Abstract
Similar to other primate species, captive cynomolgus macaques (Macaca fascicularis) are prone to becoming overweight. The relationship between body condition and feeding behaviour in group-housed animals has not been reported. This study evaluated the effect of daily feeding routines on behaviour patterns in cynomolgus macaques to determine whether overweight macaques displayed different behaviours and activity levels. In this prospective observational study, 16 macaques (m = 4, f = 12) from four separate troops (n = 4 per troop) were selected from a colony of 165 animals. Observational data were collected over six months during morning and afternoon feedings by scan sampling. Behaviours of interest included foraging, eating, aggressive and positive social interactions, inactivity and physical activities. Multivariable mixed logistic regression modelling was used for data analysis. Results indicated that overweight animals were more likely to be inactive, dominant animals had increased probabilities of eating compared with non-dominants, and aggressive behaviours were more likely to occur in the morning and before feeding, suggesting feeding anticipation. Positive social interaction before feeding was seen and may be a strategy used to avoid aggressive encounters around food resources. Individual animal caregivers had an unintentional impact on behaviour, as decreased eating and an increase in inactivity were noted when certain individuals fed animals. These findings illustrate the complexities of feeding group-housed cynomolgus macaques to avoid overweight body condition. Feeding routines may require more care and attention to distribute food in a way that ensures equitable food intake among troop animals, while not disturbing group cohesion.
Cynomolgus macaques are widely used in research as a model for many disease processes and potential therapies that affect humans, because of their similarity in physiology and genetic make-up. 1 This species of non-human primate is among several that have been observed to gain excessive weight in captivity. 2–5 In humans, the complex interplay between obesity, individual behaviour, activity levels and social networks is increasingly being recognized; 6 however, the possible effect of overweight body condition on behaviour and activity of cynomolgus macaques is unknown. Further, there is little published information on the social and behavioural feeding patterns of captive, group-housed cynomolgus macaques.
Cynomolgus macaques live within a strict dominance hierarchy system, whereby dominants may be intolerant of lower-ranking animals with respect to access to valued resources. 7,8 A review of foraging strategies among primate species indicates that in the wild, cynomolgus macaques are affected by within-group feeding competition, especially in larger groups, leading to significantly reduced reproductive success in females. 9 Observations of wild Sumatran cynomolgus macaques suggested that low-ranking females in large groups (i.e. >40 individuals) must travel more than their conspecifics to forage. 7,10 The difference in energy expenditure was significant, but the difference in food intake was not. Koenig 11 found that when wild, subordinate female cynomolgus macaques foraged away from the major feeding party, aggression from higher-ranking animals was reduced, and they obtained a similar amount of food as dominants. These findings suggest that cynomolgus macaques within troops may adjust their food-seeking behaviours based on social rank.
A recent review of husbandry procedures for non-human primates acknowledged that providing food in the captive environment is a greatly different experience from the wild, and could have a potential effect on wellbeing. 12 In general, institutions that house primates work on a structured routine and feeding times are often regulated. Some studies have shown that this predictability has both positive and negative influences on animal behaviour. Waitt and Buchanan-Smith 13 investigated the effect of varying feeding times in group-housed captive stump-tailed macaques. Delaying feeding was associated with increases in abnormal behaviour, self-directed behaviours and agonistic interactions as animals waited to be fed. Activity levels decreased during this time as well. In chimpanzees, using a variable feeding schedule that animals could not anticipate, led to a reduction in abnormal behaviour and inactivity. 14 These studies suggest that feeding time anticipation may be an important event in a captive primate's day, and may impact on animal welfare in a species-dependent fashion. There are no reported studies on the effect of various events surrounding feeding of cynomolgus macaques that may impact on animal welfare, which may provide guidance for colony management of a commonly used research species.
It seems plausible that provisioning of food to captive cynomolgus macaques may play a role in the propensity for obesity in this species, as well as creating a stressor within the social hierarchy. However, it is not known if overweight animals behave differently at the time of feeding than animals within their normal-weight range. Thus, the purpose of this study was to investigate whether overweight body condition was associated with significant differences in several behaviour categories at the time of feeding compared with normal-weight cynomolgus macaques. In addition, we wanted to examine how factors such as dominance status, time of day and individual humans influence the expression of behaviour patterns relating to daily feeding routines in mature, captive, group-housed cynomolgus macaques. We hypothesized that overweight animals would be less active with more sedentary behaviours compared with their normal-weight counterparts.
