Associations between fecal cortisol and biparental care in a pair-living primate

Abstract

Objectives:

We quantified variation in fecal cortisol across reproductive periods in Azara’s owl monkeys (Aotus azarae) to examine physiological mechanisms that may facilitate biparental care. Specifically, we evaluated evidence for the explanation that owl monkeys have hormonal mechanisms to mobilize energy during periods when each sex is investing heavily in reproduction, i.e., the gestation period for females and the infant care period for males.

Materials and Methods:

Between 2011 and 2015, we monitored 10 groups of Azara’s owl monkeys from a wild population in Formosa, Argentina and collected fecal samples from 26 adults (13 males, 13 females). Using enzyme-linked immunosorbent assays, we quantified fecal cortisol as a proxy for evaluating stress responses, including energetic demands, on both sexes during periods of reproduction and parental care.

Results:

Male cortisol was lowest during periods when they were caring for young infants (<3 months) compared to periods with older infants or no infant. Female cortisol was elevated during gestation compared to other periods. Mean fecal cortisol in both males and females was lower when an infant was present compared to when females were gestating.

Discussion:

Our results do not support the hypothesis that owl monkey males have elevated fecal cortisol during periods when they need to mobilize energy to provide intensive infant care. Our findings are also inconsistent with the Maternal Relief hypothesis. However, results from studies measuring fecal cortisol must be interpreted with care and alternative explanations, such as seasonal fluctuations in diet and thermoenergic demands, should be considered when drawing conclusions.

INTRODUCTION

Cortisol is a key hormone involved in a variety of seemingly disparate functions. It has, among others, pivotal functions related to circadian rhythms, synaptic turnover in the brain, and maintenance of homeostasis (McEwen, 2019; Sapolsky, 2015). It also mediates adaptive stress responses, an important function that unfortunately has led to measures of cortisol recurrently being interpreted as synonymous to measuring the adverse aspects of “stress” (Storey & Ziegler, 2016). Fundamentally, cortisol is a metabolic hormone that functions to liberate glucose and catabolize skeletal muscle; in other words, it generally acts as an important regulator of energy allocation (Brillon, Zheng, Campbell, & Matthews, 1995; De Feo et al., 1989; Ellison, 2017; Lukas, Campbell, & Campbell, 2005; McMahon, Gerich, & Rizza, 1988; Munck, 1971; Simmons, Miles, Gerich, & Haymond, 1984; Weiner, 1992). Therefore, rather than framing cortisol as a hormone that is associated with some loosely defined concept of “stress”, it is more useful and accurate to conceive of cortisol as a metabolic hormone that responds to different physiological challenges.

Reproductive events create unique energy requirements, and patterns of cortisol fluctuation during reproduction suggest that this steroid hormone may be involved in facilitating reproductive efforts. When female primates are pregnant and have an associated increase in energy expenditure, their cortisol levels are generally elevated, possibly indicating the need to have glucose more readily available (Carr, Parker Jr, Madden, MacDonald, & Porter, 1981; Trainer, 2002). Specifically, mid-way through gestation some primates have been shown to have a rise in cortisol that has been attributed to the development of the fetal adrenal glands and interaction of the “fetal-placental-maternal unit” (Ziegler, Washabaugh, & Snowdon, 2004). The possible association between cortisol levels and reproductive effort in males is less clear. In some biparental mammals, including both human and non-human primates, the cortisol levels of expectant fathers have been reported to increase during their mate’s pregnancy, returning to baseline after the birth of the infant (Berg & Wynne-Edwards, 2001; Gettler, Mcdade, & Kuzawa, 2011; Nunes, Fite, Patera, & French, 2001; Ziegler, Jacoris, & Snowdon, 2004; Ziegler, Washabaugh, & Snowdon, 2004). However, expectant fathers may not always experience an elevation in cortisol (Edelstein et al., 2015), and when they do, the timing of their cortisol increase is not necessarily correlated with the pattern of cortisol increase in their pregnant mate (Berg & Wynne-Edwards, 2002).

The relationship among cortisol, reproduction, and parental care may depend on the particular social and metabolic challenges that fathers of a given species confront. This is consistent with observations that hormones often vary with the amount, or type, of contact and care that fathers provide (Muller, Marlowe, Bugumba, & Ellison, 2009; Storey & Ziegler, 2016, Wynne-Edwards, 2001). Among non-human primates, differences in the social or physical environment in captive settings seem to alter the hormonal response of males and females during periods of pregnancy and infant care. For example, there is evidence that cortisol may sometimes increase after the birth of an offspring in taxa that exhibit biparental care. In black tufted-ear marmosets (Callithrix kuhlii) cortisol increased in fathers after the birth of their first litter, but decreased for already experienced ones (Nunes et al., 2001). These findings suggest that cortisol fluctuations following birth are mediated by paternal experience, with first time fathers having higher cortisol levels (Nunes et al., 2001; Ziegler, Wegner, & Snowdon, 1996). Thus, the potential functional implications of a rise in expectant father’s cortisol, if any, continues to be debated (Gettler et al., 2011; Saltzman & Ziegler, 2014).

Finally, differences in methodologies are likely to account for some of the apparent inconsistencies across studies examining cortisol and reproductive effort. At minimum, it is necessary to distinguish between studies of captive and wild populations when exploring the potential role of cortisol as a regulator of energy allocation, since captive individuals are provisioned with food and wild ones typically are not. A second important consideration is the medium from which the estimates of “cortisol” are obtained. Whereas it is possible to assess cortisol in blood, urine, or hair in captive individuals, for many non-human primates the only reliable source is feces. Each medium poses different challenges, and attention ought to be paid to this when comparing studies (Higham, 2016). Even when studies use the same medium (i.e. feces), differences in storage conditions and extraction and assay methods must be taken into consideration when comparing results (Kalbitzer & Heistermann, 2013; Khan, Altmann, Isani, & Yu, 2002; Pappano & Beehner, 2009; Wheeler, Tiddi, Kalbitzer, Visalberghi, & Heistermann, 2013).

