubated for 18 to 24 h aerobically at 37°C. After incubation, visible colonies were counted and recorded. Stress Hormones and Body Temperature Stress hormones. Plasma samples were tested for CORT concentrations using radioimmunoassay (RIA) kits obtained from IBL International (Morrisville, NC). The RIA kit instructions were followed. Plasma samples were tested for NE concentration using ELISA kits from Rocky Mountain Diagnostics (Colorado Springs, CO). The ELISA kit instructions were followed. Thermography recording and analysis. Thermal images were used to record body temperature of all experimental piglets 5 d a week, at 1700 h, for the duration of the study. In accordance with Schmidt et al. (2013), the thermal images captured the back of the ear of the piglets (Fig. 1). Thermal images were captured using a FLIR (Wilsonville, OR) thermography camera and analyzed using the FLIR Tools software. Figure 1. Open in new tabDownload slide Thermograph of the area behind the ear (+) that was used to quantify temperature of pigs given carrier only, cortisol, or norepinephrine at 0800 and 1600 h on day 0. Figure 1. Open in new tabDownload slide Thermograph of the area behind the ear (+) that was used to quantify temperature of pigs given carrier only, cortisol, or norepinephrine at 0800 and 1600 h on day 0. Statistical Methods All measures were log-transformed and analyzed using mixed models (SAS 9.4) with fixed effects of treatment and day and their interactions and either compound symmetry or unstructured covariance structures as determined by Bayesian information criterion values. Generalized linear model was used when convergence could not be obtained. Thermography body temperatures were analyzed using mixed models with fixed effects of treatment and day and compound symmetry. Microbial community profiles were log-transformed and analyzed using mixed models with fixed effects of treatment and day and either compound symmetry or unstructured covariance structures. Salmonella populations were log-transformed and analyzed using mixed models and compound symmetry covariance structures or generalized linear model. Plasma NE and CORT concentrations were log-transformed and analyzed using mixed models with fixed effects of treatment and day and compound symmetry. Results were considered significant at P value equal to or less than 0.05. Results were considered a trend when P value was less than or equal to 0.1 and greater than or equal to 0.06. RESULTS Stress Hormones Plasma CORT concentrations had treatment by day interaction (P = 0.0001; Fig. 2 top panel). On day 0, there were no differences among treatments. Mean plasma CORT concentrations were 37.2, 35.4, and 32.6 ng/mL for control, CORT, and NE piglets, respectively. On day 1 at 0800 h, CORT-treated piglets had greater (P = 0.0001) plasma concentrations of CORT than control piglets, and greater (P = 0.0003) plasma concentrations of CORT than NE-treated piglets. Norepinephrine-treated piglets also tended to have greater (P = 0.1) plasma concentrations of CORT than control piglets. Mean plasma CORT concentrations were 31.7, 163.1, and 68.1 ng/mL for control, CORT, and NE piglets, respectively. On day 1 at 1600 h, NE-treated piglets had greater (P = 0.043) concentrations of CORT than CORT-treated piglets and tended to have greater (P = 0.1) concentrations than control piglets. Mean plasma CORT concentrations were 27.8, 24.7, and 50.6 ng/mL for control, CORT, and NE piglets, respectively. On day 2, control piglets tended to have greater concentrations of plasma CORT (P = 0.1) than NE-treated piglets. Mean plasma CORT concentrations were 36.8, 30.3, and 25.3 ng/mL for control, CORT, and NE piglets, respectively. On day 7, concentrations of plasma CORT tended to be greater for control piglets (P = 0.1) than NE-treated piglets. Mean plasma CORT concentrations were 57.0, 43.0, and 35.1 ng/mL for control, CORT, and NE piglets, respectively. On day 14, there were no differences among treatments. Mean plasma CORT concentrations were 49.7, 54.4, and 59.1 ng/mL for control, CORT, and NE piglets, respectively. Figure 2. Open in new tabDownload slide Cortisol (top panel; treatment by day interaction, P = 0.0001) and norepinephrine (bottom panel, treatment by day interaction, P = 0.0001) concentrations of pigs given carrier only, cortisol, or norepinephrine at 0800 and 1600 h on day 0. Data are mean concentrations ± SE on day 0, at 0800 and 1600 h on day 1, and at 0800 h on days 2, 7, and 14. a,bP ≤ 0.05. Figure 2. Open in new tabDownload slide Cortisol (top panel; treatment by day interaction, P = 0.0001) and norepinephrine (bottom panel, treatment by day interaction, P = 0.0001) concentrations of pigs given carrier only, cortisol, or norepinephrine at 0800 and 1600 h on day 0. Data are mean concentrations ± SE on day 0, at 0800 and 1600 h on day 1, and at 0800 h on days 2, 7, and 14. a,bP ≤ 0.05. Plasma NE concentrations had a treatment by day interaction (P = 0.0001; Fig. 2, bottom panel). On day 0, there were no differences among treatments. Mean plasma NE concentrations were 1598.6, 1979.9, and 1718.7 pg/mL for control, CORT, and NE piglets, respectively. On day 1 at 0800 h, plasma NE concentrations were greater in NE-treated piglets than control (P = 0.0001) and CORT-treated piglets (P = 0.0001). Mean plasma NE concentrations were 1338.7, 2411.0, and 11,755.0 pg/mL for control, CORT, and NE piglets, respectively. On day 1 at 1600 h, NE-treated piglets had greater (P = 0.004) plasma NE concentrations than control piglets, and tended to have greater concentrations than CORT piglets (P = 0.1). Mean plasma NE concentrations were 1444.0, 2350.7, and 5129.5 pg/mL for control, CORT, and NE piglets, respectively. On day 2, plasma NE concentrations were greater for NE-treated piglets than for CORT-treated piglets (P = 0.05). Mean plasma NE concentrations were 1610.2, 1629.3, and 3393.1 pg/mL for control, CORT, and NE piglets, respectively. On day 7, control piglets tended to have greater plasma NE concentrations (P = 0.1) than NE-treated piglets. Mean plasma NE concentrations were 2637.8, 2280.3, and 1264.4 pg/mL for control, CORT, and NE piglets, respectively. On day 14, there were no differences in plasma NE concentrations among treatments. Mean plasma NE concentrations were 3449.0, 2293.0, and 2684.9 pg/mL for control, CORT, and NE piglets, respectively. Body Temperature Main effects for treatment (P = 0.006) and day (P = 0.001) were found for thermography (data not shown). No treatment by day interactions was found. Temperatures of CORT-treated piglets were greater (P = 0.002) than the temperatures of NE-treated piglets; 38.1°C and 37.5°C for CORT and NE piglets, respectively. Cortisol-treated piglets had greater temperatures (P = 0.03) than control piglets; 38.1°C and 37.7°C for CORT and control piglets, respectively. Temperatures of NE-treated piglets were not different (P = 0.24) than temperatures of control piglets; 37.5°C and 37.7°C for NE and control piglets, respectively. Intestinal Bacterial Populations Lactic acid bacteria population from rectal content had a treatment by day interaction (P = 0.0001; Fig. 3, top panel). On day 1 at 0800 h, CORT-treated piglets had greater concentrations of LAB (P = 0.03) than NE-treated piglets, and tended to have greater concentrations of LAB (P = 0.09) than control piglets. Mean LAB concentrations were 7.7, 6.5, and 6.8 log CFU/g for CORT, NE, and control piglets, respectively. On day 1 at 1600 h and day 2, there were no differences among treatments. On day 7, NE-treated piglets exhibited greater LAB concentrations (P = 0.02) than CORT piglets, and greater concentrations (P = 0.02) than control piglets. Mean LAB concentrations were 6.9, 7.7, and 7.1 log CFU/g for CORT, NE, and control piglets, respectively. On day 14, the LAB concentrations of CORT piglets were lower (P = 0.003) than control piglets, and less (P = 0.02) than NE piglets. Mean LAB concentrations were 6.8, 7.5, and 7.8 log CFU/g for CORT, NE, and control piglets, respectively. LAB populations in jejunal content, ileal content, and cecal content were not different by treatment (P = 0.81, P = 0.49, and P = 0.99, respectively). Figure 3. Open in new tabDownload slide Lactic acid bacteria concentration of rectal content (top panel, treatment by day interaction, P = 0.0001) and anaerobe concentration of ileal content (bottom panel, treatment effect P = 0.09) for pigs orally given the carrier only, cortisol, or norepinephrine at 0800 and 1600 h on day 0. Data are means ± SE at 0800 and 1600 h on day 1, and at 0800 h on days 2, 7, and 14. a,bP ≤ 0.05. Figure 3. Open in new tabDownload slide Lactic acid bacteria concentration of rectal content (top panel, treatment by day interaction, P = 0.0001) and anaerobe concentration of ileal content (bottom panel, treatment effect P = 0.09) for pigs orally given the carrier only, cortisol, or norepinephrine at 0800 and 1600 h on day 0. Data are means ± SE at 0800 and 1600 h on day 1, and at 0800 h on days 2, 7, and 14. a,bP ≤ 0.05. Anaerobe populations tended to have a treatment effect (P = 0.09; Fig. 3, bottom panel) for ileal content. Concentrations of anaerobes were greater (P = 0.05) in NE-treated piglets than CORT-treated piglets, and tended to be greater (P = 0.07) than control piglets. Mean anaerobe concentrations were 5.6, 6.5, and 5.8 log CFU/g for CORT, NE, and control piglets, respectively. Jejunal content was not different among treatments (P = 0.74). Similarly, cecum