Materials and methods
Housing and feeding routine
This study was conducted at the Health Canada Scientific Services Division (Food Directorate), in Ottawa, Canada, where a captive-bred colony of over 165 cynomolgus macaques has been maintained for over 25 years. Animals were housed in social groups (troops), consisting of one vasectomized adult male and 4–12 adult females per troop. Demographic and health characteristics of this colony have been described. 15 Troop enclosures were rooms constructed of impermeable materials and were approximately 2.9 m wide, 4.27 m long and 2.29 m high. Each room had a solid door with a small window that allowed staff to enter, with an approximately 2.9 m × 0.9 m area between the entry door and the enclosure. Enclosures were bedded with wood shavings and contained various furniture and climbing structures, such as stainless-steel swings, wood perches and plastic drums. Troops were rotated through different enclosures every two weeks during routine husbandry procedures. Rooms were maintained on a 12 h light:dark cycle with an ambient temperature of 20–22°C, and a relative humidity of 40–70%.
Standard feeding routine occurred twice daily. The morning feeding (AM) occurred between 07:30 and 09:00 h. At this time, the daily allotment of standard monkey biscuits (MNR Feeds, Arnprior, Ontario, Canada) was tossed into each enclosure. The amount of biscuits required for each enclosure depended on the number, age and size of animals in each group, but ensured that animals had biscuits available throughout the day. Although it was not possible to accurately measure the daily intake of individual animals, it was estimated that females consumed approximately 88–96 g of food and males consumed 132–144 g of food. Approximately four cups of foraging mixture were also tossed into the bedding immediately after the biscuits were dispensed. The mixture consisted of sunflower seeds, peanuts, corn, dry pasta, puffed rice cereal, oats and raisins, with the choice of ingredients varying by availability. The afternoon (PM) feeding occurred between 12:00 and 14:15 h. At this feeding, animals were offered small pieces of fresh fruit or vegetables, such as apples that had been sliced into eighths. These items were distributed such that each animal per troop could receive two or three pieces of fruit or vegetables. Feeding was conducted by one animal caregiver per session and throughout the study, three individual personnel (2 female, 1 male) rotated this task as per standard operating procedures. Water was available ad libitum from automatic waterers.
Daily operations and routines were not changed for this study. To assist the observer in identifying animals during observation sessions, registered veterinary technicians shaved a small area of hair on individual animals on either the top of the head, between the wrist and elbow of one arm, or at the base of the tail, one to two days prior to the start of the study. This technique was routinely used at this colony to identify animals within the troop setting, and was conducted without sedation using minimal restraint. Animals were acclimatized to the presence of the observer over four 30 min sessions prior to the start of the study and were not manipulated in any manner during the observation periods.
All procedures used in this study were approved by the Health Canada and the University of Guelph Animal Care Committees. The Health Canada facilities are inspected and compliant with the Animals for Research Act of Ontario and hold a certificate of Good Animal Practice issued by the Canadian Council on Animal Care.
Study animals
Demographic and health characteristics of these animals have been reported. 15 Four troops were observed for this study. Selection was based on troops that had minimal social disturbances in the six months prior to the start of the study. Social disturbances were defined as the addition or removal of animals, or occurrences of severe aggression. Within each troop, four animals were selected for observation. A total of 16 mature adult male (n = 4) and female (n = 12) cynomolgus macaques ranging in age from 14 to 24 years were selected.
The dominant (and only) male in each troop was always selected because of interest in possible behavioural differences between the sexes. The remaining females were purposively selected based on body condition using a calculated body mass index (BMI) to dichotomously classify animals as overweight or normal. BMI was calculated as (body weight [kg]/crown–rump length [cm]2) × 1000. The most recent annual crown–rump length and biweekly body weight measurement from each animal's health records were used for the calculation. A BMI of 4 and above has been associated with moderately overweight body condition in rhesus macaques. 16 Thus, a BMI ≥ 4 was used to select overweight animals for observation and a BMI < 4 was used to select normal animals. Because overweight animals were all over 10 years of age, normal-weight animals were only selected if they were >10 years old, to control for possible age-related differences in behaviour in this prospective observational study. Nine macaques were classified as overweight and seven macaques were within normal weight, according to the individual BMI calculations.