Much has been learned about the endocrinology and energetic costs of biparental care through research on captive populations (Achenbach & Snowdon, 2002; Emery Thompson, 2017; Nunes et al., 2001; Tardif, 1997). Less is known from wild populations, particularly from taxa that exhibit exclusive biparental care (i.e. no sibling care) where both males and females face energetic challenges associated with the direct care of infants. Understanding this can provide insight into the implicated life history trade-offs and their underlying hormonal mechanisms that facilitate high levels of paternal care, pair-living, and sexual monogamy (De Bruin, Ganswindt, & Le Roux, 2016; Huck, Di Fiore, & Fernandez-Duque, 2020; Saltzman & Ziegler, 2014; Storey & Ziegler, 2016).

Owl monkeys (Aotus spp.) exhibit some of the highest levels of paternal care known among non-human primates (Fernandez-Duque, Valeggia, & Mendoza, 2009), comparable only to what is observed among titi monkeys (Spence-Aizenberg, Di Fiore, & Fernandez-Duque, 2016; Van Belle, Fernandez‐Duque, & Di Fiore, 2016). This makes them a good species in which to examine potential hormonal mechanisms involved in regulating biparental care. While the evolution of biparental care has been explored from theoretical perspectives (Huck & Fernandez-Duque, 2013), the physiological correlates of biparental care in owl monkeys have yet to be examined. One proposed explanation for paternal care is that male assistance alleviates the energetic burden of the female partner so that she is able to invest more energy in future reproductive efforts with that male (Maternal Relief Hypothesis: Achenbach & Snowdon, 2002; Fite et al., 2005; Tardif, 1997; Ziegler, Prudom, Schultz-Darken, Kurian, & Snowdon, 2006). It has even been proposed that direct male care in the form of carrying young has contributed to shortening interbirth intervals in the evolution of modern humans (Gettler, 2010). After the first few weeks of life, dependent owl monkey infants are carried ~80–90% of the time by their fathers and transfer back to mothers primarily for brief nursing bouts (Dixson & Fleming, 1981; Jantschke, Welker, & Klaiber-Schuh, 1998; Rotundo, Fernandez-Duque, & Dixson, 2005). Thus, assistance by owl monkey fathers may relieve mothers of a substantial portion of the energetic burden associated with infant carrying. Throughout the manuscript, we use the term “father” to refer to the male present at the time of pregnancy and infant care, regardless of whether this male is the genetic sire of the offspring.

In this study, our main goal was to explore the associations between fecal cortisol levels and the energetic demands of reproduction and biparental care in a wild population of a pair-living and sexually monogamous primate, Azara’s owl monkey (Aotus azarae). We examine these associations under the assumption, justified by a vast literature on primate and mammal physiology, that increases in fecal cortisol may reflect increased metabolic demands (Brillon et al., 1995; McMahon et al., 1988; McEwen & Wingfield, 2003; Romero, 2004; Weiner, 1992). We work within a general model that assumes owl monkeys have hormonal mechanisms to mobilize energy during periods when each sex is investing highly in reproduction, but we refrain from formulating more specific hypotheses because there is not robust a priori knowledge to base them on. In other words, there is a plausibility argument (“common knowledge”) that warrants having expectations regarding the relationships between fecal cortisol levels and reproduction in owl monkeys (Taborsky 2008). Still, as energetic studies of any kind are lacking for wild owl monkeys, we do not yet have adequate knowledge on their reproductive physiology and energetics to examine formal hypotheses and logically derived quantitative predictions. For example, the common knowledge that lactation is more expensive than gestation may or may not apply to owl moneys, since there is considerable variation in energy expenditure during each stage of reproduction among mammalian species due to differences in life histories and behavioral compensations that may offset energy requirements (Gittleman & Thompson, 1988). If indeed the extra cost of lactation is subsidized by the male, then compared to species with no allomaternal care, owl monkey mothers may still use more energy during lactation than during gestation; because they do not have the burden of carrying the infant all the time they may have relatively more available energy to allocate to lactation. Furthermore, because cortisol can fluctuate for a multitude of reasons (McEwen, 2019; Saltzman & Ziegler, 2014; Sapolsky, 2015), and our study is subject to the limitations of observational research (Smith, 2019; Taborsky, 2010), we carefully consider several hypotheses, biases, and confounders when interpreting our findings (Betini, Avgar, & Fryxell, 2017; Chamberlin, 1965).

To begin examining the relationship between fecal cortisol values and reproduction, we characterized patterns of cortisol variation during periods when adults are investing primarily in mating, gestation, or infant care. Based on “current knowledge”, we have the general expectation that 1) males caring for infants will have higher cortisol levels during the period of infant care compared to other periods, and 2) females will have lower levels of cortisol during the infant care period than during gestation.

We want to clarify that it is not our goal to provide a robust evaluation of the Maternal Relief hypothesis. Our methods only allow us to characterize how fecal cortisol is correlated with periods of reproduction and infant care. Nonetheless, by developing a better understanding of the relationship between cortisol, reproduction, and infant care in a wild primate species with biparental care, our study will provide a basis from which more specific hypotheses for the evolution and maintenance of biparental care can be refined and evaluated.

