The authors have declared that no competing interests exist.
Conceived and designed the experiments: SF MC SLM. Analyzed the data: SF. Contributed reagents/materials/analysis tools: SF MC SLM. Wrote the paper: SF MC SLM. Collected the data: SF.
Animals facing seasonal variation in food availability experience selective pressures that favor behavioral adjustments such as migration, changes in activity, or shifts in diet. Eclectic omnivores such as many primates can process low-quality fallback food when preferred food is unavailable. Such dietary flexibility, however, may be insufficient to eliminate constraints on reproduction even for species that live in relatively permissive environments, such as moist tropical forests. Focusing on a forest-dwelling primate with a flexible diet (
Despite minimal seasonal variation in day length and temperature, most tropical habitats experience seasonal fluctuation in rainfall that mediates changes in food availability for many animals. These environmental changes require behavioral and physiological adaptations to minimize negative effects on fitness
Many primates are known especially for having evolved considerable dietary flexibility, most likely as a response to the challenge of environmental seasonality
Many cercopithecine monkeys (e.g., macaques, baboons, guenons) are generalist feeders particularly well known for their resource switching capacity, which results from long gut retention times and hindgut fermentation that facilitates the digestion of fiber-rich foods
It is well known that sufficient energy reserves in female vertebrates are required to trigger ovulation and successfully reproduce
The hypothalamic-pituitary-adrenal axis can exert control over individual reproductive decisions through its regulation of the stress response system. Glucocorticoids (GCs) are the main mediators of physiological stress responses in vertebrates
Assessing the causes and consequences of environmental seasonality in a wild social vertebrate such as blue monkeys is complicated by the fact that GC production can be influenced by many factors
Because of placental secretion of corticotrophin-releasing hormone in non-human primates
Changes in feeding efficiency as a result of variable food availability and levels of feeding effort may present considerable energetic challenges that influence female reproductive strategies. Relative changes in feeding on specific food items of high versus low energetic quality can be informative and allow inferences about changing energetic challenges
Negative social interactions such as agonism can trigger psychological stress, while affiliative social interactions can reduce the magnitude and duration of stress responses to potential psychological stressors
Although seasonal variation in temperature can be a significant metabolic stressor that affects GC excretion in wild primates
This study was purely observational, and all study animals were well habituated to human observation prior to data collection. The non-invasive collection of fecal samples for hormone analyses did not interfere with the natural behavior of the study animals. All data collection protocols were approved by Institutional Animal Care and Use Committee of Columbia University (IACUC 2885, 4307, 7232), and the Kenyan Ministry of Environmental Science and Technology (Permit #: 13/001/35C 252). All research was undertaken in accordance with the ABS/ASAB guidelines for the ethical treatment of animals in research and teaching.
The study population inhabits the Isecheno study site of the Kakamega Forest, western Kenya (0° 14′ 11″ N, 34° 52′ E, elevation 1618 m), a semi-deciduous rainforest with annual rainfall averaging approximately 2000 mm. Rain falls seasonally, with the dry season typically spanning December to March, and a shorter period of low rainfall in June/July
Females in this population give birth primarily between January and March
Subjects for this study were 21 adult females from two study groups, GN (10 females of 35–39 total members that included 10–12 adult females and 4–7 infants) and TWS (11 females of 43–52 total member that included 12–14 adult females, and 1–11 infants) that occupied adjacent home ranges (about 20 and 31 hectares, respectively). Agonistic interactions are rare in this population. Group-wide rates of agonism varied from 0.1 to 0.4 interactions per hour in GN (mean: 0.2, N = 6 monthly means across female means), and from 0.2 to 0.7 interactions per hour in TWS (mean: 0.4). Nevertheless, a linear hierarchy among adult females is well established
SF conducted 922 30-minute focal samples on all females in six non-consecutive months (with 1 or 2-month gaps) between September 2005 and October 2006. All females were observed in the same months for about equal lengths of time. Total observation time per female was 23.0±0.9 hours (range: 21.5–24.5 hours), and averaged 3.8±0.2 hrs per month. During each follow, SF recorded all activity states continuously with an approximate accuracy of ±3 seconds, using a handheld computer (Workabout MX, Psion Teklogix, Inc., Missisauga, Ontario) and Observer Software (Noldus Information Technology, Inc., Leesburg, VA). Feeding included manipulating and ingesting food items, but not simply chewing food from filled cheek pouches. Agonistic interactions were recorded on an all occurrence basis, with partner identity and context (food, space, grooming partner, infant, and other) whenever possible (74% of 431 interactions). To assess energy expenditure for travel, the distance covered during each individual movement was estimated visually to the nearest meter and then summed for each focal sample.