content was not different by treatment (P = 0.44). Additionally, rectal content was not different by treatment (P = 0.85). Escherichia coli populations had a treatment effect (P = 0.03; data not shown) for cecal content. Concentrations of E. coli were greater (P = 0.04) in CORT-treated piglets than control piglets. Norepinephrine-treated piglets also had higher concentrations of E. coli (P = 0.0126) than control piglets. Mean E. coli concentrations were 4.5, 4.8, and 3.3 log CFU/g for CORT, NE, and control piglets, respectively. Escherichia coli populations also tended (P = 0.1; Fig. 4) to have a treatment by day effect for ileal tissue. On day 1 at 0800 h, CORT-treated piglets had greater (P = 0.05) concentrations of E. coli than NE-treated piglets. Mean E. coli concentrations were 4.5, 1.5, and 2.5 log CFU/g for CORT, NE, and control piglets, respectively. On day 1 at 1600 h, NE-treated piglets tended (P = 0.08) to have greater concentrations of E. coli than CORT-treated piglets. Mean E. coli concentrations were 0, 2.8, and 1.2 log CFU/g for CORT, NE, and control piglets, respectively. There were no treatment by day interactions for days 2, 7, or 14. Escherichia coli populations in jejunal content, ileal content, rectal content, jejunal tissue, ileal tissue, and MLN were not different by treatment (P = 0.59, P = 0.50, P = 0.49, P = 0.63, P = 0.98, P = 0.77, respectively). Figure 4. Open in new tabDownload slide Escherichia coli in ileal tissue (top panel, treatment by day interaction, P = 0.10) and coliforms in ileal tissue (bottom panel, treatment by day interaction, P = 0.07) for pigs orally give the carrier only, cortisol, or norepinephrine at 0800 and 1600 h on day 0. Data are means ± SE at 0800 and 1600 h on day 1, and at 0800 h on days 2, 7, and 14. a,bP ≤ 0.05. Figure 4. Open in new tabDownload slide Escherichia coli in ileal tissue (top panel, treatment by day interaction, P = 0.10) and coliforms in ileal tissue (bottom panel, treatment by day interaction, P = 0.07) for pigs orally give the carrier only, cortisol, or norepinephrine at 0800 and 1600 h on day 0. Data are means ± SE at 0800 and 1600 h on day 1, and at 0800 h on days 2, 7, and 14. a,bP ≤ 0.05. Populations of coliforms tended to have a treatment by day interaction (P = 0.07; Fig. 4, bottom panel) for ileal tissues. On day 1 at 0800 and 1600 h, there were no treatment effects. On day 2, the control piglets had fewer coliforms than CORT-treated piglets (P = 0.01) and NE piglets (P = 0.04). Mean coliform concentrations were 3.5, 3.1, and 0.7 log CFU/g for CORT, NE, and control piglets, respectively. There were no treatment effects on days 7 and 14. There were no treatment effects for coliform populations in the jejunum content (P = 0.55), ileal content (P = 0.98), cecal content (P = 0.77), rectal content (P = 0.83), jejunal tissues (P = 0.98), MLN (P = 0.26). Populations of aerobes had a treatment by day interaction (P = 0.03; Fig. 5) for jejunal content. There were no significant treatment effects on day 1 at 0800 and 1600 h or day 1. On day 2, concentrations of aerobes tended to be greater in control piglets (P = 0.1) than NE-treated piglets. Mean aerobe concentrations were 3.2, 3.1, and 3.8 log CFU/g for CORT, NE, and control piglets, respectively. There were no treatment effects on day 7. On day 14, CORT-treated piglets had greater concentrations of aerobes (P = 0.006) than NE-treated piglets. Also, concentrations of aerobes in control piglets were greater (P = 0.001; data not shown) than concentrations in NE-treated piglets. Mean aerobe concentrations were 4.0, 2.6, and 3.9 log CFU/g for CORT, NE, and control piglets, respectively. Ileal content, cecum content, and rectal content were not different by treatment (P = 0.49, P = 0.20, and P = 0.16, respectively). Figure 5. Open in new tabDownload slide Aerobe concentrations of jejunal content (treatment by day interaction, P = 0.03) for pigs orally give the carrier only, cortisol, or norepinephrine at 0800 and 1600 h on day 0. Data are means ± SE at 0800 and 1600 h on day 1, and at 0800 h on days 2, 7, and 14. a,bP ≤ 0.05. Figure 5. Open in new tabDownload slide Aerobe concentrations of jejunal content (treatment by day interaction, P = 0.03) for pigs orally give the carrier only, cortisol, or norepinephrine at 0800 and 1600 h on day 0. Data are means ± SE at 0800 and 1600 h on day 1, and at 0800 h on days 2, 7, and 14. a,bP ≤ 0.05. There were no treatment effects for Enterococcus populations (data not shown) exhibited no significant treatment interactions in the jejunum (P = 0.58), ileum (P = 0.43), cecum (P = 0.77), or rectum (P = 0.59). Salmonella Enumeration Salmonella enterica serovar Typhimurium tended to have main effects of treatment (P = 0.09; Fig. 6) for the rectal contents. Concentrations of Salmonella were greater (P = 0.05) for control piglets than NE-treated piglets, and tended to be greater (P = 0.07) than CORT-treated piglets. Mean Salmonella concentrations were 0.3, 0, and 1.6 log CFU/g for CORT, NE, and control piglets, respectively. Salmonella populations exhibited no treatment main effects in the cecal contents (P = 0.70), ileal tissue (P = 0.18), MLN (P = 0.27), or ileal contents (P = 0.24). Figure 6. Open in new tabDownload slide Salmonella enterica serovar Typhimurium (ATCC; 4 × 108 CFU/mL) in the rectal content of pigs (treatment effect P = 0.09) orally given the carrier only, cortisol, or norepinephrine at 0800 and 1600 h on day 0 and given an internasal Salmonella enterica serovar Typhimurium challenge. Data are mean CFU ± SE at 24 h after inoculation. a,bP ≤ 0.05. Figure 6. Open in new tabDownload slide Salmonella enterica serovar Typhimurium (ATCC; 4 × 108 CFU/mL) in the rectal content of pigs (treatment effect P = 0.09) orally given the carrier only, cortisol, or norepinephrine at 0800 and 1600 h on day 0 and given an internasal Salmonella enterica serovar Typhimurium challenge. Data are mean CFU ± SE at 24 h after inoculation. a,bP ≤ 0.05. DISCUSSION In this study, the administration of CORT resulted in a spike in blood CORT concentrations and a shift in intestinal environment to favor aerobes and pathogens. The administration of NE resulted in a lasting spike in NE concentrations and a shift in intestinal environment to favor anaerobes and facultative anaerobes. This suggests increased NE takes longer to return to homeostasis and decreased blood flow to the intestines. The NE and CORT piglets each showed an increase in the blood CORT levels for their respective treatments, whereas the control piglets did not. The CORT-treated piglets had CORT concentrations that returned to baseline within 24 h, likely due to CORT’s feedback inhibition. When the adrenal cortex releases CORT, CORT interacts with the brain to inhibit the release of corticotropin-releasing factor (CRF) and adrenocorticotropic hormone (ACTH) (Sapolsky et al., 1986; Ljung et al., 1996; McEwen et al., 2015). Both CRF and ACTH are critical components in the stress response, and inhibition of these components by a spike in CORT concentrations will result in decreased CORT release. Norepinephrine piglets also exhibited a spike in CORT, which can be explained by the connection between the HPA axis and locus-ceruleus-norepinephrine (LCN) autonomic systems. The LCN is connected to the parvocellular corticotropin-releasing hormone neurons, which plays a critical role in the HPA axis. Through these connections, NE can stimulate release of CRF, which would increase CORT concentrations (Tsigos and Chrousos, 1994; Tsigos and Chrousos, 2002). Norepinephrine-treated piglets experienced a spike in NE that lasted for 24 h, longer than the spike in CORT concentrations. This may be due to the dosage of NE being greater than CORT, or because it might take more time for the body to regain homeostasis after a spike in NE. Deficits in CORT and NE, seen days after treatment, are most likely due to overcorrection as the body attempts to return to homeostasis (Saxbe, 2008). The decrease in the generally beneficial LAB 2 wk after treatment and the increase in the opportunistic pathogenic species E. coli, may point to an opportunistic infection in CORT-treated piglets. Toruner et al. (2008) showed that IBD patients treated with corticosteroids were at a greater risk of developing opportunistic infections. Additionally, CORT and glucocorticoids have a history of suppressing the immune system (Kaattari and Tripp, 1987; de Bosscher et al., 2000). Thus, due to CORT’s structural similarities to prescribed corticosteroids and history of infection, it can be concluded that CORT acted as an immunosuppressant that left the piglets open to opportunistic infections, which in turn decreased beneficial bacteria and increased opportunistic pathogens. In contrast to that, it can be argued that a short and mild stressor may lead to increased immune responses (Dhabhar, 2009). There are many factors that determine the effects of glucocorticoids and catecholamines on immunity, including whether it is an acute or chronic stress, the microenvironment of the immune response, concentration and source (endogenous or synthetic), and timing of the stressor within an immune response. But, the effect of glucocorticoids or catecholamines on the microbiota may also modulate the immune response indirectly (Freestone, 2013). Overall, it can be generalized that pathogenic bacteria are mostly aerobic and beneficial bacteria are mostly anaerobic. As well as an increase in E. coli populations, CORT-treated