MATERIALS AND METHODS

Study site and subjects

We collected data at Reserva Mirikiná, the primary site of the Owl Monkey Project (OMP) that is within the private cattle ranch, Estancia Guaycolec, in Formosa, Argentina (58˚13’W, 25˚54’S). The site is part of the Gran Chaco region, which consists of a mosaic of grasslands, savannahs, gallery forest, and isolated patches of dry forest (Juárez, 2012). Temperature and rainfall in the area vary substantially across seasons, with lows of both occurring in the austral winter months of July-August (Fernandez‐Duque, 2016). Extreme temperatures, both high and low, can frequently occur. Food availability also varies seasonally, with lows occurring during the dry season (June-August) (van der Heide, Fernandez-Duque, Iriart, & Juárez, 2012). Additional details of the climate and seasonality of the area have been described previously (Fernandez-Duque, 2009).

We collected the data presented here from 10 groups of Azara’s owl monkeys inhabiting a mapped 300-ha core study area of gallery forest along the banks of the Riacho Pilagá between 2011 and 2015. The groups in this core area have been studied since 1997 (Fernandez-Duque, Rotundo, & Sloan, 2001). On average, these groups have been contacted weekly to obtain group composition information (e.g. Group E500, contacted 63 times/year, 1997–2018, Supplementary Table 3, Fernandez-Duque et al. 2020). In contrast to most other species of owl monkeys, Azara’s owl monkeys are cathemeral (Fernandez-Duque, De La Iglesia, & Erkert, 2010; Fernandez-Duque & Erkert, 2006), which makes it possible to observe and regularly collect fecal samples from identified individuals during daylight. The Owl Monkey Project (OMP) fits some individuals with radio or bead collars, which enables researchers to consistently locate groups and reliably identify specific individuals (Juárez, Rotundo, Berg, & Fernandez-Duque, 2011). Thus, all observers who collected samples were able to discriminate all individuals sampled using a combination of collars and natural distinguishing marks.

Consistent monitoring of groups in the core study area has allowed us to estimate dates of births, deaths, dispersals, and adult replacements within a range of a few weeks (Huck & Fernandez-Duque, 2012). Pairs produce an offspring at most once per year and reproduction is highly seasonal. Observational and hormonal data indicate that most conceptions take place in the fall or early winter (May-June) with gestation occurring mainly during July-September (Fernandez-Duque, Burke, Schoenrock, Wolovich, & Valeggia, 2011). Most births (~80%) occur in October or November (Fernandez-Duque, 2012; Fernandez-Duque, Rotundo, & Ramírez-Llorens, 2002), and nearly all of them (98%; 309/317) occur between late September and December (unpublished data).

We have defined age classes for owl monkeys previously, categorizing individuals >48 months as adults (Huck, Rotundo, & Fernandez-Duque, 2011). We could not determine the exact age of many of the adults in the study because most joined the monitored groups after having dispersed from unknown natal groups. We thus classified individuals as adults if they were of typical adult size and were part of a group’s breeding pair. While owl monkeys may sometimes remain in their natal group past 48 months of age and reach sexual maturity while still in their natal groups, they do not reproduce prior to dispersing (Corley, Valeggia, & Fernandez-Duque, 2017). Thus, for the purposes of our study, we did not consider individuals who had yet to disperse to be adults nor did we include them in our analyses. We considered individuals to be infants until they reached six months of age (Huck et al., 2011).

In our population, both adult males and females face intense competition from dispersed solitary floaters, which periodically replace adults in established breeding pairs (Fernandez-Duque & Huck, 2013; Huck & Fernandez-Duque, 2017). Therefore, some of the adults in the 10 sampled groups changed during the study; in total, we sampled 26 distinct adult individuals (13 of each sex, Table 1). Group size also changed as births, dispersals, and deaths/disappearances occurred. However, as it is the norm in owl monkeys, all groups contained only one pair of reproducing adults at all times. In addition to the adult pair, groups contained up to four non-reproducing individuals at a given time; group size ranged from two to six individuals.

Table 1.

Number of fecal samples collected from adult male (M) and adult female (F) owl monkeys in each of 10 groups during each of five years at Reserva Mirikiná, Estancia Guaycolec, Formosa, Argentina.

	C0		CC		Colman		Corredor		D500		D800		E350		E500		F700		P300
	F	M	F	M	F	M	F	M	F	M	F	M	F	M	F	M	F	M	F	M
2011	0	9	0	0	0	12	0	0	0	12	0	6	0	0	0	16	0	0	0	0
2012	26	1	0	0	0	0	0	0	22	2	22	5	1	0	31	5	0	0	0	0
2013	0	0	19	17	0	0	0	0	34	32	32	35	0	0	42	43	8	12	17	13
2014	0	0	38	44	0	0	16	10	33	31	38	43	35	34	23	36	0	0	0	0
2015	28	34	18	20	0	0	17	16	3	3	21	22	29	41	26	42	17	14	0	0

Sample collection and processing

Between 2011 and 2015, we contacted weekly 10 groups to collect fecal samples from the adults. We collected feces because they are the only type of biological sample that we can reliably and regularly obtain from wild individuals (Fernandez-Duque et al., 2011). We obtained feces non-invasively by collecting them from the forest floor immediately after an identified adult individual defecated. After collection, we placed each fecal sample directly into an 8 mL screw-cap tube containing 5 mL of a 1:1 solution of ethanol and deionized (DI) water, secured the cap with a Parafilm® strip, and then shook the tube for one minute to homogenize the sample. We assigned a unique code to each sample and recorded the date, time, and GPS location, as well as the group, identity, and sex of the animal from which the sample was collected. We subsequently entered this information into the Owl Monkey Project Database (https://www.owlmonkeyproject.com/resources). After tubes were labelled, we transferred them to a freezer in the project’s station in the city of Formosa as soon as possible (< 5 days during the winter, 1–2 days during periods of warm weather).

