Sex impacts pain behaviour but not emotional reactivity of lambs following ring tail docking

Animals

In total, 96 single-born Merino lambs between 6–10 weeks of age were used for this study (pilot and main study), with an average bodyweight of 15.87 ± 0.70 kg (average ± SEM) at the time of testing. All animals were obtained from the CSIRO Chiswick farm, Armidale, New South Wales, Australia. When not part of the experimental protocol, ewe-lamb pairs were kept as a flock in a paddock assigned by the farm manager and managed under standard husbandry practice. When the lambs were required, they were transferred with their mothers into an animal housing facility in six cohorts of 16 ewe-lamb pairs/cohort. The first cohort was used in the pilot study, while the other five cohorts were used in the main study. Lambs were penned in groups containing two male and two female lambs with their respective mothers. Grouping was based on bodyweight rankings, with pen allocation assigned sequentially by bodyweight within lamb sex. Lambs were marked with symbols applied to the wool using coloured spray (Colourflow, Heiniger Australia, Western Australia) for identification for observation of behaviour through video recording. Identification marks were randomised across treatments within a pen. Pen-side health inspections were carried out assessing movement and demeanour, immediately prior to entry into the animal housing facility to confirm acceptance into the trial. Each ewe-lamb pair was fed 800 g sheep pellets and 200 g chaff daily, water was provided ad libitum. Animals were kept in the animal housing facility for a maximum of 7 days. After all lambs had completed the reactivity test the research trial was concluded. Male lambs were then castrated and Sham lambs were tail docked with the Numnuts^® device (Senesino, Brisbane, Australia) with NumOcaine™ (1.5 ml Lignocaine; Senesino, Brisbane, Australia) and all lambs received Buccalgesic OTM (10 mg/ml Meloxicam; Troy Laboratories, Glendenning, Australia) as per standard husbandry and were returned to the CSIRO Chiswick farm.

Emotional reactivity test arena and equipment

Attention bias and emotional reactivity testing for the pilot trial and the main experiment occurred in a 2.5 ×2.5 m arena with 2 m high opaque walls, located within the animal house. The arena contained a food bowl positioned between removable clear PVC barriers, to block physical access to the bowl from the side without blocking vision (Fig. 1). On the opposite wall to the food bowl was the location of the negative stimulus. Lambs entered the arena from a guillotine door on the left-hand side of the arena.

Lambs were video recorded during habituation and behavioural testing using GoPro cameras (Hero5; GoPro, San Mateo, CA, USA) mounted on the rafters above the test arena as well as a camera mounted inside a shelf located at lamb height in the left-hand corner between the guillotine door and negative stimulus location. Identical shelves were located at the same height in the remaining corners of the arena.

Pilot study

A total of 16 ewe-lamb pairs were used in the pilot study. In the first week of housing, lambs were habituated to the testing arena with no negative stimulus present, to become familiar with and consume feed within the arena. Habituation was conducted daily and involved removing lambs from their pens and moving them to the test arena (2.5 × 2.5 m), for up to 30 min before being returned to their ewes. Inside the arena, lambs were presented with four brightly coloured bowls that contained feed pellets. The bowls were positioned between removable clear PVC barriers. Across the week, lambs underwent three habituation sessions of up to 30 mins in groups of four. Lambs then underwent two sessions of up to 30 mins in randomly assigned pairs (whichever lambs were caught first). Finally, the lambs underwent two, five-minute habituation sessions individually totalling seven habituation sessions/lamb.

After the habituation stage, a negative stimulus was trialled. All lambs underwent a baseline attention bias assessment using the test configuration displayed in Fig. 1, with no threat present. Each individual lamb was placed in the arena for 5 mins while behaviours were recorded live via feed from GoPro cameras. The following day, the arena was arranged so that a photograph of a dog was positioned on the wall opposite the positive food stimulus at the beginning of the test. An individual lamb entered the arena, and the following behaviours were recorded to indicate a fearful response to the threatening stimulus compared to baseline behaviour: duration looking at the dog (s), freezing responses (yes/no), frequency of entering the zones near the dog’s location, number of high and low vocalisations, latency to eat and quantity of food eaten (g). It was found that the photograph was inadequate as a negative stimulus for the first five lambs. Subsequent lambs were therefore tested with a live dog used as a threatening stimulus. The dog’s location was that of the photograph and the same behaviours indicating a fearful response were recorded.