After selection based on body condition, social status classification was recorded for each monkey using information previously collected by the Health Canada colony managers. Within this colony, dominant and subordinate animals had been characterized by two registered veterinary technicians with a minimum of five years of experience recording social interactions and behaviour patterns in the troops. Social status information was regularly recorded in animal health records to assist with colony management decisions. Females classified as dominant had been identified as frequent victors in outcomes of aggressive interaction or through knowledge of troop matrilines. Subordinate criteria included frequent displays of submissive behaviours such as fear grimacing, rump presenting, and displacement and aggression by higher-ranking animals, and/or knowledge of troop matrilines. For the purposes of this study, social status was dichotomized as being dominant or not dominant. Therefore, eight of the 12 females and all four males were determined to be dominant.
Behavioural data collection
Ethogram scheme for behavioural categories in a study of the influence of animal-level, social, and management factors on the feeding behaviour of mature group-housed cynomolgus macaques
For the duration of the study one observer, trained in ethology, collected all behavioural data. During sessions, the observer stood outside the enclosure and did not engage in any interaction with troop members. Study animals were observed at the morning and afternoon feedings, with one troop observed per session. Data collecting sessions were conducted every four weeks over a six-month period for a total of 12 observation sessions per troop.
Information recorded for each observation session included the date, time of start and finish, identification of the animal caregiver conducting the feeding, the ingredients of the foraging mixture (AM) and the type of fruit or vegetable given (PM).
Statistical analyses
Data were imported into STATA Intercooled version 10 (StataCorp LP, College Station, TX, USA) for descriptive and statistical analysis. Initially, frequencies of behaviours were counted to assess behaviours that were observed most often. Submissive and sexual behaviours were seen infrequently and were not included. Multivariable mixed logistic regression was used to analyse the following behaviour outcomes: forage, eat, inactivity, physical, aggression, positive social and abnormal. Each behaviour category was analysed separately as a dichotomous outcome of either present or absent for each 15 s interval over the 60 min observational sessions. For the purposes of these analyses, ‘groom’ and ‘affiliative’ behaviours were combined into a single category: ‘positive social’. Categorical and dichotomous independent variables used for analyses were after-feeding (pre- or postfeeding periods), time of day (AM or PM session), person feeding (caretaker 1, 2 or 3), sex (male or female macaque), social status (dominant or not dominant), overweight status (overweight or not overweight), troop identification (A, B, C or D) and month (1–6). Continuous fixed variables included minute of observation (1–60) and age.
Linearity of continuous predictor variables was evaluated visually by assessing a locally weighted regression of the variable against the log odds of the behaviour outcome (i.e. lowess curve). For continuous variables that were non-linear, a log transform was generated and a lowess curve of the log-transformed variable and the log odds of the behaviour outcome were assessed. For continuous variables that were curvilinear, a quadratic term was generated and included with the variable in subsequent logistic regression models.
Logistic regression was used to address the issue of non-normal distribution of data. Unlike analysis of variance or linear regression, assumptions such as homogeneity of variance are not violated using this method. 19 In addition, using a generalized linear mixed model allows for control of repeated measures. The results are indicated as odds ratios.
Multivariable mixed logistic regression models were created using adaptive quadrature with the ‘xtmelogit’ command in STATA to assess the effect of the fixed variables on each behaviour category. The variables for minute and visit were used to account for the effect of the presence of the observer on behaviour patterns. To account for repeated measures, individual monkey was used as a random intercept, and the time variable of minute was handled as a random slope, as described by Brown and Prescott. 20 An unstructured correlation was used when modelling the random effects in these models. For all variables, a statistical significance level of P ≤ 0.05 was used to determine inclusion in each model.
Models could not converge when attempts were made to explore possible interacting variables so these effects could not be explored. Confounding effects were assessed by removing each non-significant variable from the model and assessing the change in the remaining coefficients. A change of >20% in remaining predictor variable coefficients was used to determine the presence of confounding and if present, the confounding variable was retained in the model. 19 Diagnostics of the final models were evaluated graphically by examining Pearson and deviance residuals.
The fixed variable for minute was significant in all of the models with the exception of aggression and abnormal behaviours. It should be noted that the relationship between minute and the behaviour categories of foraging, positive social and inactive behaviours was not linear. Thus, the log transformation of minute was used in the models for foraging and inactive behaviours, and a quadratic term for minute was used for the model for positive social behaviours.