We collected 47 ± 40 (mean ± SD) fecal samples per adult male (n=610 samples, 13 males) and 46 ± 29 (mean ± SD) per adult female (n=596, 13 females). The period during which each individual was monitored, and thus the number of fecal samples collected, varied (range: 12–123 samples; Table 1, Table S1). There are two reasons for this. First, since our ability to collect fecal samples regularly depended on the availability of an individual with a radio-collar to facilitate finding the group, the specific groups being sampled at a given time varied. Second, the identities of some of the reproducing adults in monitored groups changed because they were replaced by solitary floaters (Fernandez-Duque & Huck, 2013; Huck & Fernandez-Duque, 2017).

We transported samples to the United States following all necessary local, national, and international regulations for the collection and transportation of biological samples. We placed tubes in a −20 °C freezer once they reached a laboratory in the US. We transported samples collected in 2013, and later, directly from Argentina to the Yale Reproductive Ecology Laboratory (YREL). Samples collected prior to 2013 had been shipped to the Reproductive Ecology Lab at the University of Pennsylvania (Penn REL) and were extracted there before we shipped them to the YREL.

We maintained consistency in how samples were processed, regardless of whether they were initially shipped to the Penn REL or YREL. We utilized the same protocols for storing, extracting, and drying samples at both locations, and these protocols had previously proven successful with owl monkey fecal samples (Corley et al., 2017; Fernandez-Duque et al., 2011). Briefly, we allowed sample tubes to stand undisturbed overnight to separate the solid fecal material portion from the liquid portion. We then removed the liquid portion, which was retained in 1 mL aliquots in microcentrifuge tubes, air-dried the fecal material, and recorded its dry weight to the nearest 0.001 g. One of the 1 mL aliquots from each sample was extracted and the remaining aliquots were stored at −20 °C in a YREL freezer as backups.

We performed diethyl ether extractions using established, validated methods (Fernandez-Duque et al., 2011; Strier and Ziegler, 1994). Briefly, we added 1 mL of DI water and 5 mL of diethyl ether to a glass culture tube containing 1 mL of the liquid portion of each sample and then vortexed tubes for 1 min. After sitting undisturbed for 5 minutes, we used a Pasteur pipette to separate the top ether layer from the bottom aqueous layer, by decanting the ether into a clean glass tube. We then completely dried off the ether and re-suspended the sample in 2 mL of phosphate buffer. We stored the final extract at −20 °C in two 1 mL aliquots until we ran the cortisol assay.

We used DetectX Enzyme Immunoassay (EIA) kits from Arbor Assays (K003-H1/H5, Ann Arbor, MI) to estimate the levels of cortisol in each fecal extract following the manufacturer’s protocol. The manufacturer has validated this kit for use on fecal extracts and determined that the kit has cross reactivity of 100% with cortisol, 18.8% with dexamethasone, 7.8% with prednisolone (1-Dehydrocortisol), 1.2% with corticosterone and cortisone, and <0.1% with other reactants at the 50% binding point. We confirmed parallelism between cortisol standards and a serial dilution of pooled extracts (1:2, 1:4, 1:8, 1:16) and estimated accuracy (recovery = 96%) for the assays using pooled owl monkey fecal extracts prior to the start of experiments.

We assayed each extract in duplicate after diluting it at least 1:2 with assay buffer. For quality control, we re-ran all samples for which the duplicates had a coefficient of variation (CV) >20 in the initial assay, and we only retained for analyses samples with a CV <20. We also re-diluted and re-assayed samples for which binding was > 90% or < 20% of the maximum binding. In some cases, cortisol levels were below the limit of detection by the assay even when they were re-run using the minimum possible dilution (1:2). We assigned these samples an arbitrary low value (below the assay kit’s level of detection) rather than excluding them from the dataset in order to avoid biasing analyses against low cortisol values. Inter-assay CVs for high and low controls were 6% and 12% respectively. We corrected each sample for the dilution factor and express the mean concentration of each set of duplicates as nanograms of cortisol per gram of feces.

All research, including protocols for the capturing and collaring of owl monkeys for identification, were approved by animal care committees (IACUC) of the University of Pennsylvania (2006–2014), Yale University (2014–2020), and the Ministry of Production and the Environment of Formosa Province in Argentina. All research adhered to the Code of Ethics of the American Society of Physical Anthropologists, the American Society of Primatologists’ Principles for the Ethical Treatment of Non-Human Primates, and the Argentine Society for Mammalian Studies guidelines (Giannoni, Mera Sierra, Brengio, & Jimenez Baigorria, 2003).

Data analyses

We created a data set that combined data related to each fecal sample (e.g., date and time of collection, feces dry weight, corrected cortisol level) with demographic data extracted from the Owl Monkey Project database using R version 3.6.2 (R Core Development Team, 2016). We used the demographic data to estimate the age of infants present in the group, the number of days since the last adult replacement in the group had occurred, the number of days until the next birth in the group, and whether an individual was known to have previously reproduced at the time each sample was collected. We then used this information to define three categorical variables: pregnant, adult replacement, and previous fertility. The variable “pregnant” classified females as pregnant or not. Since gestation in A. azarae has been estimated to last 120–126 days (Fernandez-Duque et al., 2011), we considered females to be pregnant if the sample had been collected within four months prior to the birth of an infant in the group. The variable “adult replacement” coded whether one of the adults had been replaced within the last year, since the last infant birth in the group. The variable “previous fertility” coded whether an individual had previously reproduced in its current group at the time each sample was collected.