During the test with a threat present, lambs did not consistently direct their attention towards the photograph of a dog or the live dog. Lambs also did not consistently display a fear response towards the photograph of a dog or to the live dog. It therefore appeared that the dog was not an innately threatening stimulus for the young lambs used in the pilot study. The proposed method used in the pilot trial for assessing attention bias was determined not to be suitable for use with young lambs and an alternative test of emotional reactivity including an alternative assessment of attention bias (see Emotional reactivity test) was used for the main study (Erhard et al., 2004).

Main study

The study was conducted across five cohorts of 16 lambs per cohort. Within each cohort, the lambs were ranked according to weight and sequentially blocked into blocks of four. Weighted blocks were rotated between pens each week. Lambs were randomly allocated, by picking coloured marbles from an opaque bag, from within each block to each treatment;

1) FRing–Female lamb, tail docked using standard elastrator pliers without analgesia

2) MRing–Male lamb, tail docked using standard elastrator pliers without analgesia

3) FSham–Female lamb, tail manipulated to mimic ring tail docking

4) MSham–Male lamb, tail manipulated to mimic ring tail docking

This provided 20 replicates per treatment, which would allow us to define the probability distribution around the mean counts with reasonable confidence to compare those 95% CI brackets. The number of animals required per group was based on a sample size determination calculation, using anticipated mean and standard deviation values obtained from previous tail docking and attention bias studies.

During treatment application, lambs were picked at random (whichever lamb was caught first). All lambs were placed on their back in a lamb marking cradle for treatment application. Ring lambs had an elastrator ring applied to the tail at the level of the third palpable joint according to standard industry procedures (Lloyd & Playford, 2013). After treatment, lambs were placed back in their pen for behavioural observation. During the main study, lambs were not given analgesia as the aim was to compare response to pain between male and female lambs and to measure the emotional reactivity of lambs in pain. These methods comply with The Australian Animal Welfare Standards and Guidelines for Sheep (AHA, 2014). Sham lambs were placed in the lamb marking cradle and had their tail manipulated as though a ring were to be applied, but no ring was applied. All animals received analgesia (see Animals) following the end of the experiment as per the CSIRO Chiswick farm’s standard husbandry procedures before being returned to the paddock.

Behavioural assessment

Video cameras were used to continuously record the behaviour of lambs in the study. For each pen, one camera (Dahua IR Dome 4 MP network camera: DH-IPC-HDBW2431RP-ZAS-27135-S2; Zhejiang Dahua Vision Technology Co., Ltd, Hangzhou, China) was mounted on roofing rafters at each end of the pen. Each camera provided a view of the entire area available to the lambs. The cameras were connected to digital video recorders and captured by SmartGuard software from Pacific Communications (PACOM), Macquarie Park, New South Wales, Australia. The assessment of the behaviour post-treatment focused on the acute pain related behaviours and postural behaviours associated with ring tail docking. The pain related behaviour assessment was conducted from video footage, the person assessing behaviour was a female research officer with previous training and experience in assessment of pain-related behaviours and was blinded to lamb treatment application. Assessment took place for one full minute at five-minute intervals for the first 45 mins post-procedure, according to the ethogram shown in Table 1. Postural behaviours recorded were normal upright, abnormal upright, unknown upright, normal lying, abnormal lying, and unknown lying (Table 1). The most prominent posture displayed in the one-minute observation period was recorded. Acute pain behaviour was assessed every five minutes for one full minute and a count of all behaviours observed in that minute was taken. The ethogram used in this study was based on validated behaviour patterns (Molony, Kent & McKendrick, 2002) described in previous studies (Grant, 2004; Kent, Molony & Graham, 1998; Marini, Colditz & Lee, 2021) of behavioural responses of lambs following ring tail docking. Grimace scales were not applied due to their limited validity in lambs (Evangelista, Monteiro & Steagall, 2022) and the woolly faces of the Merino lambs used preventing accurate classifications.