piglets exhibited increases in aerobes. Although the increased E. coli and aerobes occurred in different sections of the intestines, the overall intestinal environment of CORT piglets appears to favor aerobes. Again, this may be due to opportunistic infections brought on by the immunosuppressant function of CORT. Norepinephrine-treated piglets exhibited an increase in anaerobes, and LAB 1 wk post-treatment. The overall intestinal environment of NE piglets appears to favor anaerobes and facultative anaerobes. However, because the LAB populations nearly match the anaerobe populations, the differences in anaerobes may be due to changes in LAB populations. Due to a lack of literature on anaerobes and NE, the mechanism that would cause these observed changes is unknown. Although, NE is known to have direct effects on microbial populations (Lyte et al., 2011) and gene expression of the microbial populations (Bearson, 2016). Additionally, it increases internalization of Salmonella and E. coli O157:H7, but not commensal E. coli into jejunal Peyer’s Patch mucosa (Green et al., 2003), demonstrating some differential effects within bacterial species. In contrast, ex-vivo treatment with NE of jejunal tissue, did not alter Salmonella recovery (Brown and Price, 2008). In our initial work with this Salmonella (Rostagno et al., 2011), we saw differences beginning at 24 h after infection in rectal contents and less than 1 log increase after that. During a challenge with S. Typhimurium, NE treatment resulted in decreased Salmonella populations in piglet rectal contents. However, we saw treatment main effects of treatments in the proximal intestinal regions (data not shown). This may be explained by NE’s role in vasoconstriction. During stress, NE causes vasoconstriction to the intestines, slowing digestion, and thus slowing the movement of intestinal contents. The differences in S. Typhimurium in NE-treated piglets were detected only in the rectal contents. Thus, it might be concluded that the slow-moving contents of the NE-treated piglets resulted in S. Typhimurium populations delayed arrival to the rectum. However, when NE was delivered orally to pigs already infected with Salmonella Typhimurium, an increase in shedding occurred, but preculture of the bacteria with NE did not affect the infection (Pullinger et al., 2010). Thus, it appears that the sequence of exposure to NE and the targeted cells (bacterial or tissue) alters the dynamics of a Salmonella infection. CONCLUSION The oral administration of NE and CORT had the desired effect of increasing concentrations of the stress hormones in the body. Administration of treatments also resulted in microbiome shifts throughout the intestines. High E. coli populations and low LAB populations indicate that the oral administration of CORT may have resulted in opportunistic infections. Piglets treated with NE exhibited an increase in anaerobes and LAB. In this, work stress hormones altered microbial communities of the intestine. Footnotes 1 Mention of trade names or commercial products in this publication is solely for providing specific information and does not imply recommendation or endorsement by the U.S. Department of Agriculture. USDA is an equal opportunity provider and employer. LITERATURE CITED Bailey , M. T. , and C. L. Coe. 1999 . Maternal separation disrupts the integrity of the intestinal microflora in infant rhesus monkeys . Dev. Psychobiol . 35 : 146 – 155 . doi: 10.1002/(SICI)1098–2302(199909)35:2 Google Scholar Crossref Search ADS PubMed WorldCat Bearson , B. L ., 2016 . In Microbial endocrinology: interkingdom signaling in Infectious Disease and Health . M. Lyte (ed.). Adv. Exp. Med. Biol . vol. 874; p. 167–182. doi: 10.1007/978-3-319-20215-0_7 Google Scholar OpenURL Placeholder Text WorldCat Crossref Brown , D. R. , and L. D. Price. 2008 . Catecholamines and sympathomimetic drugs decrease early Salmonella Typhimurium uptake into porcine Peyer’s patches . FEMS Immunol. Med. Microbiol . 52 : 29 – 35 . doi: 10.1111/j.1574-695X.2007.00348.x Google Scholar Crossref Search ADS PubMed WorldCat De Bosscher , K. , W. Vanden Berghe, and G. Haegeman. 2000 . Mechanisms of anti-inflammatory action and of immunosuppression by glucocorticoids: negative interference of activated glucocorticoid receptor with transcription factors . J. Neuroimmunol . 109 : 16 – 22 . Google Scholar Crossref Search ADS PubMed WorldCat Dhabhar , F. S . 2009 . Enhancing versus suppressive effects of stress on immune function: implications for immunoprotection and immunopathology . Neuroimmunomodulation . 16 : 300 – 317 . doi: 10.1159/000216188 Google Scholar Crossref Search ADS PubMed WorldCat Eicher , S. D. , M. H. Rostagno, and D. C. Lay. 2017 . Feed withdrawal and transportation effects on salmonella enterica levels in market-weight pigs . J. Anim. Sci . 95 : 2848 – 2858 . doi: 10.2527/jas.2017.1454 Google Scholar PubMed OpenURL Placeholder Text WorldCat Freestone , P . 2013 . Communication between bacteria and their hosts . Scientifica (Cairo) . 2013 : 361073 . doi: 10.1155/2013/361073 Google Scholar PubMed OpenURL Placeholder Text WorldCat Gareau , M. G. , E. Wine, D. M. Rodrigues, J. H. Cho, M. T. Whary, D. J. Philpott, G. Macqueen, and P. M. Sherman. 2011 . Bacterial infection causes stress-induced memory dysfunction in mice . Gut . 60 : 307 – 317 . doi: 10.1136/gut.2009.202515 Google Scholar Crossref Search ADS PubMed WorldCat Goehler , L. E. , M. Lyte, and R. P. Gaykema. 2007 . Infection-induced viscerosensory signals from the gut enhance anxiety: implications for psychoneuroimmunology . Brain. Behav. Immun . 21 : 721 – 726 . doi: 10.1016/j.bbi.2007.02.005 Google Scholar Crossref Search ADS PubMed WorldCat Green , B. T. , M. Lyte, A. Kulkarni-Narla, and D. R. Brown. 2003 . Neuromodulation of enteropathogen internalization in Peyer’s patches from porcine jejunum . J. Neuroimmunol . 141 : 74 – 82 . doi: 10.1016/S0165-5728(03)00225-X Google Scholar Crossref Search ADS PubMed WorldCat Kaattari , S. L. , and R. A. Tripp. 1987 . Cellular mechanisms of glucocorticoid mmunosuppression in salmon . J. Fish Biol . 31 : 129 – 132 . doi: 10.1111/j.1095–8649.1987.tb05304.x Google Scholar Crossref Search ADS WorldCat Knowles , S. R. , E. A. Nelson, and E. A. Palombo. 2008 . Investigating the role of perceived stress on bacterial flora activity and salivary cortisol secretion” a possible mechanism underlying susceptibility to illness . Biol Psychol . 77 : 132 – 137 . doi: 10.1016/j.biopsycho.2007.09.010 Google Scholar Crossref Search ADS PubMed WorldCat Kranendonk , G. , H. Hopster, F. van Eerdenburg, K. van Reenen, M. Fillerup, J. de Groot, M. Korte, and M. Taverne. 2005 . Evaluation of oral administration of cortisol as a model for prenatal stress in pregnant sows . Am. J. Vet. Res . 66 : 780 – 790 . Google Scholar Crossref Search ADS PubMed WorldCat Li , W. , D. Dowd, B. Scurlock, V. Acosta-Martinez, and M. Lyte. 2009 . Memory and learning behavior in mice is temporally associated with diet-induced alterations in gut bacteria . Physiol. Behav . 96 : 557 – 567 . doi: 10.1016/j.physbeh.2008.12.004 Google Scholar Crossref Search ADS PubMed WorldCat Liang , S. , T. Wang, X. Hu, J. Luo, W. Li, X. Wu, Y. Duan, and F. Jin. 2015 . Administration of Lactobacillus helveticus NS8 improves behavioral, cognitive, and biochemical aberrations caused by chronic restraint stress . Neuroscience . 310 : 561 – 577 . doi: 10.1016/j.neuroscience.2015.09.033 Google Scholar Crossref Search ADS PubMed WorldCat Ljung , T. , B. Andersson, B. A. Bengtsson, P. Björntorp, and P. Mårin. 1996 . Inhibition of cortisol secretion by dexamethasone in relation to body fat distribution: a dose-response study . Obes. Res . 4 : 277 – 282 . doi: 10.1002/j.1550–8528.1996.tb00546.x Google Scholar Crossref Search ADS PubMed WorldCat Lustri , B. C. , V. Sperandio, and C. G. Moreira. 2017 . Bacterial chat: intestinal metabolites and signals in host-microbiota-pathogen interactions . Infect. Immun . 12 : 1 – 14 . doi: 101128/IAI.00476-17 Google Scholar OpenURL Placeholder Text WorldCat Crossref Lyte , M . 2014 . Microbial endocrinology and the microbiota-gut-brain axis . Adv. Exp. Med. Biol . 817 : 3 – 24 . doi: 10.1007/978-1-4939-0897-4_1 Google Scholar Crossref Search ADS PubMed WorldCat Lyte , M. , L. Vulchanova, and D. R. Brown. 2011 . Stress at the intestinal surface: catecholamines and mucosa-bacteria interactions . Cell Tissue Res . 343 : 23 – 32 . doi: 10.1007/s00441-010-1050-0 Google Scholar Crossref Search ADS PubMed WorldCat McEwen , B. S. , N. P. Bowles, J. D. Gray, M. N. Hill, R. G. Hunter, I. N. Karatsoreos, and C. Nasca. 2015 . Mechanisms of stress in the brain . Nat. Neurosci . 18 : 1353 – 1363 . doi: 10.1038/nn.4086 Google Scholar Crossref Search ADS PubMed WorldCat Mittal , R. , L. H. Debs, A. P. Patel, D. Nguyen, K. Patel, G. O’Connor, M. Grati, J. Mittal, D. Yan, A. Eshraghi, et al. . 2017 . Neutrotrnasmitters: the critical modulators regulating gut-brain axis . J. Cell Physiol . 22 : 2359 – 2372 . doi: 10.1002/jcp.25518 Google Scholar Crossref Search ADS WorldCat Pullinger , G. D. , P. M. van Diemen, S. C. Carnell, H. Davies, M. Lyte, and M. P. Stevens. 2010 . 6-hydroxydopamine-mediated release of norepinephrine increases faecal excretion of Salmonella enterica serovar Typhimurium in pigs . Vet. Res . 41 : 68 . doi: 10.1051/vetres/2010040 Google Scholar Crossref Search ADS PubMed WorldCat Rostagno , M. H. , S. D. Eicher, and D. C. Lay , Jr. 2011 . Immunological, physiological, and behavioral effects of salmonella enterica carriage and shedding in experimentally infected finishing pigs . Foodborne Pathog. Dis . 