A field assistant (W. Adukha) assessed phenology of feeding trees at bi-weekly intervals between September 2005 and October 2006 by surveying a random selection of trees (5–20 per species depending on their abundance in the habitat; diameter at breast height, DBH ≥10 cm) from 25 known important food species (
For the food types most often consumed (fruits: 30% of all feeding time, across females and months; young leaves: 31% of feeding time), we calculated availability scores in two ways, first by summarizing across all species eaten (total availability), and second by summarizing across species that were eaten for at least 1% of the time spent feeding, on average, across all females and study months (main item availability). For flowers and ‘other’ food items, only three species met our criteria of ‘main’ food items (flowers:
SF and field assistants collected 3,322 fecal samples over 14 months between September 2005 and October 2006, of which 1,683 came from 10 females in GN and 1,639 from 11 females in TWS (12± SD 4 samples per female per month). Within a median of 87 minutes after defecation (range: 2–393 minutes; 98% within 180 minutes) we transferred fecal samples to a camping oven (Coleman Company, Inc., Wichita, KS, model no. 5010D700T) and dried them at 90±5°C. We removed samples from the field oven once per day in the evening, resulting in a mean drying time of 5.6±2.0 hrs. For our analyses, we excluded samples that dried for less than two or more than nine hours (2.1% of samples). Samples were stored in the dark at ambient temperatures (∼25°C) for a median duration of 35 days (range: 2–91) before they were transferred to a freezer (−20°C) for long-term storage. Immediate drying followed by freezing is generally regarded as one of the best preservation methods for fecal steroids
We quantified fGCs in the Endocrine Research Laboratory of the Smithsonian Conservation Biology Institute, Front Royal, VA. Detailed extraction procedures are described elsewhere
We evaluated the effects of potential confounds on sample fGC measurements, including time of day, sample storage time, and fiber content, using multivariate linear mixed models (
For all behavioral variables, we calculated average proportions and frequencies across all focal observations for a given month to obtain a monthly mean for each female. Months included in the models are those for which we had both detailed behavioral and environmental data: October 2005, January, March, June, August, and October 2006. We excluded a given female from the analysis in any month when she was observed <5 times (N = 3) or when <3 fecal samples were available (N = 3) to achieve a minimum level of reliability of estimates. To assess the influence of reproductive state, we determined the pregnancy period for each female post-hoc from the birth of an infant, assuming a gestation period of 176 days
We used General Linear Mixed Models to analyze temporal patterns in our longitudinal data, because they allow for subject-specific deviations from mean temporal response patterns, correlated errors of measurement within individuals, and the optimization of covariance structures
We entered reproductive state as a categorical fixed effect, and all other predictors as continuous covariates. As behavioral predictor variables for within-female variation in fGCs over time (i.e., across months) we included: time spent feeding on fruits, flowers, young leaves, mature leaves, insects, and other food items; fiber content; time spent grooming; rate of agonism received plus given; time spent moving and distance moved per focal observation. As environmental factors, we entered rainfall as well as estimates of the availability of main food items in the following categories: fruits, flowers, young leaves. We started with a full model including all hypothesized predictors, including first-order interactions of covariates with reproductive state, and removed parameters, one by one, that were least likely to explain fGC variation in the data until we arrived at a best model given the data. We assessed model fit with a small sample size correction of the Akaike Information Criterion (AICC) that approaches AIC values with larger sample sizes, and compared the relative fit of alternative models using Akaike weights and evidence rations
For main fruit items (those eaten for >1% of feeding time across females and months), fruit availability decreased temporarily in November/December 2005, and was generally low from April 2006 until the end of the study period (
Temporal change in food availability index (FAI) for main ripe fruits, eaten by adult female blue monkeys for at least 1% of all feeding time across months in study groups TWS and GN, and monthly rainfall recorded during the study period. Dashed horizontal line shows mean fruit availability across all months and groups.
Temporal variation in mean percentage of observation time spent feeding on different food types (top), and mean monthly food availability index (FAI) for the three most commonly eaten food types (bottom).
There were pronounced fluctuations in fGC levels that were similar in both study groups (
Horizontal lines are medians, boxes show inter-quartile range (IQR), whiskers indicate 95% confidence intervals, circles mark outliers more than 1.5x IQR, and triangles mark extreme outliers more than 3x IQR.
Our best model given the data predicted within-female variation in fGCs with reproductive state, feeding behavior, and variation in food availability (
Parameter |
Estimate | SE | df | 95% Confidence Interval | |
Lower | Upper | ||||
Intercept | −6.50 | 1.06 | 57.72 | −8.62 | −4.38 |
Second half of pregnancy |
11.36 | 2.43 | 50.50 | 6.47 | 16.24 |
First six months of lactation |
4.24 | 1.28 | 48.36 | 1.66 | 6.82 |
Feeding on insects | −0.55 | 0.16 | 43.48 | −0.88 | −0.22 |
Feeding on mature leaves | 0.22 | 0.10 | 68.12 | 0.02 | 0.43 |
Feeding on other fallback items | 0.28 | 0.12 | 57.80 | 0.05 | 0.52 |
Time spent feeding | 0.17 | 0.05 | 75.46 | 0.07 | 0.27 |
Availability of young leaves | −0.04 | 0.01 | 79.29 | −0.06 | −0.02 |
Availability of flowers | 0.12 | 0.01 | 99.66 | 0.10 | 0.15 |
Availability of food items other than fruits, flowers, or young leaves |
0.66 | 0.09 | 93.61 | 0.48 | 0.84 |
Rainfall | 0.02 | 0.01 | 38.94 | 0.01 | 0.03 |
All measures expressed relative to group mean across months (N = 6).