Emotional reactivity test

Following induction to the animal housing facility, lambs and ewes were given one day to habituate to the pens without additional disturbance. During the following two days lambs were habituated to the emotional reactivity test arena in groups of four. Lambs were kept in the arena for 10 mins, twice daily, to habituate them to the novel environment and brightly coloured feed bowls containing feed pellets.

One hour after treatment, lambs were individually moved into the test arena, in their treatment order, for behavioural testing. The behavioural test was divided into several consecutive phases: ‘Isolation’, ‘Novelty’, ‘Startle’. For the Isolation phase, individual lambs were moved into the arena; lambs were tested for 90 s without any threat present. After 90 s of isolation had passed, a novel object (brightly coloured umbrella) was introduced into the arena to begin the Novelty phase. The closed umbrella was placed through a hole in the arena on the opposite side to the food bowl location (Fig. 2), and held there until the lamb approached within approximately 20 cm. Once the lamb approached the closed umbrella, the umbrella was opened, the Novelty phase was ceased, and the next phase began. If the lamb didn’t approach the umbrella within 3 mins, the test was ceased, and the lamb was removed from the arena and returned to its pen.

If the tested lamb approached the closed umbrella and the umbrella opened, the Startle phase of the test began immediately. The lambs’ immediate startle responses were recorded, then the lamb was kept in the arena with the open umbrella for 90 s to assess their behaviour as a potential measure of attention bias. During this time, lambs were recorded for their latency to eat as well as attention towards the umbrella, assessed as duration looking towards the umbrella within the field of binocular vision. After 90 s the lamb was removed from the arena and returned to its pen.

Throughout all phases of the emotional reactivity test, the lamb’s behaviour was recorded using GoPro cameras as described in the ‘Emotional reactivity test arena and equipment’ section. Footage from the cameras was used to undertake behaviour assessments using The Observer XT software (Noldus, The Netherlands). The footage for two lambs were missing due to technical errors (one Ring Male and one Ring Female). Data from these animals were excluded from the analysis of emotional reactivity. The person performing the video observations post-treatment was blinded to treatment. Live pen side observations were also recorded for latency to eat and number of bleats during each phase of the test, as well as latency to approach the umbrella. The behaviours recorded during each phase of testing are defined in Table 2.

Statistics

Statistical analysis was conducted using R and R studio software (R version 4.0.2 “Taking Off Again”) (R Team, 2022). Data were tested for normality through visual inspection of residual plots and the Shapiro–Wilks test. A P-value of <0.05 was considered statistically significant and 0.1 >P ≥ 0.05 was considered a statistical tendency. Data are presented as means ± standard error of the mean unless otherwise stated. There were no data exclusions for the analysis of pain behaviours.

Pain behaviours

The packages lme4 (Bates et al., 2015), ggpubr (Kassambara, 2020), dplyr (Wickham et al., 2019) and car (Fox & Weisberg, 2018) were used. The frequency of occurrence for individual postures was too low for data analysis, therefore categories were combined as upright (normal upright, abnormal upright, unknown upright), lying (normal lying, abnormal lying, unknown lying) and abnormal postures (abnormal upright, abnormal lying). Lying and abnormal postural behaviours were analysed using a non-linear mixed effects model, fixed factors included Treatment (n = 20 per group), Sex, Cohort, Pen and the interaction of Treatment × Sex, with Lamb included as a random effect. Sum of all abnormal postures was not normally distributed and was analysed using a Generalized linear mixed model with Poisson distribution, fixed effects and the random effect fitted were the same as other postures. The acute pain behaviour categories were combined for analysis. Acute pain behaviour counts were analysed using a Generalized linear mixed model with a Quasi-Poisson distribution due to over dispersal. Fixed effects included Treatment (n = 20 per group), Sex, Cohort, Pen and Time as well as the interaction of Treatment × Sex, with lamb included as a random effect.