8 : 623 – 630 . doi: 10.1089/fpd.2010.0735 Google Scholar Crossref Search ADS PubMed WorldCat Sapolsky , R. M. , L. C. Krey, and B. S. McEwen. 1986 . The neuroendocrinology of stress and aging: the glucocorticoid cascade hypothesis . Endocr. Rev . 7 : 284 – 301 . doi: 10.1210/edrv-7-3-284 Google Scholar Crossref Search ADS PubMed WorldCat Saxbe , D. E . 2008 . A field (researcher’s) guide to cortisol: tracking HPA axis functioning in everyday life . Health Psychol. Rev . 2 : 163 – 190 . doi: 10.1080/17437190802530812 Google Scholar Crossref Search ADS WorldCat Schmidt , M. , K. H. Lahrmann, C. Ammon, W. Berg, P. Schön, and G. Hoffmann. 2013 . Assessment of body temperature in sows by two infrared thermography methods at various body surface locations . J. Swine Health Prod . 21 : 203 – 209 . Google Scholar OpenURL Placeholder Text WorldCat Stanton , H. C. , and R. L. Mueller. 1976 . Sympathoadrenal neurochemistry and early weaning of swine . Am. J. Vet. Res . 37 : 779 – 783 . Google Scholar PubMed OpenURL Placeholder Text WorldCat Sudo , N . 2014 . Microbiome, HPA axis and production of endocrine hormones in the gut . In: N. Lyte and J. F. Cryan, editors, Microbial endocrinology: The microbiota-gut-brain axis in health and disease . Adv. Exp. Med. Biol., Springer , New York, NY . P. 177 – 194 . doi: 10.1007/978-1-4939-0897-4_8 Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Crossref Tannock , G. W. , and D. C. Savage. 1974 . Influences of dietary and environmental stress on microbial populations in the murine gastrointestinal tract . Infect. Immun . 9 : 591 – 598 . Google Scholar PubMed OpenURL Placeholder Text WorldCat Toruner , M. , E. V. Loftus , Jr, W. S. Harmsen, A. R. Zinsmeister, R. Orenstein, W. J. Sandborn, J. F. Colombel, and L. J. Egan. 2008 . Risk factors for opportunistic infections in patients with inflammatory bowel disease . Gastroenterology . 134 : 929 – 936 . doi: 10.1053/j.gastro.2008.01.012 Google Scholar Crossref Search ADS PubMed WorldCat Tsigos , C. , and G. P. Chrousos. 1994 . Physiology of the hypothalamic-pituitary-adrenal axis in health and dysregulation in psychiatric and autoimmune disorders . Endocrinol. Metab. Clin. North Am . 23 : 451 – 466 . Google Scholar Crossref Search ADS PubMed WorldCat Tsigos , C. , and G. P. Chrousos. 2002 . Hypothalamic-pituitary-adrenal axis, neuroendocrine factors and stress . J. Psychosom. Res . 53 : 865 – 871 . Google Scholar Crossref Search ADS PubMed WorldCat Walsh , M. C. , M. H. Rostagno, G. E. Gardiner, A. L. Sutton, B. T. Richert, and J. S. Radcliffe. 2012 . Controlling Salmonella infection in weanling pigs through water delivery of direct-fed microbials or organic acids. Part I: effects on growth performance, microbial populations, and immune status . J. Anim. Sci . 90 : 261 – 271 . doi: 10.2527/jas.2010-3598 Google Scholar Crossref Search ADS PubMed WorldCat Published by Oxford University Press on behalf of the American Society of Animal Science 2018. This work is written by (a) US Government employee(s) and is in the public domain in the US. Published by Oxford University Press on behalf of the American Society of Animal Science 2018.

Hai, Xiang,;Siyu, Chen,;Hui, Zhang,;Xu, Zhu,;Dan, Wang,;Huagui, Liu,;Jikun, Wang,;Tao, Yin,;Langqing, Liu,;Minghua, Kong,;Jian, Zhang,;Hua, Li,;Xingbo, Zhao,

doi: 10.1093/jas/sky314

L, Schumacher, Loni;A, Cochrane, Roger;R, Huss, Anne;T, Gebhardt, Jordan;C, Woodworth, Jason;R, Stark, Charles;K, Jones, Cassandra;Jianfa, Bai,;G, Main, Rodger;Qi, Chen,;Jianqiang, Zhang,;C, Gauger, Philip;M, DeRouchey, Joel;D, Goodband, Robert;

Sébastien, Goumon,;Iva, Leszkowová,;Marie, Šimečková,;Gudrun, Illmann,

doi: 10.1093/jas/sky324pmid: 30102369

Abstract Farrowing pens with temporary crating have been developed as a compromise between conventional farrowing crates and pens to better accommodate the welfare of both sow and piglets during lactation. However, not much is known about the behavioral and physiological consequences of early removal of confinement on the sow and piglets during lactation. The aim of this study was to assess the effects on sow and piglet performance of temporary crating until 3-d postpartum at 2 times points, immediately after confinement removal and 25 d into lactation. Sows were crated from 5-d prepartum either to weaning (permanently crated—PC group; N =14) or to D3 (83.0 ± 1.3 h) postpartum (Temporarily crated - TC group N = 13). Sow postural changes, activity, cortisol and IgA concentrations, and piglet body weight gain and behavior were assessed on D4 and on D25 postpartum, whereas piglet mortality was assessed throughout lactation. Data were analyzed using PROC GLM and PROC GENMOD of SAS. On D4 postfarrowing, TC sows were more active (10.9% vs. 7.1%; SEM: 0.8; P = 0.002), rolled more frequently (21.3% vs. 14.4%; SEM: 1.6; P = 0.008), and had lower IgA concentrations (139.7 vs. 75.2 µg/mL; SEM: 20.3; P = 0.040) than PC sows. No effects of housing were found (P > 0.05) on standing-to-lying movement or cortisol concentrations. No differences for any variables were found (P > 0.05) on D25. Mortality, body weights, and activity levels at the udder or in the pen of pigs born to PC sows did not differ (P > 0.05) from those of piglets born to TC sows on D4 nor on D25. This study indicates that removal of confinement on the 4th-d postpartum may have had small short-term positive effects on sow behavior and stress levels (as measured by IgA), and that it did not impair piglets’ behavior and performance during lactation. Therefore, this work suggests that temporary crating limited to the first 3-d postpartum might be a feasible alternative to improve welfare under intensive production conditions. INTRODUCTION Temporary confinement provides lactating sows with the opportunity to move freely in the farrowing pen, once the crate has been opened a few days after parturition. This will allow less restrictive interactions with offspring and improve (physical) comfort, which is drastically compromised in crates (Johnson and Marchant-Forde, 2009; Baxter et al., 2011). Very little is known about the short- and long-term effects of temporary confinement on sow behavior and physiology. The removal of confinement may stimulate activity in the newly available space. Yet, relatively little research has been done to quantitatively assess the activity of temporarily crated sows once loose. Permanent confinement of sows leads to chronic stress at the end of lactation (Cronin et al., 1991; Jarvis et al., 2006; Yin et al., 2016). Moreover, prolonged stress may have deleterious effects on immunity (de Groot et al., 2001). Yet, it is not known whether temporary confinement reduces adrenal and immune reactivity of the sow in the long-term and whether opening the crate has any short-term effects on sow physiology. Reducing the confinement period during lactation may improve sow welfare. However, piglets’ survival may be put at risk as the absence of confinement allows more sow postural changes which may increase crushing events (Marchant et al., 2000; Melišová et al., 2014). In most research on temporary confinement during lactation, sows were let loose after the 4th-d pp (Lambertz et al., 2015; Chidgey et al., 2016a). Not much is known about whether earlier removal of confinement may have detrimental effects on piglet performances and behavior (3rd-d pp; Killbride et al., 2012; Condous et al., 2016; Singh et al., 2017). Therefore, the aim of this study was to investigate whether confinement until the 3rd-d pp had short- (24 h postopening of the crate) and/or long-term (D25 pp) effects on sow activity and stress levels and to assess whether it influenced piglet performance and behavior. MATERIALS AND METHODS This study was carried out from July 2015 to July 2016 at the research farm of the institute of animal science in Prague, Czech Republic, and was conducted in accordance with Czech Central Committee for Protection of Animals number 60444/2011–17214. Animals and Management A total of 27 Large White × Landrace sows (parity: 2.5 ± 0.5, range: 1 to 12) inseminated with Large White × Pietrain boar semen were used. Sows were moved from a group-housing gestation unit to a farrowing unit 5 d before estimated parturition date and immediately crated. Sows were randomly allocated over 14 batches on the basis of parity to one of the following two treatments: sows permanently crated (PC; N = 14 sows and 192 piglets born; litter size: 13.7 ± 0.7 piglets) were confined in a crate from 5 d prefarrowing until weaning (approximately 28 d postfarrowing), whereas sows temporarily crated (TC; N = 13 sows and 172 piglets born, litter size: 13.5 ± 0.7 piglets) were confined in crates from 5 d prefarrowing to day 3 (83.0 ± 1.3 h; day 0 being the day of farrowing) postpartum on which sows were let loose (opening of the crate) until weaning. A batch consisted of 2 sows, one for each treatment, housed in adjacent pens of the same type of farrowing system (Figure 1), with one treatment let loose and the other one remained confined. One sow in the TC treatment was removed due to illness. The crate was always opened around 1000 h, after completion of a nursing. There was no equalization nor cross fostering of the litters. All sows were only familiar with permanent farrowing crating prior to the experiment. Farrowing was supervised through 