Results
Description of the study population of cynomolgus macaques in a study of the influence of animal-level, social and management factors on the feeding behaviour of mature group-housed cynomolgus macaques
Results of multivariable mixed logistic regression modelling with random effects for each behaviour category in a study of the influence of animal-level and management factors on the feeding behavior of mature group-housed cynomolgus macaques
Results are expressed as (odds ratio, P value)
*Referent for troop was troop A
†Referent for person was person 1
‡Referent for month was month 1
§Natural log of minute
The period after feeding was associated with statistically significant effects on almost all behaviour categories. Foraging, eating and physical activity behaviours were increased significantly after feeding (Table 3). Significant decreases in the probability of aggression, inactivity and positive social behaviours were associated with the time after feeding compared with the prefeeding period (Table 3). Several behaviours were associated with a statistically significant effect of the time of day. The morning session was associated with significant increases in foraging, aggression, physical activity and abnormal behaviours compared with the afternoon session (Table 3). The probability of positive social behaviours was significantly decreased in the morning compared with the afternoon (Table 3). The individual person feeding impacted on the results of eating and inactive behaviours. Person 3 was associated with significant decreases in eating behaviour and significant increases in inactive behaviours when compared with person 1 (Table 3).
Significant differences between troops and probability of behaviours occurring were found in the models for positive social and abnormal behaviours. Compared with troop A, all troops showed significantly decreased positive social behaviours, and significantly increased abnormal behaviours (Table 3). With the exception of aggression, there were significant differences in probabilities across all behaviour categories over the six months, where in some months increases were seen in likelihoods of behaviours, and other months saw decreases in likelihoods of behaviours (Table 3).
Discussion
To our knowledge, this study is the first to assess the association between overweight body condition and behaviour patterns relating to provision of food to mature, male and female, group-housed cynomolgus macaques. Overweight body condition was not associated with significant differences in any of the behaviour categories modelled, except for inactive behaviours. Age was controlled for during animal selection and analysis, and was not identified as a confounding effect. Thus, these results indicate that overweight animals were more likely to be inactive compared with normal-weight animals. Although this is perhaps not surprising, these findings could raise some concern about the wellbeing of overweight macaques. The enclosures were large and equipped with several perches and climbing structures to encourage species-typical behaviours. Whether overweight animals are unwilling to engage in activities, such as climbing and jumping, due to physical discomfort or awkwardness related to excessive fat has not been formally investigated. This question might be answered through experiments designed to evaluate the motivation behind some of these physical behaviours. Alternatively, overweight animals may be less active by constitution, further exacerbating their condition.
In non-human primates, excessive weight gain has been observed to occur in captivity with little to no change in diet, particularly in the rhesus and cynomolgus macaque species. 2–4 In terms of dietary causes of obesity for animals in this study, commercial monkey biscuit is a standard, balanced food that is offered to laboratory-housed monkeys across North America and Europe, the composition of which is based on NRC guidelines, and it is deemed to be nutritionally sound. 21 Small quantities of fruit or vegetables were also offered to animals on a daily basis in this study. Finally, addition of four cups of foraging mix to a troop of 13 animals resulted in ∼1/3 cup of foraging mix per animal. Primates generally consume 2–4% of their body weight daily on a dry matter basis, or approximately 80–160 g daily for a 4 kg female and 120–240 g for a 6 kg male. The average quantities of total diet fed in this colony were well within these guidelines and there was minimal food wastage. Thus, the diet was deemed to be nutritionally complete, and amounts and types of food fed were within the recommended NRC guidelines.
Eating and foraging behaviours were not affected by overweight status in this study. There is scant published information on feed intake and meal patterns of group-housed macaques, and no known information pertaining specifically to cynomolgus macaques. Recent work by Wilson et al. described a promising technology for collecting feed intake data in group-housed rhesus macaques. A custom-built, automated feeder dispensed a pellet of food when activated by a radiofrequency chip subcutaneously implanted in the wrist of individual female rhesus macaques. 22 The feeder was used to quantify food intake in two stable social groups of monkeys. Animals were given access to a control diet, a low-fat diet and a high-fat diet over a three-week period. The results indicated that subordinate females consumed significantly more of both the low-fat and high-fat diets, and fed throughout the night as well. Technology such as this may be necessary to individualize caloric intake in macaques, without having to remove animals from their social group.