Based on the growth patterns of wild A. azarae, we considered the infant care period to include all samples collected within six months following the birth of an infant in the group (Huck et al., 2011). Since it is possible that the amount of energy that males expend caring for infants changes as infants grow, we split the six month period of infancy into two equal-length periods (<3 months and ≥3 months in age) that approximately represent an initial period of locomotor dependence and intense nursing and a second period of more locomotor independence and solid food consumption (Rotundo et al., 2005). The variable describing the category of infant (if any) that was present at the time the fecal sample was collected thus included three levels: “young infant” (individual <3 months of age was present), “older infant” (individual 3<6 months of age was present), and “no infant”. The data used for all analyses reported in this study are publicly available on Figshare (Corley et al. 2021).

To inform our data analyses and prepare the final data set, we first performed exploratory data analyses (EDA). We removed only extreme outliers and samples that were missing essential data, such as accurate dry sample weight. There were some additional samples with high cortisol values that could potentially have been considered outliers (according to Tukey fences criteria) (Tukey, 1977). We did not remove these samples because we have no biological reasons to assume that cortisol values could not vary by as many as three standard deviations from the mean during some seasons or periods (e.g. pregnancy). In fact, the EDA revealed that almost all of the samples that were automatically flagged as potential outliers by the analysis corresponded to females in the month or two before they gave birth. EDA also revealed that there was no strong association between cortisol level and group size when the sample was collected (Figure S1), so we excluded group size from subsequent analyses. Sampling effort was not even throughout the year, but all seasons were represented in the data set (Table S2).

We did not include the time that the sample was collected as a variable in the final analyses for several reasons. First, the level of cortisol detected in a fecal sample does not reflect the amount of cortisol circulating in that individual at a single time point. The rate at which fecal matter incorporates cortisol as it travels through the digestive system may be influenced by diet and a variety of other factors (Bahr, Palme, Möhle, Hodges, & Heistermann, 2000; Whitten & Stravisky, 1998), which are unknown for owl monkeys. Second, A. azarae are cathemeral, and their activity patterns change substantially throughout the lunar cycle (Fernandez-Duque, de la Iglesia, & Erkert, 2010; Fernandez-Duque & Erkert, 2006), and thus the “typical” circadian pattern of cortisol fluctuation seen in humans and other diurnal primates is not expected. Therefore, we concluded that the time at which a fecal sample was collected is not likely to be a very informative variable for the questions we pose. This suggestion is supported by an exploratory plot of the relationship between cortisol and time of day, which did not show any clear pattern of fecal cortisol increase or decrease over the course of the day (Figure S2).

We analyzed the data using an information theoretic approach that included multimodel inference and model averaging (Burnham & Anderson, 2002; Symonds & Moussalli, 2011). After describing the data (Figure 1 and Figure 2), we constructed two separate sets of a priori generalized linear mixed models (GLMM), set 1 for males and set 2 for females, to evaluate two predictions for fecal cortisol levels (Table 2). All candidate models included group and individual ID as random factors. Predictor variables in the models included “pregnant”, “adult replacement”, “previous fertility”, and “infant category”, as previously described.

Figure 1.

Sex differences in fecal cortisol concentrations (ng/g feces) of samples collected from 13 adult female and 13 adult male owl monkeys. Boxes show the median, 1^st, and 3^rd quartiles. Upper whiskers indicate values ≤3^rd quartile + 1.5*IQR (interquartile range); lower whiskers indicate minimum values.

Figure 2.

Sex differences in fecal cortisol concentrations (ng/g feces) of samples collected from adult owl monkeys with data grouped by whether or not the female in the group was pregnant (2a) and whether or not an infant was present at the time the sample was collected (2b). Boxes represent the median, 1^st, and 3^rd quartiles. Upper whiskers indicate values ≤3^rd quartile + 1.5*IQR (interquartile range); lower whiskers indicate minimum values.

Table 2.

Fixed effects in candidate models for male (Set 1, left), and female (Set 2, right) owl monkeys.

Model Set 1 (Males)	Model Set 2 (Females)
M1 = partner pregnant	M1 = pregnant
M2 = infant category	M2 = infant category
M3 = partner pregnant + adult replacement	M3 = pregnant + adult replacement
M4 = infant category + adult replacement	M4 = infant category + adult replacement
M5 = previous fertility + partner pregnant	M5 = previous fertility + pregnant
M6 = previous fertility + partner pregnant + adult replacement	M6 = previous fertility + pregnant + adult replacement
M7 = previous fertility + infant category + adult replacement	M7 = previous fertility + infant category + adult replacement
M8 = previous fertility * partner pregnant	M8 = previous fertility * pregnant
M9 = previous fertility * infant category	M9 = previous fertility * infant category

Within each model set, we calculated corrected Akaike information criteria (AICc), to account for a relatively small number of sampled units (13 males, 13 females), and utilized delta AICc and AICc weights to assess the plausibility of each candidate model (Burnham & Anderson, 2004; Long, 2012). Prior to fitting models, we set a priori criteria for selecting the model(s) from each set that would be used in making scientific inferences (Hubbard et al., 2019). We considered there to be a single best approximating model only if the model with the lowest AICc had a corrected Akaike weight (AICc weight) of > 0.9. Still, even if one best model was supported, in making scientific inferences from our models we considered all those with a delta AICc ≤ 10 and report their results, following recommendations that models within this relatively small delta AICc range should not be automatically discounted (Richards, 2005; Symonds & Moussalli, 2011). When a single best approximating model could not be identified, we made inferences utilizing multiple models by calculating model-averaged parameters (Burnham & Anderson, 2004; Symonds & Moussalli, 2011). As recommended, we report characteristics (number of parameters (K), AICc, delta AICc, likelihood, AICc weight, log-likelihood, and cumulative Akaike weights) for all candidate models, so that readers can assess the plausibility of each of the candidate models for themselves (Table 3).