24-h video recordings. Sows were fed a standard lactation diet (17% CP, 13.75 MJ DE kg−1) twice a day, and water was available ad libitum from a nipple for the sow and another for the piglets. The sows received 1 bag of chopped straw each morning and evening until they had farrowed, and during the whole lactation. Pens were cleaned once a day. All piglets received an iron injection and males were surgically castrated during the first week of farrowing and ear tagged 3 d after birth. They received creep food from day 7 postfarrowing. There was no difference in the handling of the sows and litters in crates and those housed in pens. Farrowing pens measured 5.88 m2 and were equipped with movable bars, which enabled them to be modified into crates and vice versa (Figure 1). The sow area (the part of the pen accessible to the sow) for crated sows measured 1.63 m2. When the sows were let loose, the sides of the crate were opened and placed along the sidewalls. In this configuration, the sow area measured 4.63 m2. There were no protection rails along the sidewalls or sloped wall when the sow was lying down. Solid concrete flooring was present in the whole pen. The creep area (the part of the pen accessible only to the piglets) measured 1.25 m2 and was noncovered, bedded with straw, and had 2 hanging heat lamps during the whole lactation. Figure 1. View largeDownload slide Farrowing pen equipped with movable bars, which allowed the modification from farrowing crate (left) to farrowing pen (right). A = sow area when farrowing crate is open; B = sow area when sow is crated; C = creep area for piglets; D = heating lamp; E = feed trough; F = piglet anticrushing bars. All measurements are in centimeters. Figure 1. View largeDownload slide Farrowing pen equipped with movable bars, which allowed the modification from farrowing crate (left) to farrowing pen (right). A = sow area when farrowing crate is open; B = sow area when sow is crated; C = creep area for piglets; D = heating lamp; E = feed trough; F = piglet anticrushing bars. All measurements are in centimeters. Data Collection Behavioral observations. Behaviors of sows and their litters were continuously video recorded on days 3, 4, and 25 postpartum using an overhead CCTV camera for each pen (Panasonic CCTV, WV CP 470, Osaka, Japan) and software NUUO (IP Surveillance System, NVR/DVR/NVDR, Taipei, Taiwan) (Table 1). Sows were fed twice during observation hours, at 0800 and 1500 h. No observations were recorded from 10 min before to 10 min after feeding or while staff performed husbandry tasks (piglet processing, vaccinations, or any other veterinary procedure). From video recordings, sow and piglet (all piglets in every litter) activity were sampled by 1 observer using fixed interval scan sampling of 5 min over the 24-h period following the opening of the crate (D3 to D4) and D25. In contrast, posture changes were assessed using all occurrence sampling. Behaviors were analyzed by summing the occurrences of each piglet or sow behavioral parameters during those 2 periods. Table 1. Behaviors of sows and piglets Behaviors Definitions Sow activity Inactive Sow sits or lies in lateral or sternal recumbency Active Sow is standing or moving in standing position Sow postural changes Descending movements Standing to lying Lateral movements Rolling: from a sternal to lateral recumbency and vice versa Piglet activity Active at the udder Piglet is teat seeking or massaging the udder Active in the pen Piglet is walking, standing or sitting in the pen (i.e., any area within the farrowing space that was not in the creep, or at the sow’s udder) Behaviors Definitions Sow activity Inactive Sow sits or lies in lateral or sternal recumbency Active Sow is standing or moving in standing position Sow postural changes Descending movements Standing to lying Lateral movements Rolling: from a sternal to lateral recumbency and vice versa Piglet activity Active at the udder Piglet is teat seeking or massaging the udder Active in the pen Piglet is walking, standing or sitting in the pen (i.e., any area within the farrowing space that was not in the creep, or at the sow’s udder) View Large Table 1. Behaviors of sows and piglets Behaviors Definitions Sow activity Inactive Sow sits or lies in lateral or sternal recumbency Active Sow is standing or moving in standing position Sow postural changes Descending movements Standing to lying Lateral movements Rolling: from a sternal to lateral recumbency and vice versa Piglet activity Active at the udder Piglet is teat seeking or massaging the udder Active in the pen Piglet is walking, standing or sitting in the pen (i.e., any area within the farrowing space that was not in the creep, or at the sow’s udder) Behaviors Definitions Sow activity Inactive Sow sits or lies in lateral or sternal recumbency Active Sow is standing or moving in standing position Sow postural changes Descending movements Standing to lying Lateral movements Rolling: from a sternal to lateral recumbency and vice versa Piglet activity Active at the udder Piglet is teat seeking or massaging the udder Active in the pen Piglet is walking, standing or sitting in the pen (i.e., any area within the farrowing space that was not in the creep, or at the sow’s udder) View Large Salivary cortisol and IgA concentrations. Saliva samples were obtained (1 per animal per day) on D2 (control value, 24 h before opening of the crate), D4 (24 h after opening of the crate), and D25 postpartum by allowing the pig to chew on 2 cotton swabs (cotton swabs 150 × 4 mm WA 2PL; Heinz Herenz, Hamburg, Germany) until they were thoroughly moistened. Cotton swabs were attached at the end of a wooden rod, long enough that it was possible to obtain samples without entering the pen, thereby minimizing the disturbance of the animals. Saliva collection generally took less than 2 min and samples were taken on calm and lying sows, not during nursing bouts. Samples were always collected at the same time (around 0945 h) to avoid any confounding effect of circadian rhythm (Muneta et al., 2010, 2011). Saliva was extracted from the swabs by centrifugation at 1000 g for 15 min. Samples were stored at −20 °C until required for assay. Cortisol concentrations were measured by High Sensitive SALIVARY CORTISOL Enzyme Immunoassay Kit (Salimetrics, State College, PA 16803), with concentration ranging from 0.12 to 30 ng/mL, analytical sensitivity of 0.07 ng/mL, and a sample volume of 25 μL. IgA concentrations were determined using Pig IgA ELISA Kit (Bethyl Laboratories, Inc., TX 77356), with concentration ranging from 1.37 to 1000 ng/mL and a salivary sample with a dilution of 1:1000 by provided dilution buffer. Cortisol and IgA salivary concentrations were analyzed for each sow by using the difference in concentrations between D4 and D2, and D25 and D2, respectively. Piglet mortality and body weight gain. Piglet mortality rate was assessed at the end of farrowing (after expulsion of the placenta) and every day until D25 postpartum. Piglets were individually marked by an animal marking crayon (Raidex, Germany) and were weighed at the same time (around 0930 h) on D3 (30 min before opening), D4 (24 h after opening of the crate), and D25. The 2 variables were analyzed by using, for each litter, the difference in individual piglet weight and mortality, respectively, from D3 to D4 (short-term effect) and D4 to D25 (long-term effect). Statistical Analysis All statistical analyses were performed using SAS 9.4 (SAS Institute Inc., Cary, NC). The fixed effects included in all models were 1 class variable, housing (temporary or permanent crating), and 2 continuous variables, parity and litter size. The identity of the batch was considered as a random effect in the models. The analysis of piglet mortality was done at a litter level using a generalized linear model with a binomial link (PROC GENMOD). The other response variables were analyzed by generalized linear models (PROC GLM). Results were considered statistically significant when P ≤ 0.05 and as tending to differ when 0.05 < P ≤ 0.10. Only significant results of parity and litter size effects are presented below. RESULTS Sow Hormone Stress Levels, Activity, and Postural Changes Short-term effects of removal of confinement. Housing conditions had an influence on the sow activity, IgA concentrations, and the frequency of rolling behavior (P = 0.002, P = 0.040, and P = 0.008, respectively), with TC sows being more active, having lower IgA concentrations and rolling more frequently than PC sows (Figure 2; Table 2). No effects were found on standing-to-lying movements or cortisol concentrations (P = 0.449 and P = 0.360, respectively). Litter size influenced rolling frequency (P = 0.009), with sows with bigger litters rolling more. Parity had an effect on sow activity (P = 0.04) with activity levels decreasing with increased parity. Figure 2. View largeDownload slide Cortisol (a) and IgA (b) concentrations 24 h before opening (baseline) and 24 h postopening (24 h p.o.) of the crate and on D25 in permanent and crated sows. (*P < 0.05). Figure 2. View largeDownload slide Cortisol (a) and IgA (b) concentrations 24 h before opening (baseline) and 24 h postopening (24 h p.o.) of the crate and on D25 in permanent and crated sows. (*P < 0.05). Table 2. Short-1 and long-term2 effects of housing conditions on sow activity and postural changes Permanent crating (N = 14) Temporary crating (N = 13) SEM P value Short-term effect3  Active time (%) 7.1 10.9 0.8 0.002*  Standing to lying (%) 4.9 5.6 0.6 0.449  Rolling (%) 14.4 21.3 1.6 0.008* Long-term effect3  Active time (%) 12.6 11.1 1.6 0.524  Standing to lying (%) 5.5 6.5 0.7 0.393  Rolling (%) 15.1 22.9 2.7 0.073 Permanent crating (N = 14) Temporary crating (N = 13) SEM P value Short-term effect3  Active time (%) 7.1 10.9 0.8 0.002*  Standing to lying (%) 4.9 5.6 0.6 0.449  Rolling (%) 14.4 21.3 1.6 0.008* Long-term effect3  Active time (%) 12.6 11.1 1.6 0.524  Standing to lying (%) 5.5 6.5 0.7 0.393  Rolling (%) 15.1 22.9 2.7 0.073 124-h period after opening of the crate. 