Emotional reactivity

Statistical analyses were divided by test phase. All animals completed the Isolation test phase and were given an opportunity to approach the umbrella in the Novelty phase. Latency to approach the umbrella was analysed using a Cox proportional hazards model with the survival (Therneau, 2022; Therneau & Grambsch, 2000) package, fitting Sex and Treatment as fixed effects. Latencies were deemed as censored results if the lamb failed to approach the umbrella within 180 s. Reactions to the closed umbrella entering the test arena and to the umbrella opening were analysed using Fisher’s Exact Tests. The remaining Novelty test data analysis included only those that did not approach the umbrella within 180 s (n = 30), while only lambs that approached the umbrella moved on to the Startle phase (n = 48). Reactions to the umbrella opening in the Startle phase were analysed using Fisher’s Exact Tests.

Principal component analysis (PCA) was used to condense correlated behavioural measures within each phase of testing into their principal components, using the psych (Revelle, 2022) package. Variables were normalised using the scale function in R, then the scaled variables were checked for multicollinearity by analysing the correlation matrix. Variables with any correlations above 0.90, or with most correlations below 0.30, were excluded from the analysis (Field, Miles & Field, 2012). Sampling adequacy was confirmed by conducting Bartlett’s sphericity tests (P < 0.05) and calculating Kaiser-Meyer-Olkin (KMO) measures of sampling adequacy. The KMO statistic varies between 0 and 1, where a value close to 1 is considered ideal, while variables with a score < 0.5 were deemed unacceptable (Field, Miles & Field, 2012). Factors included in the analyses were subjected to varimax rotation. The number of rotated components used in further analyses were selected based on the number needed to explain more than 75% of the total variance and by checking the PCA model residuals (less than half with absolute values > 0.05), fit (> 0.9) and communality (all communalities > 0.7).

The selected rotated components within each test phase were then used to determine the effect of Sex and Treatment on the independent dimensions underlying the lambs’ behavioural reactivity. The rotated components were analysed using linear mixed effects models with the package nlme (Pinheiro & Bates, 2022; Pinheiro & Bates, 2000), fitting fixed effects of Treatment, Sex, Cohort, Pen and the interaction of Treatment × Sex and a random effect of Lamb (Field, Miles & Field, 2012). All of the rotated components for the Isolation phase were transformed to maintain the assumption of normality of the model residuals. A log transformation was applied after adding an integer of 2.7, 2.6 and 2.3 for the 3 rotated components respectively, such that the minimum value for each rotated component was ≈ 1. The second rotated component for the Novelty phase and the third rotated component for the Startle phase were log transformed to ensure normality of model residuals after adding integers of 2.4 and 2.1 respectively. Interactions and fixed factors were removed from the models using a backward elimination approach, considering the lowest Akaike Information Criterion (AIC) as an indicator of model fit.

Pain behaviours

Only the results of mean sum of abnormal postural behaviour and lying behaviour are presented, as upright behaviour is the opposite of lying and all abnormal postures are mutually exclusive. For lying behaviours there was a treatment (F_1,69 = 19.0), cohort (F_4,69 = 7.2), and pen (F_3,69 = 3.7) effect observed (all P < 0.05), but no effect of the interaction of treatment x sex (F_1,69 = 0.5, P = 0.47). Lambs in the Sham group were observed lying fewer times than the Ring lambs (Fig. 3). Lambs in cohort 2 were observed lying more than cohort 1 (t ₆₉ = 2.3, Mean count = 5.2 ± 0.5 vs 4 ± 0.5 respectively, P = 0.02). Cohort 5 lambs lay down fewer times than cohort 1 (t ₆₉ = −2.8, Mean count = 2.5 ± 0.5 vs 4 ± 0.5 respectively, P = 0.005).