Table 3.

Nine a priori candidate models for male (a) and female (b) owl monkeys ranked by their corrected Akaike Information Criteria (AICc) values. Models that fell below a delta AICc of 10, and thus did not meet our a priori criteria for further consideration, are shaded in grey.

a.Set 1 (males)
Model	K	AICc	Delta AICc	Likelihood	AICc Weight	Log-likelihood	Cum. Weight
M2	6	5769.68	0.00	1.00	0.62	−2878.77	0.62
M4	7	5771.55	1.87	0.39	0.25	−2878.68	0.87
M7	8	5773.57	3.89	0.14	0.09	−2878.67	0.96
M9	9	5775.68	6.00	0.05	0.03	−2878.69	0.99
M1	5	5780.14	10.45	0.01	0.00	-2885.02	0.99
M3	6	5781.13	11.44	0.00	0.00	-2884.49	1.00
M5	6	5781.34	11.66	0.00	0.00	-2884.60	1.00
M6	7	5782.59	12.90	0.00	0.00	-2884.20	1.00
M8	7	5783.25	13.57	0.00	0.00	-2884.53	1.00

b.Set 2 (females)
Model	K	AICc	Delta AICc	Likelihood	AICc Weight	Log-likelihood	Cum. Weight
M6	7	6107.97	0.00	1.00	0.56	−3046.89	0.56
M8	7	6109.44	1.47	0.48	0.27	−3047.63	0.83
M5	6	6110.38	2.41	0.30	0.17	−3049.12	1.00
M3	6	6117.93	9.96	0.01	0.00	−3052.89	1.00
M1	5	6124.59	16.62	0.00	0.00	-3057.24	1.00
M7	8	6126.97	19.00	0.00	0.00	-3055.36	1.00
M9	9	6128.08	20.11	0.00	0.00	-3054.88	1.00
M4	7	6138.72	30.75	0.00	0.00	-3062.26	1.00
M2	6	6139.79	31.82	0.00	0.00	-3063.82	1.00

K = number of estimated parameters in the model

AICc = corrected Akaike Information Criteria for the model

Delta AICc = Difference between the model and the “best model” (model with lowest AICc) in the candidate set

Log-Likelihood = the relative likelihood of the model

Cum. Weight = the cumulative Akaike weights of model (when sorted from lowest to highest AICc)

We performed all statistical analyses in R version 3.6.2 (R Core Development Team, 2016). We utilized the package glmmTMB (v. 1.0.1) to construct models (Brooks et al., 2017) and AICcmodavg (v. 2.2–2) to calculate AICc and other model characteristics, as well as for calculating model-averaged parameter estimates, their standard errors, and 95% unconditional confidence intervals (Mazerolle, 2019). For descriptive statistics, we report the mean ± one standard deviation, unless otherwise specified.

A note on interpretation of results

The use of Information-Theoretic (I-T) Model Selection approaches to the analyses of data has grown steadily in biological anthropology. One of the benefits of this approach is that it encourages considering, and examining, multiple plausible models. When adopting an I-T approach, there is abundant literature discouraging the combination of such approach with the dichotomous decision process inherent to null hypothesis significance testing (e.g. Anderson et al., 2000; Burnham & Anderson, 2002; Newland 2019). Thus, our use of I-T methods has different goals than those typically pursued within a frequentist null-hypothesis significance testing approach. Our analyses focused on comparing plausible a priori models, ranking models by weighing their relative value and importance, and averaging variable estimates from multiple models when more than one model had comparable levels of support. It would be counter to the goals of the I-T approach to discuss estimates of parameters in our models in terms of statistical significance based on thresholds (i.e. p ≤ 0.05); instead we present effect sizes and odds ratios for the reader to quantitatively evaluate the results.

RESULTS

Variation in cortisol levels

Females had higher mean levels and more variability in mean fecal cortisol than males (mean ± SD ng cortisol/g feces, females: 65 ± 78, males: 42 ± 43; Figure 1). Fecal cortisol levels were higher in pregnant females than non-pregnant ones (means = 81 ± 9 and 51 ± 7 ng/g, respectively). Males whose pair mate was pregnant at the time of sample collection had higher mean cortisol levels than males whose pair mate was not pregnant (means = 46 ± 5 and 37 ± 4 ng/g, respectively) (Figure 2a).

Both males and females had lower mean fecal cortisol when an infant was present than when there was no infant in the group (females = 44.6 vs. 69.8, males = 31.3 vs. 45.5; Figure 2b). For both females and males mean cortisol was lower during the infant care period than during gestation (infant care vs gestation, females = 44 ± 6 vs. 81 ± 9, males = 31 ± 35 vs. 46 ± 5, Figure 2).

Cortisol variation and care by males: Model Set 1

Four of the nine models in Set 1 accounted for 99% of the cumulative weight (Table 3a). All four of them included “infant category” as a predictor. We used only models 2, 4 and 7 to estimate the model average for infant categories in this set; we did not include model 9 since it contained “infant category” only within an interaction term.

Fecal cortisol levels of males were lowest when caring for a young infant (<3 months) than when caring for an older one (3<6 months) or when no infant was present (Table 4). Specifically, both the presence of an older infant and no infant present were associated with increased cortisol in each of the three models in which “infant category” appeared as a fixed effect (Models 2, 4 and 7) (Table 4). The model averaged estimate (± SE) for not having an infant present was 0.5 ± 0.1 (95% Unconditional CI = 0.26 – 0.67), which was associated with an odds ratio of 1.6 (95% CI = 1.3 – 1.9) and the model averaged estimate (± SE) for having an older infant was 0.7 ± 0.3 (95% Unconditional CI = 0.15 – 1.24), which was associated with an odds ratio of 2.0 (95% CI = 1.2 – 3.5).