224-h period on D25 postpartum. 3Percentage of fixed interval scan sampling of 5 min during which the animal was seen displaying the behavior over the given period. * P < 0.05. View Large Table 2. Short-1 and long-term2 effects of housing conditions on sow activity and postural changes Permanent crating (N = 14) Temporary crating (N = 13) SEM P value Short-term effect3  Active time (%) 7.1 10.9 0.8 0.002*  Standing to lying (%) 4.9 5.6 0.6 0.449  Rolling (%) 14.4 21.3 1.6 0.008* Long-term effect3  Active time (%) 12.6 11.1 1.6 0.524  Standing to lying (%) 5.5 6.5 0.7 0.393  Rolling (%) 15.1 22.9 2.7 0.073 Permanent crating (N = 14) Temporary crating (N = 13) SEM P value Short-term effect3  Active time (%) 7.1 10.9 0.8 0.002*  Standing to lying (%) 4.9 5.6 0.6 0.449  Rolling (%) 14.4 21.3 1.6 0.008* Long-term effect3  Active time (%) 12.6 11.1 1.6 0.524  Standing to lying (%) 5.5 6.5 0.7 0.393  Rolling (%) 15.1 22.9 2.7 0.073 124-h period after opening of the crate. 224-h period on D25 postpartum. 3Percentage of fixed interval scan sampling of 5 min during which the animal was seen displaying the behavior over the given period. * P < 0.05. View Large Long-term effects of removal of confinement. Housing conditions had no effects (P > 0.05) on sow activity levels, standing-to-lying movements, or on cortisol and IgA concentrations. Rolling tended (P = 0.073) to be more frequent in TC sows than in PC sows. Piglet Mortality, Body Weight Gain, and Activity Short-term effects of removal of confinement. Because of the low incidence of piglet death over the first 24 h postopening, statistical analysis was not appropriate, and mortality data are presented in descriptive form (Table 3). Body weights and activity levels at the udder or in the pen of pigs born to PC sows did not differ from those of piglets born to TC sows (P = 0.423, P = 0.436, and P = 0.195, respectively). Litter size influenced body weight gain (P < 0.001) with decreasing weight gain in larger litters. Parity influenced activity at the udder (P = 0.016) with higher parity associated with increased piglet activity. Table 3. Short-1 and long-term2 effects of housing conditions on piglet mortality, weight gain, and activity Permanent crating (N = 14) Temporary crating (N = 13) SEM P value Short-term effect  Mortality3 (%) 1.6 1.7 0.1 .  Weight gain (g) 167.9 178.7 9.0 0.423  Active at udder4 (%) 19.1 18.2 0.8 0.436  Active in pen4 (%) 12.6 14.3 0.9 0.195 Long-term effect  Mortality (%) 8.9 8.1 2.4 0.902  Weight gain (g) 4091.1 4359.0 166.8 0.269  Active at udder4 (%) 16.5 16.8 1.3 0.871  Active in pen4 (%) 15.6 18.3 1.4 0.208 Permanent crating (N = 14) Temporary crating (N = 13) SEM P value Short-term effect  Mortality3 (%) 1.6 1.7 0.1 .  Weight gain (g) 167.9 178.7 9.0 0.423  Active at udder4 (%) 19.1 18.2 0.8 0.436  Active in pen4 (%) 12.6 14.3 0.9 0.195 Long-term effect  Mortality (%) 8.9 8.1 2.4 0.902  Weight gain (g) 4091.1 4359.0 166.8 0.269  Active at udder4 (%) 16.5 16.8 1.3 0.871  Active in pen4 (%) 15.6 18.3 1.4 0.208 124-h period after opening of the crate. 2D25 postpartum (24 h) for activity and period from D4 to D25 for mortality and weight gain. 3Descriptive statistics only. 4Percentage of fixed interval scan sampling of 5 min during which the animal was seen displaying the behavior over the given period. View Large Table 3. Short-1 and long-term2 effects of housing conditions on piglet mortality, weight gain, and activity Permanent crating (N = 14) Temporary crating (N = 13) SEM P value Short-term effect  Mortality3 (%) 1.6 1.7 0.1 .  Weight gain (g) 167.9 178.7 9.0 0.423  Active at udder4 (%) 19.1 18.2 0.8 0.436  Active in pen4 (%) 12.6 14.3 0.9 0.195 Long-term effect  Mortality (%) 8.9 8.1 2.4 0.902  Weight gain (g) 4091.1 4359.0 166.8 0.269  Active at udder4 (%) 16.5 16.8 1.3 0.871  Active in pen4 (%) 15.6 18.3 1.4 0.208 Permanent crating (N = 14) Temporary crating (N = 13) SEM P value Short-term effect  Mortality3 (%) 1.6 1.7 0.1 .  Weight gain (g) 167.9 178.7 9.0 0.423  Active at udder4 (%) 19.1 18.2 0.8 0.436  Active in pen4 (%) 12.6 14.3 0.9 0.195 Long-term effect  Mortality (%) 8.9 8.1 2.4 0.902  Weight gain (g) 4091.1 4359.0 166.8 0.269  Active at udder4 (%) 16.5 16.8 1.3 0.871  Active in pen4 (%) 15.6 18.3 1.4 0.208 124-h period after opening of the crate. 2D25 postpartum (24 h) for activity and period from D4 to D25 for mortality and weight gain. 3Descriptive statistics only. 4Percentage of fixed interval scan sampling of 5 min during which the animal was seen displaying the behavior over the given period. View Large Long-term effects of removal of confinement. Housing systems had no effects on the probability of piglet death from D4 to D25 (P = 0.902), body weight (P = 0.269), or activity levels at the udder or in the pen (P = 0.877 and P = 0.103, respectively). Litter size and parity influenced body weight gain (P = 0.030 and P = 0.041, respectively) with decreasing weight gain in larger litters and in higher parity. Nonudder related activity levels in the pen were lower (P = 0.007) for piglets from larger litters and tended to be lower for piglets from higher parity sows (P = 0.062). DISCUSSION Sow Stress Levels and Behavior Short-term effects of removal of confinement. The removal of confinement did not influence salivary cortisol concentrations 24 h after opening of the crate which suggests similar adrenal reactivity in crated and loose sows a day after opening of the crate. Research on cortisol levels of sows in early lactation is rather inconclusive as to whether confinement is a chronic stressor at that stage. Although some studies found no differences in salivary cortisol concentrations between permanently loose housed and crated sows during the first week of lactation (Cronin et al., 1991; Biensen et al., 1996), other found either greater (Oliviero et al., 2008) or lower cortisol levels (Hales et al., 2016) in crated sows. Interestingly, IgA levels of temporarily crated sows decreased (−47%) after opening of the crate, whereas those of crated sows did not vary. The valence of this response remains unclear. To the best of our knowledge, no other study has assessed the short- or long-term effect of removal of confinement postpartum on sow IgA levels. Only 2 studies (Muneta et al., 2010; Escribano et al., 2015) assessed pig salivary IgA levels in response to various stressors (snitch restraint, isolation, and mixing). They both reported an increase of IgA levels after exposure to these negative conditions. Based on these findings, the decrease found in the present study could indicate that loose sows were experiencing lower stress levels than crated sows 24 h after opening of the crate. However, contradicting this interpretation, studies on other species have reported a decrease in IgA after exposure to similar stressors (dogs: Svobodová et al. (2014); rats: Guhad and Hau (1996)). The different hormone responses in the present study may tentatively be explained by the difference in kinetics and sensibility of the response to the removal of confinement between those stress biomarkers. Cortisol is known to quickly respond (within 30 min) to different types of stressor in pigs (Muneta et al., 2010; Escribano et al, 2015). On the other hand, limited research in pigs shows that IgA has an immediate response (within 30 min) after acute physical stress (Muneta et al., 2010) but a delayed response (1–3 days) after exposure to social stress (Escribano et al., 2015). Therefore, it may be speculated that, in the present study, there was a delayed response in IgA levels and a change in cortisol levels soon after opening the crate which was only short lasting, and thus not recorded as saliva was sampled 24 h after removal of confinement. As expected, opening the crate, thus increasing the space allowance, made temporarily confined sows spend more time standing and be more active than permanently crated sows over the first 24 h after removal of confinement. However, although significant, the increase in activity is small (+3.8%) and should be interpreted with caution. This increase may have reflected a slightly greater motivation for exploration of the new space available to the temporarily crated sows. Our finding contrasts those of Lambertz et al. (2015) and Chidgey et al. (2016b) who found no short-term differences in the amount of time spent active between temporary confined sow (until D5 or until D7 and D14, respectively) and permanently crated sows. The smaller size of the pen and late opening of the crate (D7 or D14, Lambert et al., 2015) and/or the sampling period (restricted to 4 daytime periods, Chidgey et al., 2016b) may explain the differences with the present study. Removal of confinement was found to increase (+6.9 %) the frequency of rolling events over the first 24 h postopening. Similarly, Hales et al. (2016) found an increase in the number of rolling events in loose sows compared with crated