Ring lambs displayed more abnormal postures compared to Sham lambs (z = −3.8, P = .0001, Fig. 4). Each subsequent cohort (2 to 5) after cohort 1 either tended to display or displayed more abnormal postures following tail docking. Cohorts 2 and 3 displayed an additional mean of 1 ± 0.4 observed abnormal postures (z = 2.8 and 2.4, P < 0.05), and cohorts 4 and 5 displayed an additional mean of 0.7 ± 0.4 (z = 1.7 and 1.8, P < 0.08), observed abnormal postures compared to cohort 1 mean count of 0.5 ± 0.3. Pen influenced the behaviour display of lambs with animals in pen two and three displaying fewer observed abnormal postures compared to pen one, with a decrease in mean count of −0.6 ± 0.2 (z = −2.2, P = 0.02) and −0.7 ± 0.3 (z = −2.4, P = 0.01) respectively.

Acute pain behaviours were assessed every 5 mins for the 45-min period following treatment. There was an effect of cohort (χ²(4) = 119.3, P < 0.001), pen (χ²(3) = 7.9, P < 0.05) and time (χ²(8) = 678.3, P < 0.001) as well as sex (χ²(1) = 54.2, P < 0.001) and treatment (χ²(1) = 1450.3, P < 0.001) and an effect of their interaction (χ²(4) = 119.3, P < 0.05). There was a difference in overall mean count of acute behaviours displayed between male and female lambs that were tail docked (t = −4.9, P < 0.001, Fig. 5). Lambs in the Sham group displayed significantly fewer acute behaviours than the tail docked lambs (t = −7.2, P < 0.001). There were no differences in behaviour seen between sham male and female lambs (P = 0.5).

Emotional reactivity

Novelty phase

For latency to approach the umbrella (n = 78), the Cox proportional hazards model fitting treatment and sex as fixed effects was not significant (Likelihood ratio test = 3.38, df = 2, P = 0.2). There was no effect of treatment group on reaction to the closed umbrella entering the arena (number of lambs with a High reaction in FRing = 7, MRing = 5, FSham = 7, MSham = 5; Fisher’s Exact Test, P = 0.905).

For the Novelty phase including only animals that failed to approach the umbrella (n = 30), five variables fit the criteria for inclusion in a PCA. Sampling adequacy was verified as ‘good’ by the overall KMO (0.71) and values for individual variables ranging from 0.5 to 0.86. Bartlett’s test of sphericity indicated that correlations between items were sufficiently large to conduct a PCA (χ²(10) = 80, P < 0.001). Three components were retained for further analysis. Table 4 shows the factor loadings on each variable after rotation. Based on the items that cluster on each component, we labelled component 1 as high activity fearful responses, component 2 as exploration and component 3 as an active response.

Table 4:

Summary of the principal component analysis results for the Novelty phase of behavioural reactivity testing, for sheep that failed to approach the umbrella (n = 30).

Absolute factor loadings over 0.6 are written in bold.

	Varimax rotated factor loadings
Novelty behaviour	Component 1 (fearful)	Component 2 (explorative)	Component 3 (active)
Attention to umbrella	0.92	0.03	0.10
Latency to eat	0.89	0.31	0.00
Vigilance %	0.83	0.36	0.30
Sniff environment	0.24	0.94	0.18
Zones crossed	0.12	0.17	0.98
Eigenvalues	2.41	1.14	1.09
% of variance	48	23	22

DOI: 10.7717/peerj.15092/table-4

For the first rotated component, the AIC supported inclusion of a treatment by sex interaction in the model, however the interaction term was not significant (χ²(1) = 2.3, P = 0.13). There tended to be a treatment effect (χ²(1) = 3.0, P = 0.084), whereby Ring animals scored higher on the ‘fearful’ index than Sham animals (Fig. 6; Least squares means 0.30 ± 0.26 and −0.38 ± 0.26 respectively, averaging over the levels of sex). For the second rotated component, the AIC supported inclusion of sex in the model only, for which there tended to be a sex effect (χ²(1) = 3.7, P = 0.056), whereby Female animals scored higher on the ‘explorative’ index than Male animals (Least squares means 0.96 ± 0.12 and 0.67 ± 0.10 respectively). For the third rotated component, the AIC supported the null model, with no significant effect of treatment, sex or the treatment by sex interaction.