Table 4.

Summary of results for best models (those used to calculate model average estimates) from candidate model Set 1 (male) and Set 2 (female) owl monkeys.

Model Set 1 (Males)†
Term	Estimate	SE	Lower CI	Upper CI
Model 2
Intercept	3.34	0.10	3.15	3.54
Older infant	0.69	0.28	0.15	1.24
No infant	0.46	0.10	0.26	0.67
Model 4
Intercept	3.34	0.10	3.14	3.53
Older infant	0.70	0.28	0.15	1.24
No infant	0.46	0.10	0.25	0.66
Replacement	0.04	0.10	−0.15	0.23
Model 7
Intercept	3.32	0.14	3.04	3.59
Previous fertility	0.02	0.10	−0.18	0.22
Older infant	0.71	0.28	0.15	1.26
No infant	0.46	0.11	0.25	0.67
Replacement	0.04	0.10	−0.15	0.24
Model Set 2 (Females) †
Term	Estimate	SE	Lower CI	Upper CI
Model 6
Intercept	3.92	0.15	3.63	4.21
Previous fertility	−0.16	0.12	−0.40	0.08
Pregnant	0.54	0.10	0.35	0.73
Replacement	0.26	0.12	0.02	0.50
Model 8
Intercept	4.25	0.17	3.91	4.59
Previous fertility	−0.51	0.18	−0.86	−0.16
Pregnant	0.22	0.20	−0.16	0.61
Fertility * Pregnant	0.37	0.21	−0.05	0.79
Model 5
Intercept	4.05	0.14	3.77	4.32
Previous fertility	−0.27	0.11	−0.48	−0.05
Pregnant	0.52	0.10	3.25	7.09
Model 3
Intercept	3.79	0.11	3.58	4.00
Pregnant	0.57	0.10	0.38	0.75
Replacement	0.33	0.11	0.11	0.55

^†The table shows estimates (coefficients), standard errors (SE), and lower and upper bounds of the 95% confidence intervals (CI). Models in each set are listed according to their ranked AICc values. Reference categories for variables are as follows: young infant (< 3 months), not pregnant, no adult replacement, and no prior confirmed fertility.

Cortisol variation and care by females: Model Set 2

Four of the nine models in Set 2 accounted for 100% of the cumulative weight (Table 3b). All four of them included “pregnant” as a predictor.

Fecal cortisol levels of females were higher when pregnant (Table 4, Figure 2). Specifically, pregnancy was associated with increased cortisol in all four best models (Table 3b). “Adult replacement” and “previous fertility” were also associated with variation in cortisol in these best models, but the relationship of these variables to cortisol was somewhat inconsistent. Having reproduced previously in the group was associated with lower cortisol in the three models in which it occurred as a variable (Models 5, 6, and 8) and Model 3 did not contain the variable “previous fertility” at all. Similarly, having an adult replacement was associated with higher cortisol levels in the two best models in which in occurred as a variable (Models 3 and 6), but the other two best models (Models 8 and 5) did not contain this variable at all (Table 3 and Table 4).

The model averaged estimate (± SE) for pregnancy was 0.53 ± 0.10 (95% Unconditional CI = 0.34 – 0.72), which was associated with an odds ratio of 1.70 (95% CI = 1.40 – 2.05). The model averaged estimate (± SE) for having an adult replacement within the past year was 0.26 ± 0.12 (95% Unconditional CI = 0.02 – 0.50), which was associated with an odds ratio of 1.27 (95% CI = 1.02 – 1.65). The model averaged estimate (± SE) for “previous fertility” was −0.16 ± 0.12, but the confidence interval for this estimate included zero (95% Unconditional CI = −0.4 – 0.08).

DISCUSSION

Our results are consistent with previous evidence suggesting that cortisol levels of females are generally elevated during gestation (Carr et al., 1981; Trainer, 2002). However, our prediction that fecal cortisol would increase in males during periods in which they expend energy on infant care was not supported. In contrast, both males and females had lower levels of fecal cortisol during periods of infant care. Thus, our data do not provide support for the Maternal Relief hypothesis.

Because of the quite disparate findings of previous studies that examined cortisol during pregnancy and infant care in biparental primates, we consider it important to take a cautious approach in interpreting any results, particularly those derived from fecal samples. The development of methods for measuring cortisol in minimally-invasively collected samples, such as urine and feces, has allowed for the proliferation of studies examining cortisol and other hormones in wild primate populations. While important insights have undoubtedly come from this research, using excreta to assess cortisol presents challenges and may produce results that are difficult to interpret without the necessary contextual information (Higham 2016). In the discussion of our findings below we make an effort to clarify what assumptions are being made and note when alternative explanations for our results may be possible (Chamberlin, 1965; Lipton, 2005; Smith, 2019; Towner & Luttbeg, 2007).

The result that both males and females have lower fecal cortisol during infant care periods is somewhat surprising, since male owl monkeys spend substantial time carrying infants, and infant carrying has been suggested to be energetically expensive in other primate taxa (Altmann & Samuels, 1992; Sanchez, Pelaez, Gil‐Bürmann, & Kaumanns, 1999). One explanation is that increased energetic demands during the infant care period may be offset by individuals shifting other behaviors to conserve energy, as has been proposed for some primates living in captivity (Miller, Bales, Ramos, & Dietz, 2006; Nievergelt & Martin, 1998). Individuals may also shift their diets, or time spent foraging, to offset increased energetic demands associated with infant care (Vasey, 2004). There is evidence consistent with these explanations in some wild populations of primates (Guedes, Young, & Strier, 2008; Vasey, 2004), including species with biparental care, such as titi monkeys (Dolotovskaya & Heymann, 2020). We therefore cannot rule out the possibility that paternal care in owl monkeys may indeed relieve some of the energetic burden for lactating females, but males have behavioral strategies that allow them to compensate for any net energetic costs that would be associated with increases in fecal cortisol levels.