sows during the first few days postfarrowing. Greater frequency of rolling events has been shown to indicate greater restlessness or discomfort during or outside of nursing bouts (Weary et al., 1996; Harris and Gonyou, 1998; Damm et al., 2005; Bozděchová et al., 2014). Even though it cannot be completely ruled out, it may not have been the case in the present study as other indicators such as increased standing-to-lying postural changes or stress levels were absent. Thus, it seems more likely that the increase in rolling was simply linked to the increase in space allowing sows to roll more easily, whereas rolling was more prevented in crates (Damm et al., 2005; Danholt et al., 2011). Long-term effects of removal of confinement. The removal of confinement had no effects on salivary cortisol and IgA levels on D25, which may indicate that adrenal and immune reactivity levels were similar at the end of lactation in both housing conditions. The cortisol findings of the present study are in contradiction with the scarce literature on the effect of farrowing environment on cortisol concentration at the end of lactation. Comparing pens and crates, Cronin et al. (1991), Jarvis et al. (2006), and Yin et al. (2016) found higher plasma cortisol levels in crated sows compared with sows housed in pens only at the end of lactation (days 28, 29, and 35 postfarrowing, respectively), indicative of chronic stress. Differences between studies may be explained by the different sampling methods (salivary vs. blood), types of pens, parity (gilt vs. multiparous), and probably the sampling periods. In the present study, the saliva samples on D25 were collected after the peak of milk output (D21) and possibly before the point at which sows may start reducing milk output and contacts with piglets (sow-offspring conflict). Like in the other studies, higher cortisol concentrations in permanently confined sows could have speculatively been observed with later sampling as loose sows may have had more control over their investment compared with crated sows (Pajor et al., 2000, 2002). The effect of opening of the crate on IgA concentrations of temporarily crated sows seen on D4 was no longer observed at the end of lactation. This may support the assumption of a transient novelty effect of the removal of confinement. However, due to the only 2 saliva samples being collected in the present study (on D4 and D25), it is unfortunately impossible to accurately know how long the decrease of IgA in temporary crated sows lasted for. Further research should investigate in more detail the temporal dynamics of salivary IgA and cortisol responses to confinement to provide a better picture of the sow stress levels during the entirety of lactation. Moreover, other indicators such as heart rate variability may be valuable to assess the effect of temporary confinement on sow physiology. The similar activity levels found in both treatments at the end of lactation suggest that increasing the space available did not steadily motivate sows to be more active. In agreement with the findings of Lambertz et al. (2015), sows spent a low amount of time (approximately 10%) active in late lactation regardless of the farrowing system in which they were housed. The overall low activity levels may be related to sows preserving their energy levels and avoiding a catabolic state as a result of the metabolic burden of rearing a litter (Valros et al., 2003). Although it did not increase activity levels, the removal of confinement was assumed to allow a wider range of behaviors and motion. Hence, it may have enabled temporarily crated sows to interact more with their environment, which is paramount to good animal welfare as it may decrease boredom (Burn, 2017). Unfortunately, in the present study, the activity levels were roughly assessed, considering only the time spent active and standing and not providing detailed information on the behaviors of the sows during the active bouts. Thus, it would be valuable that further refined research focuses on the quality of activity (e.g., total distance walked, qualitative and quantitative assessment of interactions with environment) in order to assess the long-term potential benefits of removal of temporary confinement such as leg health or improved affective states. Piglet Performance and Behavior Similar mortality rates found in both treatments in the present study indicate that piglets were not at a higher risk of dying in temporary crating pens compared with permanent crates in the short and long term. Therefore, it is possible to achieve an acceptable overall mortality (from birth to D25: 12.5%) in regard to production standards with this system. The absence of short-term differences in piglet survival suggests that sows were as careful in pens as in crates, even though TC sows were slightly more active and changed postures more frequently. Our mortality results are consistent with the short-term and long-term findings of many other studies in which sows were loose for a few (2 to 4) days after farrowing (Stabenow and Manteuffel, 2002; Mouesten et al., 2013; Hales et al., 2015; Chidgey et al., 2016a; Condous et al., 2016; Singh et al., 2017). Confinement until D3 pp did not influence piglet short- and long-term weight gains, which is in line with other studies comparing similar crating conditions (Moustsen et al., 2013; Condous et al., 2016; Singh et al., 2017). Even though it remains to be tested, the similar weight gains in the present study may indicate that nursing and suckling behaviors may not have been disturbed shortly after removal of confinement, and possibly at the end of lactation. However, the latter case is very speculative as creep food intake may have played a role (Oostindjer et al., 2010). This would contrast the hypothesis (better nursing behavior in PC sows) of Chidgey et al., (2015) for recording a higher weaning weight gain (+0.1 kg) in piglets of sows housed in temporary crating (for 4 d) compared with those of sows housed in permanent crating. Along with mortality and weight gain, piglet activity (walking, standing, and massaging) was not modified by removal of confinement during lactation, providing additional evidence supporting that temporary crating until the 3rd-d postpartum was not detrimental to the piglets during lactation, even when sow activity and postures were moderately more frequent. Although contrasting results of other studies showing that piglet activity level increases with increased sow activity level and posture changes during the first week of lactation (Chidgey et al., 2016b), our results were in line with those reported by Singh et al. (2017), who assessed general activity in the pen on D4, 11, and 18 pp between piglets of sows in crates or temporary crating (until D3 pp). Further investigation could look at other specific behaviors indicative of positive welfare such as play, which has been shown to be increased in enriched farrowing pens with increased space and straw available for piglets and sow (Chaloupkova et al., 2007). CONCLUSION The results of this study suggest that loose-housing of sows after a short postnatal period of confinement in a crate may have small positive effects on the sow’s welfare behavior in the short term only (as reflected by activity and IgA levels). More detailed work is needed to assess the potential benefits of temporary confinement on sow welfare throughout lactation. This study also indicates that confining the sow during farrowing and until day 3 postpartum is sufficient to ensure a similar preweaning piglet survival, growth, and behavior compared with sows which were permanently crated throughout lactation. From a piglet welfare standpoint, short temporary confinement of the sow may then be considered as a safe alternative housing option to permanent confinement. Conflict of interest statement. None declared. Footnotes 1 Research supported by grant numbers NAZV QJ1610390 and MZE-RO0718 from the Ministry of Agriculture, Czech Republic. The authors would like to acknowledge Zuzanna Kocourková for her help with data collection. LITERATURE CITED Baxter , E. M. , A. B. Lawrence , and S. A. Edwards . 2011 . Alternative farrowing systems: design criteria for farrowing systems based on the biological needs of sows and piglets . Animal 5 : 580 – 600 . doi: https://doi.org/10.1017/S1751731110002272 Google Scholar Crossref Search ADS PubMed Biensen , N. J. , E. H. von Borell , and S. P. Ford . 1996 . Effects of space allocation and temperature on periparturient maternal behaviors, steroid concentrations, and piglet growth rates . J. Anim. Sci . 74 : 2641 – 2648 . Google Scholar Crossref Search ADS PubMed Bozděchová , B. , G. Illmann , I. L. Andersen , J. Haman , and R. Bhrlenbruch . 2014 . Litter competition during nursings and its effect on sow response on Day 2 postpartum . Appl. Anim. Behav. Sci . 150 : 9 – 16 . doi:10.1016/ j.applanim.2013.10.006 Google Scholar Crossref Search ADS Burn , C. C . 2017 . Bestial boredom: a biological perspective on animal boredom and suggestions or its scientific investigation . Anim. Behav . 130 : 141 – 151 . doi:10.1016/ j.anbehav.2017.06.006 Google Scholar Crossref Search ADS Chaloupková , H. , G. Illmann , K. Neuhauserová , M. Tománek , and L. Valis . 2007 . Preweaning housing effects on behavior and physiological measures in pigs during the suckling and fattening periods . J. Anim. Sci . 