For alternative models, see
Compared to females in other reproductive stages.
Primarily male cones of
An equally well fitting but less parsimonious model contained all variables included in the best model described above and two additional variables related to feeding behavior (
To test our prediction that females would be particularly sensitive to changes in food availability while they were in the energetically challenging stages of late pregnancy and early lactation, we compared fGC levels within females of a given reproductive state between periods of below- and above average food availability across all 14 study months (
Boxplot of fGC excretion in periods of below- and above-average availability of ripe fruits (main items) for females of different reproductive states. Horizontal lines are medians, boxes show inter-quartile range (IQR), whiskers indicate 95% confidence intervals, circles mark outliers more than 1.5x IQR, and triangles mark extreme outliers more than 3x IQR.
Boxplot of fGC excretion in periods of below- and above-average availability of young leaves (main items) for females of different reproductive states. Horizontal lines are medians, boxes show inter-quartile range (IQR), whiskers indicate 95% confidence intervals, circles mark outliers more than 1.5x IQR, and triangles mark extreme outliers more than 3x IQR.
Distribution of births (N = 347 for
Our study assesses the importance of potential energetic challenges in the lives of female monkeys using physiological measures of stress responses, with the goal of better understanding the adaptive value of reproductive seasonality in a rainforest environment. We found that after controlling for reproductive state, an increase in the availability of non-preferred food items as well as increases in the total feeding time and in feeding on non-preferred food items (mature leaves, other fall back items) was associated with an increase in fGC levels, despite considerable dietary flexibility in the study species and year-round availability of fallback food items. In contrast, increased feeding on foods relatively rich in easily digestible energy, such as insects, fruits, and young leaves, was associated with lower fGCs, as was the availability of young leaves that were generally more preferred than mature leaves, flowers, and other non-fruit items. Females who were in the stages of late pregnancy and early lactation were more sensitive to variation in the availability of preferred food sources, i.e. ripe fruits and young leaves, and showed greater increases in fGCs between periods of high versus low availability of these items compared to females in other reproductive states. We found no evidence that social interaction in a daily life context mediated fGC changes over time for a given female.
The results presented here indicate that temporal variation in GCs can be a useful tool to monitor variable energetic stress in female blue monkeys, a conclusion supported by a number of studies on wild social vertebrates. Decreases in GC levels coincided with periods of high fruit consumption in chimpanzees
The small but significant positive effect of rainfall on fGC variation is difficult to interpret. In some environments, rainfall may be a good indirect indicator of plant food availability and may show a negative temporal correlation with GC levels
Contrary to a population of conspecifics, indicators of feeding effort (i.e., time spent moving and distance moved) did not have measurable effects on fGC variation in female blue monkeys. Among Sykes’ monkeys at Gede Ruins, Kenya, high fGC excretion coincided with periods of significantly elevated feeding effort in the absence of more preferred energy-rich fruit resources
Temporal changes in rates of agonism, which in this population relates primarily to contest competition over fruits
Although there is some uncertainty associated with using feeding times to estimate relative changes in food intake
Across guenon populations in different parts of Africa, peak fruit abundance appears to coincide with the birth peak and earliest weeks of lactation, while the energetically most demanding period of peak lactation often occurs when fruits are less abundant
The data on physiological responses to changes in food availability presented here indicate directly that female guenons experience seasonal energetic challenges, a finding that can inform our understanding of their reproductive strategies, and may relate to the evolution of exceptionally slow life history in blue monkeys
Assessing the physiological responses to ecological variability and how these responses influence reproductive decisions
Preliminary comparative evidence supports this interpretation. In
Energetic stress among lactating females may influence another noteworthy aspect of the life history of blue monkeys, the relatively fast development of infant independence from mothers
In conclusion, we documented significant environmental stressors in the lives of tropical forest guenons and provided preliminary evidence that these stressors may have implications for female reproduction. As such, our study reveals a potential proximate mechanism that may have mediated life history evolution in blue monkeys and other forest guenons. We believe that further studies integrating behavioral, ecological, and hormonal measures can lead to a better understanding of the evolution of life history variation among extant vertebrates. Given the extent of physiological and behavioral variation across species as well as intra-specific variability based on environmental and social factors, however, it is not likely that there will be a single unifying explanation for how different populations and species respond to stressful stimuli. Careful investigation of multiple potential stressors in each population is therefore necessary to interpret the adaptive significance of GC variation.
(PDF)
(PDF)
(PDF)
(PDF)
(PDF)
(PDF)
We are grateful to the Ministry of Science, Education and Technology, and the National Council of Science and Technology, Government of Kenya, for permission to work in the Kakamega Forest, to the staff at the Kakamega Forest Station for their cooperation, and to D. Yole and the Institute of Primate Research, Nairobi, as well as Masinde Muliro University of Science and Technology, Department of Biological Sciences, for local sponsorship. Thanks go to W. Adhuka, E. Mbone, C. Oduor, M. Ogutu, D. Shabila, and C. Zacharias for help during various stages of data collection, and to G. Wahungu for sharing his data on birth intervals from