Reproduction at our study site is very seasonal, and it is possible that environmental changes associated with the seasons corresponding to gestation and parental care influence the pattern of associations we have found. For example, there may be reduced thermoenergetic challenges associated with warming ambient temperatures in the period when owl monkeys are caring for infants (Perea-Rodriguez, de la Iglesia, & Fernandez-Duque, 2022). Food availability has been implicated as one of the most important factors linked to primate birth seasonality (Di Bitetti & Janson, 2000). Thus, it is plausible that increased food availability coinciding with the period of lactation and paternal care (van der Heide et al., 2012) masked or reduced energetic challenges that would otherwise increase fecal cortisol levels in our population (Janson & Verdolin, 2005). This seasonality may also help to explain why the highest mean fecal cortisol levels of owl monkey females occurred during gestation, rather than during lactation when energetic demands may be even higher for primates (Butte & King, 2005; Dufour & Sauther, 2002).

Seasonality in the abundance of fruits or other foods may also help to explain why males’ cortisol levels were relatively high during gestation, since gestation always takes place during winter at our study site. The reason why males’ cortisol levels were somewhat higher during periods when they were caring for older infants (>3 months old) compared to younger infants, remains somewhat puzzling. The period when males are typically caring for older infants corresponds roughly to the summer/early fall, so it is possible that the increased fecal cortisol concentrations observed in this period could be attributed in part to a change in the physical environment (e.g. thermal stress or a relatively increased threat from predators due to thicker foliage) or the social environment (e.g. increased risk of attempted takeovers from conspecifics). We might expect that males would be under a greater threat of challenges from other males during the period when females are most likely to become pregnant (which overlaps with the time of year when older infants are present). However, both males and females are equally susceptible to being replaced by floaters, and expulsions can occur at any time throughout the year (Fernandez-Duque & Huck, 2013). Overall, since reproduction is highly seasonal in our population, an experimental approach using captive animals would be needed to better disentangle the potential effects of the physical and social environment from the effects of reproductive phases and paternal care.

It is also necessary to note that fecal cortisol levels can be influenced by a variety of factors not directly linked to circulating levels of cortisol. For example, the quantity and motility of gut contents can influence the amount of cortisol deposited in feces (Hau, Kalliokoski, Jacobsen, & Abelson, 2011). We did not characterize details of the diet, or how seasonal dietary fluctuations may have altered volume or patterns of excretion, and thus the hormone content of fecal samples. Overall, as in practically all other studies of wild primates, many of the factors that can influence cortisol concentrations in fecal samples were unknown to us. These limitations could potentially have obscured certain fluctuations in cortisol that may have occurred in our subjects.

It is important to acknowledge that cortisol is just one of many hormones that may help regulate metabolic challenges associated with reproduction in primates. There is evidence that prolactin, in particular, may play an important role in mediating the energetic challenge posed by infant carrying in marmoset fathers (Ziegler, Prudom, Zahed, Parlow, & Wegner, 2009). As in many studies of wild primates, we limited our examination to cortisol not because we consider other hormones unlikely to be important, but primarily because it was not possible to quantify them in the non-invasively collected fecal samples available to us. The patterns of cortisol fluctuation that we observed provide a basis for future studies to further explore the energetics of biparental care by utilizing owl monkeys in captivity to examine circulating concentrations of prolactin, insulin (or its metabolite C-peptide), and other biomarkers.

There has been a tendency in the human and non-human primate literature for elevated cortisol to be treated as synonymous with “stress”. There are two major problems with this. First, rather than being determined by, or strictly correlated, with increased cortisol levels, stress is an emergent process that depends on interactions between various physiological factors and the physical and social environment (Epel et al., 2018). Second, cortisol does much more than contribute to chronic stress and the negative consequences associated with it; cortisol has positive, adaptive effects and carries out essential functions (McEwen, 2019; Sapolsky, 2015). Thus, finding that a handful of samples from an individual have elevated cortisol levels could be indicative of a range of situations, and should not necessarily be used to conclude that an individual is experiencing chronic stress. We encourage others studying wild primates to consider alternative explanations for why cortisol may be elevated; findings based on concentrations in fecal samples in an observational study ought to be considered as a necessary first step for generating more refined hypotheses and models rather than conclusive evidence that an individual is chronically stressed rather than adaptively responding to an energetic challenge (Taborsky 2008). Support for these hypotheses can then be further evaluated by determining cortisol levels in other media and incorporating additional methods that allow for energy expenditure to be measured [e.g., doubly-labelled water (Speakman, 1998)].

In our study, we collected multiple samples from the same individual in each reproductive period; thus providing evidence that mean fecal cortisol levels are higher during gestation than in other periods. Future studies should continue to utilize this longitudinal approach to collect multiple samples across periods of gestation and parental care to characterize how other indicators of energy mobilization are associated with biparental care in males and females. Studies of captive owl monkeys have the potential to substantially increase our understanding of energy expenditure in species with biparental care by allowing researchers to measure cortisol and other biomarkers in plasma or saliva and monitor and control for dietary intake. While observational studies of wild primates have their limitations, integrating data from studies like ours with results from captive individuals will allow researchers to better understand the physiological mechanisms underlying biparental care (Fernandez-Duque, 2012).