The finding that aggression was decreased after feeding was somewhat unexpected, given the number of dominant animals in this study, and what is known about feeding behaviour of this species in the wild. Dominant animals have priority access to resources, and within-group competition can occur, leading lower-ranking animals to move away from the main group to feed. 10,11 The nature of a captive environment does not provide the same opportunities for lower-ranking animals. However, probabilities of positive social behaviours such as affiliative interactions and grooming were more likely to occur before feeding. For wild brown capuchin monkeys, group members with high dominance ranks fed significantly more at resources where competition was increased. 23 Between feedings, adult female capuchins spent considerable time grooming the dominant male. This behaviour may have served to raise his tolerance of their presence during feeding, allowing them to increase their food intake. Thus, in the current study, the decreased likelihood of aggression after feeding may be related to efforts made by the individual troop members to avoid eliciting aggressive interactions from higher-ranking animals.
In this study, dominant animals were more likely to engage in eating. This could suggest that lower-ranking monkeys may wait longer to feed in captivity, but actual latency to feed was not measured in this study. In chimpanzees, positive reinforcement training has been a successful strategy to reduce the tendency of some dominant animals to monopolize food resources. Bloomsmith et al. 24 reported on training dominant animals to sit while allowing other group members to be fed. Aggressive, submissive, and reconciliation behaviours all significantly declined specifically around meal times when the dominant animal was sitting in the targeted area. For cynomolgus macaques housed long term, perhaps positive reinforcement training such as this could be investigated as a strategy to permit lower-ranking animals to receive similar food intake as dominants, and could also be a possible strategy to minimize weight gain.
Morning observation sessions were associated with increased probabilities of aggression, inactive behaviour, as well as abnormal behaviours and decreased positive social interactions, such as grooming behaviours, compared with the afternoons. Coupled with decreased aggression after feeding, these findings suggest anticipation of the feeding event. This finding is similar to that found by Waitt and Buchanan-Smith 13 in a study examining varying feeding times in captive, group-housed stump-tailed macaques. Delaying routine feeding by 30 min to an hour was associated with increases in abnormal behaviour, agonistic interactions and decreased activity levels.
Another key finding in the present study was that the occurrence of eating and inactive behaviours in animals varied for husbandry personnel. Since captive-bred primates are not domesticated, all personnel who enter the enclosure, even if familiar, may be a source of anxiety to the macaques, which could be due to previous associations with routine procedures, such as restraint or sedation for handling. 25 This suggests that individual caregivers may affect food intake or latency to feeding in cynomolgus macaques. In a study of captive, group-housed stump-tailed macaques, Waitt et al. 26 found that animals classified as ‘caretaker friendly’ exhibited significantly higher feeding and foraging activities than animals classified as ‘unfriendly’ (those that exhibited negative behaviours such as threats, lunges, avoidance and grabbing directed towards the caretaker). These results emphasize the complexities of managing primates in captivity. The importance of nurturing human–animal bonds in the research environment has been shown to be beneficial to animal wellbeing, by reducing overall stress. 27
In conclusion, feeding is an important event in the daily routine of captive-bred, indoor group-housed cynomolgus macaques. Overweight body condition was not associated with any significant differences in foraging or eating behaviours when compared with normal-weight, age-matched conspecifics. Overweight animals were more likely to be inactive, potentially an indication that overweight body condition impairs the ability of these animals to engage in some species-typical activities such as climbing. Alternatively, overweight animals may lack motivation to be physically active, thereby maintaining this body condition. Aggressive behaviours were more likely to occur before feeding and in the morning, suggesting feeding anticipation. Positive social interaction before feeding may be a strategy used by troop members to avoid aggressive encounters around food resources. The results of this study indicate that the husbandry practices with respect to feeding group-housed cynomolgus macaques may need to be refined to quantify food intake in individual animals. Human caregivers may need to be more cognizant of methods to distribute meals such that equitable food intake is achieved while not disturbing group cohesion. Finally, these findings illustrate the complexity of patterns of cynomolgus macaques, and that there is still much to be learned about meeting their specific needs in captivity.
Footnotes
ACKNOWLEDGEMENTS
The authors gratefully acknowledge the valuable technical assistance of Sandra Stals, Scott Smith and Julie Fournier. SAB was supported by a Blythe James Chase Scholarship for animal welfare and through funding support from the Public Health Agency of Canada.