85 : 1741 – 1749 . doi: https://doi.org/10.2527/jas.2006-504 Google Scholar Crossref Search ADS PubMed Chidgey , K. L. , P. C. H. Morel , K. J. Stafford , and I. W. Barugh . 2015 . Sow and piglet productivity and sow reproductive performance in farrowing pens with temporary crating and farrowing crates on a commercial New Zealand pig farm . Livest. Sci . 173 : 87 – 94 . doi:10.1016/j.livsci.2015.01.003 Google Scholar Crossref Search ADS Chidgey , K. L. , P. C. H. Morel , K. J. Stafford , and I. W. Barugh . 2016a . The performance and behavior of gilts and their piglets is influenced by whether gilts were born and reared in farrowing crates or farrowing pens . Livest. Sci . 193 : 51 – 57 . doi:10.1016/j.livsci.2016.09.011 Google Scholar Crossref Search ADS Chidgey , K. L. , P. C. H. Morel , K. J. Stafford , and I. W. Barugh . 2016b . Observations of sows and piglets housed in farrowing pens with temporary crating or farrowing crates on a commercial farm . Appl. Anim. Behav. Sci . 176 : 12 – 18 . doi:10.1016/j.applanim.2016.01.004 Google Scholar Crossref Search ADS Condous , P. C. , K. J. Plush , A. J. Tilbrook , and W. H. van Wettere . 2016 . Reducing sow confinement during farrowing and in early lactation increases piglet mortality . J. Anim. Sci . 94 : 3022 – 3029 . doi: https://doi.org/10.2527/jas.2015-0145 Google Scholar Crossref Search ADS PubMed Cronin , G. M. , J. L. Barnett , F. M. Hodge , J. A. Smith , and T. H. McCallum . 1991 . The welfare of pigs in two farrowing/lactation environments: cortisol responses of sows . Appl. Anim. Behav. Sci . 32 : 117 – 127 . doi:10.1016/S0168-1591(05)80036-X Google Scholar Crossref Search ADS Damm , B. I. , B. Forkman , and L. J. Pedersen . 2005 . Lying down and rolling behaviour in sows in relation to piglet crushing . Appl. Anim. Behav. Sci . 90 : 3 – 20 . doi:10.1016/ j.applanim.2004.08.008 Google Scholar Crossref Search ADS Danholt , L. , V. A. Moustsen , M. B. F. Nielsen , and A. R. Kristensen . 2011 . Rolling behaviour of sows in relation to piglet crushing on sloped versus level floor pens . Livest. Sci . 141 : 59 – 68 . doi:10.1016/j.livsci.2011.05.005 Google Scholar Crossref Search ADS de Groot , J. , M. A. Ruis , J. W. Scholten , J. M. Koolhaas , and W. J. Boersma . 2001 . Long-term effects of social stress on antiviral immunity in pigs . Physiol. Behav . 73 : 145 – 158 . doi:10.1016/S0168-1591(05)80036-X Google Scholar Crossref Search ADS PubMed Escribano , D. , A. Tvarijonaviciute , F. Tecles , and J. J. Cerón . 2015 . Serum paraoxonase type-1 activity in pigs: assay validation and evolution after an induced experimental inflammation . Vet. Immunol. Immunopathol . 163 : 210 – 215 . doi: https://doi.org/10.1016/j.vetimm.2014.12.002 Google Scholar Crossref Search ADS PubMed Guhad , F. A. , and J. Hau . 1996 . Salivary iga as a marker of social stress in rats . Neurosci. Lett . 216 : 137 – 140 . doi:10.1016/0304-3940(96)13037-8 Google Scholar Crossref Search ADS PubMed Hales , J. , V. A. Moustsen , M. B. Nielsen , and C. F. Hansen . 2015 . Temporary confinement of loose-housed hyperprolific sows reduces piglet mortality . J. Anim. Sci . 93 : 4079 – 4088 . doi: https://doi.org/10.2527/jas.2015-8973 Google Scholar Crossref Search ADS PubMed Hales , J. , V. A. Moustsen , M. B. F. Nielsen , and C. F. Hansen . 2016 . The effect of temporary confinement of hyperprolific sows in Sow Welfare and Piglet protection pens on sow behaviour and salivary cortisol concentrations . Appl. Anim. Behav. Sci . 183 : 19 – 27 . doi:10.1016/ j.applanim.2016.07.008 Google Scholar Crossref Search ADS Harris , M. J. , and H. W. Gonyou . 1998 . Increasing available space in a farrowing crate does not facilitate postural changes or maternal responses in gilts . Appl. Anim. Behav. Sci . 59 : 285 – 296 . doi:10.1016/S0168-1591(98)00142-7 Google Scholar Crossref Search ADS Jarvis , S. , R. B. D’Eath , S. K. Robson , and A. B. Lawrence . 2006 . The effect of confinement during lactation on the hypothalamic-pituitary-adrenal axis and behaviour of primiparous sows . Physiol. Behav . 87 : 345 – 352 . doi: https://doi.org/10.1016/j.physbeh.2005.10.004 Google Scholar Crossref Search ADS PubMed Johnson , A. K. , and R. M. Marchant-Forde . 2009 . Welfare of pigs in the farrowing environment . In J. N. Marchant-Forde editor, The welfare of pigs . Springer Science + Business Media B.V ., Dordrecht, The Netherlands , pp. 141 – 188 . Kilbride , A. L. , M. Mendl , P. Statham , S. Held , M. Harris , S. Cooper , and L. E. Green . 2012 . A cohort study of preweaning piglet mortality and farrowing accommodation on 112 commercial pig farms in England . Prev. Vet. Med . 104 : 281 – 291 . doi: https://doi.org/10.1016/j.prevetmed.2011.11.011 Google Scholar Crossref Search ADS PubMed Lambertz , C. , M. Petig , A. Elkmann , and M. Gauly . 2015 . Confinement of sows for different periods during lactation: effects on behaviour and lesions of sows and performance of piglets . Animal 9 : 1373 – 1378 . doi: https://doi.org/10.1017/S1751731115000889 Google Scholar Crossref Search ADS PubMed Marchant , J. N. , A. R. Rudd , M. T. Mendl , D. M. Broom , M. J. Meredith , S. Corning , and P. H. Simmins . 2000 . Timing and causes of piglet mortality in alternative and conventional farrowing systems . Vet. Rec . 147 : 209 – 214 . Google Scholar Crossref Search ADS PubMed Melišová , M. , G. Illmann , H. Chaloupková , and B. Bozděchová . 2014 . Sow postural changes, responsiveness to piglet screams, and their impact on piglet mortality in pens and crates . J. Anim. Sci . 92 : 3064 – 3072 . doi: https://doi.org/10.2527/jas.2013-7340 Google Scholar Crossref Search ADS PubMed Moustsen , V. A. , J. Hales , H. P. Lahrmann , P. M. Weber , and C. F. Hansen . 2013 . Confinement of lactating sows in crates for 4 days after farrowing reduces piglet mortality . Animal 7 : 648 – 654 . doi: https://doi.org/10.1017/S1751731112002170 Google Scholar Crossref Search ADS PubMed Muneta , Y. , Y. Minagawa , T. Nakane , T. Shibahara , T. Yoshikawa , and Y. Omata . 2011 . Interleukin-18 expression in pig salivary glands and salivary content changes during acute immobilization stress . Stress 14 : 549 – 556 . doi: https://doi.org/10.3109/10253890.2011.565392 Google Scholar Crossref Search ADS PubMed Muneta , Y. , T. Yoshikawa , Y. Minagawa , T. Shibahara , R. Maeda , and Y. Omata . 2010 . Salivary iga as a useful non-invasive marker for restraint stress in pigs . J. Vet. Med. Sci . 72 : 1295 – 1300 . doi:10.1292/jvms.10-0009 Google Scholar Crossref Search ADS PubMed Oliviero , C. , M. Heinonen , A. Valros , O. Hälli , and O. A. Peltoniemi . 2008 . Effect of the environment on the physiology of the sow during late pregnancy, farrowing and early lactation . Anim. Reprod. Sci . 105 : 365 – 377 . doi: https://doi.org/10.1016/j.anireprosci.2007.03.015 Google Scholar Crossref Search ADS PubMed Oostindjer , M. , J. E. Bolhuis , M. Mendl , S. Held , W. Gerrits , H. van den Brand , and B. Kemp . 2010 . Effects of environmental enrichment and loose housing of lactating sows on piglet performance before and after weaning . J. Anim. Sci . 88 : 3554 – 3562 . doi: https://doi.org/10.2527/jas.2010-2940 Google Scholar Crossref Search ADS PubMed Pajor , E. A. , D. L. Kramer and D. Fraser . 2000 . Regulation of contact with offspring by domestic sows: temporal patterns and individual variation . Ethol . 106 : 37 – 51 . doi:10.1046/j.14390310.2000.00494.x Google Scholar Crossref Search ADS Pajor , E. A. , D. M. Weary , C. Caceres , D. Fraser and D. L. Kramer . 2002 . Alternative housing for sows and litters: part 3. Effects of piglet diet quality and sow controlled housing on performance and behaviour . Appl. Anim. Behav. Sci . 76 : 267 – 277 . doi:10.1016/S0168-1591(02)00010-2 Google Scholar Crossref Search ADS Singh , C. , M. Verdon , G. M. Cronin , and P. H. Hemsworth . 2017 . The behaviour and welfare of sows and piglets in farrowing crates or lactation pens . Animal 11 : 1210 – 1221 . doi: https://doi.org/10.1017/S1751731116002573 Google Scholar Crossref Search ADS PubMed Stabenow , B. , and G. Manteuffel . 2002 . A better welfare for nursing sows without increased piglet loss applying peri-parturition short term crating . Arch. Tierz. Dummerstorf . 45 : 53 – 60 . Svobodová , I. , H. Chaloupková , R. Končel , L. Bartoš , L. Hradecká , and L. Jebavý . 2014 . Cortisol and secretory immunoglobulin A response to stress in german shepherd dogs . PLoS ONE 9 : e90820 . doi: https://doi.org/10.1371/journal.pone.0090820 Google Scholar Crossref Search ADS PubMed Valros , A. , M. Rundgren , M. Špinka , H. Saloniemi , L. Rydhmer , F. Hultén , K. Uvnäs-Moberg , M. Tománek , P. Krejcı́ , and B. Algers . 2003 . Metabolic state of the sow, nursing behaviour and milk production . Livest. Prod. Sci . 79 : 155 – 167 . doi:10.1016/S0301-6226(02)00154-9 Google Scholar Crossref Search ADS Weary , D. M. , E. A. Pajor , D. Fraser , and A. M. Honkanen . 1996 . Sow body movements that crush piglets: a comparison between two types of farrowing accommodation . Appl. Anim. Behav. Sci . 49 : 149 – 158 . doi:10.1016/0168-1591(96)01042-8 Google Scholar Crossref Search ADS Yin , G. , H. Liu , X. Li , D. Quan and J. Bao . 2016 . Effect of farrowing environment on behaviour and physiology of primiparous sows with 35-day lactation . Intern. J. Appl. Res. Vet. Med . 14 : 159 – 169 . © The Author(s) 2018. Published by Oxford University Press on behalf of the American Society of Animal Science. All rights reserved. For permissions, please e-mail: [email protected]. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)